The Science and Applications of Synthetic and Systems Biology: Workshop Summary (2011)

Appendix A

Contributed Manuscripts

A1

COMMERCIAL APPLICATIONS OF SYNTHETIC BIOLOGY

David A. Berry^1,2

An Overview of Venture Capital

Venture capital is financial capital invested into high-potential companies. The role of venture capital is to support the entrepreneurial talent that takes basic science and breakthrough ideas to market by building companies. This risk capital ultimately supports some of the most innovative and promising companies—those that have gone on to change existing industries or create new ones altogether (Thompson Reuters, 2011).

Venture capital is a distinct asset class. Venture capital firms, which are professional, institutional capital managers, make investments by purchasing equity in a company. The stock acquired is an illiquid investment that requires the growth of the company for the investors to ultimately reap any potential return. It is this inability of venture capitalists to rapidly enter and exit investments, or “flip” them, that aligns their goals with those of the entrepreneurs. Venture capital is intrinsically a long-term investment (Thompson Reuters, 2011).

_______________

¹ Affiliation: Flagship VentureLabs, Cambridge, MA, USA.

² Key words: venture capital, biological engineering, synthetic biology, microbiome, diesel, photosynthesis, genome engineering.

Venture capitalists invest out of a fund, a vehicle that deploys capital on behalf of third-party investors. The investors in these funds, called limited partners, are often pension funds, foundations, corporations, endowments, and wealthy individuals, among others. Given the low liquidity associated with their investment into venture capital funds, limited partners expect large returns—better than those in the stock market—from the funds in which they invest. The funds represent a commitment of capital with a fixed life, typically 10 years. The general partner, a group of partners with fiduciary responsibility for the firm with the legal form of a partnership, manages the capital in the fund. The committed capital is called by the general partner from the limited partners to make a portfolio of investments. Ultimately, when investments mature and become liquid, the profits are shared, with the majority going back to the limited partners and the rest shared by the general partner.

Funding provided by venture capitalists typically takes the form of “rounds,” where a given amount of money is invested into a company at a valuation agreed upon between the management and the investors. Prior to an investment, the equity ownership is divided among the founders, management, and others. The valuation sets a share price against which the venture capital firm buys shares. At each round, the earlier investors and management team strive to increase the valuation for the subsequent round(s) of investment. The higher the valuation of a round, the less dilution (reduction in ownership) the existing shareholders take. While each round contemplates a share price that defines a paper value for an investor’s or an employee’s shares, little actual value is created. Only at a sale event or initial public offering do investors and the management team see a tangible financial return, which can take 5-8 years, if not longer.

Venture capital firms statistically see 100 business plans, take a deep look at 10 of these proposals, and invest in one. This process involves an assessment of the management team, the proposed business, its potential to exclude competition, the market being pursued, and how well the opportunity fits with the firm’s goals.

With an investment, a partner will typically get involved with a company by taking a seat on the board of directors, where he or she works closely with the management team on company strategy and growth. The venture capital industry plays an important role in the economy. Companies supported by early venture capital account for 21 percent of the U.S. gross domestic product by revenue, and 11 percent of private-sector jobs despite the fact that fewer than 1,000 new businesses get venture capital funding any given year (National Venture Capital Association, 2009).

A Brief History of the Venture Capital Industry

Venture capital is said to have originated in 1946 with the founding of the first two firms: American Research and Development Corporation (ARDC) and

J. H. Whitney & Company (Wilson, 1985). Georges Doriot, referred to as the “father of venture capitalism,” the former dean of Harvard Business School and founder of INSEAD, founded ARDC along with Karl Compton, the former president of MIT, as well as Ralph Flanders (Ante, 2008). ARDC sought to invest in businesses run by soldiers returning from World War II. The firm is most famous for investing $70,000 in Digital Equipment Corporation in 1957—a company that when it had its initial public offering in 1968 was valued at $355 million for a return to ARDC of over 1,200-fold. Employees of ARDC went on to found leading venture funds including Greylock Partners and Flagship Ventures, among others (Kirsner, 2008).

Two major government changes allowed venture capital to emerge as a fully fledged industry. First, the Small Business Investment Act of 1958 enabled the Small Business Administration to license Small Business Investment Companies to help finance and manage small entrepreneurial businesses. Second, in 1978, the Employee Retirement Income Security Act was altered to allow corporate pension funds to invest in venture capital. These two acts together supported the framework for venture capital and facilitated substantial investment in it.

Successes of the venture capital industry in the 1970s and 1980s, with companies including Digital Equipment Corporation, Apple, and Genentech resulting, led to rapid growth of the industry. With rapid growth came diminished returns. In the early 1990s the numbers of firms and managers shrank in response to declining investment performance. At the same time, the more successful firms retrenched, starting a wave of increased returns that began in 1995 and continued through the Internet bubble in 2000 (Metrick, 2007). Once again, with grossly increased returns, the investment into the sector and the number of funds skyrocketed. Beginning in March 2000, the NASDAQ crashed, and many funds suffered from a second contraction.

After the Internet bubble, the funds raised by venture firms shrank substantially. Amounts of committed capital increased through 2005 to a level much less than in 2000, and they remained flat until the economic meltdown in 2008. During the decade from 2000 to 2010, venture capital returns also fell dramatically to the point that the median 10-year return of all U.S. funds was less than the stock market (Thomson Reuters, 2011). These events resulted in another substantial contraction in the industry. The industry is currently responding to this most recent contraction. The number of funds has decreased as the average fund size has risen. This dynamic has caused venture funds to focus on either earlier-stage investments, later-stage investments (similar to private equity), or a combination. Other funds have started focusing on flipping assets by investing before or to induce specific value-creation events. This has created a new environment where only a small number of funds are focused on the earliest stage—that which venture capital is most associated with and most successful at—with several others focused on a more transactional business. This evolution is still in process, but it has been changing the nature of companies that receive investment as well.

At the earliest stage of investments, venture capitalists have returned to investing in outstanding teams and under the assumption that they can create great companies. A number of approaches have been taken to inspire innovation and support a new era of breakthrough companies. Various firms have taken different approaches. CMEA, for example, invests in proven entrepreneurs “prenapkin” (before the idea), on the belief that they will come up with ideas. Polaris’ Dogpatch Labs has created an environment where multiple entrepreneurs share a common environment and with light money attempt to prove out their concepts. Y Combinator gives entrepreneurs an education and a small amount of money to try out their ideas. Andreessen Horowitz has similarly created an infrastructure to support the earliest stages of companies and to allow them to focus their capital on the company. “Super Angels” such as Peter Thiel have also emerged to provide important early stage funding. Several companies produced from some of these efforts have emerged as important venture-backed companies. Flagship VentureLabs has created an internal infrastructure of serial entrepreneurs to co-iterate its own innovations and use that as a basis to build companies.

Flagship VentureLabs

Flagship VentureLabs was built with a focus on increasing the efficiencies of innovation and entrepreneurship. In the broadest context, both traditional entrepreneurship and venture capital have intrinsic benefits and inefficiencies. Entrepreneurs, for example, typically perform well when capital constrained, but, by the same token, avoid asking critical questions because if an undesirable answer results, they are unemployed. Venture capital has the advantage of large sample sizes and substantial funding, but it is limited in its investments to only those which it can see, and all of its investments must fundamentally go through a common set of efforts (i.e., financial infrastructure, legal, etc.). Fundamentally, Flagship VentureLabs removes the constraints from the typical elements of traditional ecosystems; that is, by harnessing the key constituents and requirements all under one roof, with the common goal of the betterment of humankind through innovation and entrepreneurship.

The focus of Flagship VentureLabs is to develop breakthrough technologies to match large unmet needs in life sciences and sustainability through the vehicle of startup companies. New companies come from a breakthrough innovation without a set utility or from work within Flagship VentureLabs identifying the intersection between the potential for technology solution and market pull. In the former case, a team is nucleated around the technology, including the inventor, to heavily iterate the concept and pressure test it against markets, intellectual property opportunities, team-building potential, and other features with the attempt at nonrationally identifying the “sweet spot.” In certain cases, this process results in the pseudolinear formation of a company focused on commercializing the technology. In others, however, through a progressive set of explorations,

the company may end up far from its origins, potentially not including the base technology. In the latter case, the defined intersection creates a hypothesis. If the hypothesis has already been manifest by others in a company or in academia (either singularly or through a combination of efforts), a simple investment may be warranted. In the absence of such a proof point, the concept then goes through heavy conceptual iteration with the attempt to prove the hypothesis wrong, and in the combination of not being able to make it fail and the generation of significant key stakeholder support (including industry and key opinion leaders), a company will be launched. Ultimately, this approach results in taking new ideas and forming companies several years before such an opportunity is likely to be compelling. The following discusses efforts in three such technology-based companies originating from Flagship VentureLabs covering both life sciences and sustainability.

Seres Therapeutics: Rethinking Drug Development

In an effort to reduce side effects, drug development has focused its efforts on target specificity, particularly on features including affinity, low off-target effects, pharmacokinetics, pharmacodynamics, and others. The Human Genome Project and systematic understandings of the functions of kinases have helped to drive this increasing target specificity. Nonetheless, the biology of diseases is complex and multi-factorial. Focusing drugs to single actors may reduce side effects but it also limits the spectrum of efficacy. The growing recognition of the nature of disease is driving the understanding of more complex biology and the development of drugs focused on the multitude of key factors.

One particular example is with the human microbiome. Microorganisms have long been thought of as independently functioning pathogens. Recently, however, the commensal and mutualistic natures of various microorganisms that inhabit the body have started to be characterized (Dethlefsen, 2011). The interactions between the multitude of organisms, as well as between the organisms and the host, play an important role in normal physiology broadly (Reid et al., 2011). Accordingly, disruptions in the microbiome, whether by antibiotics, diet, infection, or other means, can alter the microbiome and induce or simply increase the likelihood of a wide range of diseases, ranging from Clostridium difficile infection and inflammatory bowel disease to obesity and diabetes (Kau et al., 2011).

The complexity of the microbiome, including not only the interrelation between a number of species and the host but also the physical formation of the communities in specific niches (Rickard et al., 2003), is important to take into account when developing therapeutics aimed at diseases where the microbiome plays an important role. Seres Therapeutics was founded specifically to develop drugs based on the complexity of the microbiome. Probiotics and single biologics affect a limited scope of disease and, thus, have limited efficacy in complex diseases such as those involving the microbiome (Shen et al., 2009). By creat-

ing synthetic microbiomes aimed at disrupting pathogenic communities, Seres provides a therapeutics means by which a normal microbiome can be restored. Understanding biology and synthetically recapitulating conditions that can recover from a disease-associated insult enables a new class of therapeutics to be designed and developed that are focused on the etiology of underlying disease.

Sustainability

Persistently high fossil fuel prices, increasing dependency on foreign fuel supplies, and insecurity relating to the sources of petroleum have created substantial market pull for alternative solutions in the $6 trillion petrochemical industry. Outside of government-mandated markets, such as ethanol of late, fuels and chemicals are fungible products driven by price and purity, as well as supply and demand. Markets therefore require products with a known utility that meet certain industrial specifications while doing so at a competitive cost point. Consumers have not shown a willingness to pay for benefits such as greenhouse gas mitigation or domestic sourcing, so products made as alternatives must do so while competing head-to-head with the incumbents using the same metrics.

Fuels have traditionally originated from biology in some form or another. Fossil fuels are thought to ultimately derive from processing of ancient biomass through a process that takes millions to hundreds of millions of years. Historically, humans have also found faster cycle time sources of energy, namely the burning of trees for heat and energy, as well as the removal of spermaceti from whales as a source of wax. All of these resources have limited renewal potential and substantial environmental impact (Tertzakian, 2009). Given the central role biology has had in fossil fuels historically, it stands to reason that biology would be well positioned to be at the forefront of the future of sustainable fuels.

Biological engineering has evolved rapidly over the last 50-plus years. Breakthroughs in genomics research, increased genetic manipulation potential, and more complete knowledge of the inner workings of cells have set a stage for cells to be engineered to achieve desired functionalities. Moreover, the time from conception to proof of concept, and that from proof of concept to commercial viability, has reduced substantially. Historically, these periods have decreased threefold every 10 years. Given the technological potential enabled, the market needs can now drive the technological direction, thus leading toward an intersection between market pull and the potential for technology solution.

LS9: Ultraclean Renewable Diesel

In 2005, the U.S. government had built a robust market demand for ethanol by outlining a replacement timeline for methyl tert-butyl ether (MTBE), a fuel oxygenate that had been associated with groundwater contamination and potential increased cancer risks, with ethanol (U.S. Environmental Protection Agency,

2011a). Twelve billion gallons were mandated by 2012, effectively defining a market growth. This mandate was soon supplemented with the renewable fuel standard (RFS), and subsequently RFS2, ultimately requiring 36 billion gallons produced per year by 2022 (U.S. Environmental Protection Agency, 2011b). Corn ethanol was thus given ample runway to launch, and blenders were incorporating biologically based products into fuel nationwide. The intent of the MTBE replacement with ethanol, however, was replacing an oxygenate, not deeming ethanol a fuel. Nonetheless, outspoken investors were enthusiastically supportive of building the future of American renewable fuel on ethanol, asserting that it could be cheaper than and as efficient as petrochemically derived fuels (Khosla, 2006) despite the disadvantaged domestic cost structure and intrinsic lower energy density. The market was becoming well positioned for a viable alternative.

LS9 originated by asking the question, “If you could make any fuel from biology, what would you make?” The ideal fuel to be produced from biology would be diesel, given its high energy density and its use throughout the world as a primary transportation fuel. A market-acceptable biologically produced diesel must compete in a low-cost commodity market without subsidies, requiring an efficient biological pathway and process. Translating these needs into specific technological tasks required that the cell be engineerable to be feedstock agnostic (i.e., able to use any form of sugar), that the most efficient pathway of producing the product was available, that the product was to be made directly and secreted, and that it entailed both straightforward separations (a feature of the product) and no downstream processing (the final product is made by the cell).

Using the defined market constraints, various pathways to produce a straight-chain hydrocarbon were defined and evaluated. The fatty acid biosynthesis pathway not only has exceptionally high energy efficiency at over 90 percent but also produces a molecule that is chemically identical to diesel, requiring potentially fewer biological steps. Fatty acids are activated with coenzyme A (CoA) or acyl carrier protein (ACP) to make fatty acyl-CoA or fatty acyl-ACP (Zhang and Rock, 2008), which serve as the biological precursor products for fuel synthesis. These products are then modified to make a desired end product. The same pathways can be leveraged to make a series of other petrochemicals including fatty acid methyl esters, olefins, fatty alcohols, and others, in addition to alkanes (diesel) (Rude et al., 2011).

The product itself is insufficient for a commercial host and process. The identified market constraints require that the cell chassis has flexibility in feedstock, be optimized to maximize carbon flux to end product, and secrete the end product. Feedstock costs, driven by sugar prices, have risen dramatically over the past 6 years. Alternatives require the liberation of sugar from cellulosic biomass, which is done through exogenous enzymes at present. The expression of hemicellulases into the host already engineered to produce alkanes or other derivatives can enable consolidated bioprocessing, thereby reducing process costs (Magnuson et al., 1993). This is achievable, for example, with the endogenous production

of glycosyl hydrolases such as xylanase (Xsa) from Bacteroides ovatus and an endoxylanase catalytic domain (Xyn10B) from Clostridium stercorarium, which together hydrolyze hemicelluloses to xylose, which is usable in E. coli central metabolism (Adelsberger et al., 2004; Steen et al., 2010; Whitehead and Hespell, 1990). Optimizing the host requires focusing the flux of the sugar input through central metabolism to the product. Specifically, fadD and fadE knockouts block the first two steps of the ß-oxidation pathway, increasing end-product production three- to fourfold (Steen et al., 2010). Secreting the end product eliminates end-product inhibition and streamlines bioprocessing, thus increasing flux and reducing operating costs by supporting continuous operations (Berry, 2010). Expressing a leaderless version of TesA eliminates end-product inhibition, drives secretion, and notably also positively affects chain length with a natural preference for C14 fatty acids (Cho and Cronan, 1995; Jiang and Cronan, 1994; Steen et al., 2010).

Through this approach, an industrial chassis has been rationally developed to systematically meet commercial needs. By specifically including features necessary to ensure diverse feedstock utility, drop-in product synthesis, and lowest cost processing, LS9’s technology has been designed specifically to drive market pull.

Joule Unlimited: Renewable Solar Fuels

An intrinsic challenge of using a sugar-based feedstock is the price volatility associated with the commodity. Joule Unlimited was founded to develop a platform that could eliminate dependence on sugar feedstocks while still producing fuels in a way that meets market needs. A systematic exploration of sources of carbon that can be routed into central metabolism rapidly identifies photosynthesis, nature’s solution to carbon dioxide assimilation driven by solar energy, as a compelling, though insufficient, pathway. The Department of Energy’s Aquatic Species program, based on explorations of algal biofuels between 1976 and 1996, concluded that photosynthesis could support viable fuel processes, but it requires a set of key innovations to do so (Sheehan et al., 1998; Weyer et al., 2010; Zhu et al., 2008). At the outset of Joule Unlimited, the fundamental limitations of algal fuels were examined and coupled with market needs to design an entirely new and distinct approach, whose only similarity to algae was the use of photosynthesis.

A thorough exploration of market needs identified that an ideal solution would directly produce secreted fungible fuel directly from sunlight and carbon dioxide without a dependency on arable land, freshwater, or other costly reagents, while having a cost that could meet or beat fossil fuel equivalents in the absence of subsidies and at the same time scale modularly such that smaller-scale plants could be used to validate large-scale deployments. The simultaneous technological solution to meet all of these needs demands a genetically tractable cyanobacteria engineered to not need exogenous factors and to produce secreted fuel grown in a

modular bioreactor leveraging the two-dimensional scaling of light and incorporating fundamental process needs including proper mixing. A schematic of the Joule Unlimited approach, Helioculture™, is provided in Figure A1-1.

Cyanobacteria had previously been engineered to express recombinant proteins, but it had not been systematically engineered on a genome scale owing primarily to a lack of engineering tools (Alvey et al., 2011). A concerted effort using E. coli engineering over 50 years serves as a roadmap of the needs for genome engineering in cyanobacteria. Using these tools coupled with a systemic genome engineering effort allows one to overcome a theoretical maximum light use and net productivity for algal biofuels of ~2,000 gallons per acre per year by enabling photoautotrophs, for the first time, to function like industrialized heterotrophs, whose phases of growth and production are separated (Berry, 2010; Robertson et al., 2011). Systematic re-regulation of central metabolism directs 95 percent of photosynthetic activity to specific product synthesis versus up to 50 percent in un-engineered organisms, allowing for productivity ~95 percent of the light-exposed time through continuous operations versus substantial down time with batch processes, minimizing maintenance energy requirements, and limiting the energy lost to photorespiration.

At the same time, the engineered cells are grown in a reactor system designed to be low cost and linearly scalable. Low-cost product separation, cell mixing, and proper gas transfer to the cells are all incorporated into the reactor design. Coupled together, this systems approach allows for ~12 percent theoretical maximal photonic energy conversion versus 1.5 percent for traditional algal processes (Figure A1-2), which translates to unprecedented areal productivities of 25,000 gallons of ethanol per acre per year or 15,000 gallons of diesel per acre per year.

The high productivities achieved through the Joule Unlimited approach allow for cost points as low as $20/barrel for fungible diesel. Maximizing solar energy capture, carbon dioxide fixation as a replacement for sugar use, and organism productivity creates a system that can be market competitive while providing for the environmental benefits that have been sought in fossil fuel replacements. Specifically, Joule Unlimited’s technology eliminates the need for arable land, requires no freshwater, and reduces life-cycle greenhouse gases by over 90 percent by using carbon dioxide as a feedstock. By coupling the technical needs for a total solution with a market need, Joule Unlimited has uniquely developed a process that can produce a sugar-independent diesel in a highly scalable manner, overcoming the challenges of alternative approaches.

Conclusions

Systematic changes in venture capital have altered the entrepreneurial ecosystem. Flagship VentureLabs is pioneering a new approach of technology development through companies by building technologies that specifically address the intersection between the potential for technology solution with market pull

FIGURE A1-1 Schematic of the Joule Unlimited Helioculture systems approach. Specific cyanobacteria are engineered to convert sunlight, CO₂, and nonfreshwater to diesel (or other fuels and chemicals). The engineered organisms are housed within a Solar Converter, a single reactor unit designed to interconnect with others and therefore scale linearly.

FIGURE A1-2 A summation of the accumulated photon losses for algal and direct fuel processes, as well as a theoretical maximum photonic energy conversion. The losses are shown through individual contributions accumulated serially and illustrated on a logarithmic scale, beginning with the percent of photosynthetically active radiation empirically measured at the ground. Total energy conversion efficiency as a function of the losses is indicated by the green arrows.
SOURCE: Adapted from Robertson et al. (2011).

driving invention and innovation toward market needs. The resultant companies are designed by exploring cutting-edge capabilities and iterating against market needs—not just from an evolutionary standpoint, but additionally identifying the true needs of an industry across multiple facets. Synthetic biology is a new and rapidly developing tool that has particular utility in meeting broad-based and distinct market needs, particularly through its ability to create functional modules in a cell-based system. By leveraging the potential of synthetic biology with market-driven needs, Flagship VentureLabs has been able to spearhead a set of breakthrough innovations in both life sciences and sustainability. This approach now takes market potential before traditional research approaches have made for a compelling and investable technology-driven opportunity and, through heavy iteration, can bring it to bear ahead of time through broad-based collaborations with industry and academia. This approach can be broadly leveraged to develop

a series of future breakthrough technologies in a variety of important market sectors.

References

Adelsberger, H., C. Hertel, E. Glawischnig, V. V. Zverlov, and W. H. Schwarz. 2004. Enzyme system of Clostrium stercorarium for hydrolysis of arabinoxylan: Reconstitution of the in vivo system from recombinant synzmes. Microbiology 150:2257-2266.

Alvey, R. M., A. Biswas, W. M. Schluchter, and D. A. Bryant. 2011. Attachment of noncognate chromophores to CpcA of Synechocystis sp. PCC 6803 and Synechococcus sp. PCC 7002 by heterologous expression in Escherichia coli. Biochemistry 50:4890-4902.

Ante, S. E. 2008. The Birth of Venture Capital. Cambridge, MA: Harvard Business School Press.

Berry, D. A. 2010. Engineering organisms for industrial fuel production. Bioengineered Bugs 1: 303-308.

Cho, H., and J. E. Cronan, Jr. 1995. Defective export of a periplasmic enzyme disrupts regulation of fatty acid synthesis. World Journal of Biological Chemistry 270:4216-4219.

Dethlefsen, L., M. McFall-Ngai, and D. A. Relman. 2011. An ecological and evolutionary perspective on human-microbe mutualism and disease. Nature 449:811-818.

Jiang, P., and J. E. Cronan, Jr. 1994. Inhibition of fatty acid synthesis in Escherichia coli in the absence of phospholipid synthesis and release of inhibition by thioesterase action. Journal of Bacterialogy 176:2814-2821.

Kau, A. L., P. P. Ahern, N. W. Griffen, A. L. Goodman, and J. I. Gordon. 2011. Human nutrition, the gut microbiome and the immune system. Nature 474:327-336.

Khosla, V. 2006. Is ethanol controversial? Should it be? http://www.khoslaventures.com/khosla/papers.html (accessed June 8, 2011).

Kirsner, S. 2008. Venture capital’s grandfather. The Boston Globe, April 6.

Magnuson, K., S. Jackowski, C. O. Rock, and J. E. Cronan, Jr. 1993. Regulation of fatty acid biosynthesis in Escherichia coli. Microbiology Reviews 57:522-542.

Metrick, A. 2007. Venture Capital and the Finance of Innovation. New York: John Wiley & Sons.

National Venture Capital Association. 2009. Venture Impact: The Economic Importance of Venture Backed Companies to the U.S. Economy. National Venture Capital Association: Arlington, VA.

Reid, G., J. A. Younes, H. C. Van der Mei, G. B. Gloor, R. Knight, and H. J. Busscher. 2011. Micro-biota restoration: Natural and supplemented recovery of human microbial communities. Nature Reviews Microbiology 9:27-38.

Rickard, A. H., P. Gilbert, N. J. High, P. E. Kolenbrander, and P. S. Handley. 2003. Bacterial coaggregation: An integral process in the development of multi-species biofilms. Trends in Microbiology 11:94-100.

Robertson, D. E., S. Jacobson, F. Morgan, D. Berry, G. M. Church, and N. B. Afeyan. 2011. A new dawn for industrial photosynthesis: An evaluation of photosynthetic efficiencies for production of liquid fuels. Photosynthesis Research 107:269-277.

Rude, M. A., T. S. Baron, S. Brubaker, M. Alibhai, S. B. Del Cardayre, and A. Schirmer. 2011. Terminal olefin (1-alkene) biosynthesis by a novel p450 fatty acid decarboxylase from Jeotgalicoccus species. Applied and Environmental Microbiology 77:1718-1727.

Sheehan, J., T. Dunahay, J. Benemann, and P. A. Roessler. 1998. A Look Back at the U.S. Department of Energy’s Aquatic Species Program: Biodiesel from Algae. U.S. Department of Energy, Office of Fuels Development: Closeout Report TP-580-241-24190.

Shen, J., H. Z. Ran, M. H. Yin, T. Z. Zhou, and D. S. Xiao. 2009. Meta-analysis: The effect and adverse events of lactobacilli versus placebo in maintenance therapy for Crohn disease. Internal Medical Journal 39:103-109.

Steen, E. J., Y. Kang, G. Bokinsky, Z. Hu, A. Schirmer, A. McClure, S. B. del Cardayre, and J. D. Keasling. 2010. Microbial production of fatty-acid-derived fuels and chemicals from plant biomass. Nature 463:559-563.

Tertzakian, P. 2009. The End of Energy Obesity. New York: John Wiley & Sons.

Thomson Reuters. 2011. 2011 National Venture Capital Association Yearbook. New York: Thomson Reuters.

U.S. Environmental Protection Agency. 2011a. Methyl Tertiary Butyl Ether (MTBE). http://www.epa.gov/mtbe/faq.htm#concerns.

——. 2011b. Fuels and Fuel Additives. http://www.epa.gov/otaq/fuels/renewablefuels/index.htm

Weyer, K. M., D. R. Bush, A. Darzins, and B. D. Wilson. 2010. Theoretical maximum algal oil production. Bioenergy Research 3:204-213.

Whitehead, T. R., and R. B. Hespell. 1990. The genes for three xylan-degrading activities from Bacteroides ovatus are clustered in a 3.8-kilobase region. Journal of Bacteriology 172:2408-2412.

Wilson, J. W. 1985. The New Venturers: Inside the High-Stakes World of Venture Capital. Boston: Addison-Wesley.

Zhang, Y. M., and C. O. Rock. 2008. Membrane lipid homeostasis in bacteria. Nature Reviews Microbiology 6:222-233.

Zhu, X.-G., S. P. Long, and D. R. Ort. 2008. What is the maximum efficiency with which photosynthesis can convert solar energy into biomass? Current Opinion in Biotechnology 19:153-159.

A2

SYNTHETIC BIOLOGY: APPLICATIONS COME OF AGE³

Ahmad S. Khalil⁴and James J. Collins^4,5

Abstract

Synthetic biology is bringing together engineers and biologists to design and build novel biomolecular components, networks and pathways, and to use these constructs to rewire and reprogram organisms. These re-engineered organisms will change our lives over the coming years, leading to cheaper drugs, ‘green’ means to fuel our cars and targeted therapies for attacking ‘superbugs’ and diseases, such as cancer. The de novo engineering of genetic circuits, biological modules and synthetic pathways is beginning to address these crucial problems and is being used in related practical applications.

_______________

³ Originally printed as Khalil, A. S., and J. J. Collins. 2010. Synthetic biology: Applications come of age. Nature Reviews Genetics 11:367-379. Reprinted with kind permission from Nature Publishing Group.

⁴ Howard Hughes Medical Institute, Department of Biomedical Engineering, Center for BioDynamics and Center for Advanced Biotechnology, Boston University, Boston, Massachusetts 02215, USA.

⁵ Wyss Institute for Biologically Inspired Engineering, Harvard University, Boston, Massachusetts 02115, USA. Correspondence to J.J.C. e-mail: jcollins@bu.edu. doi:10.1038/nrg2775.

The circuit-like connectivity of biological parts and their ability to collectively process logical operations was first appreciated nearly 50 years ago (Monod and Jacob, 1961). This inspired attempts to describe biological regulation schemes with mathematical models (Glass and Kauffman, 1973; Savageau, 1974; Kauffman, 1974; Glass, 1975) and to apply electrical circuit analogies to biological pathways (McAdams and Arkin, 2000; McAdams and Shapiro, 1995). Meanwhile, breakthroughs in genomic research and genetic engineering (for example, recombinant DNA technology) were supplying the inventory and methods necessary to physically construct and assemble biomolecular parts. As a result, synthetic biology was born with the broad goal of engineering or ‘wiring’ biological circuitry — be it genetic, protein, viral, pathway or genomic — for manifesting logical forms of cellular control. Synthetic biology, equipped with the engineering-driven approaches of modularization, rationalization and modelling, has progressed rapidly and generated an ever-increasing suite of genetic devices and biological modules.

The successful design and construction of the first synthetic gene networks — the genetic toggle switch (Gardner et al., 2000) and the repressilator (Elowitz and Leibler, 2000) (Box A2-1) — showed that engineering-based methodology could indeed be used to build sophisticated, computing-like behaviour into biological systems. In these two cases, basic transcriptional regulatory elements were designed and assembled to realize the biological equivalents of electronic memory storage and timekeeping (Box A2-1). Within the framework provided by these two synthetic systems, biological circuits can be built from smaller, well-defined parts according to model blueprints. They can then be studied and tested in isolation, and their behaviour can be evaluated against model predictions of the system dynamics. This methodology has been applied to the synthetic construction of additional genetic switches (Gardner et al., 2000; Atkinson et al., 2003; Bayer and Smolke, 2005; Deans et al., 2007; Dueber et al., 2003; Friedland et al., 2009; Ham et al., 2006, 2008; Kramer and Fussenegger, 2005; Kramer et al., 2004), memory elements⁶ (Gardner et al., 2000; Friedland et al., 2009; Ham et al., 2006; Ajo-Franklin et al., 2005) and oscillators (Elowitz and Leibler, 2000; Atkinson et al., 2003; Fung et al., 2005; Stricker et al., 2008, Tigges et al., 2009; Danino et al., 2010), as well as to other electronics-inspired genetic devices, including pulse generators⁷ (Basu et al., 2004), digital logic gates⁸ (Anderson et al., 2007; Guet et al., 2002; Rackham and Chin, 2005; Rinaudo et al., 2007;

_______________

⁶ Memory elements – Devices used to store information about the current state of a system.

⁷ Pulse generators – Circuits or devices used to generate pulses. A biological pulse generator has been implemented in a multicellular bacterial system, in which receiver cells respond to a chemical signal with a transient burst of gene expression, the amplitude and duration of which depend on the distance from the sender cells.

⁸ Digital logic gates – A digital logic gate implements Boolean logic (such as AND, OR or NOT) on one or more logic inputs to produce a single logic output. Electronic logic gates are implemented using diodes and transistors and operate on input voltages or currents, whereas biological logic gates operate on cellular molecules (chemical or biological).

Stojanovic and Stefanovic, 2003; Win and Smolke, 2008), filters⁹ (Basu et al., 2005; Hooshangi et al., 2005; Sohka et al., 2009) and communication modules (Danino et al., 2010; Basu et al., 2005; Kobayashi et al., 2004; You et al., 2004).

Now, 10 years after the demonstration of synthetic biology’s inaugural devices (Gardner et al., 2000; Elowitz and Leibler, 2000), engineered biomolecular networks are beginning to move into the application stage and yield solutions to many complex societal problems. Although work remains to be done on elucidating biological design principles (Mukherji and van Oudenaarden, 2009), this foray into practical applications signals an exciting coming-of-age time for the field.

Here, we review the practical applications of synthetic biology in biosensing, therapeutics and the production of biofuels, pharmaceuticals and novel biomaterials. Many of the examples herein do not fit exclusively or neatly into only one of these three application categories; however, it is precisely this multivalent applicability that makes synthetic biology platforms so powerful and promising.

Biosensing

Cells have evolved a myriad of regulatory circuits — from transcriptional to post-translational — for sensing and responding to diverse and transient environmental signals. These circuits consist of exquisitely tailored sensitive elements that bind analytes and set signal-detection thresholds, and transducer modules that filter the signals and mobilize a cellular response (Box A2-2). The two basic sensing modules must be delicately balanced: this is achieved by programming modularity and specificity into biosensing circuits at the transcriptional, translational and post-translational levels, as described below.

Transcriptional biosensing. As the first dedicated phase of gene expression, transcription serves as one method by which cells mobilize a cellular response to an environmental perturbation. As such, the genes to be expressed, their promoters, rNA polymerase, transcription factors and other parts of the transcription machinery all serve as potential engineering components for transcriptional biosensors. Most synthetic designs have focused on the promoters and their associated transcription factors, given the abundance of known and characterized bacterial, archaeal and eukaryotic environment-responsive promoters, which include the well-known promoters of the Escherichia coli lac, tet and ara operons.

Both the sensory and transducer behaviours of a biosensor can be placed under synthetic control by directly engineering environment-responsive promoter sequences. In fact, this was the early design strategy adopted for establishing inducible expression systems (Brown et al., 1987; Deuschle et al., 1989; Hu and

_______________

⁹ Filters – Algorithms or devices for removing or enhancing parts or frequency components from a signal.

Davidson, 1987; Lutz and Bujard, 1997). By introducing, removing or modifying activator and repressor sites, a promoter’s sensitivity to a molecule can be tuned. Synthetic mammalian transactivation systems are generic versions of this strategy in which an environmentally sensitive transcription factor is fused to a mammalian transactivation domain to cause inducer-dependent changes in gene expression. Synthetic mammalian biosensors based on this scheme have been created for sensing signals such as antibiotics (Fussenegger et al., 2000; Gossen and Bujard, 1992; Weber et al., 2002), quorum-sensing molecules (Neddermann et al., 2003; Weber et al., 2003b), gases and metabolites (Malphettes et al., 2005; Mullick et al., 2006; Weber et al., 2006; Weber et al., 2004), and temperature changes (Boorsma et al., 2000; Weber et al., 2003a). Fussenegger and colleagues have even incorporated this transgene design into mammalian circuits, creating synthetic networks that are responsive to electrical signals (Weber et al., 2009).

Although the engineering of environment-responsive promoters has been valuable, additional control over modularity¹⁰ and specificity can be achieved by embedding environment-responsive promoters¹¹ in engineered gene networks. Achieving true modularity with genetic parts is inherently difficult because of unintended interference among native and synthetic parts and therefore requires careful decoupling of functional modules. One such modular design strategy was used by Kobayashi et al. (2004) to develop whole-cell E. coli biosensors that respond to signals in a programmable fashion. In this design, a sensory module (that is, an environment-responsive promoter and associated transcription factor) was coupled to an engineered gene circuit that functions like a central processing unit. E. coli cells were programmed to respond to a deleterious endogenous input — specifically, DNA-damaging stimuli, such as ultraviolet radiation or mitomycin C. The gene circuit, which was chosen to be the toggle switch (Box A2-1), processes the incoming sensory information and flips from an ‘OFF’ to an ‘ON’ state when a signal threshold is exceeded. Because the biosensor has a decoupled, modular nature, it can be wired to any desired output, from the expression of a standard fluorescent reporter to the activation of natural phenotypes, such as biofilm formation (for example, through expression of traA) or cell suicide (for example, through expression of ccdB).

Sometimes a single signal may be too general to characterize or define an environment. For such situations, Anderson et al. (2007) devised a transcriptional AND gate that could be used to integrate multiple environmental signals into a single genetic circuit (Box A2-2), therefore programming the desired level of biosensing specificity. Genetic biosensors of this sort could be useful for communicating the state of a specific microenvironment (for example, in an industrial

_______________

¹⁰ Modularity – The capacity of a system or component to function independently of context.

¹¹ Environment-responsive promoters – Promoters that directly transduce environmental signals (for example, heavy metal ions, hormones, chemicals or temperature) that are captured by their associated sensory transcription factors.

bioreactor) within a ‘sea’ of environmental conditions, such as temperature, metabolite levels or cell density.

Translational biosensing. RNA molecules have a diverse and important set of cellular functions (Eddy, 2001). Non-coding RNAs can splice and edit RNA, modify ribosomal RNA, catalyse biochemical reactions and regulate gene expression at the level of transcription or translation (Eddy, 2001; Doudna and Cech, 2002; Guerrier-Takada et al., 1983; Kruger et al., 1982). The regulatory subset of non-coding RNAs (Lee et al., 1993; Stougaard and Nordstrom, 1981; Wagner and Simons, 1994) is well-suited for rational design (Isaacs et al., 2006) and, in particular, for biosensing applications. Many regulatory RNA molecules are natural environmental sensors (Gelfand et al., 1999; Johansson et al., 2002; Lease and Belfort, 2000; Majdalani et al., 2002; Mandal et al., 2003; Mironov et al., 2002; Morita et al., 1999; Winkler et al., 2002; Winkler et al., 2004), and because of their ability to take on complex structures defined by their sequence, these molecules can mediate diverse modular functions across distinct sequence domains. Riboswitches (Winkler and Breaker, 2005), for instance, bind specific small-molecule ligands through aptamer¹² domains and induce conformational changes in the 5′ UTR of their own mRNA, thereby regulating gene expression. Aptamer domains that are modelled after riboswitches are versatile and widely used sensitive elements for RNA-based biosensing. The choice and number of aptamer domains can provide control over specificity. Building an entire RNA-based biosensor typically requires coupling an aptamer domain (the sensitive element) with a post-transcriptional regulatory domain (the transducer module) on a modular RNA molecule scaffold.

Antisense RNAs¹³ (Wagner and Simons, 1994; Good, 2003) are one such class of natural regulatory RNAs that can control gene expression through posttranscriptional mechanisms. By linking a riboswitch aptamer to an antisense repressor on a single RNA molecule, Bayer and Smolke (2005) engineered transacting, ligand-responsive riboregulators¹⁴ of gene expression in Saccharomyces cerevisiae (Box A2-2). Binding of the aptamer to its ligand (for example, the small molecule theophylline) induces a conformational change in the RNA sensor that either sequesters the antisense domain in a stable stem loop (ON switch) or liberates it to inhibit translation of an output gene reporter (OFF switch). As a result of the cooperative dependence on both ligand and target mRNA, this biosensor shows binary-like switching at a threshold ligand concentration, similar

_______________

¹² Aptamer – Oligonucleic acids that bind to a specific target molecule, such as a small molecule, protein or nucleic acid. Nucleic acid aptamers are typically developed through in vitro selection schemes but are also found naturally (for example, RNA aptamers in riboswitches).

