Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
5 Bioconfinement of Viruses, Bacteria, and Other Microbes INTRODUCTION The use of genetically engineered microbes can offer enormous poten- tial benefits. Because viruses, bacteria, and fungi are natural pathogens of insects and other pests, microbes can be harnessed and genetically enhanced as agents of biocontrol. Bacteria and fungi also are able to degrade some environmental pollutants, and molecular technology allows us to expand the list to include other toxic compounds as well. There is a long human history of using microbes in agriculture, food processing, waste treatment, and other beneficial capacities. Modern molecular methods allow us to broaden the range of useful applications, and all of the evidence indicates that the methods used to generate genetically engineered organisms (GEOs) are not intrinsically dangerous. Some caution is warranted, however, because information about the ecology and evolution of transgenic microbes in the wild is limited. Microbes occur in extremely large populations with short generation times, so they adapt quickly to adverse conditions. Their envi- ronments change constantly, resulting in unpredictable and variable selec- tion pressures. Bacteria also can transfer DNA into unrelated microbes, and the long-term ecological consequences of that transfer are unclear (Bushman, 2002). The consequences of releasing transgenic microbes into the environ- ment have not been evaluated adequately. As with the plants and animals discussed in the earlier chapters, it is impossible to generally predict the fitness consequences of genetically engineering a microbe. The new genetic combination could be beneficial or deleterious for a microbe's survival in the wild, depending on the ecological context (e.g., through interactions 159
160 BIOCONFINEMENT OF GENETICALLY ENGINEERED ORGANISMS with the environment and with other microbes). Most studies show that genetically engineered microbes are relatively less fit than are their nonengineered counterparts, although that can be a faulty assumption. Each case of genetic engineering must be considered on its own. This chapter is on the bioconfinement of genetically engineered microbes, especially bacteria, fungi, and viruses. The potential effects and need for bioconfinement in microbes are discussed first. Then there is a section that identifies and describes the major methods of bioconfinement for bacteria, fungi, and viruses and discusses the strengths and weaknesses of each method. Next are considered the effectiveness of those methods in different spatial and temporal scales and their potential effects on biological popula- tions and ecosystems. The needs, feasibilities, and realities of monitoring, detecting and culling genetically engineered bacteria, fungi, and viruses are discussed. Finally, the aforementioned topics are related to the bioconfine- ment of microalgae. Because the committee was not specifically asked to evaluate bio- confinement techniques for microbes, this chapter is purposefully less sub- stantive than are the chapters on plants and animals. However, the wide- spread appreciation for the usefulness of transgenic microbes warrants their treatment here. Earlier NAS reports also have dealt with genetically engi- neered microbes (NRC, 1989a, 2002b). The 1989 NAS report, Field Testing Genetically Modified Organisms, extensively evaluated environmental risks associated with the release of transgenic microorganisms. The recommendations in that publication in- fluence policy decisions today, despite the advancement of molecular tech- nology in the intervening years. Although portions of the present report discuss transgenic microbes, the subject cannot be dealt with in detail be- cause microbes were not a central focus of the committee's charge. The committee suggests that genetically engineered microbes be reconsidered on their own. POTENTIAL EFFECTS OR CONCERNS, AND NEED FOR BIOCONFINEMENT IN VIRUSES, FUNGI, AND BACTERIA The three potential areas of concern that attend the release of geneti- cally engineered bacteria, fungi, and viruses are similar to those for any other class of GEO: invasion, displacement, and transfer. Together they can be used to argue that bioconfinement measures should be considered.
VIRUSES, BACTERIA, AND OTHER MICROBES 161 Invasion into Indigenous Populations Viruses Genetically engineered viruses could infect and harm nontarget hosts. Because viruses are obligate intracellular parasites, they require metaboliz- ing (living) cells to replicate their genomes and make progeny. The reliance on host cells often produces strong selection for viruses to evolve more efficient mechanisms to exploit their hosts. In turn, selection of the host favors genotypes that excel in their ability to repel virus attack. Viruses can gain the upper hand in these coevolutionary battles simply because they can evolve more rapidly than do their hosts (Levin and Lenski, 1983). Thus, a virus can be successful by evolving a greater propensity to exploit the host, through creating more progeny per infection per unit time. Alternatively, viruses could be evolutionarily successful by adapting to infect a greater variety of hosts (DeFilippis and Villarreal, 2000). The latter adaptation exemplifies the potential consequence of releasing genetically engineered viruses into an ecological community of naïve (inexperienced) hosts. The concern is not the introduction of engineered alleles (such as transgenes) per se, but that the foreign strain of virus will harm a nontarget host species that is ill-prepared to defend against the viral attack because of its lack of resistance genes. Support for this idea comes from studies that demonstrate elevated virulence in naïve host populations (Bull, 1994; Taylor et al., 2001). Thus, releasing genetically engineered viruses into the environment might result in their becoming successfully established in nontarget hosts. Fungi The host range of fungi also can evolve. Reports of changes in the known host range of plant pathogenic fungi are common (Mundt, 1995). For example, the scabrum rust, which is pathogenic on Agropyron scabrum and Hordeum vulgare (barley) in Australia, arose from a cross between Puccinia graminis f. sp. tritici (wheat stem rust) and P. graminis f. sp. secalis (rye stem rust) (Burdon et al., 1981). In 1991 cultivated barley was found to be heavily infected with a new variety of P. coronata, a fungus that was known for more than 200 years to cause serious disease in cultivated oat. From these and other examples, it appears that the pathogen can genetically alter its host range. However, studies on the rice blast fungus showed that many genes are required for the fungus to attain the high fitness required for field survival on a new host (Valent et al., 1991). Those results indicate that survival in the wild in the face of competition from other organisms and changing environmental conditions can be far more demanding than surviving in a laboratory under optimal conditions. Field
