Summary of Proceedings
Learning from Experience
A decade ago, a genome project was a new and untested idea. Indeed, in 1986 when the Department of Energy announced it would fund a Human Genome Initiative, the idea was met with a great deal of skepticism about whether such a project could be done at a reasonable cost and whether the project, even if it could be done economically, would be of great value to science. Over the past ten years, however, such reservations have vanished, as researchers working on many different species have learned more about what goes into a genome project and more about what comes out. Today, biologists know the complete genetic sequences for a number of microorganisms, including yeast, several bacteria, and scores of viruses. And there are several programs under way that seek to acquire the much larger—and more difficult to obtain—genomes of plants and animals. Targets for these projects include a weed, a worm, the fruit fly, the mouse, and, of course, humans. Much of what has been learned from these efforts can be applied directly to an agricultural genome project.
The most straightforward borrowing will be of tools and techniques. Mapping and sequencing a genome is the same task whether the genetic material comes from yeast, people, or rose bushes—the genes of all are composed of the same chemical building blocks, just put together in different ways. This means, said Daniel Drell, a biologist with the Department of Energy's Human Genome Program, that the advanced genomics technologies developed for other species can easily be put to work on an agricultural project. "There is no point in reinventing the wheel," he said. "The Department of Energy and the National Insti-
tutes of Health have achieved quite a bit in terms of technology development, informatics, resource development. We are delighted to make available all the lessons we have learned."
How to apply these scientific tools—which information is most important to pursue, which techniques are best suited for certain types of studies, how research should be ordered and prioritized, and so on—is a more complicated issue, but it is an issue to which lessons from other genome work still apply. In this case, however, it does matter which genome efforts serve as models. The most appropriate lessons will come from genome work on multi-cellular organisms rather than bacteria and other single-celled creatures, because of the difference in the size and complexity of their genetic codes.
Microorganisms have relatively small genomes. The yeast Saccharomyces cerevisiae, for example, has 16 chromosomes with a total length of 12 million base-pairs—a base-pair being an individual letter in the genetic instructions that are encoded in an organism's DNA. Most of the bacteria whose genomes have been decoded have fewer than two million base-pairs. The genomes of plants and animals are, by comparison, huge. The common weed Arabidopsis thaliana, which is believed to have the smallest genome of any flowering plant, still boasts 120 million base-pairs. Corn's genome has 2.3 billion base-pairs, and wheat's is an oversized 16 billion. More than 100 times as large as Arabidopsis. Compared with wheat, the human genome is modest—only 3 billion base-pairs—but it is still a thousand times larger than that of the typical single-celled organism. The bigger a genome is, the more difficult it is to sequence, and not just because of the greater number of base-pairs. As the total amount of genetic material increases, the technical difficulties of handling and keeping track of all the DNA grow as well. For that reason, researchers looking to sequence the genomes of rice or corn or cows or pigs can learn most by gathering advice from their colleagues working on Arabidopsis and mice and humans.
That advice, as offered at the workshop, ranges from very broad and general to quite narrow and specific. On the broad end, there is general agreement about the right way to approach mapping and sequencing of multi-cellular organisms. The recommendations of Christopher Somerville, director of the Department of Plant Biology of the Carnegie Institution of Washington in Stanford, California, were typical: "When we began developing the Arabidopsis program, we realized the sequencing had to be preceded by other forms of information, particularly a very good map, because sequenced information is most useful when it is used in conjunction with mapping information on individual genes and mutants." In mapping a genome, researchers piece together a large-scale picture of where genes and larger chunks of the chromosomes are. A sequence, on the other hand, is a letter-by-letter compendium of the genetic code, a listing of the base-pairs in the order in which they appear on a chromosome. (See Box 1.)