¹³ Antisense RNAs – RNAs that bind segments of mRNA in trans to inhibit translation.

¹⁴ Riboregulators – Small regulatory RNAs that can activate or repress gene expression by binding segments of mRNA in trans. They are typically expressed in response to an environmental signalling event.

to the genetic toggle design. Importantly, this detection threshold can be adjusted by altering the RNA sequence and therefore the thermodynamic properties of the structure. In principle, the ‘antiswitch’ framework is modular; in other words, aptamers for different ligands and antisense stems targeting different downstream genes could be incorporated into the scaffold to create new sensors. In practice, developing new sensors by aptamer and antisense replacement often involves re-screening compatible secondary structures to create functioning switches. In the future, this platform could be combined with rapid, in vitro aptameric selection techniques (Cox et al., 2002; Ellington and Szostak, 1990; Hermann and Patel, 2000; Tuerk and Gold, 1990) for generating a suite of RNA biosensors that report on the levels of various mRNA species and metabolites in a cell. However, here it should also be noted that aptamers show specificity for a biased ligand space, and as a result aptamers for a target ligand cannot always be found.

Another method for transducing the sensory information captured by aptamer domains is to regulate translation through RNA self-cleavage (Winkler et al., 2004; Winkler and Breaker, 2005). RNA cleavage is catalysed by ribozymes, some of which naturally possess aptameric domains and are responsive to metabolites (Winkler et al., 2004). Yen et al. (2004) took advantage of this natural framework and encoded ligand-sensitive ribozymes in the mRNA sequences of reporter genes. In the absence of its cognate ligand, constitutive autocleavage of the reporter mRNA resulted in little or no signal. The RNA biosensor is flipped when the cognate ligand is present to inhibit the ribozyme’s activity. Similar to the ‘antiswitch’ framework (and with the same technical challenges), these engineered RNAs could potentially be used as endogenous sensors for reporting on a variety of intracellular species and metabolites.

Post-translational biosensing. Signal transduction pathways show vast diversity and complexity. Factors such as the nature of the molecular interactions, the number of interconnected proteins in a cascade and the use of spatial mechanisms dictate which signals are transmitted, whether a signal is amplified or attenuated and the dynamics of the response. Despite the multitude of factors and interacting components, signal transduction pathways are essentially hierarchical schemes based on sensitive elements and downstream transducer modules, and as such can be rationalized for engineering protein-based biosensors.

The primary sensitive element for most signal transduction pathways is the protein receptor. Whereas environment-responsive promoters and RNA aptamers are typically identified from nature or selected with high-throughput combinatorial methods, protein receptors can be designed de novo at the level of molecular interactions. For instance, Looger et al. (2003) devised a computational method for redesigning natural protein receptors to bind new target ligands. Starting with a ‘basis’ of five proteins from the E. coli periplasmic binding protein (PBP) superfamily, the researchers replaced each of the wild-type ligands with a new, non-native target ligand and then used an algorithm to combinatorially explore all

binding-pocket-residue mutations and ligand-docking configurations. This procedure was used to predict novel receptors for trinitrotoluene (TNT; a carcinogen and explosive), L-lactate (a medically-important metabolite) and serotonin (a chemical associated with psychiatric conditions). The predicted receptor designs were experimentally confirmed to be strong and specific in vitro sensors, as well as in vivo cell-based biosensors.

Protein receptors, such as the ones discussed above, are typically membrane-bound; they trigger protein signalling cascades that ultimately result in a cellular response. However, several synthetic methods can be used to transmit captured sensory information in a tunable and desirable manner. Skerker et al. (2008) rationally rewired the transmission of information through two-component systems¹⁵ by identifying rules governing the specificity of a histidine kinase to its cognate response regulator. Alternatively, engineered protein scaffolds can be designed to physically recruit pathway modulators and synthetically reshape the dynamical response behaviour of a system (Bashor et al., 2008) (Box A2-3). This constitutes a modular method for programming protein-based biosensors to have any desired response, including accelerated, delayed or ultrasensitive responses, to upstream signals.

Hybrid approaches. Combining synthetic transcriptional, translational and post-translational circuits into hybrid solutions and harnessing desired characteristics from each could lead to the creation of cell-based biosensors that are as robust as those of natural organisms. Using a synthetic hybrid approach, Voigt and colleagues (Levskaya et al., 2005; Levskaya et al., 2009; Tabor et al., 2009) developed E. coli-based optical sensors. A synthetic sensor kinase was engineered to allow cells to identify and report the presence of red light. As a result, a bacterial lawn of the engineered cells could faithfully ‘print’ a projected image in the biological equivalent of photographic film. Specifically, a membrane-bound photoreceptor from cyanobacteria was fused to an E. coli intracellular histidine kinase to induce light-dependent changes in gene expression (Levskaya et al., 2005) (Box A2-3). In a clever example of its use, the bacterial optical sensor was applied in image edge detection (Tabor et al., 2009). In this case, by wiring the optical sensor to transcriptional circuits that perform cell–cell communication (the quorum-sensing¹⁶ system from Vibrio fischeri) and logical functions (Box A2-3), the researchers programmed only the cells that receive light and directly neighbour cells that do not receive light to produce a pigment, allowing

_______________

¹⁵ Two-component systems – Among the simplest types of signal transduction pathways. In bacteria, they consist of two domains: a membrane-bound histidine kinase (sensitive element) that senses a specific environmental stimulus, and a cognate response regulator (transducer domain) that triggers a cellular response.

¹⁶ Quorum sensing – A cell-to-cell communication mechanism in many species of bacteria, whereby cells measure their local density (by the accumulation of a signalling molecule) and subsequently coordinate gene expression.

the edges of a projected image to be traced. This work demonstrates that complex behaviour can emerge from properly wiring together smaller genetic programs, and that these programs can lead to unique real-world applications.

Therapeutics

Human health is afflicted by new and old foes, including emergent drug-resistant microbes, cancer and obesity. Meanwhile, progress in medicine is faced with challenges at each stage of the therapeutic spectrum, ranging from the drying up of pharmaceutical pipelines to limited global access to viable medicines. In a relatively short amount of time, synthetic biology has made promising strides in reshaping and streamlining this spectrum (Box A2-4). Indeed, the rational and model-guided construction of biological parts is enabling new therapeutic platforms, from the identification of disease mechanisms and drug targets to the production and delivery of small molecules.

Disease mechanism. An electrical engineer is likely to prototype portions of a circuit on a ‘breadboard’ before printing it as an entire integrated circuit. This allows for the rigorous testing of submodules in an isolated, well-characterized environment. Similarly, synthetic biology provides a framework for synthetically reconstructing natural biological systems to explore how pathological behaviours may emerge. This strategy was used to give mechanistic insights into a primary immunodeficiency, agammaglobulinaemia, in which patients cannot generate mature B cells and as a result are unable to properly fight infections (Ferrari et al., 2007). The researchers developed a synthetic testbed by systematically reconstructing the various components of the human B cell antigen receptor (BCr) signalling pathway in an orthogonal environment.¹⁷ This allowed them to identify network topology features that trigger BCR signalling and assembly. A rare mutation in the immunoglobulin-β-encoding gene was identified in one patient and introduced into the synthetic system, in which it was shown to abolish assembly of the BCR on the cell surface, thereby linking this faulty pathway component with disease onset. Pathogenic viral genomes can similarly be reconstructed for studying the molecular underpinnings of infectious disease pandemics. For instance, synthetic reconstruction of the severe acquired respiratory syndrome (SARS) coronavirus (Becker et al., 2008) and the 1918 Spanish influenza virus (Tumpey et al., 2005) helped to identify genetic mutations that may have conferred human tropism and increased virulence.

Drug-target identification. Building up synthetic pathways and systems from individual parts is one way of identifying disease mechanisms and therapeutic

_______________

¹⁷ Orthogonal environment – A cellular environment or host into which genetic material is transplanted to avoid undesired native host interference or regulation. Orthogonal hosts are often organisms with sufficient evolutionary distance from the native host.

targets. Another is to deploy synthetic biology devices to systematically probe the function of individual components of a natural pathway. Our group, for instance, has engineered modular riboregulators that can be used to tune the expression of a toxic protein or any gene in a biological network (Isaacs et al., 2004). To achieve post-transcriptional control over a target gene, the mRNA sequence of the riboregulator 5′ UTR is designed to form a hairpin structure that sequesters the ribosomal binding site (RBS) and prevents ribosome access to it. Translational repression of this cis-repressed mRNA can be alleviated by an independently regulated transactivating RNA that targets the stem-loop for unfolding. Engineered riboregulators have been used to tightly regulate the expression of CcdB, a toxic bacterial protein that inhibits DNA gyrase,¹⁸ to gain a better understanding of the sequence of events leading to induced bacterial cell death (Dwyer et al., 2007). These synthetic biology studies, in conjunction with systems biology studies of quinolones (antibiotics that inhibit gyrase) (Dwyer et al., 2007), led to the discovery that all major classes of bactericidal antibiotics induce a common cellular death pathway by stimulating oxidative damage (Kohanski et al., 2007, 2008). This work provided new insights into how bacteria respond to lethal stimuli and paved the way for the development of more effective antibacterial therapies.

Drug discovery. After a faulty pathway component or target is identified, whole-cell screening assays can be designed using synthetic biology strategies for drug discovery. As a demonstration of this approach, Fussenegger and colleagues (Weber et al., 2008) developed a synthetic platform for screening small molecules that could potentiate a Mycobacterium tuberculosis antibiotic (Box A2-4). ethionamide, currently the last line of defence in the treatment of multidrug-resistant tuberculosis, depends on activation by the M. tuberculosis enzyme EthA for efficacy. However, due to transcriptional repression of ethA by the protein EthR, ethionamide-based therapy is often rendered ineffective. To address this problem, the researchers designed a synthetic mammalian gene circuit that featured an EthR-based transactivator of a reporter gene and used it to screen for and identify EthR inhibitors that could abrogate resistance to ethionamide. Importantly, because the system is a cell-based assay, it intrinsically enriches for inhibitors that are non-toxic and membrane-permeable to mammalian cells, which are key drug criteria as M. tuberculosis is an intracellular pathogen. This framework, in which drug discovery is applied to whole cells that have been engineered with circuits that highlight a pathogenic mechanism, could be extended to other diseases and phenotypes.

_______________

¹⁸ DNA gyrase – A type II DNA topoisomerase that catalyses the ATP-dependent supercoiling of closed-circular dsDNA by strand breakage and rejoining reactions. Control of chromosomal topological transitions is essential for DNA replication and transcription in bacteria, making gyrase an effective target for antimicrobial agents (for example, the quinolone class of antibiotics).

Therapeutic treatment. Synthetic biology devices have additionally been developed to serve as therapies themselves. Entire engineered viruses and organisms can be programmed to target specific pathogenic agents and pathological mechanisms. For instance, in two separate studies (Lu and Collins, 2007, 2009) researchers used engineered bacteriophages to combat antibiotic-resistant bacteria by endowing them with genetic mechanisms that target and thwart bacterial mechanisms for evading antibiotic action. The first study was prompted by the observation that biofilms,¹⁹ in which bacteria are encapsulated in an extracellular matrix, have inherent resistance to antimicrobial therapies and are sources of persistent infections. To more effectively penetrate this protective environment, T7 phage was engineered to express the biofilm matrix-degrading enzyme dispersin B (DspB) upon infection (Lu and Collins, 2007). The two-pronged attack of T7 expressing DspB and phage-induced lysis fuelling the creation and spread of DspB resulted in the removal of 99.997% of the biofilm bacterial cells. In the second study (Lu and Collins, 2009), it was suggested that inhibition of certain bacterial genetic programs could improve the effectiveness of current antibiotic therapies. In this case, bacteriophages were deliberately designed to be non-lethal so as not to elicit resistance mechanisms; instead, a non-lytic M13 phage was used to suppress the bacterial SOS DNA-damage response by overexpression of its repressor, lexA3. The engineered bacteriophage significantly enhanced killing by three major classes of antibiotics in traditional cell culture and in E. coliinfected mice, potentiated killing of antibiotic-resistant bacteria and, importantly, reduced the incidence of cells with antibiotic-induced resistance.

Synthetically engineered viruses and organisms that are able to sense and link their therapeutic activity to pathological cues may be useful in the treatment of cancer, in which current therapies often indiscriminately attack tumours and normal tissues. For instance, adenoviruses were programmed to couple their replication to the state of the p53 pathway in human cells (Ramachandra et al., 2001). Normal p53 production would result in inhibition of a crucial viral replication component, whereas a defunct p53 pathway, which is characteristic of tumour cells, would allow viral replication and cell killing. In another demonstration of translational synthetic biology applied to cancer therapy, Voigt and colleagues (Anderson et al., 2006) developed cancer-targeting bacteria and linked their ability to invade the cancer cells to specific environmental signals. Constitutive expression of the heterologous invasin (inv) gene (from Yersinia pseudotuberculosis) can induce E. coli cells to invade both normal human cell lines and cancer cell lines. So, to preferentially invade cancer cells, the researchers placed inv under the control of transcriptional operons that are activated by environmental signals specific to the tumour microenvironment. These engineered bacteria could be made to carry or synthesize cancer therapies for the treatment of tumours.

_______________

¹⁹ Biofilms – Surface-associated communities of bacterial cells encapsulated in an extracellular polymeric substances (EPS) matrix. Biofilms are an antibiotic-resistant mode of microbial life found in natural and industrial settings.

Therapeutic delivery. In addition to engineered therapeutic organisms, synthetic circuits and pathways can be used for the controlled delivery of drugs as well as for gene and metabolic therapy. In some cases, sophisticated kinetic control over drug release in the body may yield therapeutic advantages and reduce undesired side effects. Most hormones in the body are released in time-dependent pulses. Glucocorticoid secretion, for instance, has a circadian and ultradian²⁰ pattern of release, with important transcriptional consequences for glucocorticoid-responsive cells (Stavreva et al., 2009). Faithfully mimicking these patterns in the administration of synthetic hormones to patients with glucocorticoid-responsive diseases, such as rheumatoid arthritis, may decrease known side effects and improve therapeutic response (Stavreva et al, 2009). Periodic synthesis and release of biologic drugs can be autonomously achieved with synthetic oscillator circuits (Elowitz and Leibler, 2000; Atkinson et al., 2003; Fung et al., 2005; Stricker et al., 2008; Tigges et al., 2009) or programmed time-delay circuits (Weber et al., 2007). In other cases, one may wish to place a limit on the amount of drug released by programming the synthetic system to self-destruct after a defined number of cell cycles or drug release pulses. Our group has recently developed two variants of a synthetic gene counter (Friedland et al., 2009) that could be adapted for such an application.

Gene therapy is beginning to make some promising advances in clinical areas in which traditional drug therapy is ineffective, such as in the treatment of many hereditary and metabolic diseases. Synthetic circuits offer a more controlled approach to gene therapy, such as the ability to dynamically silence, activate and tune the expression of desired genes. In one such example (Deans et al., 2007), a genetic switch was developed in mammalian cells that couples transcriptional repressor proteins and an RNAi module for tight, tunable and reversible control over the expression of desired genes (Box A2-4). This system would be particularly useful in gene-silencing applications, as it was shown to yield >99% repression of a target gene.

Additionally, the construction of non-native pathways offers a unique and versatile approach to gene therapy, such as for the treatment of metabolic disorders. Operating at the interface of synthetic biology and metabolic engineering, Liao and colleagues (Dean et al., 2009) recently introduced the glyoxylate shunt pathway²¹ into mammalian liver cells and mice to explore its effects on fatty acid metabolism and, more broadly, on whole-body metabolism. Remarkably, the researchers found that when transplanted into mammals, the shunt actually increased fatty acid oxidation, evidently by creating an alternative cycle. Furthermore, mice expressing the shunt showed resistance to diet-induced obesity when

_______________

²⁰ Ultradian – Periods of cycles that are repeated throughout a 24-hour circadian day.

²¹ Glyoxylate shunt pathway – A two-enzyme metabolic pathway unique to bacteria and plants that is activated when sugars are not readily available. This pathway diverts the tricarboxylic acid (TCA) cycle so that fatty acids are not completely oxidized and are instead converted into carbon energy sources.

placed on a high-fat diet, with corresponding decreases in total fat mass, plasma triglycerides and cholesterol levels. This work offers a new synthetic biology model for studying metabolic networks and disorders, and for developing treatments for the increasing problem of obesity.

Finally, the discovery of drugs and effective treatments may not quickly — or ever — translate to the people who need them the most because drug production processes can be difficult and costly. As discussed below, synthetic biology is allowing rare and costly drugs to be manufactured more cost-effectively (Box A2-4).

Biofuels, pharmaceuticals and biomaterials

Recent excitement surrounding the production of biofuels, pharmaceuticals and biomaterials from engineered microorganisms is matched by the challenges that loom in bringing these technologies to production scale and quality. The most widely used biofuel is ethanol produced from corn or sugar cane (Fortman et al., 2008); however, the heavy agricultural burden combined with the suboptimal fuel properties of ethanol make this approach to biofuels problematic and limited. Microorganisms engineered with optimized biosynthetic pathways to efficiently convert biomass into biofuels are an alternative and promising source of renewable energy. These strategies will succeed only if their production costs can be made to compete with, or even outcompete, current fuel production costs. Similarly, there are many drugs for which expensive production processes preclude their capacity for a wider therapeutic reach. New synthetic biology tools would also greatly advance the microbial production of biomaterials and the development of novel materials.

Constructing biosynthetic pathways. When engineering for biofuels, drugs or biomaterials, two of the first design decisions are choosing which biosynthetic pathway or pathways to focus on and which host organism to use. Typically, these decisions begin with the search for organisms that are innately capable of achieving some desired biosynthetic activity or phenotype (Alper and Stephanopoulos, 2009). For biofuel production, for instance, certain microorganisms have evolved to be proficient in converting lignocellulosic material to ethanol, biobutanol and other biofuels. These native isolates possess unique catabolic activity, heightened tolerances for toxic materials and a host of enzymes designed to break down the lignocellulosic components. Unfortunately, these highly desired properties exist in pathways that are tightly regulated according to the host’s evolved needs and therefore may not be suitable in their native state for production scale. A longstanding challenge in metabolic and genetic engineering is determining whether to improve the isolate host’s production capacity or whether to transplant the desired genes or pathways into an industrial model host, such as E. coli or S.

cerevisiae; these important considerations and trade-offs are reviewed elsewhere (Alper and Stephanopoulos, 2009).

The example of the microbial production of biobutanol, a higher energy density alternative to ethanol, provides a useful glimpse into these design trade-offs. Butanol is converted naturally from acetyl-CoA by Clostridium acetobutylicum (Jones and Woods, 1986). However, it is produced in low yields and as a mixture with acetone and ethanol, so substantial cellular engineering of a microorganism for which standard molecular biology techniques do not apply is needed to produce usable amounts of butanol (Tummala et al., 2003; Shao et al., 2007). Furthermore, importing the biosynthetic genes into an industrial microbial host can lead to metabolic imbalances (Inui et al., 2008). In an altogether different approach, Liao and colleagues (Atsumi et al., 2008) bypassed standard fermentation pathways and recognized that a broad set of the 2-keto acid intermediates of E. coli amino acid biosynthesis could be synthetically shunted to achieve high-yield production of butanol and other higher alcohols in two enzymatic steps.

Indeed, complementary to efforts in traditional metabolic and genetic engineering is the use of engineering principles for constructing functional, predictable and non-native biological pathways de novo to control and improve microbial production. In an exemplary illustration of this, Keasling and colleagues engineered the microbial production of precursors to the antimalarial drug artemisinin to industrial levels (Martin et al., 2003; Ro et al., 2006) (Box A2-4). There are now many such examples of the successful application of synthetic approaches to biosynthetic pathway construction — such approaches have been used in the microbial production of fatty-acid-derived fuels and chemicals (such as fatty esters, fatty alcohols and waxes) (Steen et al., 2010), methyl halide-derived fuels and chemicals (Bayer et al., 2009), polyketide synthases that make cholesterol-lowering drugs (Ma et al., 2009), and polyketides made from megaenzymes that are encoded by very large synthetic gene clusters (Kodumal et al., 2004).

Optimizing pathway flux. After biosynthetic pathways have been constructed, the expression levels of all of the components need to be orchestrated to optimize metabolic flux²² and achieve high product titres. A standard approach is to drive the expression of pathway components with strong and exogenously tunable promoters, such as the P_Ltet, P_Llac, and P_BAD promoters from the tet, lac and ara operons of E. coli, respectively. To this end, there are ongoing synthetic biology efforts to create and characterize more reusable, biological control elements based on promoters for predictably tuning expression levels (Alpert et al., 2005; Ellis et al., 2009). Further to this, synthetic biologists have devised a number of alternative methods for obtaining biological pathway balance, ranging from reconfiguring network connectivity to fine-tuning individual components. A richer

_______________

²² Metabolic flux – The rate of flow of metabolites through a metabolic pathway. The rate is regulated by the enzymes in the pathway.

discussion of these topics, including the fine-tuning of parts, the application of model-guided approaches and the development of next-generation interoperable parts, is presented elsewhere (Lu et al., 2009). In Box A2-5 and Box A2-6 we detail several synthetic biology strategies that specifically pertain to the optimization of metabolic pathway flux. These strategies range from those driven by evolutionary techniques, to those driven by rational design and in silico models, to those that combine both approaches.

Programming novel functionality and materials. Beyond facilitating metabolic tasks, synthetic systems can infuse novel functionality into engineered organisms for production purposes or for building new materials. Early work in the field laid the groundwork for constructing basic circuits that could sense and process signals, perform logic operations and actuate biological responses (Voigt, 2006). Wiring these modules together to bring about reliable, higher-order functionality is one of the next major goals of synthetic biology (Lu et al., 2009), and an important application of this objective is the layering of ‘smart’ control mechanisms over metabolic engineering. For instance, circuits designed to sense the bioreactor environment and shift metabolic phases accordingly would further improve biofuel production. Alternatively, autonomous timing circuits could be used to shut down metabolic processes after a prescribed duration of time. Biological timers of this sort have been developed using genetic toggle switches that were deliberately rendered imbalanced through model-guided promoter engineering (Ellis et al., 2009). These genetic timers were used to program the time-dependent flocculation²³ of yeast cells to facilitate the separation of cells from, for instance, the alcohol produced in industrial fermentation processes.

Synthetic control systems can also be used to extract and purify the synthesized product. This is particularly important in the production of recombinant proteins, bioplastics and other large biomaterials, which can accumulate inside cells, cause the formation of inclusion bodies and become toxic to cells if they are present at high titres. To export recombinant spider-silk monomers, Widmaier et al. (2009) searched for a secretion system that would enable efficient and indiscriminate secretion of proteins through both bacterial membranes. The Salmonella type III secretion system (T3SS) not only fulfils these criteria but also possesses a natural regulatory scheme that ties expression of the protein to be secreted to the secretion capacity of the cell; as a result, the desired protein is only expressed when sufficient secretion complexes have been built. To obtain superior secretion rates of recombinant silk protein, the researchers needed only to engineer a control circuit that hitches the heterologous silk-protein-producing genes to the innate genetic machinery for environmental sensing and secretion commitment.

_______________

²³ Flocculation – A specific form of cell aggregation in yeast triggered by certain environmental conditions, such as the absence of sugars. For example, flocculation occurs once the sugar in a beer brew has been fermented into ethanol.

Finally, there is an emerging branch of synthetic biology that seeks to program coordinated behaviour in populations of cells, which could lead to the fabrication of novel biomaterials for various applications. The engineering of synthetic multicellular systems is typically achieved with cell–cell communication and associated intracellular signal processing modules, as was elegantly used by Hasty and colleagues (Danino et al., 2010) to bring about synchronized oscillations in a population of bacterial cells. Weiss and colleagues (Basu et al., 2004, 2005) have similarly done pioneering work in building biomolecular signal-processing architectures that can filter communication signals originating from ‘sender’ cells. These systems, which can be programmed to form intricate multicellular patterns from a solid-phase cellular lawn, would aid the development of fabrication-free scaffolds for tissue engineering.

Future challenges and conclusions

The future of translational synthetic biology hinges on the development of reliable means for connecting smaller functional circuits to realize higher-order networks with predictable behaviours. In a previous article (Lu et al., 2009), we outlined four research efforts aimed at improving and accelerating the overall design cycle and allowing more seamless integration of biological circuitry (Box A2-7).

Beyond the challenge of improving the design cycle, applied synthetic biology would benefit from once again summoning the original inspiration of bio-computing. The ability to program higher-level decision-making into synthetic networks would yield more robust and dynamic organisms, including ones that can accomplish many tasks simultaneously. Furthermore, as adaptive and predictive behaviours are naturally present in all organisms (including microbes) (Mitchell et al., 2009; Tagkopoulos et al., 2008), synthetic learning networks built from genetic and biological parts (Fernando et al., 2009; Fritz et al., 2007) would infuse engineered organisms with more sophisticated automation for biosensing and related applications.

Finally, the majority of synthetic biology is currently practiced in microbes. However, many of the most pressing problems, and in particular those of human health, are inherently problems with mammalian systems. Therefore, a more concerted effort towards advancing mammalian synthetic biology will be crucial for next-generation therapeutic solutions, including the engineering of synthetic gene networks for stem-cell generation and differentiation.

By addressing such challenges, we will be limited not by the technicalities of construction or the robustness of synthetic gene networks but only by the imagination of researchers and the number of societal problems and applications that synthetic biology can resolve.

Acknowledgements

We thank members of the Collins laboratory for helpful discussions and K. M. Flynn for help with artwork. We also thank the Howard Hughes Medical Institute and the US National Institutes of Health Director’s Pioneer Award Program for their financial support.

Competing Interests Statement

The authors declare competing financial interests: see Web version for details.

Databases

Entrez Gene: http://www.ncbi.nlm.nih.gov/gene ccdB | lexA3 | traA

OMIM: http://www.ncbi.nlm.nih.gov/omim agammaglobulinaemia

UniProtKB: http://www.uniprot.org CI | Cro | DspB | EnvZ | EthA | EthR

Further Information

James J. Collins’ homepage: http://www.bu.edu/abl

All links are active on the online PDF.

References

Ajo-Franklin, C. M. et al. Rational design of memory in eukaryotic cells. Genes Dev. 21, 2271–2276 (2007).

Alper, H. & Stephanopoulos, G. Engineering for biofuels: exploiting innate microbial capacity or importing biosynthetic potential? Nature Rev. Microbiol. 7, 715–723 (2009).

Alper, H., Fischer, C., Nevoigt, E. & Stephanopoulos, G. Tuning genetic control through promoter engineering. Proc. Natl Acad. Sci. USA 102, 12678–12683 (2005).

Anderson, J. C., Clarke, E. J., Arkin, A. P. & Voigt, C. A. Environmentally controlled invasion of cancer cells by engineered bacteria. J. Mol. Biol. 355, 619–627 (2006).
This study provides a potential synthetic approach to cancer therapy. Bacteria were programmed to sense environmental cues of the tumour microenvironment and respond to them by invading malignant cells.

Anderson, J. C., Voigt, C. A. & Arkin, A. P. Environmental signal integration by a modular AND gate. Mol. Syst. Biol. 3, 133 (2007).

Atkinson, M. R., Savageau, M. A., Myers, J. T. & Ninfa, A. J. Development of genetic circuitry exhibiting toggle switch or oscillatory behavior in Escherichia coli. Cell 113, 597–607 (2003).

Atsumi, S., Hanai, T. & Liao, J. C. Non-fermentative pathways for synthesis of branched-chain higher alcohols as biofuels. Nature 451, 86–89 (2008).

Bashor, C. J., Helman, N. C., Yan, S. & Lim, W. A. Using engineered scaffold interactions to reshape MAP kinase pathway signaling dynamics. Science 319, 1539–1543 (2008).

Basu, S., Mehreja, R., Thiberge, S., Chen, M. T. & Weiss, R. Spatiotemporal control of gene expression with pulse-generating networks. Proc. Natl Acad. Sci. USA 101, 6355–6360 (2004).

Basu, S., Gerchman, Y., Collins, C. H., Arnold, F. H. & Weiss, R. A synthetic multicellular system for programmed pattern formation. Nature 434, 1130–1134 (2005).
An excellent demonstration of synthetic control over a population of cells. By splitting the V. fischeri quorum-sensing circuit between ‘sender’ and ‘receiver’ cells, bacteria were programmed to communicate to generate intricate two-dimensional patterns.

Bayer, T. S. & Smolke, C. D. Programmable ligand-controlled riboregulators of eukaryotic gene expression. Nature Biotech. 23, 337–343 (2005).
This paper establishes an RNA scaffold for the ligand-dependent ON–OFF switching of gene expression.

Bayer, T. S. et al. Synthesis of methyl halides from biomass using engineered microbes. J. Am. Chem. Soc. 131, 6508–6515 (2009).

Becker, M. M. et al. Synthetic recombinant bat SARS-like coronavirus is infectious in cultured cells and in mice. Proc. Natl Acad. Sci. USA 105, 19944–19949 (2008).

Blake, W. J., Kærn, M., Cantor, C. R. & Collins, J. J. Noise in eukaryotic gene expression. Nature 422, 633–637 (2003).

Boorsma, M. et al. A temperature-regulated replicon-based DNA expression system. Nature Biotech. 18, 429–432 (2000).

Brown, M. et al. Lac repressor can regulate expression from a hybrid SV40 early promoter containing a lac operator in animal cells. Cell 49, 603–612 (1987).

Cox, J. C. et al. Automated selection of aptamers against protein targets translated in vitro: from gene to aptamer. Nucleic Acids Res. 30, e108 (2002).

Danino, T., Mondragon-Palomino, O., Tsimring, L. & Hasty, J. A synchronized quorum of genetic clocks. Nature 463, 326–330 (2010).

Dean, J. T. et al. Resistance to diet-induced obesity in mice with synthetic glyoxylate shunt. Cell Metab. 9, 525–536 (2009).

Deans, T. L., Cantor, C. R. & Collins, J. J. A tunable genetic switch based on RNAi and repressor proteins for regulating gene expression in mammalian cells. Cell 130, 363–372 (2007).

Deuschle, U. et al. Regulated expression of foreign genes in mammalian cells under the control of coliphage T3 RNA polymerase and lac repressor. Proc. Natl Acad. Sci. USA 86, 5400–5404 (1989).

Doudna, J. A. & Cech, T. R. The chemical repertoire of natural ribozymes. Nature 418, 222–228 (2002).

Dueber, J. E., Yeh, B. J., Chak, K. & Lim, W. A. Reprogramming control of an allosteric signaling switch through modular recombination. Science 301, 1904–1908 (2003).

Dueber, J. E. et al. Synthetic protein scaffolds provide modular control over metabolic flux. Nature Biotech. 27, 753–759 (2009).
The authors provide a modular framework for programming the input–output behavior of eukaryotic signalling-protein circuits, and construct synthetic switch proteins with a rich set of gating behaviours.

Dunlap, J. C. Molecular bases for circadian clocks. Cell 96, 271–290 (1999).

Dwyer, D. J., Kohanski, M. A., Hayete, B. & Collins, J. J. Gyrase inhibitors induce an oxidative damage cellular death pathway in Escherichia coli. Mol. Syst. Biol. 3, 91 (2007).

Eddy, S. R. Non-coding RNA genes and the modern RNA world. Nature Rev. Genet. 2, 919–929 (2001).

Ellington, A. D. & Szostak, J. W. In vitro selection of RNA molecules that bind specific ligands. Nature 346, 818–822 (1990).

Ellis, T., Wang, X. & Collins, J. J. Diversity-based, model-guided construction of synthetic gene networks with predicted functions. Nature Biotech. 27, 465–471 (2009).

Elowitz, M. B. & Leibler, S. A synthetic oscillatory network of transcriptional regulators. Nature 403, 335–338 (2000).²⁴

Fernando, C. T. et al. Molecular circuits for associative learning in single-celled organisms. J. R. Soc. Interface 6, 463–469 (2009).

Ferrari, S. et al. Mutations of the Igβ gene cause agammaglobulinemia in man. J. Exp. Med. 204, 2047–2051 (2007).

Fortman, J. L. et al. Biofuel alternatives to ethanol: pumping the microbial well. Trends Biotechnol. 26, 375–381 (2008).

Friedland, A. E. et al. Synthetic gene networks that count. Science 324, 1199–1202 (2009).

Fritz, G., Buchler, N. E., Hwa, T. & Gerland, U. Designing sequential transcription logic: a simple genetic circuit for conditional memory. Syst. Synth. Biol. 1, 89–98 (2007).

Fung, E. et al. A synthetic gene-metabolic oscillator. Nature 435, 118–122 (2005).

Fussenegger, M. et al. Streptogramin-based gene regulation systems for mammalian cells. Nature Biotech. 18, 1203–1208 (2000).

Gardner, T. S., Cantor, C. R. & Collins, J. J. Construction of a genetic toggle switch in Escherichia coli. Nature 403, 339–342 (2000).²⁴

Gelfand, M. S., Mironov, A. A., Jomantas, J., Kozlov, Y. I. & Perumov, D. A. A conserved RNA structure element involved in the regulation of bacterial riboflavin synthesis genes. Trends Genet. 15, 439–442 (1999).

Gibson, D. G. et al. Complete chemical synthesis, assembly, and cloning of a Mycoplasma genitalium genome. Science 319, 1215–1220 (2008).

Glass, J. I. et al. Essential genes of a minimal bacterium. Proc. Natl Acad. Sci. USA 103, 425–430 (2006).

Glass, L. Classification of biological networks by their qualitative dynamics. J. Theor. Biol. 54, 85–107 (1975).

Glass, L. & Kauffman, S. A. The logical analysis of continuous, non-linear biochemical control networks. J. Theor. Biol. 39, 103–129 (1973).

Good, L. Translation repression by antisense sequences. Cell. Mol. Life Sci. 60, 854–861 (2003).

Gossen, M. & Bujard, H. Tight control of gene expression in mammalian cells by tetracycline-responsive promoters. Proc. Natl Acad. Sci. USA 89, 5547–5551 (1992).

_______________

²⁴ [These references] describe synthetic biology’s first devices — the genetic toggle switch and the repressilator — and establish the engineering-based methodology for constructing sophisticated, dynamic behaviours in biological systems from simple regulatory elements.

Guerrier-Takada, C., Gardiner, K., Marsh, T., Pace, N. & Altman, S. The RNA moiety of ribonuclease P is the catalytic subunit of the enzyme. Cell 35, 849–857 (1983).

Guet, C. C., Elowitz, M. B., Hsing, W. & Leibler, S. Combinatorial synthesis of genetic networks. Science 296, 1466–1470 (2002).

Guido, N. J. et al. A bottom-up approach to gene regulation. Nature 439, 856–860 (2006).

Ham, T. S., Lee, S. K., Keasling, J. D. & Arkin, A. P. A tightly regulated inducible expression system utilizing the fim inversion recombination switch. Biotechnol. Bioeng. 94, 1–4 (2006).

Ham, T. S., Lee, S. K., Keasling, J. D. & Arkin, A. P. Design and construction of a double inversion recombination switch for heritable sequential genetic memory. PLoS ONE 3, e2815 (2008).

Hermann, T. & Patel, D. J. Adaptive recognition by nucleic acid aptamers. Science 287, 820–825 (2000).

Hooshangi, S., Thiberge, S. & Weiss, R. Ultrasensitivity and noise propagation in a synthetic transcriptional cascade. Proc. Natl Acad. Sci. USA 102, 3581–3586 (2005).

Hu, M. C. & Davidson, N. The inducible lac operator–repressor system is functional in mammalian cells. Cell 48, 555–566 (1987).

Inui, M. et al. Expression of Clostridium acetobutylicum butanol synthetic genes in Escherichia coli. Appl. Microbiol. Biotechnol. 77, 1305–1316 (2008).

Isaacs, F. J. et al. Engineered riboregulators enable post-transcriptional control of gene expression. Nature Biotech. 22, 841–847 (2004).

Isaacs, F. J., Dwyer, D. J. & Collins, J. J. RNA synthetic biology. Nature Biotech. 24, 545–554 (2006).

Johansson, J. et al. An RNA thermosensor controls expression of virulence genes in Listeria monocytogenes. Cell 110, 551–561 (2002).

Jones, D. T. & Woods, D. R. Acetone-butanol fermentation revisited. Microbiol. Rev. 50, 484–524 (1986).

Kauffman, S. The large scale structure and dynamics of gene control circuits: an ensemble approach. J. Theor. Biol. 44, 167–190 (1974).

Kobayashi, H. et al. Programmable cells: interfacing natural and engineered gene networks. Proc. Natl Acad. Sci. USA 101, 8414–8419 (2004).

Kodumal, S. J. et al. Total synthesis of long DNA sequences: synthesis of a contiguous 32-kb polyketide synthase gene cluster. Proc. Natl Acad. Sci. USA 101, 15573–15578 (2004).

Kohanski, M. A., Dwyer, D. J., Hayete, B., Lawrence, C. A. & Collins, J. J. A common mechanism of cellular death induced by bactericidal antibiotics. Cell 130, 797–810 (2007).

Kohanski, M. A., Dwyer, D. J., Wierzbowski, J., Cottarel, G. & Collins, J. J. Mistranslation of membrane proteins and two-component system activation trigger antibiotic-mediated cell death. Cell 135, 679–690 (2008).

Kramer, B. P. & Fussenegger, M. Hysteresis in a synthetic mammalian gene network. Proc. Natl Acad. Sci. USA 102, 9517–9522 (2005).

Kramer, B. P. et al. An engineered epigenetic transgene switch in mammalian cells. Nature Biotech. 22, 867–870 (2004).

Kruger, K. et al. Self-splicing RNA: autoexcision and autocyclization of the ribosomal RNA intervening sequence of Tetrahymena. Cell 31, 147–157 (1982).

Lartigue, C. et al. Genome transplantation in bacteria: changing one species to another. Science 317, 632–638 (2007).

Lartigue, C. et al. Creating bacterial strains from genomes that have been cloned and engineered in yeast. Science 325, 1693–1696 (2009).

Lease, R. A. & Belfort, M. A trans-acting RNA as a control switch in Escherichia coli: DsrA modulates function by forming alternative structures. Proc. Natl Acad. Sci. USA 97, 9919–9924 (2000).

Lee, R. C., Feinbaum, R. L. & Ambros, V. The C. elegans heterochronic gene lin-4 encodes small RNAs with antisense complementarity to lin-14. Cell 75, 843–854 (1993).

Levskaya, A. et al. Synthetic biology: engineering Escherichia coli to see light. Nature 438, 441–442 (2005).