162 BIOCONFINEMENT OF GENETICALLY ENGINEERED ORGANISMS studies have been conducted on a transgenic mycoinsecticide to monitor the fate of the fungi under field conditions (Hu and St. Leger, 2002). The fungus was released onto a plot of cabbage and survivorship was deter- mined in nonrhizosphere and rhizosphere soils. In nonrhizosphere soils, the fungal propagules decreased from 105 to 103 per gram after several months. However, recombinant fungi engineered only with the gene for green fluores- cent protein (GFP) remained at 105 propagules in the rhizosphere soil. Fungi that contained an additional protease gene did not persist as well in the rhizosphere as did the GFP genotypes. The observations are consistent with what has been observed for transgenic bacteria--that adding genes to cells often decreases the microbes' fitness, but that some ecological contexts (such as placement in the rhizosphere) can promote their survival. Bacteria In theory, genetically engineered bacteria introduced into the environ- ment can become established in a microbial community. A limited number of studies to assess this possibility have been done on bacteria in aquatic and terrestrial environments. In one study (Scanferlato et al., 1989) the viability of genetically engineered and wild-type strains of Erwinia carotovora were compared after their addition to an aquatic microcosm. Both declined in viability at the same rate, and within 32 days neither was detectable by viable counts. Those data suggest that the newly introduced bacteria, whether genetically engineered or wild-type, were poorly adapted to the new environment and therefore were unable to compete with indig- enous species. The observations are consistent with theory and with labora- tory experiments on resource competition. Both suggest that the competitor that grows at the lowest concentration of a limiting resource will survive and thereby displace all inferior competitors (Hansen and Hubbell, 1980; Tilman, 1982). In nature, most nutrients are present at low concentrations in terrestrial and aquatic environments (Madigan et al., 2003); therefore, only those microorganisms that can compete for those limited resources would be expected to thrive. To overcome the likelihood of introduced microbes, being poor competitors, the genetic capabilities to perform a particular function are commonly introduced into bacteria that already are adapted to a particular habitat (Glandorf et al., 2001). For example, bacteria are being used for bioremediation of oil-contaminated beaches and polluted soils. The common practice is to apply nutrients, such as nitrate and phosphate, to the contaminated area to promote growth of indigenous bacterial popula- tions, which likely will include microbes that can metabolize the pollutants. As a rule, introducing genes into indigenous bacteria to perform a specific function is preferable to introducing exotic bacteria that contain
VIRUSES, BACTERIA, AND OTHER MICROBES 163 those genes. Although bacteria initially could be maladapted to a new environment, they often can change and increase their growth rate signifi- cantly. In one study, transgenic bacteria were incubated in lake water for 15 days and then reisolated. The growth rates of the reisolates were more than 50% higher in the lake water when compared to the original strain (Sobecky et al., 1992). The researchers concluded that bacteria can adapt to oligotrophic environments and the fitness of GEOs for survival can increase in aquatic ecosystems. The ability of bacteria to adapt to new environments over time is an important concern in their release to the environment. That adaptability also applies to non-GEOs, though introduction of non- transgenic exotic bacteria into new environments has not raised strong concerns. In another study, the viability of genetically engineered Pseudomonas putida that contained a gene for the synthesis of the fungal inhibitor phenazine was compared with its wild-type parent in the rhizosphere of wheat plants for two growing seasons (Bakker et al., 2002). In both seasons, the genetically engineered and the wild-type strains decreased to below detectability within a month after the wheat harvest, indicating that the rhizosphere was essential to the survival of the introduced bacteria. In one season, within days of sowing the genetically engineered strain decreased more rapidly than did the wild-type strain. In another growing season, however, no difference in density was observed for the two strains, indicat- ing that the additional metabolic load on the GEO did not reduce its ecological fitness. Those results over successive years suggest that ecological fitness, at least in soil, depends on the variable environmental conditions encountered by the GEOs. Considerable data suggest that the increased genetic load that results from introducing additional genes into a microbe usually reduces its growth rate unless a strong selection pressure favors the added genes (Lenski and Nguyen, 1988; Milks et al., 2001; Zund and Lebek, 1980). The observed decrease in growth rate apparently can result from the additional products synthesized from the DNA rather than from the replication of the DNA (Lenski and Nguyen, 1988). However, several research groups have reported that a genetically engineered bacterium can grow at the same rate as or even faster than its parent does (Bouma and Lenski, 1988; Devanas and Stotzky, 1986; Edlin et al., 1984; Hartl et al., 1983; Marshall et al., 1988). Those latter observations are not well understood, and it is unclear whether the organisms as grown in the laboratory would be ecologically fit in a natural environment. Also, cases have been reported in which genetically engi- neered bacteria coexist with indigenous populations (Kargatova et al., 2001). If the introduced bacteria were resistant to an antibiotic present in the environment, in theory it should thrive because of reduced competition from susceptible strains. Thus, the assumption that all genetically engi-
164 BIOCONFINEMENT OF GENETICALLY ENGINEERED ORGANISMS neered bacteria are unfit in natural environments cannot be sustained. How- ever, the same would be true for exotic non-GEOs. Displacement of Indigenous Populations Viruses It is theoretically possible that genetically engineered viruses could dis- place resident species. In theory, the coevolutionary battle between viruses and their hosts leads to a never-ending arms race; hosts evolve resistance, viruses evolve counterresistance, and the cycle repeats with both species constantly running to remain in place (this is the "Red Queen hypothesis"; Clarke et al., 1994). Whereas the hosts and viruses must continuously evolve to maintain same dual-species interaction, at least this scenario pro- duces long-term stability of biodiversity in the ecosystem. In contrast, a newly introduced virus that can prove more virulent in nontarget host tips the balance in its favor (Bull, 1994; Taylor et al., 2001). Subsequent devas- tation of the nontarget host is the primary concern, but a separate concern is that the resident virus could lose out because it is less efficient (relatively less virulent). Extinction of the endemic virus could disrupt the ecological community; for instance, introduction of the relatively more virulent species could force the nontarget host to a lower equilibrium density in the commu- nity, producing a cascade effect elsewhere in a food web. Viruses might be underappreciated in terms of their influence on regulating large-scale eco- system processes (Fuhrman, 1999). Bacteria Several studies have reported on the effect of adding genetically engi- neered or wild-type bacteria to resident flora in aquatic and terrestrial environments. In one study (Scanferlato et al., 1989), genetically engineered strains of E. carotovora were added to an aquatic microcosm, and the effects were measured in some elements of the indigenous population. Thirty-two days after inoculation the number of total and proteolytic bacteria was the same in the inoculated and uninoculated microcosms. Neither did the inoculation affect the number of amylolytic and pectolytic bacteria in the water or sediment. In another study, genetically engineered Pseudomonas fluorescens were released into indigenous populations near wheat plants (De Leij et al., 1995). The results for culturable organisms can be summarized as follows: P. fluorescens and the unmodified strains produced the same results, the perturbations to the microbial population were small, and the release of bacteria had no obvious effect on either plant growth or health.