|
BOX 1 The genetic code of every living creature—plant, animal, or microorganism—is carried by long, twisting molecules of DNA. A single strand of DNA consists of a long sequence of small molecules called bases, and it is the order of these bases that encodes all genetic information. It is customary to represent the four bases that go into DNA as letters in a sort of chemical alphabet: A for adenine, C for cytosine, G for guanine, and T for thymine. In this way, a strand of DNA is written out as one long "word": AATAGCTCC ..., and so on for a hundred million or more letters in a single chromosome. In cells, DNA usually exists as a linked pair of complementary strands, with an A in one strand always paired side by side with a T in the other strand, and similarly for C and G. Thus the strand AATAGCTCC would normally be paired with its complement, TTATCGAGG. Sequencing a stretch of DNA means listing the chemical bases in the exact order that they occur. Researchers usually speak of base-pairs instead of bases, but in practice they are equivalent since knowing one member of a base-pair automatically identifies the other. Once researchers have cataloged every base-pair in an organism's DNA, they know that creature's entire genetic code. From there they can track down the particular stretches of DNA devoted to genes, identify the sections involved in turning genes on and off, and in general decode the instructions for assembling and operating that particular creature. The major obstacle to sequencing an organism's genome—its total complement of DNA—is the sheer amount of DNA involved. Even bacteria tend to have genomes that are one or two million base-pairs long, and the genomes of plants and animals are generally hundreds of millions to billions of base-pairs long. Modern sequencing techniques can handle pieces of DNA no more than a thousand base-pairs in length, so researchers must chop up the long strands of DNA into small fragments before sequencing them, and this leaves them with the problem of how to piece together these fragmentary sequences. With various tricks it is relatively straightforward to assemble sequences that are thousands or tens of thousands of base-pairs long, but this is just a drop in the bucket on a chromosome that may be hundreds of millions of base-pairs long. Furthermore, its not easy to tell where on that chromosome a particular sequence fits. This is where mapping comes in. As the name implies, a map gives researchers a way of finding locations along a chromosome. There are two basic sorts of maps, genetic and physical. A genetic map approximates where on a chromosome particular genes or markers lie. Markers are short, easily identifiable stretches of DNA that vary from person to person, so that with simple tests researchers can usually tell which parent an individual inherited a particular marker from. When a mother's and father's chromosomes break up and later recombine to form the DNA for an offspring, genes and markers that are close neighbors on a chromosome tend to be inherited together. If Gene X is only 100,000 base-pairs away from Gene Y, and Bobby inherits Gene X from his mom, chances are very good that he'll get Gene Y from her as well. By keeping careful track of how often traits and markers are inherited together, researchers can get a good idea of their relative locations on a chromosome. A genetic map of a chromosome may contain hundreds of genes and markers listed in the order they appear on the chromosome. The "genetic distance" |
|
between any two of them—how likely it is that both will be inherited together—provides an approximation to the physical distance between them in terms of numbers of base-pairs. A physical map, on the other hand, consists of a collection of overlapping stretches of the DNA itself. To produce a physical map, researchers start with many copies of a chromosome and chemically snip them into small pieces, each, say, just a few hundred thousand base-pairs long. Each piece is separated from the others and cloned, creating a set of many identical copies of this one bit of DNA, which can then be analyzed in various ways. Any given cloned stretch will likely share an overlap region with one or more other cloned stretches, and by identifying these overlaps, scientists can figure out that Clone X and Clone Y fit together with a certain common region. A complete physical map of a chromosome is a set of these clones that, together, cover the whole chromosome, along with a description of where on the chromosome the clones lie. Thus a physical map is more than a road map of the chromosome—it also includes physical bits of the chromosome that can themselves be manipulated and analyzed, and the libraries of clones that are part of a physical map are themselves very valuable to researchers. Ultimately, one can completely sequence a chromosome by getting physical maps that have finer and finer resolution, until the individual clones are small enough to be sequenced. Then it is simply a matter of sequencing these millions of small bits of DNA and assembling them into a complete genome. |
In theory, it might seem that the sequence is all one needs since all the information the genome holds is found in the order of the base pairs. But the practice is somewhat different. Maps are useful in a variety of ways. Before the entire sequence of a genome has been written out, they help researchers discover the genes responsible for particular traits or diseases. As the sequencing is going on, maps provide the physical material for the sequencing effort and allow researchers to piece together the sequences derived from small bits of DNA. Even after the sequence is complete, good maps are an indispensable guide to what is found where. One would no more throw away the maps and rely on the sequence than one would navigate a trip from Florida to California with neighborhood plots that showed every building, fence post, and fire hydrant along the way. "So," Somerville said, "much of the support during the first few years [of the Arabidopsis project] was actually directed toward mapping. I think that is a good model to follow for the other genomes."
For the Arabidopsis program, the map of choice was a "physical map," a set of overlapping stretches of chromosome that covered the entire genome. In particular, the Arabidopsis researchers used yeast artificial chromosomes, or YACs—
chunks of Arabidopsis chromosomes modified so that they were accepted by yeast organisms as their own. When the yeast multiplied, they created many copies of each stretch of Arabidopsis DNA, establishing a permanent collection for study by as many scientists as were interested.