Levskaya, A., Weiner, O. D., Lim, W. A. & Voigt, C. A. Spatiotemporal control of cell signalling using a light-switchable protein interaction. Nature 461, 997–1001 (2009).

Looger, L. L., Dwyer, M. A., Smith, J. J. & Hellinga, H. W. Computational design of receptor and sensor proteins with novel functions. Nature 423, 185–190 (2003).

Lu, T. K. & Collins, J. J. Dispersing biofilms with engineered enzymatic bacteriophage. Proc. Natl Acad. Sci. USA 104, 11197–11202 (2007).

——. Engineered bacteriophage targeting gene networks as adjuvants for antibiotic therapy. Proc. Natl Acad. Sci. USA 106, 4629–4634 (2009).²⁵

Lu, T. K., Khalil, A. S. & Collins, J. J. Next-generation synthetic gene networks. Nature Biotech. 27, 1139–1150 (2009).

Lutz, R. & Bujard, H. Independent and tight regulation of transcriptional units in Escherichia coli via the LacR/O, the TetR/O and AraC/I₁–I₂ regulatory elements. Nucleic Acids Res. 25, 1203–1210 (1997).

Ma, S. M. et al. Complete reconstitution of a highly reducing iterative polyketide synthase. Science 326, 589–592 (2009).

Majdalani, N., Hernandez, D. & Gottesman, S. Regulation and mode of action of the second small RNA activator of RpoS translation, RprA. Mol. Microbiol. 46, 813–826 (2002).

Malphettes, L. et al. A novel mammalian expression system derived from components coordinating nicotine degradation in arthrobacter nicotinovorans pAO1. Nucleic Acids Res. 33, e107 (2005).

Mandal, M., Boese, B., Barrick, J. E., Winkler, W. C. & Breaker, R. R. Riboswitches control fundamental biochemical pathways in Bacillus subtilis and other bacteria. Cell 113, 577–586 (2003).

Martin, V. J., Pitera, D. J., Withers, S. T., Newman, J. D. & Keasling, J. D. Engineering a mevalonate pathway in Escherichia coli for production of terpenoids. Nature Biotech. 21, 796–802 (2003).

McAdams, H. H. & Arkin, A. Towards a circuit engineering discipline. Curr. Biol. 10, R318–R320 (2000).

McAdams, H. H. & Shapiro, L. Circuit simulation of genetic networks. Science 269, 650–656 (1995).

Mironov, A. S. et al. Sensing small molecules by nascent RNA: a mechanism to control transcription in bacteria. Cell 111, 747–756 (2002).

Mitchell, A. et al. Adaptive prediction of environmental changes by microorganisms. Nature 460, 220–224 (2009).

Monod, J. & Jacob, F. General conclusions: telenomic mechanisms in cellular metabolism, growth, and differentiation. Cold Spring Harb. Symp. Quant. Biol. 26, 389–401 (1961).

Morita, M. T. et al. Translational induction of heat shock transcription factor σ³²: evidence for a built-in RNA thermosensor. Genes Dev. 13, 655–665 (1999).

Mukherji, S. & van Oudenaarden, A. Synthetic biology: understanding biological design from synthetic circuits. Nature Rev. Genet. 10, 859–871 (2009).

Mullick, A. et al. The cumate gene-switch: a system for regulated expression in mammalian cells. BMC Biotechnol. 6, 43 (2006).

Neddermann, P. et al. A novel, inducible, eukaryotic gene expression system based on the quorum-sensing transcription factor TraR. EMBO Rep. 4, 159–165 (2003).

Rackham, O. & Chin, J. W. Cellular logic with orthogonal ribosomes. J. Am. Chem. Soc. 127, 17584–17585 (2005).

Ramachandra, M. et al. Re-engineering adenovirus regulatory pathways to enhance oncolytic specificity and efficacy. Nature Biotech. 19, 1035–1041 (2001).

Rinaudo, K. et al. A universal RNAi-based logic evaluator that operates in mammalian cells. Nature Biotech. 25, 795–801 (2007).

_______________

²⁵ In [these references], engineered bacteriophages were used to deliver synthetic enzymes and perturb gene networks to combat antibiotic-resistant strains of bacteria.

Ro, D. K. et al. Production of the antimalarial drug precursor artemisinic acid in engineered yeast. Nature 440, 940–943 (2006).²⁶

Salis, H. M., Mirsky, E. A. & Voigt, C. A. Automated design of synthetic ribosome binding sites to control protein expression. Nature Biotech. 27, 946–950 (2009).

Savageau, M. A. Comparison of classical and autogenous systems of regulation in inducible operons. Nature 252, 546–549 (1974).

Shao, L. et al. Targeted gene disruption by use of a group II intron (targetron) vector in Clostridium acetobutylicum. Cell Res. 17, 963–965 (2007).

Skerker, J. M. et al. Rewiring the specificity of two-component signal transduction systems. Cell 133, 1043–1054 (2008).

Sohka, T. et al. An externally tunable bacterial band-pass filter. Proc. Natl Acad. Sci. USA 106, 10135–10140 (2009).

Stavreva, D. A. et al. Ultradian hormone stimulation induces glucocorticoid receptor-mediated pulses of gene transcription. Nature Cell Biol. 11, 1093–1102 (2009).

Steen, E. J. et al. Microbial production of fatty-acid-derived fuels and chemicals from plant biomass. Nature 463, 559–562 (2010).²⁶

Stojanovic, M. N. & Stefanovic, D. A deoxyribozyme-based molecular automaton. Nature Biotech. 21, 1069–1074 (2003).

Stougaard, P., Molin, S. & Nordstrom, K. RNAs involved in copy-number control and incompatibility of plasmid R1. Proc. Natl Acad. Sci. USA 78, 6008–6012 (1981).

Stricker, J. et al. A fast, robust and tunable synthetic gene oscillator. Nature 456, 516–519 (2008).

Tabor, J. J. et al. A synthetic genetic edge detection program. Cell 137, 1272–1281 (2009).
An outstanding example of how smaller circuits can be combined to realize larger genetic programs with sophisticated and predictable behaviour. Here, logic gates and cell–cell communication modules were coupled to program a population of bacteria to sense and trace the edges of a projected image.

Tagkopoulos, I., Liu, Y. C. & Tavazoie, S. Predictive behavior within microbial genetic networks. Science 320, 1313–1317 (2008).

Tigges, M., Marquez-Lago, T. T., Stelling, J. & Fussenegger, M. A tunable synthetic mammalian oscillator. Nature 457, 309–312 (2009).

Tuerk, C. & Gold, L. Systematic evolution of ligands by exponential enrichment: RNA ligands to bacteriophage T4 DNA polymerase. Science 249, 505–510 (1990).

Tummala, S. B., Welker, N. E. & Papoutsakis, E. T. Design of antisense RNA constructs for downregulation of the acetone formation pathway of Clostridium acetobutylicum. J. Bacteriol. 185, 1923–1934 (2003).

Tumpey, T. M. et al. Characterization of the reconstructed 1918 Spanish influenza pandemic virus. Science 310, 77–80 (2005).

Voigt, C. A. Genetic parts to program bacteria. Curr. Opin. Biotechnol. 17, 548–557 (2006).

Wagner, E. G. H. & Simons, R. W. Antisense RNA control in bacteria, phages, and plasmids. Annu. Rev. Microbiol. 48, 713–742 (1994).

Wang, H. H. et al. Programming cells by multiplex genome engineering and accelerated evolution. Nature 460, 894–898 (2009).
Provides a combinatorial/evolutionary approach to optimizing biosynthetic pathway components through rapid in vivo genome engineering by cycles of targeted genome modification and phenotype selection.

_______________

²⁶ [These references] provide a paradigm for the application of synthetic biology to the construction and optimization of biosynthetic pathways for cost-effective and high-yield microbial production. In these papers, the authors demonstrate industrial production of the direct precursor to the antimalarial drug artemisinin as part of a broader effort to address worldwide shortages of rare drugs.

Weber, W. et al. Macrolide-based transgene control in mammalian cells and mice. Nature Biotech. 20, 901–907 (2002).

——. Conditional human VEGF-mediated vascularization in chicken embryos using a novel temperature-inducible gene regulation (TIGR) system. Nucleic Acids Res. 31, e69 (2003a).

——. Streptomyces-derived quorum-sensing systems engineered for adjustable transgene expression in mammalian cells and mice. Nucleic Acids Res. 31, e71 (2003b).

——. Gas-inducible transgene expression in mammalian cells and mice. Nature Biotech. 22, 1440–1444 (2004).

——. A synthetic time-delay circuit in mammalian cells and mice. Proc. Natl Acad. Sci. USA 104, 2643–2648 (2007).

——. A synthetic mammalian gene circuit reveals antituberculosis compounds. Proc. Natl Acad. Sci. USA 105, 9994–9998 (2008).

——. A synthetic mammalian electro-genetic transcription circuit. Nucleic Acids Res. 37, e33 (2009).

Weber, W., Link, N. & Fussenegger, M. A genetic redox sensor for mammalian cells. Metab. Eng. 8, 273–280 (2006).

Widmaier, D. M. et al. Engineering the Salmonella type III secretion system to export spider silk monomers. Mol. Syst. Biol. 5, 309 (2009).

Win, M. N. & Smolke, C. D. Higher-order cellular information processing with synthetic RNA devices. Science 322, 456–460 (2008).

Winkler, W., Nahvi, A. & Breaker, R. R. Thiamine derivatives bind messenger RNAs directly to regulate bacterial gene expression. Nature 419, 952–956 (2002).

Winkler, W. C., Nahvi, A., Roth, A., Collins, J. A. & Breaker, R. R. Control of gene expression by a natural metabolite-responsive ribozyme. Nature 428, 281–286 (2004).

Winkler, W. C. & Breaker, R. R. Regulation of bacterial gene expression by riboswitches. Annu. Rev. Microbiol. 59, 487–517 (2005).

Yen, L. et al. Exogenous control of mammalian gene expression through modulation of RNA self-cleavage. Nature 431, 471–476 (2004).

You, L., Cox, R. S., Weiss, R. & Arnold, F. H. Programmed population control by cell–cell communication and regulated killing. Nature 428, 868–871 (2004).

A3

THE GENOME AS THE UNIT OF ENGINEERING

Andrew D. Ellington and Jared Ellefson²⁷

Introduction

Recent years have marked a dramatic increase of our capabilities to sequence and synthesize nucleic acids. Ten years ago the first human genome was sequenced, and at the time it was a monumental undertaking requiring billions of dollars in funding and legions of dedicated researchers. Today, the same genome sequence would cost a fraction of the price and soon personal genome sequencing will be commonplace. However, while our understanding of genomes has not yet

_______________

²⁷ Center for Systems and Synthetic Biology, University of Texas at Austin.

caught up with our ability to generate data, there is no question that the technology revolution has transformed and will continue to drive the biological sciences, in particular the new integrative science known as systems biology.

The same revolution is occurring in DNA synthesis, as prices for DNA drop and technologies are developed for large-scale (and ever longer) syntheses. As with sequencing, the technology revolution will lead understanding, and it can be argued that, although we can synthesize a microbial genome (Gibson et al., 2010), we don’t necessarily understand how to design one. That’s fine; we can learn by doing, and as synthetic genomes become easier to synthesize, we will undoubtedly generate customized genomes that will test the simple hypothesis of whether they function as intended.

Together, these emerging issues are the essence of what can be seen as the holy grail of both systems and synthetic biology, the genome as the unit of engineering. Synthetic biology can, in many circumstances, be defined as the abstraction of biology to the point where it can be easily engineered. This is first and foremost an operational definition, not an intellectual one. Indeed, there hides in the background of synthetic biology the usually unstated hypothesis that biology was made to be engineered by engineers (a hypothesis we take issue with below). Parallels are often drawn between electrical engineering and synthetic biology, where genetic information like promoters and proteins can be compared to components of a circuit board such as transistors or capacitors. In the view of the new breed of synthetic biologists, standardized components can almost always be used to rationally design genetic elements to perform desired tasks. This view has much merit, since it has already yielded interesting products such as biofuels, pharmaceuticals, and biomaterials. For example, Bayer and Voigt coaxed yeast into making valuable methyl halides from biomass (Bayer et al., 2009), while the Keasling lab reengineered the common lab microbe, Escherichia coli, to produce large quantities of amorphadiene, a precursor to antimalarial and anticancer drugs (Martin et al., 2003). It is not unreasonable to suspect that synthetic circuits will be introduced in human hosts at some point; for example, Fussenegger has created a synthetic circuit composed of a bacterial uric acid sensor and a fungal urate oxidase (which converts uric acid into a more tolerable compound) that can be used to control uric acid levels in mammalian hosts and thus ameliorate chronic disease states such as gout (Kemmer et al., 2010).

However, it is nonetheless still the case that many of these systems must be optimized in order for them to function as intended, in large part because the constituent parts are neither truly modular nor is their function fully predictable in new contexts. The question thus becomes whether biology is really meant to be engineered the same way a circuit board is, whether engineers will learn to make biology a circuit board, or whether some composite view is more akin to reality. Or, stated another way, the question is whether genetic tinkering (which existed well before synthetic biology) has somehow entered a new, more robust phase or if it has just been relabeled.

Systems biology has taken a different approach to how to tinker with systems. This approach proceeds from an understanding of the system as a whole instead of as an amalgam of component parts. It is top down rather than bottom up. By integrating literally millions of points of data from genomes, transcriptomes, and proteomes across the phylogenetic spectra, systems biologists can draw remarkable conclusions, up to and including the identification of nonobvious and evolutionarily repurposed subsystems. It is this sort of understanding that allows us to realize that the same interactions that govern resistance to antifungals in yeast also govern blood vessel formation in higher organisms (McGary et al., 2010). While a comparable synthetic biology approach might be to repurpose a tractable signaling pathway for blood vessel formation, this approach would require extensive empirical testing. As technology continues to develop for high-throughput analysis of DNA, RNA, and proteins, systems biologists will have the benefit of several billion years of empirical testing and, thus, will hold the intellectual high ground for understanding how organisms truly work. Quantitative modeling of the connectedness of extant systems will, in the end, be more likely to allow us to build a functional genome from scratch than the untested engineering hypothesis that organisms should work like we want them to. In greater detail, this should inevitably lead to the following discussion.

Systems Biology Eats Synthetic Biology

The fiction of synthetic biology is that it is possible to engineer biological systems in modular, composable, scalable, and programmable ways using “parts” to build circuits and eventually systems (and ecosystems) (Bromley et al., 2008; Win et al., 2009). Both the field and the fiction have emerged largely not from scientific research but from the International Genetically Engineered Machines competition, iGEM, which is constrained to rely on “BioBricks” for the construction of student projects (Smolke, 2009). In fact, long before iGEM there were numerous biological engineers, and these biological engineers would, with different degrees of success, use genes (not then known as biological parts) to construct pathways (not then known as circuits) that had particular functions.

What has changed between the long period during which biological engineering has been maturing (indeed, one could argue that biological engineering in the form of selective breeding greatly predated man’s understanding of biology itself) and today to make biological engineering into synthetic biology, besides nomenclature? It could be argued that there are two important milestones: First, many researchers in other engineering disciplines were somehow shut out of the biology because it lacked an engineering flair and was more dissective than synthetic. In this view, the influx of engineers into biology is assisted by recasting biology in terms more familiar to engineers, and this is why we often see biological circuits represented (incorrectly, as I will argue later) as electronic circuits (Khalil and Collins, 2010).

Second, there is not just a quantitative but rather a qualitative or foundational difference in being able to make large amounts of DNA. The ability to remake a whole genome (Gibson et al., 2010) is so great relative to the ability to make a mutation in that genome (or in an episome sharing a cell with that genome) that there must now be a new engineering discipline that approaches genome construction in a wholly different way than traditional biological engineering would have been capable of doing. I think there is some merit to this explanation, although it is a bit like saying that genomics is different than genetics. While we would never argue that we do not have much greater understanding of molecular detail today than we did in Mendel’s day, our fantastic increase in knowledge does not invalidate the views of Mendel; rather, this knowledge merely adds depth to the fundamental concept of inheritance. Genomics further explains genetics; it does not remake it (although individual concepts in genetics, such as the role of environment in inheritance, have certainly been remade radically). Nonetheless, it is true that techniques for the rational manipulation of whole genomes will prove to be of enormous value into the future. And while many of us who wish to reconstruct living systems will continue to have “Venter envy,” a new generation of techniques such as multiplex automated genome engineering (Wang et al., 2009) are beginning to emerge that will make such manipulations possible even for individual investigators.

In the absence of the synthetic biology “revolution” (redefinition), how might biological engineers who were attempting to manipulate larger and larger amalgams of DNA have managed? Where would they have turned for knowledge and inspiration? I contend that it would not have been electrical engineering—the frequent muse to whom many who call themselves synthetic biologists appeal—but rather to systems biology. Systems biology attempts to understand the interrelationships between all of the molecular and cellular parts of system in a quantitative way, and it has as one of its ultimate goals the modeling of biological organisms down to the molecular level. If we had the same knowledge of organisms that we do of the hardware and software accompanying a computer, the entire organism would not just be a “chassis” for the installation of a synthetic circuit; it would be the unit of engineering itself. Fortunately for both traditional biological engineers and synthetic biologists, our understanding of systems has increased dramatically in recent years, and thus there will come a point where systems biology will not only completely inform biological engineering, but it will overtake the utility of orthogonal, add-on circuits that were largely meant to operate independently of the systems in which they were embedded. Systems biology will eat synthetic biology.

This change in perspective and approach will be absolutely essential as biological engineering moves forward. The utility of orthogonal synthetic circuits is limited for a variety of reasons. First, by failing to take into account the unity of the system, the predictability of circuits must be limited. This can best be seen by thinking about a very simple synthetic circuit that has been around for

a long time: a plasmid that is used for protein overproduction. You can in fact embed a plasmid in a strain and induce protein production. From this vantage, the synthetic circuit has worked. However, it is difficult if not impossible to predict protein yield at the outset with any degree of certainty. Different proteins will express to different extents, will form aggregates to different extents, and will hamper the growth of the organism to different extents, to mention just a few variables that ultimately affect yield. This variability has nothing to do with whether a given “part” has been well characterized or not, and everything to do with the interaction of the part with the system as a whole. Protein overexpression interfaces with the cell’s machinery for transcription, translation, and degradation in myriad ways, and in the absence of a complete understanding of how the circuit interfaces with the system it will be difficult to ultimately predict how the circuit will work. Outcomes will range from making huge amounts of a protein to killing the cell outright. Of course the circuit can always be tinkered with to make more protein, but that was true before it was called a circuit and was merely a plasmid. What has synthetic biology brought to the modular, composable, scalable, and programmable overexpression of proteins that was not previously known? Nothing, because the tinkering that is now possible following the invention of the discipline is the same tinkering that was possible prior to its invention. Collins has pointed out that really tinkering is all that is necessary for engineering to prosper in biology, and he is of course correct. However, understanding and progress in this area will come from an increasingly detailed understanding of systems biology and integrated models of metabolism.

In discussions with Erik Winfree, a more charitable interpretation of the coexplosion of both systems and synthetic biology emerges, which is that these are different means to the same end—a predictable biology. The more complex system, the organism, is at present less predictable, and thus nominally orthogonal subsystems as built by synthetic biologists may allow us to initially develop better models. As those models become more sophisticated, they will of necessity take into account the discoveries and models developed through the study of systems biology.

A second, more important objection is that, by treating systems as though they were mere amalgams of components, synthetic biology ignores what I would argue is the central tenet of biology: organisms evolve. Given that an orthogonal circuit is really just an abstraction, as such circuits draw upon the metabolic, transcription, translation, and other resources of a cell, the circuit must of necessity change the fitness of a cell. To the extent that any synthetic circuit is used in the context of an evolutionary machine—a replicating cell—the possibility exists that mutations will arise that change both the cell and its added circuitry, for better or worse. While one alternative is just to not have cells replicate and to let circuits execute in preset bags of enzymes, this denies one of the great possibilities inherent in biology: the ability to generate larger amounts of complex matter from simpler substrates. An understanding of how to engineer biological systems will therefore proceed in large measure from a better understanding of

the costs and benefits of circuits to the system as a whole, again a province of systems (and evolutionary) biology.

Remaking Organismal Operating Systems

It is with this somewhat different perspective that we can move forward from thinking about how to make orthogonal circuits to thinking about how to manipulate the operating systems of organisms. Ultimately, synthetic biology is not modular, composable, scalable, or programmable because the operating system of biology does not support these features. The operating system of biology is a kludge, the result of billions of years of happenstance and compromise, and it cannot at this juncture be remade by the simple expedient of saying it’s not so.

But this does not mean that the operating system cannot be remade.

In looking at the operating system of biology, there is really only one component that is remotely akin to the synthetic biology dream of being modular, composable, scalable, and programmable, and that is DNA (or, more generally, nucleic acids). The simplicity of Watson-Crick base pairing and its implementation in a regular biopolymer meets all of these requirements. Unfortunately, this simplicity is destroyed by translation, which turns a regular biopolymer into many irregular ones. Obviously, from the point of view of organismal fitness, this is a good thing.

Thus, the question becomes, how can we remake the operating system of an organism so that the features of DNA are the features of not just one component but of all the components? My answer to this question is a limited one, but I think it suggests new directions. It seemed to me that since it is possible to have the base-pairing properties of DNA, but with different backbones, such as those seen in RNA, locked nucleic acids (LNAs), and especially peptide nucleic acids (PNAs), that it should be possible to extend base pairing to other components in biological operating systems. In particular, it should be possible to allow proteins to base pair in much the same way that PNAs would.

To this end, my lab has embarked on a journey to expand the genetic code to include four monomers that have nucleobase amino acids with side chains that have pairing properties equivalent to G, A, T, and C. Given the exploits and insights of the Schultz lab and others, this is an engineering feat that should be possible. The punctuation of the genetic code, stop codons, can be reprogrammed to code for amino acids (although ultimately with a systems-level cost), and four base codons (and now even a four base codon–reading ribosome) allow a (still uncertain) expansion of the code and the production of genetically augmented proteins.

Imposing Rationality on Biology

Whereas computers were built from the bottom up, with all of the rationality that engineers could bring to bear on both the hardware and software, biology

and its operating systems were presented to us as a fait accompli, and thus have to be engineered from the top down, in a way that can best be described as irrational, despite the fiction of synthetic biology. Biological systems evolved as kludges, and are engineered as kludges. While the underpinnings of synthetic biology seem to be to ignore these kludges, the hope of systems biology is to understand these kludges well enough that engineering short circuits and patches seem seamless.

As a small step, I envision that the addition of nucleobase amino acids to the genetic codes will allow us to begin to engineer biology in a truly rational manner. In my vision, there will be several substantive benefits to this fundamental remaking of a biological operating system. First, engineering protein–protein interactions will become more predictable, as it may be possible to uniquely code for such interactions via strings of nucleobase amino acids. Second, the same benefit will obviously accrue for protein–nucleic acid interactions. Transcription factors and RNA-binding proteins will no longer have semiprogrammable interaction domains, such as zinc fingers, but will instead be fully programmable. Third, it should be possible to rationally design cellular interactions and architectures via membrane proteins that again are tailed with nucleobase amino acids. And finally, this fundamental change in the operating system of a biological system should greatly speed the course of evolution. This conclusion comes from thinking about how transcription factors evolve to acquire new specificities. In general, there are multiple amino acid changes that must occur in order to bind a new DNA sequence, and these changes are not one-to-one, since the “code” for protein–nucleic acid interactions is three-dimensional rather than linear. By switching out the three-dimensional code for a nominally linear one, the number of sequence changes that must be acquired to bind adjacent to a promoter (or to form a new protein–protein interaction, or to make a connection to a new cell) will be proportionately reduced, and thus the acceptance of these changes (evolution) can potentially occur much more quickly.

For the longer term, I think there is the possibility that the operating systems of organisms can be changed in an even more fundamental manner. Merely adding nucleobase amino acids to the repertoire of the genetic code simplifies interactions from three-dimensional and irrational to linear and rational, but it does nothing about the way in which the associated machinery actually performs computations. Signal-transduction pathways will still pass information from the outside world to the inside of cells via a series of just-so scaffolds and complex conformational changes that will remain subject to relatively irrational engineering (and, given how I have tried to define rationality at a global level, this statement takes nothing away from the extraordinary successes of Lim and others in engineering signal transduction) (Bashor et al., 2008).

Where, then, can we find a more rational operating system for biology? In parallel with the “revolution” of synthetic biology, a true revolution in nucleic acid programming has been brewing at Caltech, led by Erik Winfree, Niles

Pierce, Peng Yin, and others (Seelig et al., 2006; Yin et al., 2008; Zhang et al., 2007). These pioneers have taken the seminal work by Adleman (1994), showing that nucleic acids can be used for computation, and further transformed it, showing that nucleic acids can be computers. To my mind, the distinction is subtle but important. An algorithm can be embedded in nucleic acids, as Adleman showed, but nucleic acids may not be the best platform for executing that algorithm (electronics beats biology, at least in terms of speed, on most computations). When instead we attempt to identify how nucleic acids can best compute—how their intrinsic properties can be exploited to do computation—then we begin to make nucleic acid computers of the sort that have emerged from the Caltech researchers. These computers rely on a very simple reaction, strand exchange, in which one nucleic acid binds to and forms a duplex with another. Duplex formation can in turn lead to the displacement of a previous duplex or the alteration of a programmed secondary structure. The displacement can be transient, with new binding sites (so-called toeholds) being revealed and paving the way to the formation of additional and/or different duplexes. Overall, extraordinarily complex executable circuits can be built in a fully rational manner, and these circuits have now been adapted to a variety of interesting computations, from amplification to determining square roots. The comparison with synthetic biology is instructive: while Weiss has pointed out that it has proven difficult to make synthetic circuits with more than six or so layers (Purnick and Weiss, 2009), Winfree has now readily generated a DNA circuit with this many layers and the obvious capability to scale much farther (Qian and Winfree, 2011).

I therefore fantasize about a biological system that is based on a rational biological computer, akin to the strand exchange circuits. As tempting as it is to suggest we should tear down biology and start afresh, this is well beyond my capabilities, and thus even my fantasies must be instantiated incrementally. It is possible to envision how strand-exchange RNA circuits could be embedded in organisms. The outcome of such circuitry could be the production of particular miRNAs that would feed back on the system, in much the same way that Benenson and Weiss have nicely shown that logical miRNAs can be embedded in the current biological operating system (Rinaudo et al., 2007). In couple with a fully realized suite of genetically augmented proteins containing nucleobase amino acids, the possibilities for superimposing a rational operating system on the current irrational one is manifest (albeit still incremental).

In moving from current biological operating systems to a fanciful new operating system built on genetically augmented proteins and strand exchange reactions, I have always been sensitive to the nature of the operating system that we already have (even while attempting to remake it). This “feeling for the organism” (to steal a phrase from McClintock) is not stealth vitalism but rather a sense of how engineering must proceed from the materials at hand. In this regard, while engineers can certainly become biologists and vice versa, each brings their preconceptions to the table in a manner that should be explicitly recognized. From

the point of view of a biologist, biological circuits are utterly unlike electronic circuits, irrespective of the analogies that can be drawn (Simpson et al., 2009). Electronic circuits operate at light speed, with spatially defined interconnectivities. Biological circuits operate via chemical diffusion, with molecular recognition and reactivity defining interconnectivity. While software is independent of hardware, and while both types of circuits can make, say, an oscillator, that’s no reason to suspect that both types of circuits will be able to operate in the same time regimes or with the same fidelity, since these features will stem in part from capabilities of the materials from which they are constructed. Thus, it may not be unreasonable to suggest that in engineering biology we must build from the capabilities of the materials involved, rather than trying to superimpose a foreign mindset—the current and likely transient Zeitgeist of synthetic biology—on those materials.

References

Adleman, L. M. 1994. Molecular computation of solutions to combinatorial problems. Science 266:1021-1024.

Bashor, C. J., N. C. Helman, S. Yan, and W. A. Lim. 2008. Using engineered scaffold interactions to reshape MAP kinase pathway signaling dynamics. Science 319:1539-1543.

Bayer, T. S., D. M. Widmaier, K. Temme, E. A. Mirsky, D. V. Santi, and C. A. Voigt. 2009. Synthesis of methyl halides from biomass using engineered microbes. Journal of the American Chemical Society 131:6508-6515.

Bromley, E. H., K. Channon, E. Moutevelis, and D. N. Woolfson. 2008. Peptide and protein building blocks for synthetic biology: From programming biomolecules to self-organized biomolecular systems. ACS Chemical Biology 3:38-50.

Gibson, D. G., J. I. Glass, C. Lartigue, V. N. Noskov, R. Y. Chuang, M. A. Algire, G. A. Benders, M. G. Montague, L. Ma, M. M. Moodie, C. Merryman, S. Vashee, R. Krishnakumar, N. Assad-Garcia, C. Andrews-Pfannkoch, E. A. Denisova, L. Young, Z. Q. Qi, T. H. Segall-Shapiro, C. H. Calvey, P. P. Parmar, C. A. Hutchison, 3rd, H. O. Smith, and J. C. Venter. 2010. Creation of a bacterial cell controlled by a chemically synthesized genome. Science 329:52-56.

Kemmer, C., M. Gitzinger, M. Daoud-El Baba, V. Djonov, J. Stelling, and M. Fussenegger. 2010. Self-sufficient control of urate homeostasis in mice by a synthetic circuit. Nature Biotechnology 28:355-360.

Khalil, A. S., and J. J. Collins. 2010. Synthetic biology: Applications come of age. Nature Reviews Genetics 11:367-379.

Martin, V. J., D. J. Pitera, S. T. Withers, J. D. Newman, and J. D. Keasling. 2003. Engineering a mevalonate pathway in Escherichia coli for production of terpenoids. Nature Biotechnology 21:796-802.

McGary, K. L., T. J. Park, J. O. Woods, H. J. Cha, J. B. Wallingford, and E. M. Marcotte. 2010. Systematic discovery of nonobvious human disease models through orthologous phenotypes. Proceedings of the National Academy of Sciences of the United States of America 107:6544-6549.

Purnick, P. E., and R. Weiss. 2009. The second wave of synthetic biology: From modules to systems. Nature Reviews Molecular Cell Biology 10:410-422.

Qian, L., and E. Winfree. 2011. Scaling up digital circuit computation with DNA strand displacement cascades. Science 332:1196-1201.

Rinaudo, K., L. Bleris, R. Maddamsetti, S. Subramanian, R. Weiss, and Y. Benenson. 2007. A universal RNAi-based logic evaluator that operates in mammalian cells. Nature Biotechnology 25:795-801.

Seelig, G., D. Soloveichik, D. Y. Zhang, and E. Winfree. 2006. Enzyme-free nucleic acid logic circuits. Science 314:1585-1588.

Simpson, Z. B., T. L. Tsai, N. Nguyen, X. Chen, and A. D. Ellington. 2009. Modelling amorphous computations with transcription networks. Journal of the Royal Society Interface 6(Suppl 4):S523-S533.

Smolke, C. D. 2009. Building outside of the box: iGEM and the BioBricks Foundation. Nature Biotechnology 27:1099-1102.

Wang, H. H., F. J. Isaacs, P. A. Carr, Z. Z. Sun, G. Xu, C. R. Forest, and G. M. Church. 2009. Programming cells by multiplex genome engineering and accelerated evolution. Nature 460:894-898.

Win, M. N., J. C. Liang, and C. D. Smolke. 2009. Frameworks for programming biological function through RNA parts and devices. Chemical Biology 16:298-310.

Yin, P., H. M. Choi, C. R. Calvert, and N. A. Pierce. 2008. Programming biomolecular self-assembly pathways. Nature 451:318-322.

Zhang, D. Y., A. J. Turberfield, B. Yurke, and E. Winfree. 2007. Engineering entropy-driven reactions and networks catalyzed by DNA. Science 318:1121-1125.

A4

SYNTHETIC BIOLOGY—A NEW GENERATION OF BIOFILM BIOSENSORS

James Chappell²⁸and Paul S. Freemont^27,29

Biofilms: A Survival Strategy

The notion that bacteria live an autonomous, independent, and planktonic lifestyle has been radically challenged with the realization of the abundance of bacterial communities known as biofilms (Hall-Stoodley et al., 2004). Biofilms are structured communities of adherent microorganisms encased in a self-produced complex extracellular polymeric substance (EPS) matrix as shown in Figure A4-1 (Costerton et al., 1999). The advent of microscopy techniques has revealed the complex nature of these structured communities, containing networks of channels for nutrient supply and organized three-dimensional structures. Moreover, bacteria within these biofilms are profoundly physiologically distinct from their planktonic counterparts, and they function in a coordinated manor as a cooperative consortium, more similar to that of multicellular organisms than a unicellular organism (Hall-Stoodley et al., 2004; Lindsay and von Holy, 2006; Parsek and Singh, 2003). The ability to form biofilms is almost ubiquitously found within the bacteria domain of life and indeed within many other classes of microorganisms, including fungi, yeasts, algae, and protoza (Van Houdt and Michiels, 2010). Evidence suggests that biofilm formation is an ancient ability,

_______________

²⁸ Centre for Synthetic Biology and Innovation, Imperial College London, South Kensington Campus, London SW7 2AZ, United Kingdom.

²⁹ Correspondence: e-mail: p.freemont@imperial.ac.uk, telephone: + 44 (0)2075945327.

FIGURE A4-1 An electron micrograph of Staphylococcus aureus bacteria biofilms on the luminal surface of an indwelling catheter.
SOURCE: Image obtained from Public Health Image Library. Provided courtesy of R. Donlan, and J. Carr, Centers for Disease Control and Prevention. Photo by J. Carr (2005).

with biofilm microcolonies being identified in fossils from 3.3 to 3.4 billion years ago (Hall-Stoodley et al., 2004; Westall et al., 2001). Taken altogether, biofilm formation is an evolutionary conserved and widespread phenomenon.

The advantages of bacteria to abandon their independent lifestyle for that of a community is simply that it provides a more protective mode of growth. Biofilms allow bacteria to colonize and sustain favorable niches independent of larger environmental changes. For example, the EPS, a key component of biofilms, can serve as storage for nutrients and water, helping to stabilize local resources (Costerton et al., 1987; Donlan, 2002; Lindsay and von Holy, 2006). In addition, biofilms provide a defensive role, acting as a physical barrier from environmental dangers such as ultraviolet light, metal toxicity, acid exposure, dehydration, and even immune defenses such as phagocytosis (Donlan, 2002; Lindsay and von Holy, 2006; Parsek and Singh, 2003). These selective advantages of biofilm formation are obvious within the context of hostile and dynamic environments such as that of ancient Earth (Hall-Stoodley et al., 2004). Yet even in today’s relatively temperate environment, the formation of biofilms is still selectively advantageous and constitutes a major component of bacterial biomass in natural environments (Costerton et al., 1978, 1999; Lindsay and von Holy, 2006). Bacteria’s attempt

to survive by the formation of biofilms extends beyond just natural environments and provides bacteria protection in many man-made environments (Hall-Stoodley et al., 2004). However, the focus of this review is the presence of pathogenic bacterial biofilms within a clinical context.

Biofilms in a Clinical Context

It is only in recent years that the importance of biofilms in clinical settings as a source of pathogenic bacteria has been realised (Lindsay and von Holy, 2006). The persistence of biofilms in hospitals is undoubtedly a significant contribution to hospital-acquired infections or nosocomial infections (Costerton et al., 1999; Donlan and Costerton, 2002; Hall-Stoodley and Stoodley, 2009; Hall-Stoodley et al., 2004; Lindsay and von Holy, 2006). Indeed it has been estimated that biofilms contribute to 65 percent of nosocomial infections (Potera, 1999; Smith and Hunter, 2008). Table A4-1 shows the most frequent pathogenic strains associated with nosocomial infections, the most frequent and problematic being Pseudomonas aeruginosa, Staphylococcus aureus, and S. epidermidis (Costerton et al., 1999; Hall-Stoodley et al., 2004). Recent studies estimate that at any one time 9 percent of all in-patients in England and Wales (United Kingdom) suffer from hospital-acquired infections, resulting in approximately 5,000 deaths per year. The costs associated with these hospital-related infections for the U.K. National Health Service are estimated at £1 billion per year (Smith and Hunter, 2008).

TABLE A4-1 The Most Common Causes of Nosocomial Infections

Although many nonpathogenic bacteria are able to form innocuous biofilms, it is the formation of pathogenic biofilms that is particularly important for contributing to the high level of hospital-acquired infections. It has been suggested that biofilm formation is analogous to a virulence factor, where biofilm formation increases the likelihood of a pathogen causing an infection.

Three general mechanisms have been proposed to explain the increase of virulence seen for pathogenic bacteria in a biofilm microenvironment (Hall-Stoodley and Stoodley, 2005). These are the survival and transmission of planktonic patho-

gens from biofilms, the phenotypic heterogeneity of biofilm populations and the potential evolution of infectious phenotypes, and the role of biofilm formation in the regulation of virulence mechanisms. Each of these mechanisms is discussed further in Box A4-1.

The seriousness of this problem has provided strong motivations for developing new antibiofilm therapies and detection methods. The development of new antibiofilm therapies is beyond the scope of this paper, although there are a number of extensive reviews (Lynch and Abbanat, 2010). Here we focus on cur-

rent methods for biofilm detection, as well as their limitations, and the prospect of synthetic biology to develop completely novel methods for the detection of pathogenic biofilms.

Current Methods of Biofilm Detection

The ability of biofilms to form on a variety of abiotic and biotic surfaces has resulted in colonization of many clinical environments. Detection of these biofilms has obvious advantages: to direct suitable sanitation protocols in biofilm niches, to indicate replacement of any contaminated medical devices such as indwelling catheters, and to direct suitable medication if patient infection ensues. It has now become common practice in hospitals to sample both environmental surfaces and patients, typically urine, for the presence of pathogenic biofilms. There are numerous methods used for detection, including conventional plate counting, phenotypic screens, microscopy methods, and genotypic methods. Two of the most established methods for biofilm detection used in laboratories and clinical settings are the genotypic and phenotypic methods that are described in more detail below.

Phenotypic methods rely on culturing bacterial samples and screening these cultures for the phenotype of biofilm producers. These methods rely on the hypothesis that biofilm formation is a marker of virulence. Three variants of phenotypic detection have been commonly used: the tissue culture plate, the tube method, and the Congo red agar (CRA) method. These methods rely on indirectly assessing biofilm producers by accumulation of biofilm components such as EPS, a polysaccharide involved in cell-cell adhesion and essential for biofilm formation.

Genotypic methods are based upon PCR amplification of specific genetic markers associated with pathogenic biofilms. The most notable example is the use of the ica cluster to identify biofilm-producing strains of Staphylococcus aureus and S. epidermidis. The ica cluster contains genes that are essential for these strains to produce EPS. In addition, PCR is commonly used to assess antibiotic resistance; for instance, the presence of the mecA gene encoding methicillin resistance can be used to identify strains of methicillin-resistant S. aureus.