VIRUSES, BACTERIA, AND OTHER MICROBES 165 Genetically engineered P. putida that contained a gene for the synthesis of the fungal inhibitor phenazine and its wild-type parent were added to the rhizosphere of wheat plants (Bakker et al., 2002). Neither the transgenic strain nor its parent affected the metabolic activity of the soil microbial population, and only transient changes were observed in the composition of the rhizosphere fungal microflora. Although the GEO had the greater effect, the authors suggested that the effect of the introduced GEOs was only minor in comparison to those that result from such common agricultural practices as plowing or crop rotation. Most studies that have assessed the influence of genetically engineered microbes on microbial populations have studied effects on culturable organ- isms alone. Yet less than 1% can be grown in culture (Madigan et al., 2003). However, polymerase chain reaction technology and measurement of ribosomal DNA (rDNA) patterns make it possible to analyze entire bacterial populations. Using those techniques, Robleto and colleagues (1998) showed that introduction of engineered strains of Rhizobium syn- thesizing a narrow-spectrum-peptide antibiotic reduced the diversity of - Proteobacteria; while the total bacterial population was not substantially affected. In another study P. putida, genetically engineered for increased activity against soilborne bacterial and fungal pathogens, were released into the rhizosphere of wheat and their effect on indigenous microflora was determined (Bakker et al., 2002). Effects of the genetically engineered bac- teria on the rhizosphere fungi and bacteria were analyzed, using amplified ribosomal DNA restriction analysis. A transient change in the composition of the rhizosphere was noted, but several soil microbial activities, such as soil nitrification and cellulose decomposition, were unaffected. The limited data from all of these experiments indicate that the introduction of geneti- cally engineered microorganisms has mostly transitory effects on indig- enous populations that are unlikely to be significant in the field. The effects of adding transgenic microbes are not likely to be any greater than are those that attend the addition of nontransgenic species. To reduce the possibility of changes in the indigenous microbial popula- tion that result from the release of genetically engineered microbes, the committee advises that strains be used that are likely to be poor competitors in the local environment. The limited available data suggest that the intro- duction of genetically engineered microbes into the environment is unlikely to have significant long-lasting effects on microbial communities. Horizontal Genetic Transfer into Local Populations A third concern of introducing genetically engineered microbes is the potential consequence of horizontal transfer of engineered genes from introduced microbes into local populations.
166 BIOCONFINEMENT OF GENETICALLY ENGINEERED ORGANISMS Viruses Viruses have two distinct mechanisms for the exchange of genetic material. When two or more DNA or RNA viruses infect the same host cell, recombination can lead to hybrid progeny that contain genetic information from both parents (Hershey and Rotman, 1949). In contrast, some RNA viruses have genomes that are split into several smaller segments, and co- infection can produce hybrids that contain a random reassortment of the segments found in the infecting parent viruses. Such exchanges have peri- odically led to new strains of influenza virus that have caused human pandemics (Palese, 1984; Webby and Webster, 2003). Recombination in viruses, can promote linkage equilibrium (free asso- ciation of alleles) to create the potential for engineered alleles to enter and circulate within a local gene pool. Laboratory experiments show that gene exchange can profoundly affect virus evolution (Rambaut et al., 2004; Turner, 2003; Turner and Chao, 1998), and it is generally accepted that viral recombination in natural infections is a major force in the evolution of new viruses (e.g., Goldbach, 1986). Many viruses also can recombine with host chromosomes, thus introducing virus-derived genes into the host genome and the host gene pool. The fitness effects of engineered genes in the origi- nal virus background are likely to be assessed before strains are released into target populations of hosts. The concern is that those genes could have unanticipated effects when they transfer horizontally from the engineered background into new ones. Epistasis (gene interaction) between introduced alleles and those in the gene pools of other species can hamper the fitness of individuals in those groups. Hammond and colleagues (1999) studied transfer of engineered genes through virus recombination and considered the likelihood that the process would harm natural populations. The greatest concern identified by those authors was the creation of new virus types as a result of recombination between wild-type viruses and unrelated transgenes in genetically engi- neered plants. However, the rate at which this happens is unlikely to exceed that in naturally occurring mixed infections of viruses of nonengineered plants. More data are needed from field trials to evaluate the benefits and risks associated with release of transgenic viruses. The most extensive survey to date (Thomas et al., 1998) studied interactions between transgenes derived from potato leafroll virus and viruses to which transgenic plants were exposed. The experiments revealed no evidence of recombination, altered transmission, or altered virus properties, suggesting that such phe- nomena are extremely rare.