Future genome projects may want to use a slightly different type of physical map, suggested Neal Copeland, a mouse researcher at the National Cancer Institute. "As people [working on the mouse genome] have started using these YAC maps, it is clear that they are not the ultimate maps." In particular, he said, because YACs are frequently unstable or chimeric, the resulting YACs "are no good for sequencing." Instead of YACs, he recommended making maps with BACs—bacteria artificial chromosomes—which can be used for sequencing.
By borrowing technology and techniques from other genome work, researchers in an agricultural project should have everything necessary to map and sequence the genomes they're interested in. "We do not have any technical problems that I am aware of," Somerville said of the Arabidopsis effort. "One of our largest problems is just acquiring the financial resources to do it." The only problem—if ''problem'' is the right word for it—is that the technology is improving so rapidly that most technical advice will be outdated in a few years. "In going back and looking at the projections for what would happen in the mouse and the human genome projects," Copeland said, "what we have learned is that all these projections have been way underestimated. We have gone much beyond where anybody could have even dreamed that we would be right now."
That technical proficiency is the root of the one major weakness that a number of workshop participants identified in current genome projects. As researchers in various laboratories collect more and more genetic data, some way must be found to transform this mass of often disconnected information into a unified whole—to collect it in one or a few places, organize it, catalog it, annotate it, and make it available to whoever wishes to use it. But too often this "bioinformatics" side of a genome project has failed to keep up with the generation of the data.
"Scientists who do genome research often neglect informatics," said DOE's Drell. "They just figure someone else is going to take care of it. As a consequence, they neither ask for nor get the kind of financial support that is necessary to curate data for deposition into a database. This is a major problem because what you get is a large collection of individual Web sites, and it is extremely difficult to find the data. The data officially are public. They are available if you know where to look for them, but they are very hard to get."
"There is a whole variety of informatics issues," Drell continued. "They have been a running headache for the genome projects, and they are too important to neglect. You should think about how to address them from the start. The most difficult and expensive solution of all is to not work them out first and then have to play catch-up afterwards."
Many of the scientists at the workshop agreed with Drell that any agricul-
tural genome project should have plans in place from the beginning to take care of the data. The project should include large, central databases that are compatible with each other and easily accessible to researchers. Because the data will be valuable for years, if not decades, to come, it will be vital to plan for—and pay for—long-term curating of those databases. Furthermore, as David Galas, president of the Darwin Molecular Corporation and a member of the NRC's Board on Biology noted, the databases need to be more than just a collection of maps and sequences. "Sequence is fine, but what is really interesting is the biology that is in it." When biologists discover information about a particular gene in a database—what it does, how it is regulated, the structure of the protein that it encodes, and so on—they should be able to add that information to the database. "That is really a major challenge," he said, "and it is not cheap. There are a lot of different ways of going at it, but our ability to annotate, while getting better, is still a long way from being what we would like it to be."
Researchers must keep in mind that they will be collecting data not just on one species but on dozens. "We've got real problems that go well beyond what we have dealt with in the Human Genome Project, and that is in comparative genetics—genomics among a very large number of species," said David Galas. If an agricultural genome project is to be successful, the databases for the various species should be interconnected and compatible with one another, so that a researcher studying, say, tomatoes can easily find relevant information from Arabidopsis and other species. "That is a whole other level of difficulty," Galas said, as researchers on mouse and human genomes have discovered. Integrating the databases from those two species would be valuable for the same reasons that integration among agricultural genomes would be valuable, but so far such coordination is more hope than reality. "Quite frankly," he said, "we're still catching up.''
Finally, the early consideration of social and ethical issues for an agricultural genome project is a step that other genome projects have proven to be valuable (see Box 2: Considering Social and Ethical Issues).