Each of these methods has its own advantages and limitations. The major limitations of these are described below:

• Expertise and cost. Although techniques such as PCR and microscopy techniques have become commonplace in molecular biology laboratories, in hospitals, particularly in developing countries, access to the equipment and expertise required for analysis can be limited (Bose, 2009).
• In situ and point-of-care detection. Sampling of biofilms on hospital surfaces can be abrasive to attached cells and may result in damage to

• cells, rendering them nonculturable (Lindsay and von Holy, 2006). In addition, sampling of indwelling medical devices such as catheters requires the removal of catheters, whether a biofilm is present or not. This can result in unnecessary stress for patients and wastefulness of such devices (Donlan and Costerton, 2002). Finally, the recovery efficiency of sampling methods commonly used is largely unknown, leaving the possibility of not detecting a representative population.

• Reliability and reproducibility. The genotypic methods are generally considered more reliable and sensitive. However, they can be subject to false positives. This is due to the ica cluster not being absolutely required for biofilm formation. It has been shown that polysaccharide intercellular adhesin (PIA) synthesis from the ica operon alone is not sufficient to produce biofilms and that biofilms can form without producing PIA in staphylococci (Chokr et al., 2006).

  For the phenotypic detection methods there is limited sensitivity, reliability, and reproducibility compared to genetic detection. There have been numerous independent studies comparing the different phenotypic methods, which have shown varying success rates using the different phenotypic methods. For example, studies assessing the sensitivities of the CRA method have shown a detection success ranging from 5.26 to 89 percent (Arciola et al., 2006; Bose, 2009; Knobloch et al., 2002; Mathur et al., 2006; Oliveira and Cunha, 2010). In addition, the phenotypic methods are commonly prone to false positives and nondetection of “weak” biofilm-forming strains.

• Time of response. The phenotypic detection methods rely upon the culturing of microorganisms collected from samples. These incubation periods are typically 24 to 48 hours and cultures are often analyzed offsite, increasing time for biofilms to develop (Bose and Ghosh, 2011).
• Informative. Neither the genotypic nor the phenotypic methods elicit an in situ biofilm structure. Although detection of the pathogenic biofilms indicates the potential for the presence of biofilms, it is not an accurate reflection of the actual microenvironment. This can lead to an inaccurate detection in the in situ environment and false diagnosis.

Whole-Cell Biosensors: Bridging the Detection Gap

As discussed above there is a strong motivation for the development of in situ detection methods with increased response time, specificity, and more informative outputs. A potential solution is the development of whole-cell biosensors, using genetically engineered biological cells to directly detect pathogenic biofilms. An essential ability for cells to function is the ability to sense and respond to a diverse range of environmental signals from environmental conditions such as temperature change to the detection of nutrients in the environment. This ability can be

broken down to two subfunctions, that of sensing and that of response (signal transduction). Whole-cell biosensors are essentially composed of the same two elements integrated into microorganisms: sensing elements detecting the analyte of interest and transducing elements that are usually coupled to the production of a detectable output, for example, expression of reporter proteins such as natural fluorescent proteins.

Some of the advantages of whole-cell biosensors are that they detect a variety of biological, organic, and nonorganic compounds. In addition, they can produce a number of detectable outputs that can be easily quantifiable, for example, fluorescent proteins measured by a fluorometer, dyes and pigments measured by spectroscopy and by eye, and electrochemical signals such as pH measured by litmus paper or a pH meter. These easily measured outputs require little expertise to use or to interpret. In addition, the timescale for biosensors is generally rapid, with signals being transduced in seconds to generate detectable signals within minutes to hours. In addition, compared to other in vitro biosensors like widely used enzymes or antibodies, whole-cell biosensors are easily produced by cell culture, in contrast to the costly and time-consuming purification of enzymes and antibodies.

Another advantage of whole-cell biosensors can be highlighted using the example of detecting pathogenic biofilms. Genotypic methods detect at a microscopic level the DNA genetic elements that are required for the formation of biofilms rather than the products of these genes, which can therefore lead to false positives. Phenotypic detection detects the products of biofilms but at a macroscopic level. Whole-cell biosensors offer a way to bridge the gap between these two methods, allowing sensing and evaluation at the level of microbial habitats. Whole-cell biosensors could allow the detection of new markers of biofilms to provide more effective detection methods. In recent years, there has been increased interest in the development of whole-cell biosensors, particularly within the field of synthetic biology. In the remainder of this review, the potential of synthetic biology for creating more effective biosensors for the detection of pathogenic biofilms is discussed.

Synthetic Biology: A New Generation of Biofilm Biosensors?

Synthetic biology is an application-driven field attempting to apply a rational engineering approach to the redesign of biological systems, to produce valuable and novel biological functions. Synthetic biology can be thought of as a natural evolution of biotechnology as opposed to being a separate field, although it does offer a novel and exciting approach. In the past 40 years biotechnology has undoubtedly produced many valuable biological products, yet current approaches tend to be lengthy due to the ad hoc nature of the approach. In addition, there is little lateral transfer of knowledge between projects, meaning knowledge gained in research and development in one area may not assist the post hoc development of projects in other, unrelated areas.

Synthetic biology aims to move away from this ad hoc design by providing a conceptual framework based on systematic design and rational engineering, so that redesigning biological systems would become more equivalent to that of redesigning a plane or a car. Indeed, it is the adoption of an engineering framework that has revolutionized the design and output of applications in the fields of mechanical, electrical, civil, and aeronautic engineering, and it is proposed that such an approach will do the same for biotechnology. The focus of this paper is to discuss the potential of whole-cell biosensors and the synthetic biology approach both in general and for the specific application of detection of pathogenic biofilms, the need of which was highlighted earlier.

Three major advantages of the synthetic biology approach for biosensor designs are highlighted: the advantages of the engineering approach, the development of novel sensing elements, and the complexity synthetic biology could enable.

Putting the Engineering into Genetic Engineering

Modern engineering relies on designing systems from catalogues of well-defined standard parts and higher-level devices that can be assembled into larger systems in a standardized manner. For electrical engineers, catalogues such as RS Components and Farnell provide well-defined parts and devices, each associated with data sheets of the functional characteristics. System designs can then be assessed by applying mathematical models in combination with known behavior characteristics, to computationally simulate their overall functions. This allows rapid assessment in silico of designs that meet the required specifications of design. Through iterations of this engineering approach or engineering cycle, efficient production of well-characterized and robust end products, from electronics to buildings, can be produced.

Synthetic biology is trying to adopt this engineering cycle and develop the tools required for high-throughput engineering of biological systems (Figure A4-2) (Gulati et al., 2009; Kitney et al., 2007). However, living systems are not computer chips and thus it is important to note that synthetic biology aims to provide a conceptual engineering approach that will allow biological engineering to be easier and more predictable. Living systems are intricate, formed from a complex network of interacting chemical components driven by the ability to self-replicate, adapt, and survive. Therefore, in living systems, context dependency and stochastic behavior are common features and therefore the engineering of predictable new cellular systems with defined functions is challenging. Nevertheless, the field of synthetic biology aims to provide a series of foundational technologies and a defined framework that will enable robust biological design and overcome these clear challenges. At present much emphasis in synthetic biology is focused on engineering single-cell organisms like Escherichia coli or yeast, given that our understanding of these organisms is more mature than for

FIGURE A4-2 The engineering cycle as an approach for synthetic biology. This adaption includes the four stages of defining specifications, design of biological devices, in silico modeling of the potential designs, and characterisation of performances.

multicellular systems. With this in mind, a synthetic biology engineering cycle is discussed below.

Engineering cycle

1. Specification The first stage in the engineering cycle is to define the specifications required. For example, a previous whole-cell biosensor designed to detect arsenic in polluted water defined the detection range as defined by the World Health Organization recommendation of safe drinking water at 10 ppb arsenic. Other specifications can be tunable, for example, threshold, time of response, particular microbial organisms, or host chassis to be used.
2. Design Once the specifications have been defined, genetic devices have to be designed to give the functions of interest to host cells, often microbes. For the design of a genetic device, the principle of abstraction is applied to help separate out unnecessary layers of complexity (Endy, 2005). At the most complex layer, the design of a biological device involves the writing of a DNA nucleotide sequence to give desired func-

3. tion when implanted in a cell. However, an abstract view can be taken and these sequences modularized into functional parts such as promoters, gene coding sequences, ribosome binding sites, and terminators. These modules of DNA allow the functional separation of biological parts (BioParts) and allow a higher level of design, where the underlying complexity of the DNA sequence can be hidden (Endy, 2005).

   Collaborative and international efforts have been made to make open-source registries to document and physically store a huge variety of BioParts and devices, most notably being the BioBricks registry (http://partsregistry.org) and the International Open Facility Advancing Biotechnology (BIOFAB). Two important criteria of BioParts in these registries are the adherence of standards and the documentation of characterisation data. The BioBricks foundation has proposed a “request for comment” process, encouraging the synthetic biology community to list standards for physically joining DNA and protocols such as for characterisation of BioParts.

   These catalogues are beginning to provide a huge toolbox of biological parts for the engineering of biological systems, from basic BioParts such as promoters, ribosome binding sites, genes, and terminators to more complex composite devices such as toggle switches, oscillators, logic gates, and cell-to-cell communication systems. Moreover, the standards required for submission to these registries ensure compatibility and interchangeability. For example, the BioBricks registry has defined standard restriction sites flanking all parts and devices, which has allowed standardized methods of assembly where any two parts can be combined by following use of a standardized protocol. Finally, the documentation of characterisation data allows transfer of knowledge and experience gained from separate projects to be used in the evaluation of potential designs (Arkin, 2008).

4. Modeling After the design stage, in silico computational simulation of mathematical models representative of the genetic devices is performed. In the 1960s the mathematical logic in gene regulation of the lac operon was demonstrated (Andrianantoandro et al., 2006; Monod and Jacob, 1961). Since then, systems biology has transformed our understanding of biology using representative mathematical models based on quantitative data, most notably stochastic and ordinary differential equations. These modeling techniques, coupled with the use of experimentally defined characteristics defining model parameters, can predict the functional characteristics of a design without the need to undergo cloning or de novo synthesis, or experimental characterisation both of which increase overall time and cost of development. Ellis et al. (2009) demonstrated the use of computational modeling to predict the behavior of feedforward loops in yeast. To do so, a library of inducible promoters

4. with known minimum and maximum outputs (expression levels) were used in the device design. The characterisation data were incorporated into the modeling, the predictions of which were shown to accurately reflect the experimentally derived behaviors of these devices (Ellis et al., 2009).

5. Quality control The idea that a device’s behavior can be fully predicted by modeling the known characteristics of individual parts is obviously an oversimplification. It is inevitable that unexpected emergent properties will result when individual parts are put together that have been characterised in isolation and different contexts such as a switch in host chassis of E. coli to Bacillus subtilis (Arkin, 2008; Serrano, 2007). Unexpected interactions can occur from the context of the other BioParts in the device or even the host chassis. For example, in mRNA the untranslated regions of promoters can form secondary structures with ribosome binding sites of mRNA sequences modifying the function of this BioPart (Arkin, 2008).

   These factors necessitate a quality control step, first to assess how well the device meets the specification but also to gain a greater understanding of the device being designed. Although approaches such as BioPart characterisation and modeling might not give an absolute prediction of device function as seen in other fields such as electronics, it nonetheless reduces the number of potential designs that need to be explored and therefore increases the overall efficiency.

Novel Biosensor Targets

As mentioned earlier, biosensors are composed of two subfunctions, that of sensing and that of signal transduction. The example of quorum sensing has already been discussed, where bacteria sense the local population densities based on the concentration of autoinducers and, once a threshold is reached, gene expression is induced. In this example, transcription factors act as the sensing elements binding to autoinducers and the quorum-sensing responsive promoters act as signal-transducer elements, defining the threshold of response and converting the autoinducer concentration to transcriptional output.

The bacterial toolbox of sensing elements is mostly mRNA and protein based, both of which can form complex tertiary structures able to bind a myriad of targets. Coupled to these sensing elements, the signal-transducing elements are generally transcriptional, translational, and posttranslational. Sensing modules and transducer modules can be separate biological components such as in quorum sensing, or within the same biological molecule such as riboswitches that contain an mRNA aptamer element able to sense and convert this sensing to translational output by sequestering or releasing a proximal ribosome binding site.

When considering the design of biosensors, it is essential to identify both

the sensing and the transducing elements. In general, synthetic biologists have employed three strategies to use these elements: first, to import exogenous sensing elements into new biological host chassis; second, to use and rewire existing endogenous sensing elements to detect new signals; and finally, in the production of novel elements not found in nature.

Three examples are described in Box A4-2 to explain each of these strategies. In addition, each example provides an illustration of biosensors functioning at the level of transcription, translation, and posttranslation.

Increasing Complexity

Within the field of synthetic biology, a variety of synthetic genetic devices have been developed. Many of these devices have been inspired by those found in electronics and include toggle switches (Atkinson et al., 2003; Bayer and Smolke, 2005; Deans et al., 2007; Dueber et al., 2003; Friedland et al., 2009; Gardner et al., 2000; Ham et al., 2006, 2008; Kramer and Fussenegger, 2005; Kramer et al., 2004), memory elements (Ajo-Franklin et al., 2007; Basu et al., 2004; Friedland et al., 2009; Ham et al., 2006, 2008), pulse generators (Basu et al., 2004), time-delayed circuits (Ellis et al., 2009; Weber et al., 2007), oscillators (Atkinson et al., 2003; Danino et al., 2010; Elowitz and Leibler, 2000; Fung et al., 2005; Stricker et al., 2008; Tigges et al., 2009), and logic gates (Anderson et al., 2007; Guet et al., 2002; Rackham and Chin, 2005; Rinaudo et al., 2007; Win and Smolke, 2008). These synthetic genetic devices are aiding the design of more complex biological systems the likes of which have not been previously seen. For the application of whole-cell biosensors, the implementation of these genetic devices could allow more complex and informative biosensors to be developed. For example, consider the use of logic gates for biosensor design. Logic gates are devices that perform a logical operation based on one or more inputs to produce a signal output. A frequently featured logic gate implemented in synthetic biology is the AND gate. An AND gate integrates two or more signals and will only give an output when all inputs are present. The principle and a biological example of an AND gate is shown in Figure A4-3.

The example of logic gates highlights the potential complexity of design that synthetic biology could offer for biosensor design. In general, multiple signal integration would allow fewer false positives by relying on more than one signal for detection and thus increasing the reliability. For example, the use of NOT gates could be used to remove known signals contributing to false positives. In addition, by being able to integrate multiple signals, logic gates could offer more complex and informative detection.

A more relevant illustration with regard to developing biosensors for detecting biofilms can be highlighted when we take into account mixed-species biofilms. Consider the case of biofilms on indwelling urinary catheters. A study of 106 biofilms found on catheters showed 14 species of bacteria to be commonly present (Macleod and Stickler, 2007; Stickler, 2008). Furthermore, although

single-species biofilms were found, most were mixed-species biofilms—most frequently containing Enterococcus faecalis, Escherichia coli, Pseudomonas aeruginosa, and Proteus mirabilis. With the use of logic gates and multiple signal integration, synthetic biology could develop biosensors to respond not only to multiple targets but also with the ability to distinguish between species. This would allow general detection of biofilms as well as species-dependent detection that could aid correct treatment should the infection worsen.

Summary

Synthetic biology is a newly developing field that aims to provide an engineering framework for the predictable construction of new biological systems. The most immediate exemplars relate to biosensors where synthetic biology has already produced several living cell biosensors that act robustly and predictably. The application of synthetic biology design tools will enable the development of a new generation of biosensors to detect pathogenic biofilms. The implementation of these designs will have a profound impact in many areas including the control of hospital-acquired infections.

FIGURE A4-3 The principle and biological example of an AND gate (A) The symbolic representation of an AND gate and (B) a truth table describing the logic of an AND gate with regard to two inputs. Only when both inputs are present is an output produced. (C) An example of a biological AND gate previously described by Anderson et al. (2007). The AND gate is based around the interplay between two genes. One of these genes is the T7 RNA polymerase containing two amber stop codons (denoted by asterisks), meaning that translation of full-length T7 polymerase is inhibited (T7ptag). Only when the supD gene is transcribed are the amber stop codons translated into serines allowing full-length T7 polymerase to be translated and drive expression of the output from a T7 polymerase regulated promoter. This example represents an AND gate taking two transcriptional inputs and integrating these to a single transcriptional output.

References

Ajo-Franklin, C. M., D. A. Drubin, J. A. Eskin, E. P. Gee, D. Landgraf, I. Phillips, and P. A. Silver. 2007. Rational design of memory in eukaryotic cells. Genes & Development 21:2271-2276.

Aleksic, J., F. Bizzari, Y. Cai, B. Davidson, K. De Mora, S. Ivakhno, S. L. Seshasayee, J. Nicholson, J. Wilson, A. Elfick, C. French, L. Kozma-Bognar, H. Ma, and A. Millar. 2007. Development of a novel biosensor for the detection of arsenic in drinking water. IET Synthetic Biology 1(1-2):87-90.

Anderson, J. C., C. A. Voigt, and A. P. Arkin. 2007. Environmental signal integration by a modular AND gate. Molecular Systems Biology 3:133.

Andrianantoandro, E., S. Basu, D. K. Karig, and R. Weiss. 2006. Synthetic biology: New engineering rules for an emerging discipline. Molecular Systems Biology 2:2006.0028.

Arciola, C. R., D. Campoccia, L. Baldassarri, M. E. Donati, V. Pirini, S. Gamberini, and L. Montanaro. 2006. Detection of biofilm formation in Staphylococcus epidermidis from implant infections. Comparison of a PCR-method that recognizes the presence of ica genes with two classic phenotypic methods. Journal of Biomedical Materials Research Part A 76:425-430.

Arkin, A. 2008. Setting the standard in synthetic biology, Nature Biotechnology 26:771-774.

Asad, S., and S. M. Opal. 2008. Bench-to-bedside review: Quorum sensing and the role of cell-to-cell communication during invasive bacterial infection. Critical Care 12:236.

Atkinson, M. R., M. A. Savageau, J. T. Myers, and A. J. Ninfa. 2003. Development of genetic circuitry exhibiting toggle switch or oscillatory behavior in Escherichia coli. Cell 113:597-607.

Basu, S., R. Mehreja, S. Thiberge, M. T. Chen, and R. Weiss. 2004. Spatiotemporal control of gene expression with pulse-generating networks. Proceedings of the National Academy of Sciences of the United States of America 101:6355-6360.

Baumgartner, J. W., C. Kim, R. E. Brissette, M. Inouye, C. Park, and G. L. Hazelbauer. 1994. Trans-membrane signalling by a hybrid protein: Communication from the domain of chemoreceptor Trg that recognizes sugar-binding proteins to the kinase/phosphatase domain of osmosensor EnvZ. Journal of Bacteriology 176:1157-1163.

Bayer, T. S., and C. D. Smolke. 2005. Programmable ligand-controlled riboregulators of eukaryotic gene expression. Nature Biotechnology 23:337-343.

Boles, B. R., M. Thoendel, and P. K. Singh. 2004. Self-generated diversity produces “insurance effects” in biofilm communities. Proceedings of the National Academy of Sciences of the United States of America 101:16630-16635.

Bose, S. 2009. Detection of biofilm producing staphylococci: Need of the hour. Journal of Clinical and Diagnostic Research 3:1915-1920.

Bose, S., and A. K. Ghosh. 2011. Biofilms: A challenge to medical science. Journal of Clinical and Diagnostic Research 5:127-139.

Brooks, S. E., M. A. Walczak, R. Hameed, and P. Coonan. 2002. Chlorhexidine resistance in antibiotic-resistant bacteria isolated from the surfaces of dispensers of soap containing chlorhexidine. Infection Control & Hospital Epidemiology 23:692-695.

Chokr, A., D. Watier, H. Eleaume, B. Pangon, J. C. Ghnassia, D. Mack, and S. Jabbouri. 2006. Correlation between biofilm formation and production of polysaccharide intercellular adhesin in clinical isolates of coagulase-negative staphylococci. International Journal of Medical Microbiology 296:381-388.

Colman-Lerner, A., A. Gordon, E. Serra, T. Chin, O. Resnekov, D. Endy, C. G. Pesce, and R. Brent. 2005. Regulated cell-to-cell variation in a cell-fate decision system. Nature 437:699-706.

Costerton, J. W., G. G. Geesey, and K. J. Cheng. 1978. How bacteria stick. Scientific American 238:86-95.

Costerton, J. W., K. J. Cheng, G. G. Geesey, T. I. Ladd, J. C. Nickel, M. Dasgupta, and T. J. Marrie. 1987. Bacterial biofilms in nature and disease. Annual Review of Microbiology 41:435-464.

Costerton, J. W., P. S. Stewart, and E. P. Greenberg. 1999. Bacterial biofilms: A common cause of persistent infections. Science 284:1318-1322.

Danino, T., O. Mondragon-Palomino, L. Tsimring, and J. Hasty. 2010. A synchronized quorum of genetic clocks. Nature 463:326-330.

Davies, D. G., M. R. Parsek, J. P. Pearson, B. H. Iglewski, J. W. Costerton, and E. P. Greenberg. 1998. The involvement of cell-to-cell signals in the development of a bacterial biofilm. Science 280:295-298.

Deans, T. L., C. R. Cantor, and J. J. Collins. 2007. A tunable genetic switch based on RNAi and repressor proteins for regulating gene expression in mammalian cells. Cell 130:363-372.

Donlan, R. M. 2002. Biofilms: Microbial life on surfaces. Emerging Infectious Diseases 8:881-890.

Donlan, R. M., and J. W. Costerton. 2002. Biofilms: Survival mechanisms of clinically relevant microorganisms. Clinical Microbiology Reviews 15:167-193.

Drenkard, E., and F. M. Ausubel. 2002. Pseudomonas biofilm formation and antibiotic resistance are linked to phenotypic variation. Nature 416:740-743.

Dueber, J. E., B. J. Yeh, K. Chak, and W. A. Lim. 2003. Reprogramming control of an allosteric signaling switch through modular recombination. Science 301:1904-1908.

Ellis, T., X. Wang, and J. J. Collins. 2009. Diversity-based, model-guided construction of synthetic gene networks with predicted functions. Nature Biotechnology 27:465-471.

Elowitz, M. B., and S. Leibler. 2000. A synthetic oscillatory network of transcriptional regulators. Nature 403:335-338.

Endy, D. 2005. Foundations for engineering biology. Nature 438:449-453.

Friedland, A. E., T. K. Lu, X. Wang, D. Shi, G. Church, and J. J. Collins. 2009. Synthetic gene networks that count. Science 324:1199-1202.

Fung, E., W. W. Wong, J. K. Suen, T. Bulter, S. G. Lee, and J. C. Liao. 2005. A synthetic gene-metabolic oscillator. Nature 435:118-122.

Fuqua, W. C., S. C. Winans, and E. P. Greenberg. 1994. Quorum sensing in bacteria: The LuxR-LuxI family of cell density-responsive transcriptional regulators. Journal of Bacteriology 176: 269-275.

Gander, S. 1996. Bacterial biofilms: Resistance to antimicrobial agents. Journal of Antimicrobial Chemotherapy 37:1047-1050.

Gardner, T. S., C. R. Cantor, and J. J. Collins. 2000. Construction of a genetic toggle switch in Escherichia coli. Nature 403:339-342.

Gilbert, P., D. G. Allison, and A. J. McBain. 2002. Biofilms in vitro and in vivo: Do singular mechanisms imply cross-resistance? Journal of Applied Microbiology 92(Suppl):98S-110S.

Guet, C. C., M. B. Elowitz, W. Hsing, and S. Leibler. 2002. Combinatorial synthesis of genetic networks. Science 296:1466-1470.

Guinto, C. H., E. J. Bottone, J. T. Raffalli, M. A. Montecalvo, and G. P. Wormser. 2002. Evaluation of dedicated stethoscopes as a potential source of nosocomial pathogens. American Journal of Infection Control 30:499-502.

Gulati, S., V. Rouilly, X. Niu, J. Chappell, R. I. Kitney, J. B. Edel, P. S. Freemont, and A. J. deMello. 2009. Opportunities for microfluidic technologies in synthetic biology. Journal of the Royal Society Interface 6(Suppl 4):S493-S506.

Hall-Stoodley, L., and P. Stoodley. 2005. Biofilm formation and dispersal and the transmission of human pathogens. Trends in Microbiology 13:7-10.

——. 2009. Evolving concepts in biofilm infections. Cellular Microbiology 11:1034-1043.

Hall-Stoodley, L., J. W. Costerton, and P. Stoodley. 2004. Bacterial biofilms: From the natural environment to infectious diseases. Nature Reviews Microbiology 2:95-108.

Ham, T. S., S. K. Lee, J. D. Keasling, and A. P. Arkin. 2006. A tightly regulated inducible expression system utilizing the fim inversion recombination switch. Biotechnology and Bioengineering 94:1-4.

——. 2008. Design and construction of a double inversion recombination switch for heritable sequential genetic memory. PLoS One 3:e2815.

Jefferson, K. K. 2004. What drives bacteria to produce a biofilm? FEMS Microbiology Letters 236:163-173.

Kirisits, M. J., L. Prost, M. Starkey, and M. R. Parsek. 2005. Characterization of colony morphology variants isolated from Pseudomonas aeruginosa biofilms. Applied and Environmental Microbiology 71:4809-4821.

Kitney, R. I., P. S. Freemont, and V. Rouilly. 2007. Engineering a molecular predation oscillator. IET Synthetic Biology 1:68-70.

Knobloch, J. K., M. A. Horstkotte, H. Rohde, and D. Mack. 2002. Evaluation of different detection methods of biofilm formation in Staphylococcus aureus. Medical Microbiology and Immunology 191:101-106.

Kramer, B. P., and M. Fussenegger. 2005. Hysteresis in a synthetic mammalian gene network. Proceedings of the National Academy of Sciences of the United States of America 102:9517-9522.

Kramer, B. P., A. U. Viretta, M. Daoud-El-Baba, D. Aubel, W. Weber, and M. Fussenegger. 2004. An engineered epigenetic transgene switch in mammalian cells. Nature Biotechnology 22:867-870.

Legewie, S., H. Herzel, H. V. Westerhoff, and N. Bluthgen. 2008. Recurrent design patterns in the feedback regulation of the mammalian signalling network. Molecular Systems Biology 4:190.

Lesic, B., F. Lepine, E. Deziel, J. Zhang, Q. Zhang, K. Padfield, M. H. Castonguay, S. Milot, S. Stachel, A. A. Tzika, R. G. Tompkins, and L. G. Rahme. 2007. Inhibitors of pathogen intercellular signals as selective anti-infective compounds. PLoS Pathogens 3:1229-1239.

Lindsay, D., and A. von Holy. 2006. Bacterial biofilms within the clinical setting: What healthcare professionals should know. Journal of Hospital Infection 64:313-325.

Looger, L. L., M. A. Dwyer, J. J. Smith, and H. W. Hellinga. 2003. Computational design of receptor and sensor proteins with novel functions. Nature 423:185-190.

Lynch, A. S., and D. Abbanat. 2010. New antibiotic agents and approaches to treat biofilm-associated infections. Expert Opinion on Therapeutic Patents 20:1373-1387.

Macleod, S. M., and D. J. Stickler. 2007. Species interactions in mixed-community crystalline biofilms on urinary catheters. Journal of Medical Microbiology 56:1549-1557.

Mahajan-Miklos, S., L. G. Rahme, and F. M. Ausubel. 2000. Elucidating the molecular mechanisms of bacterial virulence using non-mammalian hosts. Molecular Microbiology 37:981-988.

Marchisio, M. A., and F. Rudolf. 2011. Synthetic biosensing systems. International Journal of Biochemistry & Cell Biology 43:310-319.

Mathur, T., S. Singhal, S. Khan, D. J. Upadhyay, T. Fatma, and A. Rattan. 2006. Detection of biofilm formation among the clinical isolates of staphylococci: An evaluation of three different screening methods. Indian Journal of Medical Microbiology 24:25-29.

Monod, J., and F. Jacob. 1961. Teleonomic mechanisms in cellular metabolism, growth, and differentiation. Cold Spring Harbor Symposia on Quantitative Biology 26:389-401.

Oliveira, A., and M. de L. Cunha. 2010. Comparison of methods for the detection of biofilm production in coagulase-negative staphylococci. BMC Research Notes 3:260.

O’Toole, G. A. 2002. Microbiology: A resistance switch. Nature 416:695-696.

Parsek, M. R., and P. K. Singh. 2003. Bacterial biofilms: An emerging link to disease pathogenesis. Annual Review of Microbiology 57:677-701.

Popat, R., S. A. Crusz, and S. P. Diggle. 2008. The social behaviours of bacterial pathogens. British Medical Bulletin 87:63-75.

Potera, C. 1999. Forging a link between biofilms and disease. Science 283:1837, 1839.

Rackham, O., and J. W. Chin. 2005. Cellular logic with orthogonal ribosomes. Journal of the American Chemical Society 127:17584-17585.

Rinaudo, K., L. Bleris, R. Maddamsetti, S. Subramanian, R. Weiss, and Y. Benenson. 2007. A universal RNAi-based logic evaluator that operates in mammalian cells. Nature Biotechnology 25:795-801.

Rumbaugh, K. P., J. A. Griswold, B. H. Iglewski, and A. N. Hamood. 1999. Contribution of quorum sensing to the virulence of Pseudomonas aeruginosa in burn wound infections. Infection and Immunity 67:5854-5862.

Sato, T., and Y. Kobayashi. 1998. The ars operon in the skin element of Bacillus subtilis confers resistance to arsenate and arsenite. Journal of Bacteriology 180:1655-1661.

Serrano, L. 2007. Synthetic biology: Promises and challenges, Molecular Systems Biology 3:158.

Sinha, J., S. J. Reyes, and J. P. Gallivan. 2010. Reprogramming bacteria to seek and destroy an herbicide. Nature Chemical Biology 6:464-470.

Smith, K., and I. S. Hunter. 2008. Efficacy of common hospital biocides with biofilms of multi-drug resistant clinical isolates. Journal of Medical Microbiology 57:966-973.

Stewart, P. S., and M. J. Franklin. 2008. Physiological heterogeneity in biofilms. Nature Reviews Microbiology 6:199-210.

Stickler, D. J. 2002. Susceptibility of antibiotic-resistant Gram-negative bacteria to biocides: A perspective from the study of catheter biofilms. Journal of Applied Microbiology 92(Suppl): 163S-170S.

——. 2008. Bacterial biofilms in patients with indwelling urinary catheters. Nature Clinical Practice Urology 5:598-608.

Stoodley, P., R. Cargo, C. J. Rupp, S. Wilson, and I. Klapper. 2002. Biofilm material properties as related to shear-induced deformation and detachment phenomena. Journal of Industrial Microbiology and Biotechnology 29:361-367.

Stricker, J., S. Cookson, M. R. Bennett, W. H. Mather, L. S. Tsimring, and J. Hasty. 2008. A fast, robust and tunable synthetic gene oscillator. Nature 456:516-519.

Tigges, M., T. T. Marquez-Lago, J. Stelling, and M. Fussenegger. 2009. A tunable synthetic mammalian oscillator. Nature 457:309-312.

Van Houdt, R., and C. W. Michiels. 2010. Biofilm formation and the food industry, a focus on the bacterial outer surface. Journal of Applied Microbiology 109:1117-1131.

Weber, W., J. Stelling, M. Rimann, B. Keller, M. Daoud-El Baba, C. C. Weber, D. Aubel, and M. Fussenegger. 2007. A synthetic time-delay circuit in mammalian cells and mice. Proceedings of the National Academy of Sciences of the United States of America 104:2643-2648.

Westall, F., M. J. de Wit, J. Dann, S. van der Gaast, C. E. J. de Ronde, and D. Gerneke. 2001. Early Archean fossil bacteria and biofilms in hydrothermally-influenced sediments from the Barberton greenstone belt, South Africa. Precambrian Research 106:93-116.

Win, M. N., and C. D. Smolke. 2008. Higher-order cellular information processing with synthetic RNA devices. Science 322:456-460.

Yu, R. C., C. G. Pesce, A. Colman-Lerner, L. Lok, D. Pincus, E. Serra, M. Holl, K. Benjamin, A. Gordon, and R. Brent. 2008. Negative feedback that improves information transmission in yeast signalling. Nature 456:755-761.

A5

SYNTHETIC BIOLOGY AND THE ART OF BIOSENSOR DESIGN

Christopher E. French,³⁰Kim de Mora,³¹Nimisha Joshi,³²Alistair Elfick,³³James Haseloff,³⁴and James Ajioka³⁵

Introduction

The term “biosensor” refers to a wide variety of devices. The common element is that a biological component provides highly specific recognition of a certain target analyte, and this detection event is somehow transduced to give an easily detectable, quantifiable response, preferably one that can be easily converted to an electrical signal so that the result can be fed to an electronic device

_______________

³⁰ School of Biological Sciences, University of Edinburgh.

³¹ School of Biomedical Sciences, University of Edinburgh.

³² School of Geosciences, University of Edinburgh.

³³ School of Engineering & Electronics, University of Edinburgh.

³⁴ Department of Plant Sciences, University of Cambridge.

³⁵ Department of Pathology, University of Cambridge.

for signal processing, data storage, etc. The biological component in many biosensors is either an enzyme, as in the glucose-oxidase–based biosensors used for blood glucose monitoring, or an antibody, as in most optical biosensors. Another class of biosensor, sometimes also referred to as a bioreporter, uses living cells as a component. These cells detect the target analyte via some more-or-less specific receptor and generate a detectable response, most commonly by induction of a reporter gene. Many such devices have been reported in the scientific literature, with detection of mercury and arsenic in the environment being particularly common applications. However, very few such devices are commercially available, the best-known examples being the mutagen-detecting devices such as the SOS-Chromotest (Environmental Bio-Detection Products, Inc.) system. Here we discuss the reasons for this gap between promise and delivery, and ways in which the emerging discipline of synthetic biology may lead to a new generation of whole-cell biosensors.

Whole-Cell Biosensors

Whole-cell biosensors, or bioreporters, are living cells that indicate the presence of a target analyte. The most commercially successful by far are nonspecific toxicity sensors based on naturally occurring luminescent bacteria such as Vibrio harveyi, Vibrio (Photobacterium) fischeri, and Photobacterium phosphoreum. Examples include MicroTox (Strategic Diagnostics, SDIX) and BioTox (Aboatox). In these mainly marine organisms, above a certain population density, light is produced continuously by the action of bacterial luciferase (LuxAB), which oxidizes a long-chain aldehyde such as tetradecanal in the presence of FMNH₂ and oxygen. Regeneration of the reduced flavin (catalyzed by LuxG) and the aldehyde substrate (catalyzed by LuxCDE) requires NADPH and ATP, so any toxic substance that interferes with metabolism will reduce light emission, which is easily detected using a luminometer.

However, these systems are nonspecific and are only useful for preliminary screening of environmental samples to determine whether or not a toxic substance is present. The potentially more useful class of bioreporter consists of genetically modified microorganisms in which the presence of a specific target analyte is linked to a detectable response. The genetic modification involved in these cases consists of linking the receptor for the target analyte to induction of an easily detectable reporter gene. Commonly used reporter genes are shown in Table A5-1. For recent reviews of such systems, see Belkin (2003), Daunert et al. (2000), Tecon and van der Meer (2008), and van der Meer and Belkin (2010).

Synthetic Biology and Whole-Cell Biosensors

Like “biosensor,” the term “synthetic biology” is widely used by different authors to mean different things. In this context, we are using it to refer to a

TABLE A5-1 Reporter Genes Commonly Used in Whole-Cell Biosensors

systematic approach to rationalizing genetic modification to make it more like other engineering disciplines, in terms of the use of standardized parts that can be assembled in a modular way to make a variety of different constructs. Since whole-cell biosensors are intrinsically modular, consisting of a recognition element coupled to an arbitrarily chosen reporter, synthetic biology seems well suited to the development of such devices.

This approach to synthetic biology is associated particularly with MIT, BioBricks, the Registry of Standard Biological Parts,³⁶ and iGEM (the International Genetically Engineered Machine competition).³⁷ BioBricks (Knight, 2003) are a type of standardized biological “part,” consisting of pieces of DNA which conform to a certain standard (defined by a document known as RFC10, available

_______________

³⁶ See http://partsregistry.org/Main_Page.

³⁷ For more information, see the following: iGEM 2006, University of Edinburgh, http://parts.mit.edu/wiki/index.php/University_of_Edinburgh_2006; iGEM 2007, University of Cambridge, http://parts.mit.edu/iGEM07/index.php/Cambridge; iGEM 2007, University of Glasgow, http://parts.mit.edu/igem07/index.php/Glasgow; iGEM 2007, University of Science and Technology, China, http://parts.mit.edu/igem07/index.php/USTC; iGEM 2008, Harvard University, http://2008.igem.org/Team:Harvard; iGEM 2009, University of Cambridge, http://2009.igem.org/Team:Cambridge; iGEM 2010, Bristol Centre for Complexity Studies, http://2010.igem.org/Team:BCCS-Bristol; iGEM 2010, Imperial College, London, http://2010.igem.org/Team:Imperial_College_London; and iGEM 2010, Peking University, http://2010.igem.org/Team:Peking.

from the BioBricks Foundation³⁸) specifying certain characteristics of the ends. Each BioBrick may be a protein-coding region, or some other component such as a promoter, ribosome binding site, transcription termination sequence, or any other piece of DNA that may be useful in making genetic constructs. The essential point of this format is that any BioBrick can, through a standardized procedure, be combined with any other BioBrick to form a new BioBrick, which can then be combined with any other BioBrick, and so on. In this way, quite large and complex constructs can be built up fairly quickly. The Registry of Standard Biological Parts, currently hosted at MIT, was established to store both the DNA of these parts and also associated information such as DNA sequence, performance characteristics, and user experience. The intention was, and is, that this library of BioBricks and associated information should become a valuable resource for synthetic biologists.

To demonstrate the potential of this approach to synthetic biology, the iGEM competition was established in 2005. Each year, interdisciplinary teams of undergraduates consisting of a mixture of biologists, engineers, and computer scientists compete over the summer vacation period to conceive, design, mathematically model, construct, and test novel genetically modified systems made using BioBricks from the Registry, as well as new BioBricks created specifically for the project. All new BioBricks made are deposited in the Registry and are available for use by future teams. iGEM projects, completed on a short timescale by undergraduate students, are generally not nearly as well characterized as systems reported in the peer-reviewed literature; however, they are often based on highly creative ideas and generally include a mathematical modeling component far in excess that usually found in biological publications. They can therefore be a very interesting way to follow possible future application areas in synthetic biology.