VIRUSES, BACTERIA, AND OTHER MICROBES 167 Bacteria Most species of bacteria have several mechanisms for horizontal gene transfer: DNA-mediated transformation, in which "naked" DNA is trans- ferred to recipient cells; generalized or specialized transduction, in which donor DNA is enclosed in the coat of a bacteriophage; and conjugation, in which DNA--primarily through plasmids--is transferred from donor to recipient cells after contact between the two. Bacterial gene exchange can affect the persistence of a strain or its engineered alleles, and a genetically engineered bacterium can be the donor or the recipient of genetic informa- tion by horizontal gene transfer. Only a new genetic combination with higher fitness than the indigenous genotype under the specific environ- mental conditions has a high likelihood of persisting (Koonin et al., 2001). In the laboratory, the transfer of genetic material is most efficient between members of the same species. However, transfer in nature has been observed between widely different genera, and even domains, of bacteria. Comparative analysis of bacterial, archeal, and eukaryotic genomes sug- gests that a significant fraction of the genes in prokaryotic genomes have been involved in horizontal transfer over evolutionary time (Koonin et al., 2001). At least one bacterium, Agrobacterium, can transfer DNA into plants, and it is the workhorse of plant genetic engineering (Chilton et al., 1977). The broad-host-range plasmid RSF1010, when in Agrobacterium, can mediate its own transfer into plants as well as into other Gram-negative bacteria (Buchanan-Wollaston et al., 1987). Thus, plants have ready access to the gene pool of Gram-negative bacteria, thereby expanding the possi- bilities of horizontal gene transfer from prokaryotes to eukaryotes. Many varieties of tobacco (Nicotiana tabacum) contain genes transferred from Agrobacterium over evolutionary time (Furner et al., 1986). In the labora- tory, Agrobacterium can transfer DNA into a variety of fungi (Bundock et al., 1995; de Groot et al., 1998; Piers et al., 1996) and because Agrobacterium and many fungi occupy the same habitat in soil, transfer between the bac- terium and fungi might also occur in nature. Escherichia coli (E. coli) can transfer plasmid DNA into yeast in the laboratory (Heinemann and Sprague, 1989). It also has been reported that Agrobacterium can transfer DNA into mammalian cells (Kunik et al., 2001) and that E. coli can transfer plasmids into mammalian cells (Waters, 2001). Several studies have examined the possibility of gene transfer from transgenic plants to bacteria (Gebhard and Smalla, 1999; Schlüter et al., 1995). The conclusion has been that such an occurrence would be extremely rare, although plant DNA can persist in the soil under field conditions for up to 2 years. Nielsen and colleagues (1998) emphasized that, although gene transfer can be a rare event, it is critical to understand the selective forces that act on the outcome of any transfer.
168 BIOCONFINEMENT OF GENETICALLY ENGINEERED ORGANISMS Plasmids are a common vector for cloning and moving transgenes from one organism to another, and conjugation in particular would allow these genes to move easily into other bacteria in the environment. Some conjuga- tive plasmids are highly promiscuous in their ability to transfer horizontally between unrelated bacteria, and that has contributed to widespread dis- semination of antibiotic-resistance genes, which often reside on plasmids. In contrast, other plasmids are nonconjugative and thus unable to be trans- ferred. However, they can be transferred if another plasmid in the same cell provides the missing functions. The probability of horizontal transfer can be reduced by cloning genes into nonconjugative plasmids and by using bacteria that contain no other plasmids. The risk of transfer can be further minimized if the engineered genes are integrated into the bacterial chromo- some rather than remaining on a plasmid. Further, the enzyme transposase-- required for gene movement inside a cell, and frequently used in construct- ing transgenic bacteria--should be disabled. All introduced plasmids should be defective in conjugation functions. If any antibiotic resistance loci are used to mark strains, the resistance loci should not involve antibiotics that currently are in clinical use. A sensible choice should be made for the bacterial strain introduced into the environment to carry out a specific function. For example, it would be unwise to introduce a close relative of a disease-causing bacterium because nonpathogenic relatives conceivably could differ only by the presence of a genetic element (plasmid, transposon, prophage) on which virulence genes reside. Thus, in theory, inadvertent acquisition of one or more func- tions might convert the introduced strain into a dangerous pathogen of humans, animals, or plants. This is because bacteria, unlike viruses, are not obligate parasites. That is, bacterial pathogenicity depends on an array of characteristics that relatively few bacteria have acquired through extended coevolution with a particular host (Salyers and Whitt, 2002). Therefore, it is unlikely that minor genetic modifications would convert a nonpathogenic strain into a pathogen, and laboratory experiments to achieve it--at least with E. coli--have thus far been unsuccessful (S. Moseley, University of Washington, personal communication, 2003). Fungi Evidence is weaker for horizontal gene transfer between fungi than it is between bacteria (Rosewich and Kistler, 2000). In one case, however, sequence analysis of a gene for chymotrypsin synthesis in a fungus in which chymotrypsins had never been observed revealed that the sequence is related to that of a soil bacterium (Screen and St. Leger, 2000). As more sequences become available, it should become clearer whether horizontal gene trans- fer can occur.