Dealing with a Multitude of Genomes
An agricultural genome project will have much in common with the Human Genome Project. Much of the technology will be the same, the agricultural effort will demand the same massive databases and careful coordination between laboratories that the human program does, and both projects have the promise of potentially revolutionary payoffs. But, as Daniel Drell of the Department of Energy said, there will be a major difference, one that makes the agricultural project much more challenging. "There is only one species of human, so the human genome project is easier to define, and it only has to be done once," he said. Agriculture, by contrast, involves hundreds of species, dozens of them
|
BOX 2 In the early 1980s, four drug companies announced that they would soon begin selling farmers a product to increase milk production in cows: recombinant bovine growth hormone, or rBGH. Except for being produced in bacteria, rBGH was the same molecule that cows themselves produced to trigger milk production, so the companies anticipated no problems with acceptance, and indeed, in 1984, the Food and Drug Administration pronounced milk from rBGH-treated cows safe for human consumption. Soon after, however, the artificial hormone became mired in controversy. Proponents of family farms claimed rBGH would shrink the nation's dairy herds and spell the end of small producers. Environmentalists and some scientists began raising safety issues they claimed the FDA had glossed over, including charges that drinking rBGH milk might increase the risk of breast cancer in women and that the increased use of antibiotics on rBGH-treated cows might lead to the development of antibiotic-resistant bacteria. In response, the FDA asked the National Institutes of Health to study the issue, and in 1990 an NIH panel gave rBGH a clean bill of health. Still, the controversy continued. Ben & Jerry's ice cream stores announced they would not buy from any dairy using the still-experimental treatment. Several grocery store chains followed suit. Maine and Vermont passed labeling laws intended to identify milk products from cows given rBGH. It was not until 1994 that the FDA gave the final go-ahead for commercial sales of the recombinant hormone. Since then, the concerns have subsided, and rBGH is now used regularly in many herds. The rBGH controversy offers a cautionary tale about the pitfalls of genetic engineering in agriculture. Although neither the cows nor the milk were altered, and although the only thing "unnatural" about the rBGH was that it had been produced in bacteria rather than cows, a significant part of the public was nevertheless uneasy about the use of the hormone. In light of this history and the nervous reception given to the few genetically engineered food products to reach the market, such as the slow-ripening Flavr Savr tomatoes, biological researchers need to think about the social and ethical implications of an agricultural genome project, said Daniel Drell, a biologist with the Department of Energy's Human Genome Program. "I think it would be naive for the agricultural community to imagine you are not going to get challenged." This is particularly true because the whole point of an agricultural genome project is to make possible wholesale manipulation of the genes of crops and livestock. Yet, as Nina Fedoroff of Pennsylvania State University pointed out, ''People continue to have problems with the concept of designer foods. People continue to have problems with genetically engineered animals and plants." Instead of assuming, as the makers of rBGH did, that the public will automatically welcome their efforts to improve agriculture, workshop participants urged those planning an agricultural genome project to consider the social and ethical implications of the work ahead of time and try to anticipate objections and difficulties. At a minimum, a task force on social and ethical issues should be an official part of the program. Scientists need to acknowledge that some of the concerns of the public are legitimate and that more technical information is not necessarily the answer. "The ultimate regulation," Drell said, ''is going to be whether John and Jane Q. Citizen buy your product. If they are made nervous by what they imagine you have done to it, you will find out about it in the worst way." |
economically important. The key issue for any agricultural genome program will be how to deal with this multitude of genomes.
In theory, of course, it would be nice to have the complete genome of every animal and every plant that is important to the food and fiber industries: cows, pigs, sheep, chickens, turkeys, corn, wheat, rice, soybeans, potatoes, tomatoes, strawberries, cotton, apple trees, orange trees, pine trees, oak trees, and many more. In practice, that isn't going to happen, at least not any time soon. Genome projects are expensive. The Human Genome Project is expected to cost $3 billion by the time it is finished, and even the very short genome of Arabidopsis will take as much as $75 million to decipher. As a practical matter, it will be possible to go after only a small fraction of all the agricultural genome information one would ideally like to have.
Fortunately, it's possible to make a small fraction go a long way, as long as the fraction is chosen correctly. Most of the agriculturally important crops are genetically related—distant cousins perhaps, but still cousins—and so information obtained about one can often be put to use with another. The same is true for farm animals. So the major question facing an agricultural genome project becomes: Which fraction of all the genetic information available should be tackled first to get the most bang for the buck?
A number of the participants in the workshop agreed that there is no better way to get bang for the buck than to sequence the entire genome of a representative organism, as is now being done for humans and for the weed Arabidopsis thaliana (see Box 3: Saved by the Weed). As David Galas, president of the Darwin Molecular Corporation, said, "The actual value of having a single species worked out in some detail is absolutely enormous and will be catalytic, independent of what is done beyond that." Having an entire genome to study—not just the sections coding for the genes, but the regions that turn the genes on and off and also the vast stretches of DNA that seem to serve no function—will allow researchers to learn details about the design and functioning of living organisms that can now only be guessed at.
But perhaps the greatest value of a complete genome is that it offers a catalog of all the genes at work in an organism. Humans have, for instance, an estimated 50,000 to 100,000 genes, but only about 10,000 of them have been identified and sequenced. Arabidopsis has approximately 20,000 genes, of which only about 2,000 have been fully sequenced. When their respective genome projects are complete, researchers will have a list of every gene as well as its sequence and its exact position on a chromosome. Furthermore, these catalogs will provide researchers with complete or nearly complete lists of genes for other animals and plants, even those that are only distantly related to humans or Arabidopsis.