Since biosensors are conceptually simple devices with a clear real-world application, and many opportunities for elaboration in terms of novel input and output modalities, in vivo signal processing, and other aspects, they are a popular choice of project, and a number of interesting innovations have been reported as a result of iGEM projects. We refer to a number of these later in this report. All information relating to previous iGEM projects is available via the relevant websites.

Arsenic Biosensors

Arsenic is a particularly attractive target for whole-cell biosensors, in that it is a major groundwater contaminant in Bangladesh, West Bengal, and a number of other regions (Meharg, 2005; Smith et al., 2000). This only came to light in the 1980s, and it is a major and increasing public health issue. The problem initially arose when, to combat waterborne diarrheal diseases caused by consumption of contaminated surface water, nongovernmental organizations drilled some millions of tube wells to supply clean drinking water. Unexpectedly, it was

_______________

³⁸ See http://www.biobricks.org/.

discovered some years later that many of these produced water with unacceptably high levels of arsenic. The current recommended World Health Organization (WHO) limit for drinking water is 10 ppb arsenic, while Bangladesh and some other countries maintain an earlier limit of 50 ppb, but many wells exceed this by a large margin. Chronic consumption of water with high arsenic concentrations leads to arsenicosis, resulting in skin lesions and various cancers. Thus, there is a clear and present need for cheap and simple tests for monitoring arsenic levels in drinking water. Current field test kits are based around the Gutzeit method, which involves reduction of arsenate to toxic arsine gas, and the detection of this as a color spot following reaction with mercuric salts. Such tests reportedly are unreliable at low but still significant arsenic concentrations, and disposal of mercuric salts poses its own environmental issues. Thus there is a clear potential niche for a simple arsenic biosensor device (Diesel et al., 2009; M. Owens, Engineers Without Borders, personal communication).

Most early arsenic biosensors were based on the arsenic detoxification operons of Staphylococcus plasmid pI258 and Escherichia coli plasmid R773. The former consists simply of the ars promoter controlling genes arsR, arsB, and arsC, encoding the repressor, arsenite efflux pump, and arsenate reductase, respectively, whereas the latter has a relatively complex structure, arsRDABC, with two separate repressors, ArsR and ArsD, which control the operon with different affinities (Oremland and Stolz, 2003). ArsD is also reported to act as a metallochaperone, carrying arsenite to the efflux pump formed by ArsAB (Lin et al., 2006). Later systems were based on the simpler E. coli chromosomal arsenic detoxification operon (Cai and DuBow, 1996; Diorio et al., 1995), which consists of the ars promoter followed by arsR, arsB, and arsC, and the similar operon of Bacillus subtilis (Sato and Kobayashi, 1998), which is discussed further below. In either case, the preparation of the biosensor organism is straightforward—the reporter gene is simply inserted adjacent to the controlled promoter so that induction of the promoter results in expression of the reporter gene, giving an easily detectable signal (usually a color change, luminescence, or fluorescence). A number of systems are reportedly at or near commercialization; for example, the Aboatox BioTox Heavy Metal Assay kits, developed at the University of Turku, use E. coli as host, with firefly luciferase as the reporter gene. Stocker et al. (2003) described production of a set of E. coli–based arsenic bioreporters using ß-galactosidase with a chromogenic substrate, bacterial luciferase, or green fluorescent protein (GFP) as reporter; one of these, based on the luciferase reporter gene, has been field tested in Vietnam (Trang et al., 2005) and was reported to give good results in comparison to chemical field tests. Whole-cell arsenic biosensors have been reviewed recently by Diesel et al. (2009).

The Edinburgh Arsenic Biosensor

One example which we consider in some detail is the first biosensor project to be submitted to iGEM: the arsenic biosensor submitted for iGEM 2006 by the

team from the University of Edinburgh, under the supervision of C. French and A. Elfick. (K. de Mora was a student member of this iGEM team.) The intention was to develop a device that would be suitable for field use in developing countries. It should therefore be cheap, simple to use, and should deliver an output that could be assessed by eye, without requiring expensive electrical equipment, but should also give a quantifiable response using cheap equipment where this was required. For this reason the standard reporters based on luminescence and fluorescence were not considered appropriate. Instead, it was decided that the output should be in the form of a pH change. This would give a bright and easily assessed visible response using a pH indicator chemical, and it could also give a quantitative electrical response using a cheap glass pH electrode or similar solid-state device (ISFET).

For practical reasons, it was decided to use a standard laboratory host strain of Escherichia coli. Such organisms are easy to manipulate, carry multiple disabling mutations that make them harmless to humans and unable to propagate in the environment, and grow well at temperatures up to 45°C. E. coli and related organisms naturally ferment a variety of sugars, including lactose, via the mixed acid pathway, resulting in the production of acetic, lactic, and succinic acids, which can rapidly lower the pH of the medium below 4.5. Many laboratory strains carry a deletion in the gene lacZ, encoding ß-galactosidase, the initial enzyme of lactose degradation, so that a nonfunctional truncated LacZ protein, missing the first few amino acids, is produced. This can be complemented by a short peptide known as the alpha peptide, consisting of the first 50 to 70 amino acids of LacZ. This is encoded by a short open reading frame known as lacZ′α. Thus, to generate an acid response, it was only necessary to use such a lactose-defective host strain, such as E. coli JM109, and place the lacZ′α gene under the control of the ars promoter in a standard Registry multicopy plasmid, pSB1A2. Thus, in the presence of arsenate or arsenite, expression of the LacZ alpha peptide would be induced, complementing the truncated LacZ and allowing rapid fermentation of lactose to acids, lowering the pH. To generate an alkaline response, the urease genes, ureABC, of Bacillus subtilis (Kim et al., 2005) were chosen. (Uropathogenic strains of E. coli also possess urease genes, but these are longer and more complex than those of B. subtilis.) Expression of these genes allows conversion of urea to ammonia and carbon dioxide and can raise the pH of the medium above 10. Both acid- and alkali-producing systems were tested and were found to work well (Aleksic et al., 2007).

The original design of the arsenic biosensor submitted for iGEM 2006 was a complex system, with a multistage output. This is discussed further below. However, the practical demonstration provided consisted only of the acid-generating system, which was found to give robust and reliable responses to arsenic concentrations as low as 2.5 ppb, with the time of pH change being related to the arsenic concentration in a simple and reproducible way (Aleksic et al., 2007; Joshi et al., 2009). This construct is available from the Registry of Standard Biological Parts (BBa_J33203), as are its components, the ars promoter and associated arsR gene

(BBa_J33201) and the lacZ′α reporter gene (BBa_J33202). Interestingly, during early testing to determine whether buffer ions expected to be present in groundwater might interfere with the pH response, we found that bicarbonate ions actually increase the sensitivity of the response, leading to induction at much lower arsenate levels (Joshi et al., 2009). The reason for this is not clear, but it may be due to altered speciation or uptake of arsenate (de Mora et al., 2011).

The original concept for this biosensor system involved use of a universal pH indicator solution which gives a strong color response—blue in alkaline conditions, green in neutral conditions, and red in acidic conditions—coupled with quantitation via a glass pH electrode. However, it became apparent that the red component of the universal pH indicator, as well as pure methyl red, were rapidly bleached in the presence of living cells under the conditions used. This was therefore replaced with bromothymol blue, which is blue under alkaline conditions and yellow under acid conditions, with pKa around 7.3 (Figure A5-1).

For quantitative monitoring of multiple samples simultaneously, as might be useful in a local or regional testing laboratory, an inexpensive system was developed based on the use of freeze-dried cells together with a webcam; following aseptic addition of groundwater samples to freeze-dried cells and sterile medium, the webcam would monitor the color of multiple tubes simultaneously, and software would extract the pixels representing the tubes and monitor the color of each over time. From these data, the time of color change could be extracted, and this was found to correlate well with arsenic levels in model groundwaters

FIGURE A5-1 Demonstration of the Edinburgh pH-based arsenic biosensor, Escherichia coli JM109/pSB1A2-BBa_J33203 with bromothymol blue as pH indicator, following static overnight incubation. From left to right: arsenic-free control; 5, 10, 25, 50, and 100 ppb arsenic as sodium arsenate; and cell-free control with 100 ppb arsenate. Note the increasing size of the cell pellet in tubes with increasing arsenic concentrations. Color change occurs more rapidly in samples with increasing arsenic concentration (not shown).
SOURCE: C. French, unpublished.

and also in real arsenic-contaminated groundwater samples from Hungary (de Mora et al., 2011).

This system seems well suited for use in water quality laboratories at a local or regional level. However, our original aim was to develop a system that could be used by relatively unskilled users in the field, so that local people could easily monitor the quality of their own well water. This poses several further issues. One technical issue is that any contamination with lactose-degrading bacteria will lead to false-positive results. To avoid this, water samples must be sterilized prior to introduction into the test device. To overcome this problem, we envisage a disposable plastic device containing freeze-dried cells and medium components, with an integral sterile filter through which the water sample is introduced. A second potential complicating factor is the temperature dependence of the rapidity of color change. This requires further investigation. Storage lifetime under relevant conditions also requires further research. However, we are confident that these issues can be overcome, and that this system can form the basis of a simple, cheap, sensitive, and reliable field test for arsenic concentration in groundwater.

One nontechnical issue which must first be addressed is the regulatory and safety issues associated with use of live genetically modified bacteria outside of a contained laboratory context. As noted above, the host strains used are disabled and are unable to colonize humans or to propagate in the environment in competition with wild-type bacteria (Chart et al., 2001). Nevertheless, depending on the jurisdiction, there are many regulatory hurdles associated with the use of genetically modified mciroorganisms in poorly contained applications. This is discussed further below. In the meantime, we are focusing our attention on development of a device suitable for use in local or regional laboratories. A company, Lumin Sensors, has been formed to explore these possibilities (www.luminsensors.com).

In the remainder of this paper, we consider some of the ways synthetic biology can improve the performance of whole-cell biosensors and lead to the development of a new generation of devices with improved performance.

Better Biosensors Through Synthetic Biology

Alternative Host (Chassis) Organisms

Synthetic biology generally involves the introduction of a new genetic system into a host organism, which in the context of synthetic biology is often called a “chassis.” The great majority of whole-cell biosensors reported in the literature have used E. coli as a host organism, due to familiarity, ease of manipulation, and availability of a wide variety of vector systems and other tools. However, E. coli has a number of characteristics which may mean that it is not necessarily the ideal chassis for any given purpose. For example, as a Gram-negative bacterium, it produces and sheds lipopolysaccharide (endotoxin), a powerful activator of the innate immune system; hence, it is generally unsuitable for any in vivo uses. Its

outer membrane prevents large analytes such as peptides from approaching the cell membrane. Unlike many other bacteria, it generally does not secrete most proteins effectively into the medium, unless specific modifications are introduced to allow it. Most significantly in terms of biosensor applications, E. coli does not naturally produce dormant states such as spores, meaning that freeze drying is likely to be required for storage and distribution. While freeze drying of bacterial cells is a well-understood process, this adds an extra level of complexity and expense to the manufacturing process. Some other potential hosts have much better characteristics in this regard. Most attention has been paid to Bacillus subtilis, a low-GC Gram-positive soil bacterium. B. subtilis is used as a model Gram-positive bacterium; its physiology is therefore well studied, and a number of vector systems and other tools are available, though in both characteristics it lags well behind E. coli. In contrast to E. coli, B. subtilis is naturally competent during a certain stage of its life cycle, and, unlike E. coli, will happily take up large pieces of linear DNA and integrate them onto its chromosome by homologous recombination. Also, and more importantly, B. subtilis naturally forms a dormant resting state known as endospores (Errington, 2003). When conditions become unfavorable for growth, each cell undergoes an asymmetrical cell division resulting in a large “mother cell” and a small “forespore.” The mother cell engulfs the forespore and produces layers of protein coats to surround it, while the forespore produces small acid-soluble proteins and calcium dipicolinate, which act to protect its nucleic acids. When this process is complete, the mother cell lyses to release the mature endospore. The spores can simply be harvested and dried for storage and distribution. Endospores are extremely tolerant to heat, drying, and other stresses. There are well-attested reports of endospores surviving for hundreds and even thousands of years in dry conditions (Nicholson et al., 2000), as well as more controversial reports of survival for many millions of years in unusual contexts such as in the guts of insects preserved in amber (Cano and Borucki, 1995).

B. subtilis possesses an arsenic detoxification operon similar to that found on the E. coli chromosome (Sato and Kobayashi, 1998); thus, arsenic biosensors can be prepared in a similar way to that described above, either by adding the ars promoter and arsR regulatory gene to a reporter gene on a multicopy plasmid, or by introduction of such a reporter gene to the chromosome downstream of the ars promoter. As a demonstration of principle, we have constructed such a system, designated a “Bacillosensor,” using the plasmid vector pTG262 and the reporter gene xylE of Pseudomonas putida, which encodes catechol-2,3-dioxygenase. This enzyme acts on the cheap substrate catechol to produce a bright yellow compound, 2-hydroxy-cis,cis-muconic semialdehyde. This system was found to be sensitive to arsenic levels well below the WHO recommended limit of 10 ppb (Figure A5-2). Spores could be boiled for 2 minutes prior to use in the assay; this would not only kill competing organisms but also activate the spores for rapid germination. The ability to remove most contaminating organisms from

FIGURE A5-2 Detection of arsenic by B. subtilis 168/pTG262-arsR-xylE: absorbance at 377 nm vs. arsenic concentration (ppb arsenic as sodium arsenate). The vector and BioBrick components used to make this device (BBa_J33206, BBa_J33204) are available from the Registry of Standard Biological Parts. Conditions of the assay were as described by Joshi et al. (2009).
SOURCE: L. Montgomery and C. French, unpublished.

the sample simply by heat treatment offers a considerable potential advantage for field use of such devices, eliminating the need for filter sterilization or similar treatments.

Several reports in the literature have also described bioreporters based on B. subtilis and related organisms. Tauriainen et al. (1997, 1998) reported the use of B. subtilis as a host for firefly luciferase-based bioreporters for arsenic, antimony, cadmium, and lead but did not specifically describe the use of endospores in the assays. More recently, Date et al. (2007) reported the construction of bioreporters for arsenic and zinc based on endospores of B. subtilis and Bacillus megaterium, with ß-galactosidase plus a chemiluminescent substrate, or enhanced green fluorescent protein, as reporter genes. The genetically modified spores could be stored at room temperature for at least 6 months. The same authors later described incorporation of such endospore-based bioreporters into microfluidic devices (Date et al., 2010, discussed further below). Fantino et al. (2009) reported the construc-

tion of a device designated “Sposensor,” incorporating B. subtilis endospores engineered to produce ß-galactosidase in response to the target analyte. Systems responsive to zinc and bacitracin were demonstrated using a chromogenic substrate with spores dried on filter paper discs.

Another potentially interesting host is the yeast Saccharomyces cerevisiae. Again, this is a model organism, well studied, and for which numerous vector systems and other genetic modification tools are available. It can be stored and distributed in a “dry active” state (Baronian, 2003). However, as a eukaryote, S. cerevisiae has more complex regulatory systems than bacteria such as E. coli and B. subtilis, and to date there have been relatively few reports of its use as a host for bioreporter applications. One example is a nonspecific toxicity reporter described by Välimaa et al. (2008), using S. cerevisiae modified to produce firefly luciferase; as in the MicroTox system described above, toxic substances reduced the level of luminescence observed. S. cerevisiae has also been used as a platform for analyte-specific biosensors. For example, Leskinen et al. (2003) reported construction of a yeast-based bioreporter for copper ions, with firefly luciferase under control of the copper-responsive CUP1 promoter. This was used in environmental analysis for bioavailable copper (Peltola et al., 2005). Some further examples are described by Baronian (2003). With further development, analyte-specific yeast-based biosensors could be a useful addition to the biosensor toolkit.

Detection of Extracellular Analytes

In the examples discussed so far, the signal has been generated internally, either as a stress response (as in the case of the SOS-Chromotest) or else by binding of an intracellular protein, such as ArsR, to an analyte that has been internalized (arsenate is probably taken up in error by the phosphate uptake machinery). For medical applications, it would be advantageous to be able to detect and respond to analytes such as peptides, which do not naturally enter bacterial or fungal cells. However, bacteria are able to sense and respond to extracellular analytes via “two-component” systems. In these cases, one component is a sensor kinase that spans the cell membrane, and the second is a response regulator protein that binds DNA to activate or repress transcription from a given promoter. The extracellular analyte binds to the extracellular domain of the sensor kinase, and this increases or decreases the kinase activity of the intracellular domain. This alters the tendency of the kinase domain to phosphorylate the response regulator protein, which in turn alters its propensity to bind to and activate or repress the promoter(s) in question. Most bacteria possess multiple two-component sensor systems responding to a variety of different stimuli, including osmotic strength of the extracellular medium, extracellular phosphate levels, and the presence of various small molecules such as sugars and amino acids. One interesting subgroup consists of two-component sensor systems in Gram-positive bacteria such as Bacillus, Streptococcus, and Enterococcus, which sense and respond to the pres-

ence of short-peptide pheromones produced by other cells of the same species, a phenomenon analogous to the more familiar N-acyl homoserine lactone–based quorum-sensing systems of Gram-negative bacteria. One might imagine that such systems could be modified, perhaps by rational engineering or directed evolution, to respond instead to some peptide of analytical interest.

There are two interesting points regarding two-component sensor systems. One is that these systems show a surprising degree of modularity. A number of reports have described cases where the extracellular domain of one sensor kinase has been fused to the intracellular domain of another, giving a hybrid protein that responds to the normal stimulus of the first sensor kinase by activating the normal response regulator of the second. One well-known example is the report of Levskaya et al. (2005) describing fusion of the extracellular domain of the light-sensing domain of a cyanobacterial phytochrome, Cph1, to the intracellular domain of an E. coli osmotic stress sensor, EnvZ; the hybrid protein, Cph8, responded to red light by activating the response regulator, OmpR, which normally responds to EnvZ. When a promoter controlled by OmpR was fused to a pigment-producing gene, the resulting genetically modified cells responded to red light by producing a pigment, allowing “bacterial photographs” to be made by focusing images onto a plate of the bacteria (Levskaya et al., 2005). Another example is fusion of the extracellular domain of a chemotaxis receptor, Trg, which responds to the presence of ribose, among other chemoattractants, by controlling the cell’s motility apparatus, to the intracellular domain of EnvZ (Baumgartner et al., 1994). (In this case, the interaction is indirect: ribose binds to periplasmic ribose binding protein, which then interacts with the extracellular domain of Trg.) In the presence of the hybrid protein, designated Trz, an OmpR-controlled promoter (specifically, the ompC promoter) was found to respond to the presence of ribose. Recent structural studies have begun to offer some insight into the basis of communication between extracellular and intracellular domains in two-component sensor kinases (Casino et al., 2010), which may allow more rational engineering of such hybrid proteins.

The second point is that it is possible to engineer the recognition elements of such sensors to respond to nonnatural target molecules. For example, Looger et al. (2003) reported rational reengineering of the ribose binding protein, which binds ribose and interacts with the extracellular domain of the hybrid Trz sensor kinase mentioned above, so that it would respond to lactate or to 2,4,6-trinitro-toluene (TNT). In the presence of these molecules, OmpR-controlled promoters were reported to be activated. This opens the possibility that it might be possible to use such a platform as a “universal bioreporter” by generating a library of reengineered sensor kinases or associated binding proteins to respond to any analyte of interest. If such rational reengineering proved challenging in the general case, another interesting possibility would be to attempt fusion of the extracellular domain of such a sensor to some binding molecule such as a nanobody (variable region of a camelid heavy-chain-only antibody) or scFv fragment of an antibody,

so that any analyte to which an antibody could be raised could be detected by a bioreporter. Other protein-based scaffolds for specific recognition are also under development (Hosse et al., 2006; Nuttall and Walsh, 2008) and might also be suitable for such applications. Alternatively, extracellular domains able to bind a desired target analyte might be selected from a library of mutants by a technique such as phage display or cell surface display (Benhar, 2001); this might also enable simple screening for clones which were able not only to bind to the analyte of interest but also to generate the appropriate response on binding.

An interesting, but rather application-specific, approach to the detection of extracellular analytes via two-component sensing systems was proposed by the 2010 iGEM team of Imperial College, London. In this case, the objective was to detect cercaria (larvae) of the parasite Schistosoma in water. It was proposed that a protease produced by the parasite could be detected by cleavage of an engineered substrate to release the autoinducer peptide of Streptococcus pneumoniae; this peptide would then be detected by the two-component sensor system ComDE of S. pneumoniae expressed in B. subtilis. A similar system might be used to detect other proteases or hydrolytic enzymes of biological interest.

Another possible platform for detection of extracellular peptides would be the yeast mating peptide sensing system. In Saccharomyces cerevisiae, mating peptides (a and α) are detected by G-protein coupled receptors, Ste2 and Ste3, which initiate an intracellular signalling pathway (Bardwell, 2004; Naider and Becker, 2004). As noted above, S. cerevisiae is arguably an underexploited platform for biosensor development, with much potential for investigation. One interesting example of a yeast-based sensor system was reported by Chen and Weiss (2005). In this case, a cytokinin (plant hormone) was detected by yeast cells expressing Arabidopsis thaliana cytokinin receptor Atcre1, a two-component-type sensor kinase, which apparently is fortuitously able to interact with the endogenous yeast response regulator Ypd1 in the absence of Ypd1’s normal sensor kinase partner, Sln1 (Inoue et al., 2001). The engineered “receiver” cells expressed GFP from a Ypd1-activated promoter in the presence of cytokinin.

Animal cells have numerous and varied extracellular receptors but, with current techniques, probably lack the necessary robustness to be generally useful as a biosensor platform except possibly for specialized clinical uses.

Modulation of Sensitivity and Dynamic Range

The critical parameters in the performance of a sensor are the sensitivity (the lowest analyte concentration to which a detectable response is seen) and the dynamic range (in this context, the range of analyte concentrations over which the analyte concentration can be estimated based on the response; that is, between the sensitivity limit and the concentration at which the response saturates). In simple whole-cell biosensors such as those described above, a hyperbolic or sigmoidal response is seen, and it may be that the response curve initially obtained is not in

the desired range for the planned application. Obviously, one critical parameter is the affinity of the receptor for the analyte; however, in practice, many other factors are important, making analysis rather complicated (for a more detailed discussion of such issues, see van der Meer et al., 2004). For example, in the case of arsenic biosensors, arsenate is first taken up by the cell (probably mistakenly, via the phosphate uptake system). The characteristics of uptake determine the ratio between extracellular and intracellular arsenate concentrations. It then interacts with the ArsR repressor with a certain affinity, and interferes with the binding of the repressor to the ars promoter; both of these interactions have a characteristic affinity. Within a given cell, the numbers of arsenate anions, ArsR repressor molecules, and promoter sites are also all limited, so the relative numbers of these will also strongly affect the interaction. Finally, when the repressor is not bound to the promoter, and the promoter is thus “active,” RNA polymerase will bind the promoter with a certain affinity, and begin transcription with a certain efficiency. The actual level of arsenate repressor and reporter protein molecules generated will be strongly affected by the rate of messenger RNA synthesis, the rate of mRNA degradation, the affinity of ribosomes for the ribosome binding sites on the mRNA, and the degradation rate of the proteins. Thus, many steps intervene between the extracellular arsenate concentration and the level of the reporter protein. This means that, in practice, it is possible to modulate the sensitivity and dynamic range of the sensor without actually altering the receptor at all, simply by making minor changes to factors such as the strength of the ribosome binding sites.

This type of experiment is greatly facilitated by the modular, composable nature of BioBricks and the availability of a library of ribosome binding sites of different strengths in the Registry of Standard Biological Parts. To give one trivial example, we assembled an alternative version of the simple arsenic biosensor construct BBa_J33203, the differences being that the reporter gene was moved to a position between the promoter and the arsR gene encoding the repressor, and the native ribosome binding site of arsR was replaced by a strong synthetic ribosome binding site, BBa_J15001. The response characteristics of this modified construct were quite different from those of the original construct (Figure A5-3). While the mechanism of this was not investigated, one plausible explanation would be an increased level of expression of the ArsR repressor due to the stronger ribosome binding site.

More profound reorganizations of the arsenic recognition system have also been investigated. For example, a second copy of the ArsR-binding site was introduced between arsR and the reporter gene, with the aim of decreasing background expression in the absence of arsenic. This led to considerably improved induction characteristics (Stocker et al., 2003). It was further reported that modification of the activity or synthesis rate of the reporter enzyme (cytochrome c peroxidase or ß-galactosidase) led to strong changes in the system response to given arsenic concentrations (Wackwitz et al., 2008), allowing the generation of an array of

FIGURE A5-3 Altered response characteristics of a whole-cell arsenic biosensor through reassembly of the components. (A) Original Edinburgh arsenic biosensor, consisting of the E. coli chromosomal ars promoter and arsR gene (BBa_J33201) followed by lacZ′α (BBa_J33202). Both arsR and lacZ′α have their native ribosome binding site. (B) Reassembled operon consisting of ars promoter (BBa_J15301), strong synthetic ribosome binding site (BBa_J15001), lacZ′α coding sequence (BBa_J15103), ribosome binding site (BBa_J15001), and arsR coding sequence (BBa_J15101). Diamonds, time zero; squares, 6 hours; triangles, 24 hours. The vector was pSB1A2 and host was E. coli JM109 in all cases. Assay conditions were as described by Joshi et al. (2009). All components and assembled constructs are available from the Registry of Standard Biological Parts.
SOURCE: X. Wang and C. French, unpublished.

reporter strains with different response characteristics (discussed further below). In a similar investigation, using mercury biosensors, the 2010 Peking University iGEM team investigated the effects of placing the regulatory gene merR under the control of a variety of promoters of different strengths and found that this resulted in a wide variety of different sigmoidal response curves. A similar effect was achieved by screening a library of mutants with altered MerR-binding sites in the mercury-responsive promoter. Thus, a variety of simple modifications to the system can be used to achieve alterations in the sensitivity and dynamic range of such sensors. The composable nature of BioBricks, together with other assembly strategies used in synthetic biology, make it easy to generate and screen a large number of such systems to find a set with the desired characteristics.

It is also possible to use rational design principles to modify the dynamic range of a sensor, for example, by amplifying a weak transcriptional signal. One way to do this is through the use of genetic “amplifiers.” One set of such devices, submitted and tested by iGEM teams from the University of Cambridge in 2007 and 2009, consists of bacteriophage activator-promoter pairs. Rather than the

analyte-responsive promoter deriving the reporter gene directly, the promoter drives expression of an activator protein, which activates a second promoter, which controls the reporter gene. This gives a genetic “amplification” effect. A small library of cross-reactive activators and promoters allows a mix-and-match approach to select a pair that gives the desired response characteristics. In these projects, the promoters came from bacteriophages P2 (promoters P_F, P_O, P_P, and P_V) and P4 (promoters P_sid and P_LL), and the activators from bacteriophages P2 (Ogr), P4 (δ protein), PSP3 (Pag), and φR73 (δ protein) (Julien and Calendar, 1996). Fifteen promoter-activator combinations were characterized, allowing the biosensor designer to choose a pair with the desired response characteristics.

The availability of a number of similar biosensors with different response characteristics, as described above, allows the preparation of “bar graph”-like arrays of sensors to obtain a quantitative output over a wider range of analyte concentrations than any single sensor could achieve. This was proposed by Wackwitz et al. (2008), who used the term “traffic light biosensors” to describe such devices. In principle, such devices should not require calibration (van der Meer and Belkin, 2010).

Another issue with induction-based whole-cell biosensors is the time required for induction; few systems in the literature show detectable responses in less than 30 minutes or so, and several hours is more typical. For many applications, it would be advantageous to obtain a faster response. One ingenious approach to this problem was presented by the Imperial College iGEM team of 2010. In this case, the reporter enzyme is presynthesized in the cell in an inactive form. The analyte-responsive promoter drives expression of a protease, which cleaves and activates the reporter. Since each molecule of protease can rapidly activate multiple molecules of the reporter protein, this can potentially give a much faster response. In the case of the Imperial College iGEM project, the reporter, catechol-2,3-dioxygenase (XylE), was synthesised as an inactive fusion with GFP, joined by a linker which could be cleaved by site-specific TEV protease. Detection of the target analyte, in this case a peptide released by a protease of the parasite Schistosoma, led to induction of TEV protease expression and consequent cleavage of the linker, allowing rapid formation of active catechol-2,3-dioxygenase tetramers.

In Vivo Signal Processing and Multiplex Output

In addition to tuning of response characteristics, it is possible to introduce more complex forms of in vivo signal processing. One simple example is a genetic inverter, in which a promoter that would normally activate transcription is instead used to repress it. This is accomplished by having the analyte-responsive promoter drive production of a repressor, which represses expression from a second promoter driving the reporter gene. Three well-characterized repressor-promoter pairs are widely used in such systems: the LacI/lac promoter pair, bac-

teriophage λ cI/P_L or P_R pair, and the TetR-tet promoter pair (see, for example, Elowitz and Leibler, 2000). These repressor-promoter pairs are analogous to insulated wires used in an electronic circuit to communicate between different parts of the device. For complex devices, it would clearly be useful to have more than three such “wires” available. One interesting approach to this was reported by the USTC iGEM team of 2007 (following work described by Sartorius et al., 1989). In this case, mutations were made to the bases of the lac operator site involved in binding of the repressor LacI, and also to the amino acids of LacI involved in this binding. The libraries of mutant lac promoters and LacI repressors were then analyzed to determine which pairs interacted efficiently. From this experiment, multiple repressor-promoter pairs were chosen that did not crosstalk. Furthermore, these were used to generate biological equivalents of several logic gates. The simplest of these is the NOT gate, in which activation of one promoter leads to repression of another. This is simply achieved by having the first promoter drive expression of a repressor which represses the second, as described above. More complex gates include NAND and NOR. In the former case, binding of two different repressors is required to inactivate the second promoter; in the latter case, binding of either of two repressors is sufficient to achieve this effect. From combinations of such gates, more complex circuits can be assembled. Such logic systems can also be extended to systems consisting of several different types of engineered cell, with the cells communicating via quorum-sensing signals (Tamsir et al., 2010).

The term “traffic-light sensors,” discussed above, is also applied to a class of devices that have been proposed and modeled, but, so far as we know, never demonstrated in practice, in which discrete, different outputs are activated at different analyte levels by in vivo signal processing within a single bioreporter organism. The originally proposed iGEM 2006 Edinburgh arsenic biosensor fell into this category, giving an alkaline response at very low arsenate concentrations, a neutral pH at moderate arsenate concentrations, and an acidic response at dangerously high arsenate levels (Aleksic et al., 2007). This was to be achieved through the use of two separate repressors, with different affinities for arsenate, controlling two different reporters: urease for an alkaline response, and ß-galactosidase for the acidic response. Response of the high-affinity repressor-promoter pair was inverted via a repressor, so that the presence of a low concentration of arsenate led to production of a repressor that switched off production of urease, whereas higher levels of arsenate switched on production of ß-galactosidase. A similar arrangement of two different repressor systems can be used to generate a genetic “band detector,” which responds only to analyte concentrations within a certain concentration range.

Visual Outputs

While the majority of reported arsenic sensors use luminescence or fluorescence as output, it might be advantageous in some cases to have an output that can be easily detected by eye. The use of enzymes such as ß-galactosidase and catechol-2,3-dioxygenase, together with chromogenic substrates, offers one route to achieving this end. An alternative is the pH-based approach used in the Edinburgh arsenic biosensor described above; this is useful in that pH changes, together with standard pH indicators, give very strong and easily detected color changes, but they can also be quantified using an inexpensive pH electrode. In some cases it might be preferable to have cells produce an endogenous pigment in response to the target analyte. Several such examples have been reported. Fujimoto et al. (2006) described a system in which the photosynthetic bacterium Rhodovulum sulfidophilum was engineered to place the endogenous carotenoid pigment gene crtA under the control of the E. coli ars promoter, so that the presence of arsenite led to a change in cell pigmentation from yellow to red. Subsequently, Yoshida et al. (2008) reported a similar system based on Rhodopseudomonas palustris. The combinatorial nature of synthetic biology components, together with the possibility of multistage outputs discussed above, opens the possibility of systems in which a range of discrete colors is produced for different levels or combinations of analytes. To facilitate the construction of such devices, the University of Cambridge iGEM team, 2009, presented a set of modular BioBrick components that could generate a variety of carotenoid and indole-based pigments using different combinations of components from the carotenoid and violacein biosynthetic pathways (Figure A5-4). As with all of the iGEM entries discussed in this paper, these genetic modules are freely available from the Registry of Standard Biological Parts.

FIGURE A5-4 Escherichia coli cells producing a variety of pigments. From left to right, the first four tubes are derived from the carotenoid biosynthesis pathway, the last three from the violacein biosynthesis pathway. All of these pigment-producing pathways are available in BioBrick format from the Registry of Standard Biological Parts.
SOURCE: University of Cambridge, iGEM 2009.

Integration of Biological and Electronic Components

For data processing and storage, it is advantageous if biosensor outputs can easily be converted into electrical signals that can be read by a computer. A number of integrated biological-electrical devices have been reported. One example is the Bioluminescent Bioreporter Integrated Circuit (Nivens et al., 2004; Vijayaraghavan et al., 2007), in which luminescence emitted by bacterial luciferase is detected by a Complementary Metal-Oxide Semiconductor (CMOS) microluminometer incorporated with signal processing circuitry and other relevant components on a “biochip.” Several recent reports have described the integration of bioreporter cells into microfluidic devices. For example, Diesel et al. (2009) reported incorporation of an E. coli GFP-based arsenic bioreporter into a microfluidic device. Fluorescence from approximately 200 cells within the detection cavity was imaged using a camera. Date et al. (2010) reported the integration of Bacillus endospore-based biosensors for arsenic and zinc (described above) into a centrifugal microfluidic platform, in which pumping is provided by rotation of a disc-shaped substrate. Luminescence and fluorescence measurements were performed using separate laboratory instruments with fiber-optic probes held above the detection chambers of the microfluidic device. Storage and germination of endospores, and sensitive detection of the analytes, were reported.

Whereas these devices use “conventional outputs (fluorescence and luminescence), other reporter systems can give an electrical output directly, which may be more convenient for incorporation into simple devices for field use. One example is the Edinburgh arsenic biosensor described above. The pH response can easily be converted to a voltage signal using a standard glass pH electrode, or an equivalent solid-state device (ion selective field effect transistor). An alternative is the generation of an amperometric signal via a device similar to a microbial fuel cell. Bacteria respire by transferring electrons from a donor (such as a sugar) to an acceptor (such as oxygen). This electron transfer occurs at the cell membrane. Many bacteria can instead transfer electrons to or from an electrode, generating an electrical current; others can do so in the presence of electron shuttle molecules known as mediators (Lovley, 2006). An amperometric signal can be generated by placing synthesis of such a mediator, or of an essential component of the electron transfer apparatus, under the control of an analyte-responsive promoter. Both of these approaches have been described by iGEM teams. The iGEM 2007 entry from the University of Glasgow described a system for biosynthesis of pyocyanin, a redox-active metabolite of Pseudomonas aeruginosa, which could act as a mediator to transport electrons between the bacterial respiratory chain and an electrode. The iGEM 2008 entry from Harvard University took advantage of a naturally “electricigenic” bacterium, Shewanella oneidensis, by controlling synthesis of one of the outer membrane proteins, MtrB, required for efficient transfer of electrons to an electrode. Either of these approaches can allow controllable generation of an electrical current induced by the presence of a target analyte.

Regulatory Issues Related to Field Use of Whole-Cell Biosensors

The Edinburgh arsenic biosensor described above was designed to be used in the field in the form of a contained disposable device. Some other bioreporters have been designed for more direct use in the field. One of the earliest examples to be tested was that of Ripp et al. (2000), reportedly the first genetically modified microorganisms to be approved for field testing for a bioremediation application in the United States, in which a strain of Pseudomonas fluorescens modified to produce a luminescent signal in the presence of aromatic hydrocarbons was introduced into the soil in a contaminated site, and its persistence monitored over a 2-year period. While this organism was designed to persist in the environment for a prolonged period, there are other cases where the bioreporter only needs to survive for a few hours—just long enough to report on the conditions in which it finds itself. One well-known example is the Microbial Mine Detection System developed at Oak Ridge National Laboratories. This consisted of a strain of Pseudomonas modified to produce GFP in the presence of explosives leaking from land mines (discussed by Habib, 2007). The positions of the land mines could then be mapped using an ultraviolet lamp. This system was reportedly field tested with good results. Another example was reported by the iGEM team of the Bristol Centre for Complexity Studies (BCCS) in 2010. In this case, the bioreporter consisted of E. coli cells modified to indicate the presence of nitrate by producing a fluorescent response. These cells, encapsulated in a hydrophilic gel, were to be spread over the soil and left for several hours to express the fluorescent protein, after which the nitrate levels in the field could be mapped by a mobile device, providing information useful in agriculture.

On consideration of the examples described above, we can distinguish between four different cases:

1. bioreporters designed to be used within the confines of a laboratory (e.g., the SOS-Chromotest);
2. bioreporters designed to be used in a contained device, but outside of a laboratory (e.g., the Edinburgh arsenic biosensor);
3. bioreporters designed to be exposed directly to the environment, but not to survive longer than required to report the levels of the target analyte (e.g., the MMDS and Bristol nitrate sensor); and
4. bioreporters designed to survive and persist in the field for long periods (e.g., the bioremediation monitoring system of Ripp et al., 2000).

Unfortunately, current regulatory regimes in the United Kingdom and European Union do not appear to distinguish between cases 2, 3, and 4. Case 1 is unproblematic; such “contained use” applications are dealt with in the United Kingdom by the Health and Safety Executive under the Genetically Modified Organisms (Contained Use) Regulations 2000, whereas “uncontained uses” are dealt with by the Department for the Environment, Food and Rural Affairs un-

der the Genetically Modified Organisms (Deliberate Release) Regulations 2002 (which implement EU Directive 2001/18/EC). Whereas this is wholly appropriate for case 4, where such organisms are designed to persist in the environment, no provision seems to be made for cases 2 or 3, where organisms may be specifically designed not to survive in the environment, thereby minimizing any threat to ecosystems. This is a serious inhibiting factor in the use of genetically modified whole-cell biosensors or bioreporters outside of the laboratory, as expensive field trials with extensive postrelease monitoring are required. We feel that it would be useful if future legislation took this into account. In the United States, the situation is less clear to us; the relevant federal legislation appears to be the Toxic Substances Control Act (Sayre and Seidler, 2005). In the absence of specific legal and regulatory expertise, we will refrain from further comment. However, it is clear that until these regulatory issues are addressed, it will not be possible to use genetically modified bioreporter organisms on a large scale in the field, despite their obvious potential.