VIRUSES, BACTERIA, AND OTHER MICROBES 169 To reduce the possibility of horizontal transfer of genetically engi- neered alleles, it is advisable to use microbial strains in which transgenes are integrated into chromosomes. The enzyme transposase, required for gene movement inside a cell and frequently used in constructing transgenic bac- teria, should be disabled. All introduced plasmids should be defective in conjugation functions. If any antibiotic resistance loci are used to mark strains, the resistance loci should not involve antibiotics that are currently in clinical use. BIOCONFINEMENT OF BACTERIA, VIRUSES, AND FUNGI Because they are small, easily dispersed, and numerous, genetically engineered microbes will require bioconfinement approaches that are dif- ferent from those for other GEOs. Control centers on fitness reduction. Fitness Reduction Phenotypic Handicapping One potential consequence of releasing transgenic microbes to the envi- ronment is that they could perpetuate by invading or displacing natural populations in competition for resources. The limited experimental data suggest that, in general, genetically engineered bacteria and viruses will be competitively less fit than their wild-type counterparts because of burdens associated with carrying and expressing additional functions coded by transgenes. As noted already, microbes generally fare poorly when intro- duced into a new environment, although numerous cases could be cited in which the genetically engineered microbe did not appear to be significantly handicapped compared with the parental strain (Bouma and Lenski, 1988; Devanas and Stotzky, 1986; Hartl et al., 1983; Marshall et al., 1988). One solution to the uncertainty of whether transgenic microbes or their genes will persist is to use strains with phenotypic handicaps, such as reduced survival capability, reduced reproductive capacity, low resistance to a pre- dictable change in the environment (such as seasonal heat or cold), or a tendency to lose the specific function of concern (NRC, 1989a). Viruses Phenotypic handicapping is widely applied in the design of live viruses for use as vaccines (Murphy and Chanock, 2001). They do not cause dis- ease but they stimulate the host's immune system to produce antibodies against wild-type viruses to fight subsequent infection. An ideal live vaccine is an attenuated (weakened) form of the virus that is phenotypically handi-
170 BIOCONFINEMENT OF GENETICALLY ENGINEERED ORGANISMS capped, thus ensuring that its competitive inferiority will prevent it from persisting in nature. That is, phenotypic handicapping hampers the vaccine or its engineered alleles from influencing the evolution of wild populations through entry of those elements into the natural gene pool. Similarly, attenuated viruses can be engineered to carry proteins of unrelated virus pathogens, and those chimeric vectors elicit immune responses without posing the threat of long-term persistence (Rose et al., 2001). Recent studies have attempted to engineer herpes simplex virus-1 and other well-described viruses for therapeutic interventions, such as combat- ing cancer through gene therapy (e.g., Advani et al., 2002). Overall, few data exist regarding the long-term persistence of genetically engineered viruses. The effectiveness of phenotypic handicapping of viruses as a con- finement measure is not clear. Bacteria Phenotypically handicapped live bacteria also have been used in induc- ing cellular immune response (Stocker, 1990). When a gene that codes for a particular epitope (a short, linear peptide sequence that is a portion of a larger protein antigen) was inserted into a gene that codes for flagellin in Salmonella auxotrophic for aromatic acids, a cellular immune response to the epitope was generated (Verma et al., 1995). The bacteria did not multi- ply because of a lack of required compounds in the environment. Pheno- typic handicapping of bacteria as a confinement measure already has been alluded to: One form involves the rapid decline of nonindigenous microbial strains (including genetically altered ones) after they are introduced into soil or aquatic environments (e.g., Glandorf et al., 2001; Scanferlato et al., 1989). The data support the widely accepted view that long-established microbial communities are able to resist invasion by foreign organisms (Liang et al., 1982). However, one challenge to phenotypic handicapping is the evidence that genetically engineered strains can persist for long periods by quickly adapting to a local environment. For instance, Kargatova and colleagues (2001) observed that recombinant E. coli strains can persist for one year or more in aquatic microcosms, and that they can coexist with indigenous microflora. The strains adapted by decreasing their expression of cloned genes, suggesting that the genetically engineered bacteria tended to lose the genes of concern. Similarly, addition of plasmids does not necessarily lead to a long-lived handicap to the bacterial host. Bacteria can adapt through the mutation of genes that are not associated with the plasmid and thereby restore their growth rate to that of the original parental strain (e.g., Bouma and Lenski, 1988; Hartl et al., 1983).
VIRUSES, BACTERIA, AND OTHER MICROBES 171 Most experiments on phenotypic handicapping have been performed in the laboratory. But in natural environments such as lake water, the fitness consequences of added genetic material are more difficult to evaluate and apparently depend on many factors associated with the genetics of the bacteria and the environment, with its usually obscure and variable selec- tion pressures. For instance, in a study involving prototrophic strains of P. putida, one with and the other without a plasmid, the plasmid-bearing strain was maintained in a lake system over a period of 2 months (Sobecky et al., 1992). The plasmid was lost within 24 hours if the strains were amino acid auxotrophs. Variations in weather, such as rainfall and tem- perature, can affect the population density of transgenic microbes intro- duced into the soil (Glandorff et al., 2001). In addition, it is possible that an altered microbe can become immediately more fit than the wild-type if the phenotypic change increases resistance to a noxious substance in the envi- ronment or increases the ability of the microorganism to metabolize a substrate in the environment. Expression of additional functions and their effects on fitness reduction have not been clearly defined, and a greater effort to examine this phenomenon in field tests is warranted (e.g., Palmer et al., 1997). A second identifiable complication of phenotypic handicapping is that indigenous bacteria generally do not exist as free-living, individual cells. Rather, in their natural environment, bacteria often form highly structured clumps, called biofilms, with properties that are quite different from those of bacteria growing in the laboratory. For this reason, phenotypic handi- capping of transgenic bacteria growing in liquid medium in the laboratory might not be relevant to performance in a natural setting. In particular, bacteria in biofilms are far more resistant to noxious chemicals, including antibiotics and heavy metals, than are individual cells (Madigan et al., 2003). Biofilms attach to inanimate objects in the environment and their formation requires the action of several genes. It could be undesirable for genetically engineered bacteria to persist long term, but they must live long enough to perform the intended function. Unless those bacteria form biofilms by attaching to such objects as rocks, soil particles, and teeth, they could be washed away by rain or other fluids. Several genes have been identified in E. coli and other microorganisms that are necessary for biofilm formation, so mutations in those genes could debilitate the organisms to the extent that they would not persist even for short periods with the indig- enous flora (O'Toole et al., 2000; Pratt and Kolter, 1998). A third difficulty of phenotypic handicapping is that the GEO could be so handicapped that it is not practical to use. Perhaps the best example is strain 1776 (see Box 5-1).
172 BIOCONFINEMENT OF GENETICALLY ENGINEERED ORGANISMS BOX 5-1 1776 The most extreme case of phenotypically handicapping a microbe was carried out in the laboratory of Roy Curtiss after the 1975 International Conference on Recombinant DNA Molecules. The idea was to disable the K-12 strain of E. coli, considered safest for use in cloning experiments, and make it even safer, against the possibility of its escape from the laboratory. The thinking that went into the disabling process serves as a guideline for genetically handicapping other organisms. Curtiss and his colleagues (1977) introduced mutations that precluded coloni- zation of and survival in the intestinal tract, prevented biosynthesis of the rigid layer of the bacterial cell wall in nonlaboratory conditions, led to the degradation of DNA in any organisms that escaped the laboratory environment, permitted moni- toring of strains, and inactivated DNA repair mechanisms. The key mutations were deletions or independent mutations. Thus, the traits were stable and unlikely to revert. The strain had a generation time twice to four times longer than the wild- type E. coli K-12 strain and likely would not compete well with healthy microbes in the environment. The strain also was resistant to most known E. coli-transducing phages and defective in inheriting many conjugative plasmids. Thus, the strain could not transmit genetic information by transduction or conjugation at detectable frequencies. To celebrate the nation's bicentennial, the strain was named 1776. It was used for the industrial production of insulin after the cloning of the eukaryotic insulin gene into the nonconjugative plasmid pSC101. However, the strain, with its multi- ple auxotrophic markers, sensitivity to detergents, and increased generation time, proved so difficult to grow that widespread use was clearly impractical. Accordingly, as studies with recombinant DNA became more routine and the guidelines for biocontainment were relaxed, wild-type strains of E. coli K-12 or HB101 were used as the hosts for DNA cloning. 1776 is now just a memory of a bygone era. Fungi Two methods have been proposed to phenotypically handicap fungi. One is to isolate auxotrophic mutants that can exist on the pest host, in the case of a biocontrol agent, but that would not survive outside the host. Such mutants should be isolated using a physical mutagen, such as gamma or neutron radiation, that fragments genes and thereby prevents reversion. Another proposed technique is to render the fungi asporogenic (unable to produce spores), thereby helping not only to prevent their spread but also inhibiting the formation of dormant resting structures that resist heat, cold, desiccation, and other harsh environmental conditions (Gressel, 2001). Asporogenic mutants would be handicapped both in persistence and in the
VIRUSES, BACTERIA, AND OTHER MICROBES 173 major structures dispersed by wind, water, or animals. Spores of some fungi, however, are required for pathogenesis. Thus, if the fungus is to be used in biocontrol, an asporogenic mutant would not be suitable, and this form of handicapping would be inappropriate. Suicide Genes Suicide genes can be used to confine bacteria and fungi under two circumstances. The first ensures that bacteria growing in a closed container (such as a vat) are unable to survive if they escape. The strain should die as quickly as possible after escape. In many situations this is best achieved by chemical sterilants. The second circumstance is to combat a perceived threat to the environment should released microorganisms persist beyond the intended period of usefulness, for example, for bioremediation of Superfund sites or for biocontrol of plant pests in agriculture. In that case, the GEO must be able to carry out its function before it expires. In either scenario, the microorganism would carry a suicide gene that is repressed when the microbe is at work and becomes active immediately thereafter (Curtiss, 1988; Molin et al., 1987). Bacteria The key to designing an effective suicide containment system rests on regulating gene expression from one of a variety of controllable promoters. They can be divided into two categories: those that function when a trigger is present and those that function until repressed (see Molin et al., 1993 for review). Systems that have been devised in the first category include the PL promoter of phage lambda and a thermosensitive lambda repressor (Ahrenholz et al., 1994) and the lac promoter from E. coli. The PL promoter is induced by raising the temperature to inactivate the lambda repressor; the lac pro- moter is activated by the chemical isopropyl--D-thiogalactopyranoside (IPTG) (Bej et al., 1988; Knudsen and Karlstrom, 1991; Knudsen et al., 1995). Although such systems work in the laboratory under controlled conditions, it is unrealistic to apply heat to fields or to irrigate fields with a chemical inducer such as IPTG. As a solution, Molin and colleagues (1993) suggest manipulating the regulated system such that growth of the cells in the laboratory leads to the synthesis of a compound that is toxic to the microbe. When cells are introduced into the environment, several genera- tions of growth would be needed before the repressor would be diluted and the toxin synthesized. Because generation times in the wild can be just days or weeks long, the engineered cells should survive only long enough to achieve the goal. The approach is speculative, but it seems promising. For bioremediation, a possible approach involves repressing transcription from
174 BIOCONFINEMENT OF GENETICALLY ENGINEERED ORGANISMS a promoter that functions only in the presence of the target substrate. Once the substrate is exhausted, transcription from the promoter ensures that suicide is induced (Contreras et al., 1991). The systems that function when the activator is absent include the trp promoter from E. coli, which represses the synthesis of a toxic compound when tryptophan is present. If the microorganisms escape, their new envi- ronment would likely contain insufficient supplies of this amino acid, result- ing in transcription of the toxin gene from the active trp promoter and synthesis of the toxin. Two goals are paramount in the design of suicide systems. First, the gene product should extend beyond mere growth inhibition; the best candi- dates are killing functions whose targets are likely to be found in essentially all bacteria. Second, the toxicity of the gene product should be high, so that a high efficiency of killing is achieved at a low concentration. Therefore, putative killing functions should be assayed in various bacteria at several ranges of induction (Molin et al., 1993). The strengths of suicide genes in bioconfinement are their high specificity and the variety of potential targets and activators. Molin and colleagues (1993) reviewed the major systems of suicide genes developed in bacteria. The hok/sok (host killing/suppression of killing) system originally was observed in bacterial plasmids (Gerdes et al., 1986). That system and others in the gef gene family consist of genes that encode for a toxic polypeptide that both attacks the cytoplasmic membrane and is the antidote to that polypeptide. If a cell spontaneously loses the plasmid (or if the gene inserted in the chromosome mutates) it dies because the leftover mRNA is translated into a toxic protein that degrades the bacterial membrane. One advantage of using this class of toxic factors is that their target, the cytoplasmic membrane, is similar in structure in many bacteria. Other killing systems include nucleases, which target destruction of genetic material (e.g., Ahrenholz et al., 1994). Those systems are highly promising. Not only would they kill the engineered bacterium but they also destroy its DNA, which might otherwise be transferred from the dead organism to living cells via transformation. Genes that code for nucleases from Serratia marcescens and Staphylococcus aureus have been fused to an inducible lac promoter to create such killing systems. It is unclear to what extent the DNA repair systems in cells would make it difficult for the nucleases to degrade the DNA to the extent necessary to kill the cells. Lysis genes from bacterial viruses also have been considered as a source of killing genes. Those genes have been cloned and fused with regulated gene expres- sion systems, with promising results (Molin et al., 1993). It is noteworthy that research in the development of suicide genes as a means of bioconfine- ment appears to have stopped about ten years ago for reasons that are not clear to the committee.