Conclusions

Sensitive and highly specific response to various molecules is one of the core functions of biological systems. As such, whole-cell biosensors offer a versatile and widely applicable method for detecting the presence of a wide range of analytes. The techniques of synthetic biology offer numerous possible improvements in terms of response tuning, in vivo signal processing, and direct interface with electronic devices for further signal processing and output. However, regulatory issues will need to be clarified before such devices can fulfill their true potential as highly sensitive, inexpensive sensors for field use.

Acknowledgments

The original Edinburgh arsenic biosensor concept was developed and tested by the 2006 University of Edinburgh iGEM team, consisting of Jelena Aleksic, Farid Bizzari, Yizhi Cai, Briony Davidson, Kim de Mora, Sergii Ivakhno, Judith Nicholson, Sreemati Lalgudi Seshasayee, and Jennifer Wilson. Valuable further contributions were made by Lucy Montgomery and Xiaonan Wang, formerly of the University of Edinburgh. Contaminated groundwater samples were provided by Dr. Balint L. Balint of the Department of Biochemistry and Molecular Biology, Medical and Health Science Centre, University of Debrecen.

Disclosures

The Edinburgh 2006 iGEM team was generously supported by the Gatsby Foundation, the Royal Commission for the Exhibition of 1851, and Synbiocomm, an initiative of the European Union. Funding for further development of the

Edinburgh arsenic biosensor was provided by the Biotechnology and Biological Sciences Research Council via the U.K. Synthetic Biology Standards Network, of which A. Elfick and J. Haseloff are codirectors.

References

Aleksic, J., F. Bizzari, Y. Z. Cai, B. Davidson, K. de Mora, S. Ivakhno, S. L. Seshasayee, J. Nicholson, J. Wilson, A. Elfick, C. E. French, L. Kozma-Bognar, H. W. Ma, and A. Millar. 2007. Development of a novel biosensor for detection of arsenic in drinking water. IET Synthetic Biology 1:87-90.

Bardwell, L. 2004. A walk-through of the yeast mating pheromone response pathway. Peptides 25(9):1465-1476.

Baronian, K. H. R. 2003. The use of yeast and moulds as sensing elements in biosensors. Biosensors and Bioelectronics 19:953-962.

Baumgartner, J. W., C. Kim, R. E. Brisette, M. Inoue, C. Park, and G. L. Hazelbauer. 1994. Trans-membrane signaling by a hybrid protein: Communication from the domain of chemoreceptor Trg that recognizes sugar-binding proteins to the kinase/phosphatase domain of osmosensor EnvZ. Journal of Bacteriology 176(4):1157-1163.

Belkin, S. 2003. Microbial whole-cell sensing systems of environmental pollutants. Current Opinion in Microbiology 6:206-212.

Benhar, I. 2001. Biotechnological applications of phage and cell display. Biotechnology Advances 19(1):1-33.

Cai, J., and M. S. DuBow. 1996. Expression of the Escherichia coli chromosomal ars operon. Canadian Journal of Microbiology 42:662-671.

Cano, R. J., and M. K. Borucki. 1995. Revival and identification of bacterial spores in 25 million year old to 40 million year old Dominican amber. Science 268(5213):1060-1064.

Casino, P., V. Rubio, and A. Marina. 2010. The mechanism of signal transduction by two-component systems. Current Opinion in Structural Biology 20:763-771.

Chart, H., H. R. Smith, R. M. La Ragione, and M. J. Woodward. 2001. An investigation into the pathogenic properties of Escherichia coli strains BLR, BL21, DH5α and EQ1. Journal of Applied Microbiology 89:1048-1058.

Chen, M. T., and R. Weiss. 2005. Artificial cell-cell communication in yeast Saccharomyces cerevisiae using signaling elements from Arabidopsis thaliana. Nature Biotechnology 23(12):1551-1555.

Date, A., P. Pasini, and S. Daunert. 2007. Construction of spores for portable bacterial whole-cell biosensing systems. Analytical Chemistry 79:9391-9397.

——. 2010. Integration of spore-based genetically engineered whole-cell sensing systems into portable centrifugal microfluidic platforms. Analytical and Bioanalytical Chemistry 398:349-356.

Daunert, S., G. Barret, J. S. Feliciano, R. S. Shetty, S. Shrestha, and W. Smith-Spencer. 2000. Genetically engineered whole-cell sensing systems: Coupling biological recognition with reporter genes. Chemical Reviews 100:2705-2738.

de Mora, K., N. Joshi, B. L. Balint, F. B. Ward, A. Elfick, and C. E. French. 2011. A pH-based biosensor for detection of arsenic in drinking water. Analytical and Bioanalytical Chemistry 400(4):1031-1039.

Diesel, E., M. Schreiber, and J. R. van der Meer. 2009. Development of bacteria-based bioassays for arsenic detection in natural waters. Analytical and Bioanalytical Chemistry 394:687-693.

Diorio, C., J. Cai, J. Marmor, R. Shinder, and M. S. DuBow. 1995. An Escherichia coli chromosomal ars operon homolog is functional in arsenic detoxification and is conserved in Gram negative bacteria. Journal of Bacteriology 177(8):2050-2056.

Elowitz, M. B., and S. Leibler. 2000. A synthetic oscillatory network of transcriptional regulators. Nature 403(6767):335-338.

Errington, J. 2003. Regulation of endospore formation in Bacillus subtilis. Nature Reviews Microbiology 1(2):117-126.

Fantino, J. R., F. Barras, and F. Denizot. 2009. Sposensor: A whole-cell bacterial biosensor that uses immobilized Bacillus subtilis spores and a one-step incubation/detection process. Journal of Molecular Microbiology and Biotechnology 17:90-95.

Fujimoto, H., M. Wakabayashi, H. Yamashiro, I. Maeda, K. Isoda, M. Kondoh, M. Kawase, H. Miyasaka, and K. Yagi. 2006. Whole cell arsenite biosensor using photosynthetic bacterium Rhodovulum sulfidophilum. Applied Microbiology and Biotechnology 73(2):332-338.

Habib, M. K. 2007. Controlled biological and biomimetic systems for landmine detection. Biosensors and Bioelectronics 23:1-18.

Hosse, R. J., A. Rothe, and B. E. Power. 2006. A new generation of protein display scaffolds for molecular recognition. Protein Science 15(1):14-27.

Inoue, T., M. Higuchi, Y. Hashimoto, M. Seki, M. Kobayashi, T. Kato, S. Tabata, K. Shinozaki, and T. Kakimoto. 2001. Identification of CRE1 as a cytokinin receptor from Arabidopsis. Nature 409(6823):1060-1063.

Joshi, N., X. Wang, L. Montgomery, A. Elfick, and C. E. French. 2009. Novel approaches to biosensors for detection of arsenic in drinking water. Desalination 248(1-3):517-523.

Julien, B., and R. Calendar. 1996. Bacteriophage PSP3 and φR73 activator proteins: Analysis of promoter specificities. Journal of Bacteriology 178(19):5668-5675.

Kim, J. K., S. B. Mulrooney, and R. P. Hausinger. 2005. Biosynthesis of active Bacillus subtilis urease in the absence of known urease accessory proteins. Journal of Bacteriology 187(20):7150-7154.

Knight, T. 2003. Idempotent vector design for standard assembly of Biobricks. http://web.mit.edu/synbio/release/docs/biobricks.pdf (accessed October 20, 2011).

Leskinen, P., M. Virta, and M. Karp. 2003. One-step measurement of firefly luciferase activity in yeast. Yeast 20(13):1109-1113.

Levskaya, A., A. A. Chevalier, J. J. Tabor, Z. Booth Simpson, L. A. Lavery, M. Levy, E. A. Davidson, A. Scouras, A. D. Ellington, E. M. Marcotte, and C. A. Voigt. 2005. Engineering Escherichia coli to see light. Nature 438:441-442.

Lin, Y.-F., A. R. Walmsley, and B. P. Rosen. 2006. An arsenic metallochaperone for an arsenic detoxification pump. Proceedings of the National Academy of Sciences of the United States of America 103(42):15617-15622.

Looger, L. L., M. A. Dwyer, J. J. Smith, and H. W. Hellinga. 2003. Computational design of receptor and sensor proteins with novel functions. Nature 423:185-190.

Lovley, D. 2006. Bug juice: Harvesting electricity with microorganisms. Nature Reviews Microbiology 4(7):497-508.

Meharg, A. 2005. Venomous Earth: How Arsenic Caused the World’s Worst Mass Poisoning. Basing-stoke, UK: Macmillan.

Naider, F., and J. M. Becker. 2004. The α-factor mating pheromone of Saccharomyces cerevisiae: A model for studying the interaction of peptide hormones and G protein-coupled receptors. Peptides 25(9):1441-1463.

Nicholson, W. L., N. Munakata, G. Horneck, H. J. Melosh, and P. Setlow. 2000. Resistance of Bacillus endospores to extreme terrestrial and extraterrestrial environments. Microbiology and Molecular Biology Reviews 64(3):548-572.

Nivens, D. E., T. E. McKnight, S. A. Moser, S. J. Osbourn, M. L. Simpson, and G. S. Sayler. 2004. Bioluminescent bioreporter integrated circuits: Potentially small, rugged and inexpensive whole-cell biosensors for remote environmental monitoring. Journal of Applied Microbiology 96:33-46.

Nuttall, S. D., and R. B. Walsh. 2008. Display scaffolds: Protein engineering for novel therapeutics. Current Opinion in Pharmacology 8(5):609-615.

Oremland, R. S., and J. F. Stolz. 2003. The ecology of arsenic. Science 300(5621):939-944.

Peltola, P., A. Ivask, M. Astrom, and M. Virta. 2005. Lead and Cu in contaminated urban soils: Extraction with chemical reagents and bioluminescent bacteria and yeast. Science of the Total Environment 350(1-3):194-203.

Ripp, S., D. E. Nivens, Y. Ahn, C. Werner, J. Jarrell, J. P. Easter, C. D. Cox, R. S. Burlage, and G. S. Sayler. 2000. Controlled field release of a bioluminescent genetically engineered microorganism for bioremediation process monitoring and control. Environmental Science and Technology 34(5):846-853.

Sartorius, J., N. Lehming, B. Kisters, B. von Wilcken-Bergmann, and B. Müller-Hill. 1989. lac repressor mutants with double or triple exchanges in the recognition helix bind specifically to lac operator variants with multiple exchanges. EMBO Journal 8(4):1265-1270.

Sato, T., and Y. Kobayashi. 1998. The ars operon in the skin element of Bacillus subtilis confers resistance to arsenate and arsenite. Journal of Bacteriology 180(7):1655-1661.

Sayre, P., and R. J. Seidler. 2005. Application of GMOs in the US: EPA research and regulatory considerations related to soil systems. Plant and Soil 275(1-2):77-91.

Smith, A. H., E. O. Lingas, and M. Rahman. 2000. Contamination of drinking-water by arsenic in Bangladesh: A public health emergency. Bulletin of the World Health Organization 78(9): 1093-1103.

Stocker, J., D. Balluch, M. Gsell, H. Harms, J. Feliciano, S. Daunert, K. A. Malik, and J. R. van der Meer. 2003. Development of a set of simple bacterial biosensors for quantitative and rapid measurements of arsenite and arsenate in potable water. Environmental Science and Technology 37:4743-4750.

Tamsir, A., J. J. Tabor, and C. A. Voigt. 2010. Robust multicellular computing using genetically encoded NOR gates and chemical “wires.” Nature 469(7329):212-215.

Tauriainen, S., M. Kapr, W. Chang, and M. Virta. 1997. Recombinant luminescent bacteria for measuring bioavailable arsenite and antimonite. Applied and Environmental Microbiology 63(11):4456-4461.

——. 1998. Luminescent bacterial sensor for cadmium and lead. Biosensors and Bioelectronics 13:931-938.

Tecon, R., and J. R. van der Meer. 2008. Bacterial biosensors for measuring availability of environmental pollutants. Sensors 8(7):4062-4080.

Trang, P. T. K., M. Berg, P. H. Viet, N. V. Mui, and J. R. van der Meer. 2005. Bacterial bioassay for rapid and accurate analysis of arsenic in highly variable groundwater samples. Environmental Science and Technology 39:7625-7630.

Välimaa, A.-L., A. Kivistö, M. Virta, and M. Karp. 2008. Real-time monitoring of non-specific toxicity using a Saccharomyces cerevisiae reporter system. Sensors 8:6433-6447.

van der Meer, J. R., and S. Belkin. 2010. Where microbiology meets microengineering: Design and applications of reporter bacteria. Nature Reviews Microbiology 8:511-522.

van der Meer, J. R., D. Tropel, and M. Jaspers. 2004. Illuminating the detection chain of bacterial bioreporters. Environmental Microbiology 6(10):1005-1020.

Vijayaraghavan, R., S. K. Islam, M. Zhang, S. Ripp, S. Caylor, N. D. Bull, S. Moser, S. C. Terry, B. J. Blalock, and G. S. Sayler. 2007. Bioreporter bioluminescent integrated circuit for very low-level chemical sensing in both gas and liquid environments. Sensors and Actuators B 123:922-928.

Wackwitz, A., H. Harms, A. Chatzinotas, U. Breuer, C. Vogne, and J. R. van der Meer. 2008. Internal arsenite bioassay calibration using multiple reporter cell lines. Microbial Biotechnology 1:149-157.

Yoshida, K., K. Inoue, Y. Takahashi, S. Ueda, K. Isoda, K. Yagi, and I. Maeda. 2008. Novel carotenoid-based biosensor for simple visual detection of arsenite: Characterization and preliminary evaluation for environmental application. Applied and Environmental Microbiology 74:6730-6738.

A6

SYSTEMS ANALYSIS OF ADAPTIVE IMMUNITY BY UTILIZATION OF HIGH-THROUGHPUT TECHNOLOGIES³⁹

Sai T. Reddy^40,41and George Georgiou^40,41,42,43

A new generation of high-throughput technologies for quantitative and clonal analysis of adaptive immune responses have been developed. Functional analysis of lymphocyte populations has been accomplished via microfluidic assay systems. Additionally, lymphocyte receptor repertoires have been characterized on proteomic and genomic levels with multiplexed protein microarrays and high-throughput DNA sequencing. These tools are providing an unprecedented level of information depth on the distribution of adaptive immune cell (B and T cell) functionalities and repertoires, which develop upon activation following vaccination, pathogenic infection, or in disease states. These various high-throughput technologies have unlocked the potential to transform immunology into an information-rich science that will enable rapid expansion of the field of experimental systems immunology.

Introduction

Adaptive immunity plays an indispensible role in maintaining vertebrate host protection against a constant barrage of pathogens. Adaptive immunity is fundamentally reliant upon the generation of an incredible diversity of lymphocyte

_______________

³⁹ Reprinted from Current Opinion in Biotechnology, 22/4, Reddy, S. T., and G. Georgiou, Systems analysis of adaptive immunity by utilization of high-throughput technologies, 584-589, 2011, with permission from Elsevier.

⁴⁰ Department of Chemical Engineering, University of Texas at Austin, Austin, TX 78712, USA.

⁴¹ Department of Biomedical Engineering, University of Texas at Austin, Austin, TX 78712, USA.

⁴² Institute for Cellular and Molecular Biology, University of Texas at Austin, Austin, TX 78712, USA.

⁴³ Section of Molecular Genetics and Microbiology, University of Texas at Austin, Austin, TX 78712, USA.
Corresponding author: Georgiou, George (gg@che.utexas.edu)
Current Opinion in Biotechnology 2011, 22:1–6
This review comes from a themed issue on systems biology
Edited by Roy Kishony and Vassily Hatzimanikatis
0958-1669/$ – see front matter
# 2011 Elsevier Ltd. All rights reserved.
DOI 10.1016/j.copbio.2011.04.015

receptors (B cell receptors, or BCRs; and T cell receptors, or TCRs); furthermore, the development of lymphocytes subsets, which display a plethora of functions, act synergistically to counteract pathogenic invasion. Receptor diversity in lymphocyte populations is generated by the genetic recombination of noncontiguous germline elements: variable (V), diversity (D), and joining (J) gene segments. Primary and secondary diversification via germline and junctional recombination, combinatorial pairing, and somatic hypermutation (for B cells only) combine to generate theoretical diversities of human antibodies and TCRs of ~10¹³ and ~10¹⁸, respectively (Janeway, 2005), far outnumbering the physiological number of lymphocytes in an average human (~10¹⁰). In addition to molecular diversity, there is an assortment of lymphocyte cellular subclasses (e.g., CD4+ and CD8+ T cells, Tregs, Th17 cells) and differentiation states (e.g., naïve, memory, and plasma B cells) that synergize to produce a multiprong attack on pathogens, leading to infection resolution and long-term immunological memory. Finally, the anatomical compartments (e.g., lymph nodes, blood, bone marrow) in which lymphocyte activation, homing, and trafficking take place add to the complex interconnected nature of adaptive immunity.

Conventional approaches for measuring adaptive immune responses have so far relied on cursory biological measurements, such as antibody titers to an antigen or lymphocyte assays that measure a single (ELISPOT) or multiple cytokines (intracellular staining, ICS) expressed by individual cells (Janetzki et al., 2004; Papagno et al., 2007). However, these techniques are surrogate measures and provide only a partial picture of the state of the immune system; for example, antibody titers do not provide specific molecular information on the monoclonal antibody components that comprise polyclonal responses in serum or secretory fluids. Likewise, ELISPOT or ICS assays can interrogate only a subset of lymphocytes based on functional properties; furthermore, they are sample-destructive thus precluding the simultaneous cloning and analysis of the respective BCR and TCR repertoires. Very importantly, existing assays typically capture the mean of a particular response or cellular phenotype and afford limited quantification. The understanding of immunological processes at greater resolution is not only important from a scientific point of view but can have major ramifications for human health. Detailed and quantitative deconvolution of the humoral immune response can assist in human immunotherapy by the identification of neutralizing or cytotoxic antibodies or of antigens overexpressed in a disease state (Gnjatic et al., 2010; Law et al., 2008; Yu et al., 2008). High-resolution information on adaptive immune responses to immunization may also greatly aid in assessing the efficacy of experimental vaccines (Germain et al., 2010; Querec et al., 2009), a process that currently relies on longitudinal studies of neutralizing antibody titers as the primary readout. In this review, we highlight recent advances in the high-throughput analysis of functional properties, proteomic interactions, and genomic repertoires of lymphocyte populations following an adaptive immune response (Table A6-1).

TABLE A6-1 High-throughput technologies used to study adaptive immunity and the scale of information generated.

Single Cell Analysis of Lymphocyte Populations

Single cell PCR-cloning of immunoglobulin genes

Until recently, the determination of the surface-anchored immunoglobulins/antibodies (BCRs) expressed by animal or human cells has relied on B cell immortalization techniques such as hybridoma technology or viral transduction (Becker et al., 2010; Kwakkenbos et al., 2009; Lanzavecchia and Sallusto, 2009); however, these techniques are compatible with only some stages of B cell maturation and furthermore, due to their low efficiencies (merely 1–3% of B cells interrogated) they are unable to generate a comprehensive picture of humoral repertoires. For example, terminally differentiated antibody-secreting plasma cells are not amenable to immortalization. Consequently, the antibodies isolated by immortalization of memory B cells may neither correspond to nor exhibit the same therapeutic potency as the population of circulating antibodies.

These problems can be addressed by single-cell cloning. Briefly, a desired B cell population is plated at limiting dilution in microtiter well plates and

then the variable heavy (V_H) and light (V_L) chains of surface immunoglobulins (BCRs) or of secreted antibodies from antibody secreting cells are amplified by PCR using primer sets. The V_H and V_L genes are then inserted into suitable vectors for expression from heterologous hosts, resulting in recombinant antibody fragments (e.g., scFv, FAB) or full-length IgGs. Subsequent antigen-specificity and binding affinity of antibodies are determined by standard ELISA techniques (Tiller et al., 2008). In one notable example, flow cytometry was used to isolate human plasmablasts from the peripheral blood of patients immunized with the influenza vaccine. The cells were sorted into 96 well plates at limiting dilution to yield approximately one cell per well. Nested RT-PCR and cDNA amplification of V_L and V_H genes was performed to amplify V genes in wells occupied by cells, followed by cloning into eukaryotic expression vectors (Smith et al., 2009). This approach was used to study the dynamics and antigen specificity of transient plasmablasts that arise three to seven days after vaccination and decay within a month. It was observed that at the peak of the response, nearly 80% of the plasmablasts expressed influenza-specific antibodies and the respective antibody repertoires were pauciclonal, as they were dominated by only a few sequences (Wrammert et al., 2008). Alternative techniques have been developed that rely on single B-cell cloning by linking V_L and V_H genes during RT-PCR followed by subsequent expression and screening in bacterial systems. For example, peripheral blood plasma cells were isolated from patients receiving tetanustoxiod (TT) immunizations; single-cell PCR and screening was performed which was then followed by extensive bioinformatic analysis of V_L and V_H pairing (Meijer et al., 2006). In another study, Andersen and co-workers isolated tetanus toxoid-specific antibodies from immunized patients and showed by biochemical characterization that they display affinity and kinetic binding constants close to the proposed limits for protective immunity (Poulsen et al., 2007). These single-cell cloning methods have been used by the company Symphogen (Denmark) to develop recombinant polyclonal antibodies, potentially valuable as human therapeutics. Multiple neutralizing antibodies were expressed under carefully controlled bioprocess conditions to produce a ‘recombinant polyclonal’ mixture that mimics the diversity, specificity, and binding affinity of natural polyclonal human immune responses (Wiberg et al., 2006).

Single cell microarrays for functional profiling of lymphocytes

Recently, large-scale analysis of lymphocyte populations has been enabled by methods in soft lithography and microengraving. These techniques have generated lymphocyte functional profiles. For example, microfluidic devices were deployed for the large-scale detection of antigen-specific B lymphocytes at the single-cell level and the subsequent cloning of their respective V genes. Kishi and co-workers used deep reactive ion etching on silicon to produce chips containing arrays of 10 mm diameter and 20 mm deep wells to trap single B lympho-

cytes at a density >200,000 cells per chip. The well surfaces were coated with anti-immunoglobulin antibodies, which allowed for capture of antibodies from antibody-secreting cells isolated from mouse or human B cells (Tajiri et al., 2007; Tokimitsu et al., 2007). Next, antigen was added to discover cells producing specific antibodies, which was followed by single-cell RT-PCR and cloning into a eukaryotic system for the generation of recombinant monoclonal antibodies (Jin et al., 2009). Alternatively, Love et al. used photolithography and intaglio printing to generate microarrays from poly(dimethylosiloxane). The dimensions of the arrays were compatible for imaging on a microscope slide, therefore allowing fluorescence-based imaging of B lymphocytes expressing antigen-specific antibodies (Love et al., 2006; Ogunniyi et al., 2009). Microengraved devices were used to assay the molecular profile of individual B cells isolated from mice following immunization; multiparametric analysis was performed to determine antibody specificities, isotypes, and apparent affinities (Story et al., 2008). The high throughput afforded by the microwell chip arrays described above notwithstanding, these methods rely on single-cell PCR and thus suffer from practical cloning limitations, therefore making it challenging to obtain a complete picture of the repertoire.

In addition to B cells, the functional profiles of T cells can also be measured by single cell microarray technologies (Song et al., 2010). Multidimensional activation profiles were generated from human peripheral blood T cells; cytokine secretion rates from individual cells were measured with very high sensitivity (0.5–4 molecules/s) (Han et al., 2010). In another recent study, single cell gene expression profiling was performed on CD8 T cell populations following various prime-boost HIV DNA vaccine schemes in mice. Multiparameter transcriptional analysis from single T cells was used to measure mRNA expression of 91 different genes related to phenotype, regulation, survival, and function, allowing discrimination of central memory and effector memory T cell subsets; furthermore providing quantitative parameters associated with vaccine efficacy and protective immunity (Flatz et al., 2011). The Heath group has developed a system for detecting antigen-specific T cells using nucleic acid cell sorting. Peptide-major histocompatibility complex tetramers (p/MHC) were site-specifically attached to DNA oligomers, which were annealed to complimentary strands printed on a glass slide. These p/MHC microarrays allowed for multiplexed sorting of antigen-specific T cells from heterogenous cell suspensions; this system can be exploited for clinical lymphocyte detection, such as detection of T cells from cancer patients (Kwong et al., 2009). These examples illustrate that single-cell methods are rapidly expanding the ability to generate high-throughput data for quantitative molecular analysis and functional profiling of adaptive immune responses.

Microarrays for Studying Protein–Protein Interactions in the Adaptive Immune Response

Systems-level analysis of gene expression using DNA microarrays has had a tremendous impact on cellular and molecular biology; the potential now exists

for proteomic microarrays to have a similar influence on immunological research (Michaud et al., 2003; Prechl et al., 2010). A substantial amount of work has been done to apply multiplexed high-throughput protein microarrays for the serological profiling and analysis of antibody responses; this research has ranged from infectious disease to cancer to autoimmune disease (Kingsmore, 2006).

Technologies for high-throughput printing of pathogen proteomes (with each protein expressed individually) onto protein microarrays have been developed. A typical approach consists of high-throughput PCR amplification of each predicted open-reading frame, which is followed by in vivo homologous recombination into a T7 expression vector. An Escherichia coli-based cell-free in vitro expression system is used to generate crude lysates possessing recombinant proteins, which are then printed onto nitrocellulose membranes resulting in a pathogen proteome microarray. In one example, the 185 viral proteins that comprise the vaccinia virus proteome were expressed individually and used to prepare protein arrays, which then were used to analyse the specificity of serum antibodies from vaccinia virus-vaccinated subjects; this allowed for high-throughput evaluation of humoral immunity (Davies et al., 2005). In an even more impressive study, protein microarrays comprising the 1205 proteins encoded by Burkholderia pseudomallei were constructed and probed with sera from infected and healthy patients. Such antibody profiling studies can reveal immunodominant antigens, which can be exploited as biomarkers for diagnostic purposes or to guide vaccine development (Felgner et al., 2009). Utilizing this same protein microarray methodology, it was found that in tuberculosis patients, circulating antibodies react with 10% of the Myobacterium tuberculosis proteome. Further analysis identified that during active infection the humoral response was highly biased towards only 0.5% of the proteome and particularly towards extracellular proteins, suggesting that the humoral immune response to M. tuberculosis correlates with the evolution of the infection (Kunnath-Velayudhan et al., 2010).

In addition to pathogenic proteins, autologous cellular antigens expressed by malignant cells or in autoimmune disorders can also elicit robust antibody responses. These autoantibodies serve as potential immunological biomarkers valuable for diagnosis and possibly for treatment. In an early study, a protein array was constructed by the immobilization of 196 different human proteins onto a glass slide and used for the detection of autoantibodies (Robinson et al., 2002). More recently, Old and co-workers constructed an array of 329 cancer antigens that was used to detect antibodies produced in cancer patients but not in healthy volunteers (Gnjatic et al., 2009). In a follow-up study, they expanded their analysis and used an array consisting of >8000 human proteins to probe the antibody specificities in sera from over 150 individuals, including healthy volunteers and ovarian and pancreatic cancer patients; this led to the identification of autoantigen signatures for these two malignancies (Gnjatic et al., 2010). In a recent study, a combinatorial library of unnatural synthetic molecules was used to screen for ligands reactive with antibodies abundant in serum (Reddy et al., 2011). The underlying concept of this study was that synthetic molecules act as a ‘shape library’, such that they are outside of the chemical space range occupied

by natural biomolecules. Microarrays of peptoids (N-substituted oligoglycines) were used to identify reactive IgGs in a mouse model of multiple sclerosis and in human patients with Alzheimer’s disease, thus demonstrating the potential to identify immunological serum antibody biomarkers without previous knowledge of antigen. In a similar approach, a microarray consisting of 10,000 random peptides was used to determine the serological antibody binding profile in mice and humans immunized with influenza. Although peptides were random sequences, this analysis still allowed for discrimination of pre-immune and post-immune samples, suggesting a novel platform for the discovery of immunological signatures present in antibody responses (Legutki et al., 2010).

High-Throughput DNA Sequencing of Adaptive Immune Repertoires

The emergence of ‘next generation’ high-throughput DNA sequencing platforms has been applied to the analysis of immune repertoires. The sequencing of immune repertoires at unprecedented levels of depth has opened up a realm of possibilities; these include answering basic questions on repertoire diversity and selection, monitoring repertoire diversity for clinical applications, and advancing strategies in monoclonal antibody discovery and engineering. The challenge of sequencing variable regions of immune receptors is that there is significant redundancy in the sequences encoding the four framework regions but extensive diversity in the three hypervariable complementarity determining regions (CDRs). As a result, sequencing platforms such as SOLiD (Applied Biosystems) that generate short reads (<100 bp) cannot be used to assemble the complete variable genes. Therefore, typically, the 454 (Roche) technology which yields reads >350 bp has been used to sequence the entire variable genes but at a relatively lower throughput (up to 10⁶ reads per plate). Alternatively, the Illumina technology which currently has a maximum read length of ~200 bp (using paired-end sequencing) has been used to sequence the most hypervariable domain of the V gene (CDR3) at high depth (>10⁶) (Fischer, 2011).

In one of the first examples, Quake and co-workers used 454 GS-FLX to generate 640 million bp of antibody V gene sequences from zebrafish, thus providing the first ever glimpse of the complete antibody diversity in an organism. Zebra-fish were shown to utilize between 50% and 86% of all possible germline VDJ combinations. Additionally convergence was observed such that several fish had identical antibody sequences (Weinstein et al., 2009). In a follow-up work, Jiang et al. showed that in early stages of zebrafish development, the VDJ repertoire is highly stereotypical, but becomes more diversified as it matures, suggesting both deterministic and stochastic elements shaped the antibody repertoire (Jiang et al., 2011). Boyd, Fire, and colleagues recently explored the application of antibody repertoire sequencing for clinical studies. In this work, the B lymphocyte clonality of V_H genes was determined from the blood of healthy patients and from various tissues (e.g., bone marrow, lymph nodes) of patients with haematological

malignancies. This study demonstrated that high-resolution analysis of immunoglobulin repertoires and their clonality offers a potential strategy for diagnosis and therapeutic monitoring in cancer (Boyd et al., 2009).

Similarly the human TCR repertoire has been analyzed by Illumina-based high-throughput sequencing; high-resolution analyses were performed on CDR3 length distribution frequency and germline usage in TCRb chains from peripheral blood T cells (Freeman et al., 2009; Robins et al., 2009). In an expanded study, the TCRb CDR3 regions from naïve and memory CD8⁺ T cells isolated from seven patients were sequenced revealing a strong preferential usage of certain V–J combinations, with a bias towards a reduction in nontemplated nucleotide insertions. A computational analysis indicated a substantial overlap in the CD8⁺ CDR3 repertoire between any two individuals, approaching ~7000 fold greater than what would be expected based on the theoretical diversity of the TCR repertoire (Robins et al., 2010). In another similar study, the TCRb CDR3 regions from three different individuals were sequenced; comparisons between these repertoires (at the amino acid level) also revealed substantial overlap (e.g., two donors had 14.4% common sequences) (Warren et al., forthcoming). These studies suggest that overlap may be indicative of CDR3 TCRs that share specificity to common pathogens; the presence of such sequences could potentially be used as biomarkers for detection and diagnosis in infectious and autoimmune disease.

High-throughput DNA sequencing and analysis of immune repertoires has also started to make an impact on the field of monoclonal antibody discovery and engineering. For example, the antibody repertoire diversity in a phage display human combinatorial library was precisely quantified using 454 sequencing and bioinformatic analysis; it was found that nearly all germline families were present and that a substantial amount of diversity was generated by somatic mutations in CDR1 and 2 (Glanville et al., 2009). Similarly, a synthetic antibody library constructed by polynucleotide annealing and extension (amplification, PCR-free) was subjected to deep-sequencing for accurate determination of diversity and quality control (Ge et al., 2010). The Illumina platform was used to sequence CDR3 regions from an antibody fragment phage display library following rounds of antigen panning; this sequence-based approach allowed for identification and rescue of clones that were lost by subsequent rounds of screening (Ravn et al., 2010). In addition to in vitro libraries, high-throughput sequencing has been utilized for sequencing of antibody repertoires in immunized animals. Recently, 454 sequencing was performed on the antibody repertoires derived from bone marrow plasma cells of protein-immunized mice; it was found that V_H and V_L repertoires were highly polarized with the most abundant sequences representing 1–10% of the entire repertoire (Reddy et al., 2010). The most abundant V_H and V_L genes were paired simply by relative frequency and then constructed by high-throughput automated gene synthesis. Following recombinant expression, it was found that >75% (21/27) of antibodies were antigen-specific, representing the first ever example of monoclonal antibody discovery without screening.

Conclusions

The development of high-throughput technologies has led to a substantial amount of progress in the understanding of adaptive immune responses, which has allowed for a shift towards information-rich, systems-based approaches to immunological research (Table A6-1). Techniques that utilize single-cell analysis have generated high content genetic and functional data from various lymphocyte populations in mice and humans. Additionally, protein microarray technologies have been extensively used for humoral serological profiling and identification of antigens from a variety of pathogens and disease states. Finally, high-throughput DNA sequencing of adaptive immune repertoires has facilitated substantial advances in our understanding of immunological diversity and its subsequent exploitation for clinical and technological applications. These are only the beginning stages; the next steps will be to leverage these high-throughput technologies for the integration of biophysical, biochemical, genomic, and proteomic data sets, which will require the incorporation of statistical and mathematical modelling. This rapidly expanding field of experimental systems immunology will enable a more comprehensive understanding of adaptive immunity.

Acknowledgements

We thank Gregory C. Ippolito for helpful discussions and review of our manuscript. This work was funded by grants from the Cancer Prevention Research Institute of Texas (CPRIT) and Defense Advanced Research Projects Agency (DARPA).

References

Becker PD, Legrand N, van Geelen CMM, Noerder M, Huntington ND, Lim A, Yasuda E, Diehl SA, Scheeren FA, Ott M et al.: Generation of human antigen-specific monoclonal IgM antibodies using vaccinated ‘human immune system’ mice. PLoS ONE 2010, 5:e13137.

Boyd SD, Marshall EL, Merker JD, Maniar JM, Zhang LN, Sahaf B, Jones CD, Simen BB, Hanczaruk B, Nguyen KD et al.: Measurement and clinical monitoring of human lymphocyte clonality by massively parallel VDJ pyrosequencing. Sci Trans Med 2009, 1:12ra23.

Davies DH, Liang X, Hernandez JE, Randall A, Hirst S, Mu Y, Romero KM, Nguyen TT, Kalantari-Dehaghi M, Crotty S et al.: Profiling the humoral immune response to infection by using proteome microarrays: high-throughput vaccine and diagnostic antigen discovery. Proc Natl Acad Sci U S A 2005, 102:547-552.

Felgner PL, Kayala MA, Vigil A, Burk C, Nakajima-Sasaki R, Pablo J, Molina DM, Hirst S, Chew JSW, Wang D et al.: A Burkholderia pseudomallei protein microarray reveals serodiagnostic and cross-reactive antigens. Proc Natl Acad Sci U S A 2009, 106:13499-13504.

Fischer N: Sequencing antibody repertoires: the next generation. mAbs 2011, 3:17-20.

Flatz L, Roychoudhuri R, Honda M, Filali-Mouhim A, Goulet J-P, Kettaf N, Lin M, Roederer M, Haddad EK, Sékaly RP et al.: Single-cell gene-expression profiling reveals qualitatively distinct CD8 T cells elicited by different gene-based vaccines. Proc Natl Acad Sci U S A 2011, 108:5724-5729.

Freeman JD, Warren RL, Webb JR, Nelson BH, Holt RA: Profiling the T-cell receptor beta-chain repertoire by massively parallel sequencing. Genome Res 2009, 19:1817-1824.

Ge X, Mazor Y, Hunicke-Smith SP, Ellington AD, Georgiou G: Rapid construction and characterization of synthetic antibody libraries without DNA amplification. Biotechnol Bioeng 2010, 106:347-357.

Germain RN, Meier-Schellersheim M, Nita-Lazar A, Fraser IDC: Systems biology in immunology: a computational modeling perspective. Annu Rev Immunol 2010, 29:527-585.

Glanville J, Zhai W, Berka J, Telman D, Huerta G, Mehta GR, Ni I, Mei L, Sundar PD, Day GMR et al.: Precise determination of the diversity of a combinatorial antibody library gives insight into the human immunoglobulin repertoire. Proc Natl Acad Sci U S A 2009, 106:20216-20221.

Gnjatic S, Wheeler C, Ebner M, Ritter E, Murray A, Altorki NK, Ferrara CA, Hepburne-Scott H, Joyce S, Koopman J et al.: Seromic analysis of antibody responses in non-small cell lung cancer patients and healthy donors using conformational protein arrays. J Immunol Methods 2009, 341:50-58.

Gnjatic S, Ritter E, Büchler MW, Giese NA, Brors B, Frei C, Murray A, Halama N, Zörnig I, Chen Y-T et al.: Seromic profiling of ovarian and pancreatic cancer. Proc Natl Acad Sci U S A 2010, 107:5088-5093.

Han Q, Bradshaw EM, Nilsson B, Hafler DA, Love JC: Multidimensional analysis of the frequencies and rates of cytokine secretion from single cells by quantitative microengraving. Lab Chip 2010, 10:1391-1400.

Janetzki S, Schaed S, Blachere NEB, Ben-Porat L, Houghton AN, Panageas KS: Evaluation of Elispot assays: influence of method and operator on variability of results. J Immunol Methods 2004, 291:175-183.

Janeway C: Immunobiology. Garland Pub.; 2005.

Jiang N, Weinstein JA, Penland L, White RA, Fisher DS, Quake SR: Determinism and stochasticity during maturation of the zebrafish antibody repertoire. Proc Natl Acad Sci U S A 2011.

Jin A, Ozawa T, Tajiri K, Obata T, Kondo S, Kinoshita K, Kadowaki S, Takahashi K, Sugiyama T, Kishi H et al.: A rapid and efficient single-cell manipulation method for screening antigen-specific antibody-secreting cells from human peripheral blood. Nat Med 2009, 15:1088-1092.

Kingsmore SF: Multiplexed protein measurement: technologies and applications of protein and antibody arrays. Nat Rev Drug Discov 2006, 5:310-320.

Kunnath-Velayudhan S, Salamon H, Wang H-Y, Davidow AL, Molina DM, Huynh VT, Cirillo DM, Michel G, Talbot EA, Perkins MD et al.: Dynamic antibody responses to the Mycobacterium tuberculosis proteome. Proc Natl Acad Sci U S A 2010, 107:14703-14708.

Kwakkenbos MJ, Diehl SA, Yasuda E, Bakker AQ, van Geelen CMM, Lukens MV, van Bleek GM, Widjojoatmodjo MN, Bogers WMJM, Mei H et al.: Generation of stable monoclonal antibody-producing B cell receptor-positive human memory B cells by genetic programming. Nat Med 2009, 16:123-128.