VIRUSES, BACTERIA, AND OTHER MICROBES 175 Fungi The same methods that are being used to control the spread of trans- genic bacteria are being applied to fungi. The object is to prevent fungal persistence and spread through the formation of various kinds of spores. If the spore is necessary for the fungus to execute its intended function, such as the infection of an insect, it is necessary to suppress sporulation after that goal is accomplished. Genes that inhibit sporulation could be put under the control of an inducible promoter and then engineered into the fungus. The spores would be treated with a chemical or environmental inducer before they are applied to the target pest. Some fungi are being genetically engineered to contain genes that increase their virulence in specific insect pests (Hu and St. Leger, 2002; St. Leger et al., 1996). To prevent the creation of hypervirulent organisms, it has been proposed that the genes be flanked by antisense forms so as not to affect the virulence of the strain but to target genes in the recipient cells that might inadvertently receive the virulence genes. Such targets could involve reproduction, spore formation, and spore germination. To prevent vegetative spread of the mycelium, suicide genes could be engineered into cells under the control of an inducible promoter. A major weakness of suicide genes in fungi and bacteria is the occur- rence of mutations that prevent the system from operating. In large part, their usefulness has been demonstrated only in laboratory studies and it is not clear how they will function in the field. Some suicide systems are intriguing ideas that have yet to work even in the laboratory. Laboratory experiments show that killing by suicide gene systems is never absolute; a surviving subpopulation can continue to grow even in the presence of inducer (Molin et al., 1993). The survivors result from mutations in the killing gene, mutations in the expression system that inactivate the suicide function, or mutations in other parts of the cell that confer resistance to the action of the killing agent (Knudsen et al., 1995). Suicide systems also can be lost from the cell if they are located on a plasmidthe plasmid can be lost after transfer to only one daughter cell during cell division. One way to reduce the problem is through redundancy, provided by the use of two identical systems or the combining of different suicide systems (Jensen et al., 1993; Knudsen et al., 1995). Those efforts will lower, but not eliminate, the probability of mutations, resulting in resistance. Thus, suicide systems can reduce, but not eliminate, a genetically engineered population. Viruses To the committee's knowledge, suicide gene systems per se have not been applied in the production of genetically engineered viruses. However,
176 BIOCONFINEMENT OF GENETICALLY ENGINEERED ORGANISMS viruses can mutate spontaneously into temperature-sensitive or ts mutants such that the mutant genotype cannot replicate at some temperatures. If a genetically engineered virus featured a ts mutation, that system could be harnessed in a fashion similar to the physical control of bacterial suicide genes. However, ts mutants typically are much less fit than are wild-type viruses, and using them might be more accurately described by the fitness reduction method known as phenotypic handicapping. Failed or Inappropriate Methods of Microbial Bioconfinement Chapters 3 and 4 describe the major bioconfinement methods used in higher organisms, and most of them are inappropriate for use in microbes. Because bacterial reproduction is strictly asexual, for instance, sterilization cannot be used to confine transgenic bacteria. Although some viruses can reproduce through reassorting chromosomal segments when multiple virus particles infect the same cell, the same viruses also can reproduce clonally. Thus, unlike most eukaryotes, they are not bound to obligate sexual repro- duction. As a consequence, confinement of genetically engineered microbes must be limited to fitness reduction methods such as the induction of suicide genes or phenotypic handicapping. Effectiveness of Methods at Different Temporal and Spatial Scales Temporal scales could influence the effectiveness of bioconfinement in bacteria, fungi, and viruses. Although those microbes can grow rapidly under ideal laboratory conditions (up to one generation per hour), typically they grow much more slowly in nature (Madigan et al., 2003). Nutrients in the wild usually are limiting, and their scarcity can prevent microbes from achieving rapid exponential growth. In nature, bacteria often experience "feast or famine;" periods of rapid growth are interspersed with longer periods of retarded growth. In the transition, bacteria undergo dramatic changes in physiology and morphology, which adapt them to poor growth conditions. In periods of slow growth bacteria are much more resistant to environmental assault than are rapidly growing cells (Siegele and Kolter, 1992). The process of bacterial sporulation, in which bacteria enter a dor- mant, nonmetabolizing, highly resistant state, is an extreme example. Fungi develop spores in the course of sexual or asexual reproduction. The spores, which are readily dispersed, are hardier than mycelia but not nearly as resistant to harsh environmental conditions as bacterial spores can be. Viruses are nonmetabolizing entities, so they do not have the luxury of regulating metabolism as do bacteria and fungi. Rather, as obligate parasites, viruses are at the mercy of their hosts (the biotic environment), and their ability to grow in adverse conditions (the abiotic environment)
VIRUSES, BACTERIA, AND OTHER MICROBES 177 depends on host metabolism. Because the ideal growth conditions of host cells are likely to be separated in time, many viruses can infect their hosts latently until the hosts resume growth or active metabolism, which then would allow productive infection to occur again. Fitness reduction methods are designed to hamper the reproductive potential of genetically engineered strains of bacteria, fungi, and viruses, placing them at a growth disadvantage relative to wild-type strains. These debilitated strains should fare no better and likely far worse than their wild- type counterparts in terms of ability to survive in the natural environment. However, insufficient field testing has been done in a variety of environ- ments with different organisms and genotypes to confirm that expectation. A more important consideration for the effects of temporal scale on fitness reduction methods in any microbe is the possibility that the microbe will become latent (for viruses) or sporulate (for bacteria and fungi). Effects of spatial scale on the confinement of bacteria and viruses are difficult to gauge; relatively little is known about dispersal of microbes in the wild. Most current data concern dispersal of pathogenic bacteria and viruses through physical processes (such as flow of water) or geographic movement of their host organisms (such as air travel by infected humans). Phenotypic handicapping and other fitness reduction methods are designed to reduce local survival of introduced bacteria, fungi, and viruses. Should