Kwong GA, Radu CG, Hwang K, Shu CJ, Ma C, Koya RC, Comin-Anduix B, Hadrup SR, Bailey RC, Witte ON et al.: Modular nucleic acid assembled p/MHC microarrays for multiplexed sorting of antigen-specific T cells. J Am Chem Soc 2009, 131:9695-9703.

Lanzavecchia A, Sallusto F: Human B cell memory. Curr Opin Immunol 2009, 21:298-304.

Law M, Maruyama T, Lewis J, Giang E, Tarr AW, Stamataki Z, Gastaminza P, Chisari FV, Jones IM, Fox RI et al.: Broadly neutralizing antibodies protect against hepatitis C virus quasispecies challenge. Nat Med 2008, 14:25-27.

Legutki JB, Magee DM, Stafford P, Johnston SA: A general method for characterization of humoral immunity induced by a vaccine or infection. Vaccine 2010, 28:4529-4537.

Love JC, Ronan JL, Grotenbreg GM, van der Veen AG, Ploegh HL: A microengraving method for rapid selection of single cells producing antigen-specific antibodies. Nat Biotechnol 2006, 24:703-707.

Meijer P-J, Andersen PS, Haahr Hansen M, Steinaa L, Jensen A, Lantto J, Oleksiewicz MB, Tengbjerg K, Poulsen TR, Coljee VW et al.: Isolation of human antibody repertoires with preservation of the natural heavy and light chain pairing. J Mol Biol 2006, 358:764-772.

Michaud GA, Salcius M, Zhou F, Bangham R, Bonin J, Guo H, Snyder M, Predki PF, Schweitzer BI: Analyzing antibody specificity with whole proteome microarrays. Nat Biotechnol 2003, 21:1509-1512.

Ogunniyi AO, Story CM, Papa E, Guillen E, Love JC: Screening individual hybridomas by microengraving to discover monoclonal antibodies. Nat Protoc 2009, 4:767-782.

Papagno L, Almeida JR, Nemes E, Autran B, Appay V: Cell permeabilization for the assessment of T lymphocyte polyfunctional capacity. J Immunol Methods 2007, 328:182-188.

Poulsen TR, Meijer P-J, Jensen A, Nielsen LS, Andersen PS: Kinetic, affinity, and diversity limits of human polyclonal antibody responses against tetanus toxoid. J Immunol (Baltimore, MD) 2007, 179:3841-3850.

Prechl J, Papp K, Erdei A: Antigen microarrays: descriptive chemistry or functional immunomics? Trends Immunol 2010, 31:133-137.

Querec TD, Akondy RS, Lee EK, Cao W, Nakaya HI, Teuwen D, Pirani A, Gernert K, Deng J, Marzolf B et al.: Systems biology approach predicts immunogenicity of the yellow fever vaccine in humans. Nat Immunol 2009, 10:116-125.

Ravn U, Gueneau F, Baerlocher L, Osteras M, Desmurs M, Malinge P, Magistrelli G, Farinelli L, Kosco-Vilbois MH, Fischer N: By-passing in vitro screening—next generation sequencing technologies applied to antibody display and in silico candidate selection. Nucleic Acids Res 2010, 38:e193.

Reddy MM, Wilson R, Wilson J, Connell S, Gocke A, Hynan L, German D, Kodadek T: Identification of candidate IgG biomarkers for Alzheimer’s disease via combinatorial library screening. Cell 2011, 144:132-142.

Reddy ST, Ge X, Miklos AE, Hughes RA, Kang SH, Hoi KH, Chrysostomou C, Hunicke-Smith SP, Iverson BL, Tucker PW et al.: Monoclonal antibodies isolated without screening by analyzing the variable-gene repertoire of plasma cells. Nat Biotechnol 2010, 28:965-969.

Robins HS, Campregher PV, Srivastava SK, Wacher A, Turtle CJ, Kahsai O, Riddell SR, Warren EH, Carlson CS: Comprehensive assessment of T-cell receptor beta-chain diversity in alphabeta T cells. Blood 2009, 114:4099-4107.

Robins HS, Srivastava SK, Campregher PV, Turtle CJ, Andriesen J, Riddell SR, Carlson CS, Warren EH: Overlap and effective size of the human CD8+ T cell receptor repertoire. Sci Transl Med 2010, 2:47ra64.

Robinson WH, DiGennaro C, Hueber W, Haab BB, Kamachi M, Dean EJ, Fournel S, Fong D, Genovese MC, de Vegvar HEN et al.: Autoantigen microarrays for multiplex characterization of autoantibody responses. Nat Med 2002, 8:295-301.

Smith K, Garman L, Wrammert J, Zheng N-Y, Capra JD, Ahmed R, Wilson PC: Rapid generation of fully human monoclonal antibodies specific to a vaccinating antigen. Nat Protoc 2009, 4:372-384.

Song Q, Han Q, Bradshaw EM, Kent SC, Raddassi K, Nilsson B, Nepom GT, Hafler DA, Love JC: On-chip activation and subsequent detection of individual antigen-specific T cells. Anal Chem 2010, 82:473-477.

Story CM, Papa E, Hu C-CA, Ronan JL, Herlihy K, Ploegh HL, Love JC: Profiling antibody responses by multiparametric analysis of primary B cells. Proc Natl Acad Sci U S A 2008, 105:17902-17907.

Tajiri K, Kishi H, Tokimitsu Y, Kondo S, Ozawa T, Kinoshita K, Jin A, Kadowaki S, Sugiyama T, Muraguchi A: Cell-microarray analysis of antigen-specific B-cells: single cell analysis of antigen receptor expression and specificity. Cytometry 2007, 71:961-967.

Tiller T, Meffre E, Yurasov S, Tsuiji M, Nussenzweig MC, Wardemann H: Efficient generation of monoclonal antibodies from single human B cells by single cell RT-PCR and expression vector cloning. J Immunol Methods 2008, 329:112-124.

Tokimitsu Y, Kishi H, Kondo S, Honda R, Tajiri K, Motoki K, Ozawa T, Kadowaki S, Obata T, Fujiki S et al.: Single lymphocyte analysis with a microwell array chip. Cytometry 2007, 71:1003-1010.

Warren RL, Freeman JD, Zeng T, Choe G, Munro S, Moore R, Webb JR, Holt RA: Exhaustive T-cell repertoire sequencing of human peripheral blood samples reveals signatures of antigen selection and a directly measured repertoire size of at least 1 million clonotypes. Genome Res. Advanced Online Publication.

Weinstein JA, Jiang N, White RA, Fisher DS, Quake SR: High-throughput sequencing of the zebrafish antibody repertoire. Science (New York, NY) 2009, 324:807-810.

Wiberg FC, Rasmussen SK, Frandsen TP, Rasmussen LK, Tengbjerg K, Coljee VW, Sharon J, Yang CY, Bregenholt S, Nielsen LS et al.: Production of target-specific recombinant human polyclonal antibodies in mammalian cells. Biotechnol Bioeng 2006, 94:396-405.

Wrammert J, Smith K, Miller J, Langley WA, Kokko K, Larsen C, Zheng N-Y, Mays I, Garman L, Helms C et al.: Rapid cloning of high-affinity human monoclonal antibodies against influenza virus. Nature 2008, 453:667-671.

Yu X, Tsibane T, McGraw PA, House FS, Keefer CJ, Hicar MD, Tumpey TM, Pappas C, Perrone LA, Martinez O et al.: Neutralizing antibodies derived from the B cells of 1918 influenza pandemic survivors. Nature 2008, 455:532-536.

A7

THE NEW SCIENCE OF SOCIOMICROBIOLOGY AND THE REALM OF SYNTHETIC AND SYSTEMS ECOLOGY

E. Peter Greenberg⁴⁴

The Emergence of Sociomicrobiology as an Interest Area in Microbiology

Up until the later part of the past century, microbiologists believed that with rare exceptions bacteria did not practice sociality; this was in spite of the fact that growing bacteria as colonies on agar plates was and still remains a cornerstone technology for the discipline. Just the terminology “colonial” or “colony growth” implies some sort of social interactions. But with the few exceptional scientists working on myxobacteria, a special group of social bacteria, the eyes of microbiology did not see social interactions among individual bacteria. The longstanding view that bacteria were by-and-large asocial creatures began to crumble in the 1990s with rapid developments in the areas of quorum sensing and biofilm research. Now it seems apparent that the same selective pressures that led to the evolution of sociality in animals are forces for evolution of sociality in bacteria, and we see and appreciate social behavior in bacteria (Camilli and Bassler, 2006; Parsek and Greenberg, 2005). I have no proof but I believe that contributing to our past resistance to sociomicrobiology is the fact that bacteria are at once single cells and single individuals. They have served biology as wonderful models for cellular activities. We have not focused as much on the bacterial cell as an individual member of a species as we have on their existence as single-celled organisms.

My laboratory has focused on three areas of sociomicrobiology. The first is quorum sensing, which can be defined as a cell-to-cell communication system that enables individuals to sense the local population density and regulate expression of specific genes in response to population density. This is the primary topic

_______________

⁴⁴ Department of Microbiology, University of Washington School of Medicine.

of this paper. The second is biofilm biology. Biofilms are organized assemblages of bacteria embedded in a self-produced extracellular matrix. As a consequence of the biofilm lifestyle, the assembled bacteria exist in a heterogeneous environment and bacteria in different regions of a biofilm have different functions, which confer different traits to the group. As a consequence of the biofilm lifestyle, at least a subpopulation of the assemblage is protected from environmental stresses and in fact biofilm infections are difficult or impossible to cure by antibiotic treatment (Costerton et al., 1999). Because biofilm infections are very prevalent medical problems there is considerable interest in developing novel therapies that are effective in killing biofilm bacteria. To name just a few, infections of any implanted medical device are biofilm infections, cystic fibrosis lung infections are biofilm infections, heart valve infections are biofilms, and at least some chronic middle ear infections are biofilms (Costerton et al., 1999). Third, we have also worked on conspecific territoriality exhibited by swarms of bacteria moving over surfaces (Gibbs and Greenberg, 2011; Gibbs et al., 2008).

The emergence of sociomicrobiology as a new subdiscipline is important for several reasons. First, there is an idea that we might be able to develop novel antivirulence therapeutics, which target quorum sensing in bacterial species that control virulence gene expression by cell-to-cell signaling. In fact there are many investigators working to identify potent quorum-sensing and biofilm inhibitors, and we ourselves have undertaken such programs together with collaborators in the pharmaceutical industry (Banin et al., 2008; Muh et al., 2006a,b), but this is not the primary motivation of our work. This is in part because it is my belief that we do not yet understand enough about the biology of quorum sensing to predict how, when, or where its inhibition might be of therapeutic value. Second, it is now clear that if one strives to really understand bacteria, social aspects of their biology cannot be ignored. Third, now that we understand that bacteria are social and we understand sociality at an unprecedented level of molecular detail in bacteria like Pseudomonas aeruginosa, these model organisms have become wonderful tools to study fundamental questions about the costs and benefits of cooperation, the selective pressures that lead to cooperative behaviors, and the advantages of controlling cooperative behaviors by communication. Finally, our detailed understanding of quorum-sensing signal synthesis and signal reception, as is discussed below, has led to the widespread use of quorum-sensing regulatory circuits in synthetic biology (Marguet et al., 2010, and references therein).

Quorum Sensing in Proteobacteria

In the late 1960s and early 1970s there was a very modest literature describing pheromone production and activity in bacteria (Eberhard, 1972; Tomasz, 1965). It was not until the 1980s that work on quorum sensing in marine luminescent bacteria led to the idea that these sorts of gene regulatory activities function as intercellular communication systems that coordinate group activities.

Not until the 1990s did we begin to understand the prevalence of quorum sensing in bacteria. Although there are many different types of bacterial cell-to-cell signaling systems considered as quorum-sensing systems, our work focuses on acyl-homoserine lactone quorum-sensing systems prevalent but not universal among the Proteobacteria. Our original model system was a marine bacterium, Vibrio fischeri, which controls a small set of about 25 or fewer genes, including genes for light production, by a transcriptional activator called LuxR, and 3-oxo-hexanoyl-homoserine lactone (3OC6-HSL), which is produced by the luxI gene product (Antunes et al., 2007; Engebrecht et al., 1983). This quorum-sensing system allows V. fischeri to discriminate between its free-living low-population-density lifestyle and its high-density host-associated lifestyle (Figure A7-1). It exists in the light organs of specific marine animals where it produces light, which serves the mutualistic symbiosis. Of note, 3OC6-HSL moves in and out of cells by passive diffusion. In this way the environmental signal concentration is a reflection of cell density. The luxI gene itself is controlled by quorum sensing—it is activated by LuxR and 3OC6-HSL. In terms of biology, this provides hysteresis to the system. The population density required to activate quorum-controlled genes is much higher than the density required to shut down an activated system.

We later turned our attention to the opportunistic pathogen P. aeruginosa. We learned that there were two acyl-HSL circuits, the C4-HSL-RhlI-RhlR circuit and the 3OC12-HSL-LasI-LasR circuits, which together are required for activation of about 325 genes (Figure A7-2) (Pearson et al., 1994, 1995; Schuster et al., 2003). Other investigators showed that at least under certain experimental conditions quorum-sensing mutants were impaired in virulence (Pearson et al., 2000; Tang et al., 1996), and thus the belief was that, as for V. fischeri quorum sensing, P. aeruginosa quorum sensing allowed discrimination between host and free-living states. However, it is not clear from the evidence that this is in fact the case. Among the 300-plus quorum-controlled genes, those coding for extracellular products like exoenzymes or coding for functions required for the production of exoproducts like hydrogen cyanide or pyocyanin are grossly overrepresented.

In a social context these sorts of extracellular products can be considered public goods or resources produced by individuals and shared by all members of the group. Thus we believe quorum sensing serves to coordinate cooperative behaviors. Investigators have devised experiments where growth of P. aeruginosa requires quorum sensing. This can be accomplished by providing protein as the sole source of carbon and energy. Growth requires quorum-sensing-induced production of the exoprotease elastase. In this setting, LasR quorum-sensing mutants will emerge and become a stable, significant minority of the overall population. These cheaters or freeloaders do not bear the cost of contributing to the public good (elastase in this case), but they benefit from use of the public goods (Sandoz et al., 2007). It is obvious that light production by V. fischeri is a shared behavior and we believe this also represents quorum control of cooperation. As far as I am aware, there are no biological systems sensitive enough to detect the light pro-

FIGURE A7-1 Quorum sensing in Vibrio fischeri. (Top) The lux gene cluster. The luxR gene encodes a 3OC6-HSL (AHL)-dependent transcriptional activator of the luxI-G operon. The luxI product is the AHL synthase; luxC, D, and E form a complex responsible for generation of one of the substrates for the luciferase reaction, luxA and B encode the two subunits of luciferase, and the function of luxG remains unknown. (Bottom) Cartoon of a V. fischeri cell producing the diffusible AHL signal. At low cell densities the luminescence operon is transcribed at a basal level. At high cell densities the AHL signal can reach a sufficient concentration and bind to the cellular LuxR protein, which will then activate transcription of the luminescence operon. The substrates for AHL synthesis are shown, as is the molecule. Note the positive autoregulation of luxI, a common feature of R-I circuits.

FIGURE A7-2 Diagram of the acyl-HSL quorum-sensing regulatory circuit in P. aeruginosa. The lasI and R genes are linked on the chromosome, as are the rhlR and I genes. The qscR gene is not linked tightly to any other quorum-sensing regulator. LasI-R sit atop a quorum-sensing cascade. The rhl quorum-sensing system is among the functions controlled by LasI-R and transcription of qscR is positively autoregulated but requires production of 3OC12-HSL by LasI. Pink ball, 3OC12-HSL; orange ball, C4-HSL.

duced by one or a few V. fischeri cells (maximum emission is about 1,000 photons per second per cell). But the light produced by large groups of cells can easily be observed. For example, we can see the light produced by colonies of V. fischeri growing on a Petri plate. So here are two examples where it seems apparent that quorum-sensing functions control and coordinate cooperative behaviors. Is this universal for LuxR-LuxI-acyl-HSL-type systems? Of course we do not know the answer to this question and one might imagine that such regulatory elements have been adapted by different bacteria for different purposes.

As a general principle, the LuxR-LuxI type of regulatory circuits appears to represent elements of true communication. That is, the signal and the signal receptor are coevolved (Keller and Surette, 2006). One finds that a disturbing anthropomorphic trend has crept into the field and the term “bacterial language” is sometimes used to describe quorum-sensing systems. I do not believe any linguist would find evidence for syntax and grammar in the bacterial world.

With the discovery that quorum-sensing-controlled virulence of bacteria like P. aeruginosa, the number of scientists studying acyl-HSL signaling grew exponentially. Generally speaking, studies of quorum sensing and control of social activity in bacteria are important for the several reasons listed earlier in this paper. We might be able to develop novel antivirulence therapeutics that target

quorum sensing in bacteria like P. aeruginosa. Sociality is an important aspect of bacterial biology. Bacteria are now viewed as experimental models for fundamental studies about the costs and benefits of cooperation, the selective pressures that lead to cooperative behaviors, and the advantages of controlling cooperative behaviors by communication. In addition, our detailed understanding of signal synthesis by Lux-type proteins and signal reception by LuxR-type transcription factors has led to their widespread use in the area of synthetic biology (Marguet et al., 2010). This was a largely unintended consequence of our work, and I believe there are three reasons for the popularity of these systems in building synthetic regulatory circuits. First, the acyl-HSLs are diffusible signaling ligands (Kaplan and Greenberg, 1985). Thus, the complication of transport systems for the ligand need not concern the synthetic biologist. Second, the V. fischeri system (Engebrecht et al., 1983), the P. aeruginosa systems, and many other acyl-HSL quorum-sensing circuits have positive autoregulatory loops. Signal production proceeds at a low level until it has accumulated sufficiently to activate quorum-controlled genes, one of which is the gene for its own synthesis. The signal concentration then increases very rapidly. Work in the area of synthetic biology has focused on this phenomenon as a method to reduce noise in regulatory circuits or build gene expression oscillators. Finally, we have identified many different systems and we know of dozens of acyl-HSL signals—the signal specificity of a given LasR homolog resides in the nature of the acyl side group. Thus, multiple systems can be strung together in a single bacterium with each controlling, for example, expression of fluorescent proteins that emit at different wavelengths. Examples showing the diversity of acyl-HSL signals identified in different bacterial species are shown in Figure A7-3.

New Twists to the Acyl-Homoserine Lactone Signaling Story

The P. aeruginosa quorum-sensing circuitry described in Figure A7-2 shows, in addition to LasR-LasI and RhlR-RhlI, a LuxR homolog, QscR. There is no cognate acyl-HSL synthase gene for QscR and it has been termed an orphan quorum-sensing signal receptor (Chugani et al., 2001). These orphans, which have also been called solos (Subramoni and Venturi, 2009), are quite commonly found in sequenced proteobacterial genomes. We know that QscR responds to the LasI-produced 3OC12-HSL, and it controls a set of genes that partially overlaps with the genes controlled by LasR and RhlR (Lee et al., 2006; Lequette et al., 2006). There has been recent interest in studying orphans (Subramoni and Venturi, 2009). There are orphan LuxR homologs in Salmonella and in pathogenic E. coli, neither of which have any luxI homologs. Evidence indicates that the E. coli and Salmonella LuxR orphans respond to acyl-HSLs made by other bacterial species in mixed microbial communities (Soares and Ahmer, 2011; Sperandio, 2010).

Most of the acyl-HSL signals that have been identified are fatty acyl-HSLs with fatty acyl groups of varying carbon length and with a limited number of

FIGURE A7-3 Some examples of acyl-HSL quorum-sensing signals. From top to bottom; the P. aeruginosa RhlI signal butanoyl-HSL, the V. fischeri LuxI signal, 3-oxo-hexanoyl-HSL, The B. mallei BmaI signal, 3-hydroxy-decanoyl-HSL, and the Rhodopseudomonas palustris RpaI signal, para-coumaryl-HSL.

substitutions on the fatty acyl carbon chain. We recently described examples of two bacteria that use LuxI-LuxR homologs to produce and respond to aryl-HSLs. Both are photosynthetic; Rhodopseudomonas palustris uses p-coumaroyl-HSL as a quorum-sensing signal and photosynthetic baradyrhizobia uses cinnamoyl-HSL (Ahlgren et al., 2011; Schaefer et al., 2008). These discoveries broaden the possibilities in terms of signal molecules. Furthermore, the photosynthetic bradyrhizobia produce nanomolar amounts of cinnamoyl-HSL and respond to picomolar

amounts. This is of interest because the fatty acyl-HSL systems that have been described involve responses to nanomolar signal levels and the signals are usually produced at levels in the micromolar range. The bradyrhizobium system is shifted to much lower levels for production and detection. The significance of the low production and ultrasensitivity is not clearly understood. However, it is clear that, by assuming acyl-HSLs will be produced at higher levels, we likely have overlooked at least some acyl-HSL signaling systems.

Concluding Comments

This paper seeks to introduce an emerging field, sociomicrobiology, and to point out opportunities for synthetic biology and therapeutic development in the area of sociomicrobiology. The paper emphasizes quorum sensing in Proteobacteria and provides examples where quorum sensing serves to allow individuals of a species to communicate. This rudimentary form of chemical communication serves to coordinate certain cooperative behaviors. Whether this is a universal characteristic of acyl-HSL signaling systems remains to be seen. There are also a variety of other small-molecule signals produced by bacteria and many of these are considered quorum-sensing signals (Camilli and Bassler, 2006). If and how these intercellular signals are controlling cooperation also remains to be seen. Finally, this Forum has covered topics related to synthetic and systems biology, and opportunities in the area of human health and disease. We need to remain cognizant that bacterial cells are individual organisms involved in complex social interactions. So not only do we need to think about systems and synthetic biology from a cellular perspective, but we need to think about systems and synthetic ecology too.

References

Ahlgren, N. A., C. S. Harwood, A.L. Schaefer, E. Giraud, and E. P. Greenberg. 2011. Aryl-homoserine lactone quorum sensing in stem-nodulating photosynthetic bradyrhizobia. Proceedings of the National Academy of Sciences of the United States of America 108:7183-7188.

Antunes, L. C., A. L. Schaefer, R. B. R. Ferreira, N. Qin, A. M. Stevens, E. G. Ruby, and E. P. Greenberg. 2007. Transcriptome analysis of the Vibrio fischeri LuxR-LuxI regulon. Journal of Bacteriology 189:8387-8391.

Banin, E., A. Lozinski, et al. 2008. The potential of desferrioxamine-gallium as an anti-Pseudomonas therapeutic agent. Proceedings of the National Academy of Sciences of the United States of America 105:16761-16766.

Camilli, A., and B. L. Bassler 2006. Bacterial small-molecule signaling pathways. Science 311: 1113-1116.

Chugani, S. A., M. Whiteley, K. M. Lee, D. D’Argenio, C. Manoil, and E. P. Greenberg. 2001. QscR, a modulator of quorum-sensing signal synthesis and virulence in Pseudomonas aeruginosa. Proceedings of the National Academy of Sciences of the United States of America 98:2752-2757.

Costerton, J. W., P. S. Stewart, and E. P. Greenberg. 1999. Bacterial biofilms: A common cause of persistent infections. Science 284:1318-1322.

Eberhard, A. 1972. Inhibition and activation of bacterial luciferase synthesis. Journal of Bacteriology 109:1101-1105.

Engebrecht, J., K. H. Nealson, and M. Silverman. 1983. Bacterial bioluminescence: Isolation and genetic analysis of functions from Vibrio fischeri. Cell 32:773-781.

Gibbs, K. A., and E. P. Greenberg. 2011. Territoriality in Proteus: Advertisement and aggression. Chemical Reviews 111(1):188-194.

Gibbs, K. A., M. L. Urbanowski, and E. P. Greenberg. 2008. Genetic determinants of self identity and social recognition in bacteria. Science 321:256-259.

Kaplan, H. B., and E. P. Greenberg. 1985. Diffusion of autoinducer is involved in regulation of the Vibrio fischeri luminescence system. Journal of Bacteriology 163:1210-1214.

Keller, L., and M. G. Surette. 2006. Communication in bacteria: An ecological and evolutionary perspective. Nature Reviews Microbiology 4:249-258.

Lee, J. H., Y. Lequette, and E. P. Greenberg. 2006. Activity of purified QscR, a Pseudomonas aeruginosa orphan quorum-sensing transcription factor. Molecular Microbiology 59:602-609.

Lequette, Y., J. H. Lee, F. Ledgham, A. Lazdunski, and E. P. Greenberg. 2006. A distinct QscR regulon in the Pseudomonas aeruginosa quorum-sensing circuit. Journal of Bacteriology 188: 3365-3370.

Marguet, P., Y. Tanouchi, E. Spitz, C. Smith, and L. C. You. 2010. Oscillations by minimal bacterial suicide circuits reveal hidden facets of host-circuit physiology. PLoS One 5:e11909.

Muh, U., B. J. Hare, B. A. Duerkop, M. Schuster, B. L. Hanzelka, R. Heim, E. R. Olson, and E. P. Greenberg. 2006a. A structurally unrelated mimic of a Pseudomonas aeruginosa acyl-homoserine lactone quorum-sensing signal. Proceedings of the National Academy of Sciences of the United States of America 103:16948-16952.

Muh, U., M. Schuster, R. Heim, A. Singh, E. R. Olson, and E. P. Greenberg. 2006b. Novel Pseudomonas aeruginosa quorum-sensing inhibitors identified in an ultra-high-throughput screen. Antimicrobial Agents and Chemotherapy 50:3674-3679.

Parsek, M. R., and E. P. Greenberg. 2005. Sociomicrobiology: The connections between quorum sensing and biofilms. Trends in Microbiology 13:27-33.

Pearson, J. P., K. M. Gray, L. Passador, K. D. Tucker, A. Eberhard, B. H. Iglewski, and E. P. Greenberg. 1994. Structure of the autoinducer required for expression of Pseudomonas aeruginosa virulence genes. Proceedings of the National Academy of Sciences of the United States of America 91:197-201.

Pearson, J. P., L. Passador, B. H. Iglewski, and E. P. Greenberg. 1995. A second N-acylhomoserine lactone signal produced by Pseudomonas aeruginosa. Proceedings of the National Academy of Sciences of the United States of America 92:1490-1494.

Pearson, J. P., M. Feldman, B. H. Iglewski., and A. Prince. 2000. Pseudomonas aeruginosa cell-to-cell signaling is required for virulence in a model of acute pulmonary infection. Infection and Immunity 68:4331-4334.

Sandoz, K. M., S. M. Mitzimberg, and M. Schuster. 2007. Social cheating in Pseudomonas aeruginosa quorum sensing. Proceedings of the National Academy of Sciences of the United States of America 104:15876-15881.

Schaefer, A. L., E. P. Greenberg, C. M. Oliver, Y. Oda, J. J. Huang, G. Bittan-Banin, C. M. Peres, S. Schmidt, K. Juhaszova, J. R. Sufrin, and C. S. Harwood. 2008. A new class of homoserine lactone quorum-sensing signals. Nature 454:595-599.

Schuster, M., C. P. Lostroh, T. Ogi, and E. P. Greenberg. 2003. Identification, timing, and signal specificity of Pseudomonas aeruginosa quorum-controlled genes: A transcriptome analysis. Journal of Bacteriology 185:2066-2079.

Soares, J. A., and B. M. Ahmer. 2011. Detection of acyl-homoserine lactones by Escherichia and Salmonella. Current Opinions in Microbiology 14:188-193.

Sperandio, V. 2010. SdiA sensing of acyl-homoserine lactones by enterohemorrhagic E. coli (EHEC) serotype O157:H7 in the bovine rumen. Gut Microbes 1:432-435.

Subramoni, S., and V. Venturi. 2009. LuxR-family ‘solos’: Bachelor sensors/regulators of signalling molecules. Microbiology 155:1377-1385.

Tang, H. B., E. DiMango, R. Bryan, M. Gambello, B. H. Iglewski, J. B. Goldberg, and A. Prince. 1996. Contribution of specific Pseudomonas aeruginosa virulence factors to pathogenesis of pneumonia in a neonatal mouse model of infection. Infection and Immunity 64:37-43.

Tomasz, A. 1965. Control of the competent state in Pneumococcus by a hormone-like cell product: An example for a new type of regulatory mechanism in bacteria. Nature 208:155-159.

A8

CREATION OF A BACTERIAL CELL CONTROLLED BY A CHEMICALLY SYNTHESIZED GENOME⁴⁵

Daniel G. Gibson,⁴⁶John I. Glass,⁴⁶Carole Lartigue,⁴⁶Vladimir N. Noskov,⁴⁶Ray-Yuan Chuang,⁴⁶Mikkel A. Algire,⁴⁶Gwynedd A. Benders,⁴⁷Michael G. Montague,⁴⁶Li Ma,⁴⁶Monzia M. Moodie,⁴⁶Chuck Merryman,⁴⁶Sanjay Vashee,⁴⁶Radha Krishnakumar,⁴⁶Nacyra Assad-Garcia,⁴⁶Cynthia Andrews-Pfannkoch,⁴⁶Evgeniya A. Denisova,⁴⁶Lei Young,⁴⁶Zhi-Qing Qi,⁴⁶Thomas H. Segall-Shapiro,⁴⁶Christopher H. Calvey,⁴⁶Prashanth P. Parmar,⁴⁶Clyde A. Hutchison III,⁴⁷Hamilton O. Smith,⁴⁷J. Craig Venter^46,47,48

We report the design, synthesis, and assembly of the 1.08–mega–base pair Mycoplasma mycoides JCVI-syn1.0 genome starting from digitized genome sequence information and its transplantation into a M. capricolum recipient cell to create new M. mycoides cells that are controlled only by the

_______________

⁴⁵ From Daniel G. Gibson, et al. 2010. Creation of a Bacterial Cell Controlled by a Chemically Synthesized Genome. Science 329:52-56. Reprinted with permission from AAAS.

⁴⁶ The J. Craig Venter Institute, 9704 Medical Center Drive, Rockville, MD 20850, USA.

⁴⁷ The J. Craig Venter Institute, 10355 Science Center Drive, San Diego, CA 92121, USA.

⁴⁸ To whom correspondence should be addressed. E-mail: jcventer@jcvi.org
We thank Synthetic Genomics, Inc. for generous funding of this work. We thank J. B. Hostetler, D. Radune, N. B. Fedorova, M. D. Kim, B. J. Szczypinski, I. K. Singh, J. R. Miller, S. Kaushal, R. M. Friedman, and J. Mulligan for their contributions to this work. Electron micrographs were generously provided by T. Deerinck and M. Ellisman of the National Center for Microscopy and Imaging Research at the University of California at San Diego. J.C.V. is chief executive officer and co-chief scientific officer of SGI. H.O.S. is co-chief scientific officer and on the Board of Directors of SGI. C.A.H. is chairman of the SGI Scientific Advisory Board. All three of these authors and JCVI hold SGI stock. JCVI has filed patent applications on some of the techniques described in this paper.
Supporting Online Material www.sciencemag.org/cgi/content/full/science.1190719/DC1
Materials and Methods
Figs. S1 to S6
Tables S1 to S7
References
9 April 2010; accepted 13 May 2010
Published online 20 May 2010; 10.1126/science.1190719.
Include this information when citing this paper.

synthetic chromosome. The only DNA in the cells is the designed synthetic DNA sequence, including “watermark” sequences and other designed gene deletions and polymorphisms, and mutations acquired during the building process. The new cells have expected phenotypic properties and are capable of continuous self-replication.

In 1977, Sanger and colleagues determined the complete genetic sequence of phage ϕX174 (Sanger et al., 1977), the first DNA genome to be completely sequenced. Eighteen years later, in 1995, our team was able to read the first complete genetic sequence of a self-replicating bacterium, Haemophilus influenzae (Fleischmann et al., 1995). Reading the genetic sequence of a wide range of species has increased exponentially from these early studies. The ability to rapidly digitize genomic information has increased by more than eight orders of magnitude over the past 25 years (Venter, 2010). Efforts to understand all this new genomic information have spawned numerous new computational and experimental paradigms, yet our genomic knowledge remains very limited. No single cellular system has all of its genes understood in terms of their biological roles. Even in simple bacterial cells, do the chromosomes contain the entire genetic repertoire? If so, can a complete genetic system be reproduced by chemical synthesis starting with only the digitized DNA sequence contained in a computer?

Our interest in synthesis of large DNA molecules and chromosomes grew out of our efforts over the past 15 years to build a minimal cell that contains only essential genes. This work was inaugurated in 1995 when we sequenced the genome of Mycoplasma genitalium, a bacterium with the smallest complement of genes of any known organism capable of independent growth in the laboratory. More than 100 of the 485 protein-coding genes of M. genitalium are dispensable when disrupted one at a time (Hutchison et al., 1999; Glass et al., 2006; Smith et al., 2008). We developed a strategy for assembling viral-sized pieces to produce large DNA molecules that enabled us to assemble a synthetic M. genitalium genome in four stages from chemically synthesized DNA cassettes averaging about 6 kb in size. This was accomplished through a combination of in vitro enzymatic methods and in vivo recombination in Saccharomyces cerevisiae. The whole synthetic genome [582,970 base pairs (bp)] was stably grown as a yeast centromeric plasmid (YCp) (Gibson et al., 2008b).

Several hurdles were overcome in transplanting and expressing a chemically synthesized chromosome in a recipient cell. We needed to improve methods for extracting intact chromosomes from yeast. We also needed to learn how to transplant these genomes into a recipient bacterial cell to establish a cell controlled only by a synthetic genome. Because M. genitalium has an extremely slow growth rate, we turned to two faster-growing mycoplasma species, M. mycoides subspecies capri (GM12) as donor, and M. capricolum subspecies capricolum (CK) as recipient.

To establish conditions and procedures for transplanting the synthetic ge-

nome out of yeast, we developed methods for cloning entire bacterial chromosomes as centromeric plasmids in yeast, including a native M. mycoides genome (Lartigue et al., 2009; Benders et al., 2010). However, initial attempts to extract the M. mycoides genome from yeast and transplant it into M. capricolum failed. We discovered that the donor and recipient mycoplasmas share a common restriction system. The donor genome was methylated in the native M. mycoides cells and was therefore protected against restriction during the transplantation from a native donor cell (Lartigue et al., 2007). However, the bacterial genomes grown in yeast are unmethylated and so are not protected from the single restriction system of the recipient cell. We overcame this restriction barrier by methylating the donor DNA with purified methylases or crude M. mycoides or M. capricolum extracts, or by simply disrupting the recipient cell’s restriction system (Lartigue et al., 2009).

We now have combined all of our previously established procedures and report the synthesis, assembly, cloning, and successful transplantation of the 1.08-Mbp M. mycoides JCVI-syn1.0 genome, to create a new cell controlled by this synthetic genome.

Synthetic Genome Design

Design of the M. mycoides JCVI-syn1.0 genome was based on the highly accurate finished genome sequences of two laboratory strains of M. mycoides subspecies capri GM12 (Lartigue et al., 2009; Benders et al., 2010).⁴⁹ One was the genome donor used by Lartigue et al. [GenBank accession CP001621] (2007). The other was a strain created by transplantation of a genome that had been cloned and engineered in yeast, YCpMmyc1.1-ΔtypeIIIres [GenBank accession CP001668] (Lartigue et al., 2009). This project was critically dependent on the accuracy of these sequences. Although we believe that both finished M. mycoides genome sequences are reliable, there are 95 sites at which they differ. We began to design the synthetic genome before both sequences were finished. Consequently, most of the cassettes were designed and synthesized based on the CP001621 sequence.⁴⁹ When it was finished, we chose the sequence of the genome successfully transplanted from yeast (CP001668) as our design reference (except that we kept the intact typeIIIres gene). All differences that appeared biologically significant between CP001668 and previously synthesized cassettes were corrected to match it exactly.⁴⁹ Sequence differences between our synthetic cassettes and CP001668 that occurred at 19 sites appeared harmless and so were not corrected. These provide 19 polymorphic differences between our synthetic genome (JCVI-syn1.0) and the natural (nonsynthetic) genome (YCpMmyc1.1) that we have cloned in yeast and use as a standard for genome transplantation from yeast (Lartigue et al., 2009). To further differentiate between the synthetic

_______________

⁴⁹ Supporting material is available on Science Online.

genome and the natural one, we designed four watermark sequences (fig. S1) to replace one or more cassettes in regions experimentally demonstrated [watermarks 1 (1246 bp) and 2 (1081 bp)] or predicted [watermarks 3 (1109 bp) and 4 (1222 bp)] to not interfere with cell viability. These watermark sequences encode unique identifiers while limiting their translation into peptides. Table S1 lists the differences between the synthetic genome and this natural standard. Figure S2 shows a map of the M. mycoides JCVI-syn1.0 genome. Cassette and assembly intermediate boundaries, watermarks, deletions, insertions, and genes of the M. mycoides JCVI syn1.0 are shown in fig. S2, and the sequence of the transplanted mycoplasma clone sMmYCp235-1 has been submitted to GenBank (accession CP002027).

Synthetic Genome Assembly Strategy

The designed cassettes were generally 1080 bp with 80-bp overlaps to adjacent cassettes.⁵⁰ They were all produced by assembly of chemically synthesized oligonucleotides by Blue Heron (Bothell, Washington). Each cassette was individually synthesized and sequence-verified by the manufacturer. To aid in the building process, DNA cassettes and assembly intermediates were designed to contain Not I restriction sites at their termini and recombined in the presence of vector elements to allow for growth and selection in yeast (Gibson et al., 2008b).⁵⁰ A hierarchical strategy was designed to assemble the genome in three stages by transformation and homologous recombination in yeast from 1078 1-kb cassettes (Fig. A8-1) (Gibson, 2009; Gibson et al., 2008a).

Assembly of 10-kb synthetic intermediates

In the first stage, cassettes and a vector were recombined in yeast and transferred to Escherichia coli.⁵⁰ Plasmid DNA was then isolated from individual E. coli clones and digested to screen for cells containing a vector with an assembled 10-kb insert. One successful 10-kb assembly is represented (Fig. A8-2A). In general, at least one 10-kb assembled fragment could be obtained by screening 10 yeast clones. However, the rate of success varied from 10 to 100%. All of the first-stage intermediates were sequenced. Nineteen out of 111 assemblies contained errors. Alternate clones were selected, sequence-verified, and moved on to the next assembly stage.⁵⁰

Assembly of 100-kb synthetic intermediates

The pooled 10-kb assemblies and their respective cloning vectors were transformed into yeast as above to produce 100-kb assembly intermediates.⁵⁰

_______________

⁵⁰ Supporting material is available on Science Online.

FIGURE A8-1 The assembly of a synthetic M. mycoides genome in yeast. A synthetic M. mycoides genome was assembled from 1078 overlapping DNA cassettes in three steps. In the first step, 1080-bp cassettes (orange arrows), produced from overlapping synthetic oligonucleotides, were recombined in sets of 10 to produce 109 ~10-kb assemblies (blue arrows). These were then recombined in sets of 10 to produce 11 ~100-kb assemblies (green arrows). In the final stage of assembly, these 11 fragments were recombined into the complete genome (red circle). With the exception of two constructs that were enzymatically pieced together in vitro (Gibson et al., 2009) (white arrows), assemblies were carried out by in vivo homologous recombination in yeast. Major variations from the natural genome are shown as yellow circles. These include four watermarked regions (WM1 to WM4), a 4-kb region that was intentionally deleted (94D), and elements for growth in yeast and genome transplantation. In addition, there are 20 locations with nucleotide polymorphisms (asterisks). Coordinates of the genome are relative to the first nucleotide of the natural M. mycoides sequence. The designed sequence is 1,077,947 bp. The locations of the Asc I and BssH II restriction sites are shown. Cassettes 1 and 800-810 were unnecessary and removed from the assembly strategy. (Supporting material is available on Science Online.) Cassette 2 overlaps cassette 1104, and cassette 799 overlaps cassette 811.

FIGURE A8-2 Analysis of the assembly intermediates. (A) Not I and Sbf I double restriction digestion analysis of assembly 341-350 purified from E. coli. These restriction enzymes release the vector fragments (5.5 and 3.4 kb) from the 10-kb insert. Insert DNA was separated from the vector DNA on a 0.8% E-gel (Invitrogen). M indicates the 1-kb DNA ladder (New England Biolabs; NEB). (B) Analysis of assembly 501-600 purified from yeast. The 105-kb circles (100-kb insert plus 5-kb vector) were separated from the linear yeast chromosomal DNA on a 1% agarose gel by applying 4.5 V/cm for 3 hours. S indicates the BAC-Tracker supercoiled DNA ladder (Epicentre). (C) Not I restriction digestion analysis of the 11 ~100-kb assemblies purified from yeast. These DNA fragments were analyzed by FIGE on a 1% agarose gel. The expected insert size for each assembly is indicated. λ indicates the lambda ladder (NEB). (D) Analysis of the 11 pooled assemblies shown in (C) following topological trapping of the circular DNA and Not I digestion. One-fortieth of the DNA used to transform yeast is represented.

Our results indicated that these products cannot be stably maintained in E. coli, so recombined DNA had to be extracted from yeast. Multiplex polymerase chain reaction (PCR) was performed on selected yeast clones (fig. S3 and table S2). Because every 10-kb assembly intermediate was represented by a primer pair in this analysis, the presence of all amplicons would suggest an assembled 100-kb

intermediate. In general, 25% or more of the clones screened contained all of the amplicons expected for a complete assembly. One of these clones was selected for further screening. Circular plasmid DNA was extracted and sized on an agarose gel alongside a supercoiled marker. Successful second-stage assemblies with the vector sequence are ~105 kb in length (Fig. A8-2B). When all amplicons were produced following multiplex PCR, a second-stage assembly intermediate of the correct size was usually produced. In some cases, however, small deletions occurred. In other instances, multiple 10-kb fragments were assembled, which produced a larger second-stage assembly intermediate. Fortunately, these differences could easily be detected on an agarose gel before complete genome assembly.

Complete genome assembly

In preparation for the final stage of assembly, it was necessary to isolate microgram quantities of each of the 11 second-stage assemblies.⁵¹ As reported (Devenish and Newlon, 1982), circular plasmids the size of our second-stage assemblies could be isolated from yeast spheroplasts after an alkaline-lysis procedure. To further purify the 11 assembly intermediates, they were treated with exonuclease and passed through an anion-exchange column. A small fraction of the total plasmid DNA (1/100) was digested with Not I and analyzed by field-inversion gel electrophoresis (FIGE) (Fig. A8-2C). This method produced ~1 mg of each assembly per 400 ml of yeast culture (~10¹¹ cells).

The method above does not completely remove all of the linear yeast chromosomal DNA, which we found could substantially decrease the yeast transformation and assembly efficiency. To further enrich for the 11 circular assembly intermediates, ~200 ng samples of each assembly were pooled and mixed with molten agarose. As the agarose solidifies, the fibers thread through and topologically “trap” circular DNA (Dean et al., 1973). Untrapped linear DNA can then be separated out of the agarose plug by electrophoresis, thus enriching for the trapped circular molecules. The 11 circular assembly intermediates were digested with Not I so that the inserts could be released. Subsequently, the fragments were extracted from the agarose plug, analyzed by FIGE (Fig. A8-2D), and transformed into yeast spheroplasts.⁵¹ In this third and final stage of assembly, an additional vector sequence was not required because the yeast cloning elements were already present in assembly 811-900.

To screen for a complete genome, multiplex PCR was carried out with 11 primer pairs, designed to span each of the 11 100-kb assembly junctions (table S3). Of 48 colonies screened, DNA extracted from one clone (sMmYCp235) produced all 11 amplicons. PCR of the wild-type positive control (YCpMmyc1.1) produced an indistinguishable set of 11 amplicons (Fig. A8-3A). To further demonstrate the complete assembly of a synthetic M. mycoides genome, intact DNA

_______________

⁵¹ Supporting material is available on Science Online.

FIGURE A8-3 Characterization of the synthetic genome isolated from yeast. (A) Yeast clones containing a completely assembled synthetic genome were screened by multiplex PCR with a primer set that produces 11 amplicons; one at each of the 11 assembly junctions. Yeast clone sMmYCp235 (235) produced the 11 PCR products expected for a complete genome assembly. For comparison, the natural genome extracted from yeast (WT, wild type) was also analyzed. PCR products were separated on a 2% E-gel (Invitrogen). L indicates the 100-bp ladder (NEB). (B) The sizes of the expected Asc I and BssH II restriction fragments for natural (WT) and synthetic (Syn235) M. mycoides genomes. (C) Natural (WT) and synthetic (235) M. mycoides genomes were isolated from yeast in agarose plugs. In addition, DNA was purified from the host strain alone (H). Agarose plugs were digested with Asc I or BssH II, and fragments were separated by clamped homogeneous electrical field (CHEF) gel electrophoresis. Restriction fragments corresponding to the correct sizes are indicated by the fragment numbers shown in (B).

was isolated from yeast in agarose plugs and subjected to two restriction analyses: Asc I and BssH II.⁵² Because these restriction sites are present in three of the four watermark sequences, this choice of digestion produces restriction patterns that are distinct from that of the natural M. mycoides genome (Figs. A8-1 and A8-3B). The sMmYCp235 clone produced the restriction pattern expected for a completely assembled synthetic genome (Fig. A8-3C).

Synthetic Genome Transplantation

Additional agarose plugs used in the gel analysis above (Fig. A8-3C) were also used in genome transplantation experiments.⁵² Intact synthetic M. mycoides

_______________

⁵² Supporting material is available on Science Online.

genomes from the sMmYCp235 yeast clone were transplanted into restriction-minus M. capricolum recipient cells, as described (Lartigue et al., 2009). Results were scored by selecting for growth of blue colonies on SP4 medium containing tetracycline and X-gal at 37°C. Genomes isolated from this yeast clone produced 5 to 15 tetracycline-resistant blue colonies per agarose plug, a number comparable to that produced by the YCpMmyc1.1 control. Recovery of colonies in all transplantation experiments was dependent on the presence of both M. capricolum recipient cells and an M. mycoides genome.

Semisynthetic Genome Assembly and Transplantation

To aid in testing the functionality of each 100-kb synthetic segment, semisynthetic genomes were constructed and transplanted. By mixing natural pieces with synthetic ones, the successful construction of each synthetic 100-kb assembly could be verified without having to sequence these intermediates. We cloned 11 overlapping natural 100-kb assemblies in yeast by using a previously described method (Leem et al., 2003). In 11 parallel reactions, yeast cells were cotransformed with fragmented M. mycoides genomic DNA (YCpMmyc1.1) that averaged ~100 kb in length and a PCR-amplified vector designed to overlap the ends of the 100-kb inserts. To maintain the appropriate overlaps so that natural and synthetic fragments could be recombined, the PCR-amplified vectors were produced via primers with the same 40-bp overlaps used to clone the 100-kb synthetic assemblies. The semisynthetic genomes that were constructed contained between 2 and 10 of the 11 100-kb synthetic subassemblies (Table A8-1). The

TABLE A8-1 Genomes that have been assembled from 11 pieces and successfully transplanted. Assembly 2-100, 1; assembly 101-200, 2; assembly 201-300, 3; assembly 301-400, 4; assembly 401-500, 5; assembly 501-600, 6; assembly 601-700, 7; assembly 701-799, 8; assembly 811-900, 9; assembly 901-1000, 10; assembly 1001-1104, 11. WM, watermarked assembly.

production of viable colonies produced after transplantation confirmed that the synthetic fraction of each genome contained no lethal mutations. Only one of the 100-kb subassemblies, 811-900, was not viable.

Initially, an error-containing 811-820 clone was used to produce a synthetic genome that did not transplant. This was expected because the error was a single–base pair deletion that creates a frameshift in dnaA, an essential gene for chromosomal replication. We were previously unaware of this mutation. By using a semisynthetic genome construction strategy, we pinpointed 811-900 as the source for failed synthetic transplantation experiments. Thus, we began to reassemble an error-free 811-900 assembly, which was used to produce the sMmYCp235 yeast strain. The dnaA-mutated genome differs by only one nucleotide from the synthetic genome in sMmYCp235. This genome served as a negative control in our transplantation experiments. The dnaA mutation was also repaired at the 811-900 level by genome engineering in yeast (Noskov et al., 2010). A repaired 811-900 assembly was used in a final-stage assembly to produce a yeast clone with a repaired genome. This yeast clone is named sMmYCP142 and could be transplanted. A complete list of genomes that have been assembled from 11 pieces and successfully transplanted is provided in Table A8-1.

Characterization of the Synthetic Transplants

To rapidly distinguish the synthetic transplants from M. capricolum or natural M. mycoides, two analyses were performed. First, four primer pairs that are specific to each of the four watermarks were designed such that they produce four amplicons in a single multiplex PCR reaction (table S4). All four amplicons were produced by transplants generated from sMmYCp235, but not YCpMmyc1.1 (Fig. A8-4A). Second, the gel analysis with Asc I and BssH II, described above (Fig. A8-3C), was performed. The restriction pattern obtained was consistent with a transplant produced from a synthetic M. mycoides genome (Fig. A8-4B).

A single transplant originating from the sMmYCp235 synthetic genome was sequenced. We refer to this strain as M. mycoides JCVIsyn1.0. The sequence matched the intended design with the exception of the known polymorphisms, eight new single-nucleotide polymorphisms, an E. coli transposon insertion, and an 85-bp duplication (table S1). The transposon insertion exactly matches the size and sequence of IS1, a transposon in E. coli. It is likely that IS1 infected the 10-kb subassembly following its transfer to E. coli. The IS1 insert is flanked by direct repeats of M. mycoides sequence, suggesting that it was inserted by a transposition mechanism. The 85-bp duplication is a result of a nonhomologous end joining event, which was not detected in our sequence analysis at the 10-kb stage. These two insertions disrupt two genes that are evidently nonessential. We did not find any sequences in the synthetic genome that could be identified as belonging to M. capricolum. This indicates that there was a complete replacement of the M. capricolum genome by our synthetic genome during the trans-

FIGURE A8-4 Characterization of the transplants. (A) Transplants containing a synthetic genome were screened by multiplex PCR with a primer set that produces four amplicons, one internal to each of the four watermarks. One transplant (syn1.0) originating from yeast clone sMmYCp235 was analyzed alongside a natural, nonsynthetic genome (WT) transplanted out of yeast. The transplant containing the synthetic genome produced the four PCR products, whereas the WT genome did not produce any. PCR products were separated on a 2% E-gel (Invitrogen). (B) Natural (WT) and synthetic (syn1.0) M. mycoides genomes were isolated from M. mycoides transplants in agarose plugs. Agarose plugs were digested with Asc I or BssH II and fragments were separated by CHEF gel electrophoresis. Restriction fragments corresponding to the correct sizes are indicated by the fragment numbers shown in Fig. A8-3B.

plant process. The cells with only the synthetic genome are self-replicating and capable of logarithmic growth. Scanning and transmission electron micrographs (EMs) of M. mycoides JCVI-syn1.0 cells show small, ovoid cells surrounded by cytoplasmic membranes (Fig. A8-5, C to F). Proteomic analysis of M. mycoides JCVI-syn1.0 and the wild-type control (YCpMmyc1.1) by two-dimensional gel electrophoresis revealed almost identical patterns of protein spots (fig. S4) that differed from those previously reported for M. capricolum (Lartigue et al., 2007). Fourteen genes are deleted or disrupted in the M. mycoides JCVI-syn1.0 genome; however, the rate of appearance of colonies on agar plates and the colony morphology are similar (compare Fig. A8-5, A and B). We did observe slight differences in the growth rates in a color-changing unit assay, with the JCVI-syn1.0 transplants growing slightly faster than the MmcyYCp1.1 control strain (fig. S6).

FIGURE A8-5 Images of M. mycoides JCVI-syn1.0 and WT M. mycoides. To compare the phenotype of the JCVI-syn1.0 and non-YCp WT strains, we examined colony morphology by plating cells on SP4 agar plates containing X-gal. Three days after plating, the JCVI-syn1.0 colonies are blue because the cells contain the lacZ gene and express β-lactosidase, which converts the X-gal to a blue compound (A). The WT cells do not contain lacZ and remain white (B). Both cell types have the fried egg colony morphology characteristic of most mycoplasmas. EMs were made of the JCVI-syn1.0 isolate using two methods. (C) For scanning EM, samples were postfixed in osmium tetroxide, dehydrated and critical point dried with CO₂, and visualized with a Hitachi SU6600 SEM at 2.0 keV. (D) Negatively stained transmission EMs of dividing cells with 1% uranyl acetate on pure carbon substrate visualized using JEOL 1200EX CTEM at 80 keV. To examine cell morphology, we compared uranyl acetate–stained EMs of M. mycoides JCVI-syn1.0 cells (E) with EMs of WT cells made in 2006 that were stained with ammonium molybdate (F). Both cell types show the same ovoid morphology and general appearance. EMs were provided by T. Deerinck and M. Ellisman of the National Center for Microscopy and Imaging Research at the University of California at San Diego.

Discussion

In 1995, the quality standard for sequencing was considered to be one error in 10,000 bp, and the sequencing of a microbial genome required months. Today, the accuracy is substantially higher. Genome coverage of 30 to 50× is not unusual, and sequencing only requires a few days. However, obtaining an error-free genome that could be transplanted into a recipient cell to create a new cell controlled only by the synthetic genome was complicated and required many quality-control steps. Our success was thwarted for many weeks by a single–base

pair deletion in the essential gene dnaA. One wrong base out of more than 1 million in an essential gene rendered the genome inactive, whereas major genome insertions and deletions in nonessential parts of the genome had no observable effect on viability. The demonstration that our synthetic genome gives rise to transplants with the characteristics of M. mycoides cells implies that the DNA sequence on which it is based is accurate enough to specify a living cell with the appropriate properties.

Our synthetic genomic approach stands in sharp contrast to various other approaches to genome engineering that modify natural genomes by introducing multiple insertions, substitutions, or deletions (Itaya et al., 2005; Itaya, 1995; Mizoguchi et al., 2007; Chun et al., 2007; Wang et al., 2009). This work provides a proof of principle for producing cells based on computer-designed genome sequences. DNA sequencing of a cellular genome allows storage of the genetic instructions for life as a digital file.

The synthetic genome described here has only limited modifications from the naturally occurring M. mycoides genome. However, the approach we have developed should be applicable to the synthesis and transplantation of more novel genomes as genome design progresses (Khalil and Collins, 2010). We refer to such a cell controlled by a genome assembled from chemically synthesized pieces of DNA as a “synthetic cell,” even though the cytoplasm of the recipient cell is not synthetic. Phenotypic effects of the recipient cytoplasm are diluted with protein turnover and as cells carrying only the transplanted genome replicate. Following transplantation and replication on a plate to form a colony (>30 divisions or >10⁹-fold dilution), progeny will not contain any protein molecules that were present in the original recipient cell (Lartigue et al., 2007).⁵³ This was previously demonstrated when we first described genome transplantation (Lartigue et al., 2007). The properties of the cells controlled by the assembled genome are expected to be the same as if the whole cell had been produced synthetically (the DNA software builds its own hardware).

The ability to produce synthetic cells renders it essential for researchers making synthetic DNA constructs and cells to clearly watermark their work to distinguish it from naturally occurring DNA and cells. We have watermarked the synthetic chromosome in this and our previous study (Gibson, et al., 2008b).

If the methods described here can be generalized, design, synthesis, assembly, and transplantation of synthetic chromosomes will no longer be a barrier to

_______________

⁵³ A mycoplasma cell, with a mass of about 10^–13 g, contains fewer than 10⁶ molecules of protein. (If it contains 20% protein, this is equivalent to 2 × 10^–14 g of protein per cell. At a molecular mass of 120 daltons per amino acid residue, each cell contains (2 × 10^–14)/120 = 1.7 × 10^–16 mol of peptide residues. This is equivalent to (1.7 × 10^–16) × (6 × 10²³) = 1 × 10⁸ residues per cell. If the average size of a protein is 300 residues, then a cell contains about 3 × 10⁵ protein molecules.) After 20 cell divisions the number of progeny exceeds the total number of protein molecules in the recipient cell. So, following transplantation and replication to form a colony on a plate, most cells will contain no protein molecules that were present in the original recipient cell.

the progress of synthetic biology. We expect that the cost of DNA synthesis will follow what has happened with DNA sequencing and continue to exponentially decrease. Lower synthesis costs combined with automation will enable broad applications for synthetic genomics.

We have been driving the ethical discussion concerning synthetic life from the earliest stages of this work (Cho et al., 1999; Garfinkel et al., 2007). As synthetic genomic applications expand, we anticipate that this work will continue to raise philosophical issues that have broad societal and ethical implications. We encourage the continued discourse.

References

G. A. Benders et al., Nucleic Acids Res. 38, 2558 (2010).

M. K. Cho, D. Magnus, A. L. Caplan, D. McGee, Science 286, 2087, 2089 (1999).

J. Y. Chun et al., Nucleic Acids Res. 35, e40 (2007).

W. W. Dean, B. M. Dancis, C. A. Thomas Jr., Anal. Biochem. 56, 417 (1973).

R. J. Devenish, C. S. Newlon, Gene 18, 277 (1982).

R. D. Fleischmann et al., Science 269, 496 (1995).

M. S. Garfinkel, D. Endy, G. L. Epstein, R. M. Friedman, Biosecur. Bioterror. 5, 359 (2007).

D. G. Gibson et al., Proc. Natl. Acad. Sci. U.S.A. 105, 20404 (2008a).

——, Science 319, 1215 (2008b).

——, Nat. Methods 6, 343 (2009a).

D. G. Gibson, Nucleic Acids Res. 37, 6984 (2009b).

J. I. Glass et al., Proc. Natl. Acad. Sci. U.S.A. 103, 425 (2006).

C. A. Hutchison III et al., Science 286, 2165 (1999).

M. Itaya, FEBS Lett. 362, 257 (1995).

M. Itaya, K. Tsuge, M. Koizumi, K. Fujita, Proc. Natl. Acad. Sci. U.S.A. 102, 15971 (2005).

A. S. Khalil, J. J. Collins, Nat. Rev. Genet. 11, 367 (2010).

C. Lartigue et al., Science 317, 632 (2007).

——, Science 325, 1693 (2009).

S.-H. Leem et al., Nucleic Acids Res. 31, e29 (2003).

H. Mizoguchi, H. Mori, T. Fujio, Biotechnol. Appl. Biochem. 46, 157 (2007).

V. N. Noskov, T. H. Segall-Shapiro, R. Y. Chuang, Nucleic Acids Res. 38, 2570 (2010).

F. Sanger et al., Nature 265, 687 (1977).

H. O. Smith, J. I. Glass, C. A. Hutchison III, J. C. Venter, in Accessing Uncultivated Microorganisms: From the Environment to Organisms and Genomes and Back, K. Zengler, Ed. (American Society for Microbiology, Washington, DC, 2008), p. 320.

J. C. Venter, Nature 464, 676 (2010).

H. H. Wang et al., Nature 460, 894 (2009).

A9

SYNTHETIC BIOLOGY “FROM SCRATCH”

Gerald F. Joyce⁵⁴

The Darwinian Basis of Life

The previous presentations in this session described a mix of top-down and bottom-up approaches to synthetic biology. This presentation is intended to take us all the way down to the very bottom as we discuss synthetic biology “from scratch,” that is, life from scratch. Life is something we all know when we see it, right? Consider the giant bacterium Titanospirillum velox, which we all would agree is alive. Yet in a recent publication by Richard Hoover that appeared in the online Journal of Cosmology (Hoover, 2011), he describes what he believes is an extraterrestrial analog of Titanospirillum, found within a certain class of carbonaceous meteorites. Much of the popular media fell for it, abetted by a Fox News report showing a micrograph of Titanospirillum, rather than the mineralic artifacts within the meteorite that most experts agree have nothing to do with life.

But surely the experts must know life when they see it, right? And therefore the experts must know how to define life. Indeed some scientists have advanced a definition, or at least a working definition, of life. Some, including me, have pointed to the so-called NASA working definition of life, which is a self-sustained chemical system that is capable of undergoing Darwinian evolution (Joyce, 1994). In several of the presentations at this workshop, evolution has been referred to as the distinguishing, if not defining, feature of life. The Darwinian paradigm is the only one we know that explains how biological complexity can sustain itself against the vagaries of a changing environment. Darwinian evolution has a rigorous scientific foundation, but the word “life” is not a scientific term. As Andrew Ellington would say, it is a term for poets and philosophers, and scientists, let alone a government agency such as NASA, should not be in the business of trying to define it.

Without becoming bogged down in definitions, we can agree that life is all about Darwinian evolution, and that scientists understand what Darwinian evolution is all about. The key principles of Darwinian evolution are, first, heritable variation of form and function among a population of individuals; second, competition for finite resources by those individuals; and third, preferential reproduction of variants that operate most effectively in the competitive environment. In The Origin of Species, Charles Darwin made the observation: “Owing to this struggle for life, any variation…if it be in any degree profitable to an individual of any species, in its infinitely complex relations to other organic beings and to external nature…will generally be inherited by its offspring” (Darwin, 1859;

_______________

⁵⁴ The Scripps Research Institute.

italics added). If life has a slogan, if not a formal definition, it would be inherit profitable variation. I, for one, would be comfortable abiding by this slogan.

In considering synthetic biology from scratch, the focus is on the evolution of functional molecules, rather than organisms. The principles of directed molecular evolution are the same as those for the Darwinian evolution of organisms, again captured by the slogan inherit profitable variation. In chemical terms, Darwinian evolution involves three processes: (1) reproduction of information-carrying molecules (inherit), (2) selection of molecules that meet some fitness criteria (profitable), and (3) maintenance of chemical diversity among the population of molecules (variation).

Over the past two decades, the technology of direct molecular evolution has become very powerful, but also routine. There are many methods for introducing molecular variation, both for generating initial combinatorial libraries and for maintaining variation in a population. Individuals that meet some fitness test can be separated based on a high-throughput screen or a selection procedure. Through screening, one or more highly advantageous variants can be identified and subsequently mutagenized to provide a second-generation combinatorial library. This process of iterative high-throughput screening can be a powerful discovery tool, although it does not fully capture the power of Darwinian evolution, which requires that the population as a whole be subject to repeated rounds of selection and randomization. Maintaining a diverse population is key to exploring the fitness landscape because it allows both dominant and subdominant individuals to give rise to novel variants. The subdominant individuals, following the acquisition of a few beneficial mutations, may yield descendants that are more advantageous compared to the previously dominant individuals. Ideally, the loss of variation due to selection should be compensated by the introduction of novel variation throughout the course of evolution.

There are many methods for selecting profitable molecules. For example, molecules can be selected based on their specific chromatographic mobility, their ability to withstand exposure to some physical condition, their ability to bind to a target ligand (aptamers), or their ability to catalyze a particular chemical transformation (enzymes). More complex selection criteria can be imposed, such as a combination of both positive and negative selection, conditional selection, and the selection for multiple attributes.

Finally, there are various methods for reproducing the profitable molecules in order to bring about the inheritance of selectively advantageous traits. If the selected molecules are DNA or RNA, then it is straightforward to achieve their amplification by using the appropriate polymerase enzyme(s), resulting in large numbers of progeny. If the selected molecules are proteins, which cannot be amplified directly, then one must amplify nucleic acid molecules that encode and are physically linked to the corresponding proteins. Techniques such as phage display, ribosome display, and compartmentalized self-replication make it possible to amplify an ensemble of genes that encode a corresponding set of proteins. The same principle of genetic encoding can be used to amplify other

informational macromolecules, such as peptide analogs, polysaccharides, and even multicomponent organic molecules.

All of the amplification methods discussed above are not part of a self-sustained evolving system because they rely on informational macromolecules that are not themselves subject to evolution within the system. Protein polymerases, phage particles, and ribosomes all are the products of Darwinian evolution in biology. They are employed as operators in laboratory evolution systems, but their information content is not subject to evolution within those systems.

The experimental pursuit of life from scratch began in 1953 with the work of Stanley Miller, then a graduate student at the University of Chicago, who sought to cook up a prebiotic soup from entirely abiotic ingredients (Miller, 1953). None of those ingredients had information content derived from Darwinian processes. It has been almost 60 years since Miller’s classic experiments, and life has yet to be produced starting from a prebiotic soup. However, considerable progress has been made toward synthesizing life using other methods, and it now appears that the production of life from scratch will be achieved in the near future. Some research efforts along these lines have attempted, as Miller did, to recapitulate the historical origins of life on Earth. Other efforts take inspiration from the origins of life on Earth, but aim for a second origin that would occur under decidedly artificial laboratory conditions. Considerable attention has been directed toward the critical role that RNA is thought to have played in the early history of life on Earth, during an era referred to as the “RNA world” (Atkins et al., 2011). RNA continues to play a central role in contemporary biology, which further motivates the goal of constructing RNA-based life from scratch.

Self-Sustained Darwinian Evolution

An explicit aim of my own research program is to construct a system of RNA molecules that undergo self-sustained Darwinian evolution. In fact this goal was recently achieved, although the system still lacks the complexity and inventiveness of what one might regard as life. The self-sustained evolving system employs populations of RNA enzymes that catalyze the RNA-templated joining of RNA substrates. The enzymes contain ~55 essential nucleotides and can be made to join pairs of RNA substrates of almost any sequence (Rogers and Joyce, 2001). If the substrates, once joined, form additional copies of the enzymes, then self-replication can be achieved. The newly formed enzymes behave similarly, resulting in exponential growth (Paul and Joyce, 2002). However, the process cannot be sustained indefinitely and is informationally restricted by the requirement that the original and newly formed enzymes must have the same sequence.

An improved version of the replication system employs two different RNA enzymes that catalyze each other’s synthesis, enabling their cross-replication and sustained exponential growth (Kim and Joyce, 2004; Lincoln and Joyce, 2009). Each enzyme of the cross-replicating pair contains two substrate-binding domains that recognize corresponding oligonucleotide substrates through Watson-Crick pairing.

During cross-replication, the “Watson” enzyme joins two pieces of RNA to form the “Crick” enzyme, while the “Crick” enzyme joins two pieces of RNA to form the “Watson” enzyme. Information is passed back and forth between these two enzymes in the form of particular sequences within the two substrate-binding domains.

Following optimization of the cross-replication system, it now is possible to achieve 100-fold amplification in just a few hours at a constant temperature and in the absence of any biological materials (Lincoln and Joyce, 2009). The only informational macromolecules in the system are the enzymes and their components, which themselves are subject to Darwinian evolution within the system. The only other components are MgCl₂, a buffer to maintain pH, and H₂O. Evolution can occur because there are many potential variants of the cross-replicating enzymes that must compete for a finite supply of substrates and can undergo mutation through recombination of the two substrate-binding domains.

Beginning with a small seed of the cross-replicating enzymes, amplification occurs with exponential growth, limited only by the amount of substrates that are available. The amplification profile follows the logistic growth equation [enzyme]_t = a / (1 + be^–ct), where a is the maximum extent of amplification, b is the degree of sigmoidicity, and c is the exponential growth rate. This equation also describes population growth for biological organisms constrained by the carrying capacity of their local environment.

Cross-replication of the RNA enzymes can be sustained indefinitely by continuing to supply the necessary substrates. This is most conveniently achieved through a serial transfer procedure, whereby a small aliquot is taken from a spent reaction mixture and transferred to a new reaction vessel that contains a fresh supply of substrates. The new reaction mixture contains only those enzymes that were carried over in the aliquot, and these enzymes immediately resume exponential amplification in the new mixture. Within a period of 24 hours, an overall amplification factor of >10⁹ can be achieved (Lincoln and Joyce, 2009).

Self-sustained exponential amplification provides the growth engine for Darwinian evolution, but it is the diversity of enzymes in the population, their differential reproductive fitness, and their capacity for mutation that provides the opportunity to evolve macromolecular information. Profitable variation in the system emerges through particular combinations of substrates that form cross-replicators with high reproductive fitness. The genetic basis for this variation is represented by the two “loci”—the two substrate-binding domains—that can exist as any of a large number of possible “alleles.” Each allele can be made to encode a different corresponding phenotypic trait, embodied by the functional domain that is physically linked to that allele. If there are n potential variants of the first allele and m potential variants of the second allele, then the combinatorial complexity of the system is n × m.

As an example, a population of cross-replicating enzymes was constructed with 12 different alleles at each of the two loci, providing a combinatorial complexity of 12 × 12 = 144. Each variant allele was linked to a different form of the catalytic center of the enzyme, which resulted in differential reproductive fitness

for various combinations of alleles at the two loci (Lincoln and Joyce, 2009). The evolution process was seeded with 12 different cross-replicating pairs that, due to mutation, could give rise to any of the other 132 combinations. Evolution was allowed to proceed in a self-sustained manner for 100 hours, with an overall amplification factor of 10²⁵. During this time the starting 12 replicators decreased in abundance as novel variants arose and came to dominate the population. Three of these novel variants together accounted for about half of the population members after 100 hours. The basis for their selective advantage was shown to be their relatively fast amplification rate and their propensity to cross-mutate to form additional copies of each other.

Toward Inventive Evolution

The capacity for the invention of novel function within the context of Darwinian evolution depends on both the genetic complexity of the system and the functional richness of the corresponding phenotypes. A population of 144 replicators is represented by only ~7 bits of genetic information. This is much less than the genetic complexity of even the simplest biological systems, which have an information content of 2 bits per base pair of genetic material. In principle, the synthetic RNA-based evolving system could have an information content of 30 bits, considering the 15 total base pairs within the two genetic loci. This would provide a molecular diversity of 4¹⁵ = 10⁹. However, it would not be possible to manage such high diversity because of the vast number of substrate molecules that would need to be present in the reaction mixture. These would slow replication as each enzyme must find its cognate substrates from among the complex mixture.

Current research efforts in our laboratory aim to maximize the genetic complexity of the self-sustained evolving system within the practical limits of both generating and harvesting molecular diversity. We constructed a population of cross-replicating enzymes with 64 different alleles for each of the two genetic loci, providing a combinatorial complexity of 64 × 64 = 4,096. In this test population, each variant allele was linked to the same functional sequence. This was done to assess the extent to which differences in genotype alone would result in differential fitness, something that should be minimized to allow the broadest exploration for novel phenotypes. Starting with this library, 10⁶-fold selective amplification was carried out; then individuals were cloned from the population and sequenced. Indeed two sources of genotype-related bias were identified, and these biases were eliminated from subsequent populations that were constructed.

Although the replicative function is the most important aspect of phenotype, replication can be made contingent on other functions so that fitness reflects the ability to execute those other functions. There is a stem-loop region of the RNA enzyme that supports the structure of the catalytic center and is generic in sequence, so long as it forms a stable secondary structure. This stem loop can be replaced by a ligand-binding (aptamer) domain, configured so that in the absence

of the ligand the domain is unstructured, whereas in the presence of the ligand the domain adopts a folded state that supports the active structure of the enzyme. In this way replication can be made contingent upon recognition of the target ligand (Lam and Joyce, 2009).

As two examples, the supporting stem loop of the enzyme was replaced by an aptamer that recognizes either theophylline or FMN (Lam and Joyce, 2009). In the absence of the ligand there is no replication, but in the presence of the ligand there is sustained exponential growth. Furthermore, the exponential growth rate depends on the concentration of the ligand relative to the K_d (equilibrium binding constant) of the ligand-binding domain. This method for quantitative, ligand-dependent exponential amplification may have applications in biosensing and molecular diagnostics. It is analogous to quantitative PCR for the measurement of nucleic acid targets, but it operates at a constant temperature and can be generalized to non–nucleic acid targets, including proteins, drugs, and metabolites that can be recognized by an aptamer (Lam and Joyce, 2011).

The self-sustained evolution system could, in principle, be used to discover novel aptamers. A genetically encoded random-sequence domain could be placed adjacent to the catalytic center such that ligand recognition would result in selective amplification of the functional molecules. This would require sufficient genetic complexity to encode a population containing enough variants to include molecules with the desired function. Our most recently constructed populations of cross-replicating enzymes have 256 different alleles for each of the two loci, providing a combinatorial complexity of 256 × 256 = 65,536. This still is likely to be insufficient to derive novel aptamers within the context of self-sustained evolution, unless the target ligands are compounds that have a strong propensity to bind to RNA.

The populations with a combinatorial complexity of 256 × 256 were constructed by two different methods. The first involved serial production of all 256 + 256 = 512 allelic variants, synthesizing individual DNA templates to link each genotype-phenotype combination, then transcribing the templates to generate the corresponding RNAs. The advantage of this approach is that each variant is stored in a separate location, allowing one to prepare custom sublibraries. The disadvantage is that, at cost of synthesis of ~$10 per variant, the method cannot be extended to much more complex populations. The second approach for constructing the populations involved a novel split-and-pool method that makes it possible to synthesize the entire population in parallel. The parallel method enables the construction of populations with as many as 10⁹ different members, although, as discussed above, it would be difficult to manage such high-diversity populations throughout the course of a self-sustained evolution experiment.

With increasing population complexity it becomes important to consider the nature of the code that relates particular genetic sequences to their corresponding functional sequences. This code can be chosen arbitrarily, but evolutionary optimization likely will benefit if more closely related genetic sequences correspond to more closely related functional sequences. The code need not have a collinear

3:1 relationship, as is the case for the genetic code in biology which relates trinucleotide codons within mRNA to individual amino acids within proteins. Furthermore, the same code need not apply for all genotype positions, although this simplification clearly has great selective advantage for natural biology. In synthetic biology, the experimenter must decide what genetic code to employ based on theoretical and practical considerations.

In preparing the populations of 256 × 256 combinatorial complexity, two different “sparse” codes were implemented, whereby each nucleotide within the genotype region encodes one or more nucleotides within the phenotype region. For the serially constructed population, each of four genetic nucleotides encodes three noncontiguous nucleotides in the functional region, with a different codon relationship for each genetic position. For the parallel library, each of four genetic nucleotides encodes either one or two contiguous nucleotides in the functional region, again with a different codon relationship for each genetic position. Experimental studies are under way to assess the operational characteristics of these different codes.

When practicing synthetic biology from scratch the experimenter makes the rules and allows Darwinian evolution to play the game. The capacity of the system to invent novel function is the most important measure of its robustness. Inventiveness should be as broad as possible so that the system can adapt to unanticipated changes in its environment. Life on Earth, although vulnerable to extreme changes of environmental conditions, has demonstrated extraordinary resiliency and inventiveness in adapting to highly disparate niches. Perhaps the most significant invention of life is a genetic system that has an extensible capacity for inventiveness, something that likely will not be achieved soon for synthetic biological systems. However, once informational macromolecules are given the opportunity to inherit profitable variation through self-sustained Darwinian evolution, they just may take on a life of their own.

Acknowledgments

Jeff Rogers developed the RNA enzyme that provides the basis for replication, Natasha Paul provided the first demonstration of a self-replicating RNA enzyme, Dong-Eun Kim converted the system to a cross-replication format, Tracey Lincoln first achieved self-sustained Darwinian evolution of RNA enzymes, Bianca Lam made replication contingent upon recognition of a target ligand, and Michael Robertson and Jonathan Sczepanski constructed increasingly complex populations of the cross-replicating RNA enzymes. This work was supported by grants from the National Aeronautics and Space Administration (NNX10AQ91G), the National Institutes of Health (GM065130), the National Science Foundation (MCB-0948161), and the Defense Advanced Research Projects Agency (DSO BAA-09-63).

References

Atkins, J. F., R. F. Gesteland, and T. R. Cech (eds). 2011. RNA Worlds: From Life’s Origins to Diversity in Gene Regulation. Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press.

Darwin, C. R. 1859. On the Origin of Species by Means of Natural Selection, or the Preservation of Favoured Races in the Struggle for Life. London: John Murray. P. 61.

Hoover, R. B. 2011. Fossils of cyanobacteria in CI1 carbonaceous meteorites. Journal of Cosmology 13, published online March 2011.

Joyce, G. F. 1994. Foreword. In Origins of Life: The Central Concepts, edited by D. W. Deamer and G. R. Fleischacker. Boston: Jones and Bartlett. Pp. xi-xii.

Kim, D. E., and G. F. Joyce. 2004. Cross-catalytic replication of an RNA ligase ribozyme. Chemistry and Biology 11:1505-1512.

Lincoln, T. A., and G. F. Joyce. 2009. Self-sustained replication of an RNA enzyme. Science 323: 1229-1232.

Lam, B. J., and G. F. Joyce. 2009. Autocatalytic aptazymes enable ligand-dependent exponential amplification of RNA. Nature Biotechnology 27:288-292.

——. 2011. An isothermal system that couples ligand-dependent catalysis to ligand-independent exponential amplification. Journal of the American Chemical Society 133:3191-3197.

Miller, S. L. 1953. A production of amino acids under possible primitive Earth conditions. Science 117:528-529.

Paul, N., and G. F. Joyce. 2002. A self-replicating ligase ribozyme. Proceedings of the National Academy of Sciences of the United States of America 99:12733-12740.

Rogers, J., and G. F. Joyce. 2001. The effect of cytidine on the structure and function of an RNA ligase ribozyme. RNA 7:395-404.

A10

MANUFACTURING MOLECULES THROUGH METABOLIC ENGINEERING⁵⁵

Jay D. Keasling^56,57,58