those microbes become dispersed to distant locales, one might assume that the methods would be effective there as well. But this is not necessarily true, especially if the microbes are dispersed to different kinds of environments. Some viruses can inflict very different degrees of damage (becoming more virulent) when the host population is naïve to virus attack because of an absence of resistance alleles or antibodies in the host population (Bull, 1994; Taylor et al., 2001). Although highly speculative, a potential concern is that migration of genetically engineered bacteria, fungi, and viruses to new places could release them from phenotypic handicapping and from other mechanisms designed to hinder fitness. Because the effects of sporulation and germination traditionally have not been evaluated in field tests, it would be wise to avoid, if possible, genetic modification and release of sporulating microbes as a way to mini- mize the risk of long-term survival and dispersal and allay the fear that those strains would transfer their genetic material to local populations. Ecosystem and Population Effects The committee has identified phenotypic handicapping and suicide sys- tems as the primary methods that could be used for bioconfinement of bacteria, fungi, and viruses. Although more field data are needed, it is unlikely that the methods themselves would damage ecosystems and natural
178 BIOCONFINEMENT OF GENETICALLY ENGINEERED ORGANISMS populations. However, because they have been studied only under labora- tory conditions, three conceivable consequences can be foreseen that echo the motivations for bioconfinement outlined above. All relate to the envi- ronmental consequences of failure under natural conditions. Although there are no supporting data, the possibility exists that release of transgenic bacteria, fungi, and viruses can damage indigenous microbial populations. For instance, although the effects of introduced viruses on intended hosts can be gauged accurately through laboratory experiments, the introduced viruses could attack nontarget hosts that are ill prepared to defend themselves. Similarly, survival of bacteria in the wild could be under- estimated from laboratory results. The second concern involves displace- ment of resident species. Indigenous viruses could be less virulent than introduced ones, creating the opportunity for engineered viruses to severely reduce the population size of local hosts. Finally, genes from introduced bacteria, fungi, and viruses could be transferred horizontally into resident species. Because those genes could have unanticipated effects when they migrate to new backgrounds, they could reduce fitness at one or more loci through negative epistasis. Although selection should act to remove the introduced genes from the local gene pool, it could take a long time for dangerous alleles to be completely removed as a result of weakened selec- tion as the genes become rarer in the population (Hartl and Clarke, 1997). The committee believes that the ecological consequences of using fitness reduction methods, such as phenotypic handicapping and suicide systems in genetically engineered microbes, are likely to be minimal, because those methods are designed to employ genotypes that are competitively inferior indigenous strains. However, because the methods have not been evaluated in the field, it is not possible to state with certainty that they will have the desired effect in confinement. Monitoring, Detection, and Culling: Needs, Feasibility, and Realities The frequency of genetically engineered microbes in natural environ- ments can be estimated rather straightforwardly if natural populations are extensively sampled and screened with modern molecular techniques, espe- cially if the engineered organisms contain easily detected phenotypic markers, for example, that are visible on a selective agar medium. However, it is virtually impossible to completely eliminate specific genotypes in natural populations of microbes (Salyers and Whitt, 2002). This needs to be consid- ered when deciding whether a genetically engineered microbe should be released into the environment.
VIRUSES, BACTERIA, AND OTHER MICROBES 179 Microalgae Although this chapter focuses on bioconfinement of transgenic bacteria, fungi, and viruses, the possibility of bioconfinement also should be evalu- ated for genetically engineered microalgae. Microalgae have already been successfully engineered (see review by Minocha, 2003). The best results have been obtained with Chlamydomo- nas reinhardtii, which has long served as a model system for physiological and molecular studies (e.g, Cerruti et al., 1997; Dunahay, 1993). In particu- lar, genetic engineering of C. reinhardtii could be useful for bioremediation of heavy-metal pollution, a pervasive environmental problem because trace metals cannot be decomposed but must be sequestered from the environ- ment. Cai and colleagues (1999) demonstrated that the trace-metal-binding properties of Chlamydomonas can be enhanced in transgenic genotypes that express a foreign-metal-binding protein, without slowing their growth rate relative to wild-type cells. In addition, stable nuclear transformation has been achieved in the colonial green alga Volvox carteri (Hallman and Sumper, 1994; Schiedlmeier et al., 1994). A few diatoms also have been successfully transformed, including the widely studied model system Phaeodactylum tricornutum (e.g., Apt et al., 1996). This is promising because diatoms have commercial uses as feed in aquaculture and as poten- tial sources of useful pharmaceuticals. Most commercial-scale cultivation of microalgae is performed in large, open outdoor ponds. Zaslavskaia and colleagues (2001) identified several disadvantages of this approach, including invasion of ponds by contami- nants and reduction in biomass production resulting from seasonal and diurnal variations in temperature and light. Thus, improved efficiency and reduced cost of micro-algal biomass production could be achieved if the microbes were engineered to grow as heterotrophs in conventional micro- bial fermenters (in the absence of light). Zaslavskaia and colleagues (2001) introduced a gene that encodes a glucose transporter into the obligate photosynthetic microalga P. tricornutum, allowing the diatom to thrive on glucose in the absence of light. The approach seems promising because fermentation technology eliminates contamination by microbes, which is an important criterion for maintaining food industry standards. Microalgae are biologically similar to bacteria (especially photosynthetic bacteria) that grow in aquatic environments, and they have similar mechanisms for horizontal gene transfer. Therefore, the same consequences and concerns would apply to their bioconfinement. However, most transgenic microalgae have been cultivated in closed-system indoor tanks and are not intended for release into natural environments. Because of their similarity to bacteria, phenotypic handicapping and suicide systems should provide effective bio- confinement if necessary. The committee did not find any reports in the literature of efforts to test the feasibility of those methods in microalgae.