"Our impression is that basically all plant genes are represented in Arabidopsis," said Chris Somerville of the Carnegie Institute of Washington. The exact structure of the genes varies from plant to plant—which accounts for the differences between species—but all plants seem to make do with essentially the
|
BOX 3 Ironically enough, the most important plant in agricultural research over the next decade may be a common weed. Arabidopsis thaliana is a small, nondescript member of the mustard family. Even when its white flowers are in bloom, it is not a plant that attracts much attention—except from biologists. Over the past fifteen years, Arabidopsis has become the plant world's equivalent to the laboratory mouse, and now it is poised to become the first plant to have its entire genome sequenced and made available for study. In the early 1980s, a group of plant biologists decided to choose one plant that could, like the mouse, serve as a model for exploring the workings of many species. They settled on Arabidopsis for many of the same reasons that the mouse became a model animal: it is small and easy to maintain, it reproduces rapidly (its growing cycle is six weeks), and it has a variety of mutants. Furthermore, Arabidopsis has an exceptionally small genome—only 120 million base-pairs, believed to be the smallest of any flowering plant. This made it easier for researchers to track genetic mutations to a particular spot on one of the plant's five chromosomes, and it also made Arabidopsis a natural target for a genome project. In the fall of 1996, three U.S. government agencies announced funding for an Arabidopsis sequencing effort, the first—and so far only—plant genome project in the United States. The National Science Foundation and the Departments of Energy and Agriculture awarded $12.7 million over three years to three sequencing groups headquartered at Cold Spring Harbor, Stanford University, and The Institute for Genome Research. At about the same time, those three U.S. teams announced that they would cooperate with two other Arabidopsis sequencing groups, one at a laboratory in Japan and the other a consortium of 17 European labs. By agreeing to divide up the sequencing work and share the results, the five groups expect to delineate the entire Arabidopsis genome by 2004. Like work on the mouse, the ultimate value of research on Arabidopsis lies in what it teaches scientists about other, more important species. All flowering plants are very similar genetically, so insights garnered from Arabidopsis can be applied to everything from roses and sunflowers to corn and tomatoes. Indeed, Christopher Somerville, an Arabidopsis researcher at the Carnegie Institution of Washington, said that essentially every gene found in any flowering plant has a counterpart in Arabidopsis. Thus if a corn researcher wished to study, say, the uptake of phosphorus from the soil by the roots of a corn plant, he would not have to search blindly through the corn genome to find the relevant genes. He could instead identify those genes first in Arabidopsis—a much easier process, given the size of its genome and its eventual complete sequencing—and from there track them to corn. Although there is not agreement on such possibilities, the Arabidopsis Genome Initiative promises to be important for corn and many other crops, said Tony Cavalieri of Pioneer Hi-Bred International, the major supplier of corn seed in the United States. "You might say that the most important commitment [by the government] to corn genetics to date has been not the genomics work in corn that has been publicly funded, but the commitment to the Arabidopsis sequencing program." |
same set of genes. "For example," Somerville said, "we find the same genes controlling floral morphology [the shape and structure of flowers] in Arabidopsis as in snapdragon, which has a very different flower, and in maize, which has an even more distant, different-looking flower." The story is the same for other genes.
"I think that Arabidopsis research is going to culminate in understanding what every plant gene does," Somerville said. "The significance of that is going to be an ability to take that knowledge and map it on to all of the plants that we care about." In other words, to study a particular trait in any plant, it should be enough to figure out which gene controls that trait in Arabidopsis and then look for the corresponding gene in the plant of interest. By pinpointing the genes that make Arabidopsis resistant to various diseases, for instance, researchers could track down those genes in other plants, especially strains with a strong resistance to particular diseases. Those genes could then be inserted into the DNA of commercial crops to give them the same resistance.
Leveraging the Arabidopsis genome in this way depends on an exciting recent discovery that during the evolution of the angiosperms, plants whose seeds are within a fruit, the order of genes on chromosomes appears to change slowly. Thus, closely related species such as green pepper and tomato have large regions of chromosomes with the same gene order. Similarly, rice, maize, wheat, barley, millet and sorghum have been found to have very similar gene order. This means that if the genome of one of the angiosperms were completely characterized it would be possible to extrapolate much of that information to all of the other plants in that group with a high degree of accuracy. Thus, for instance, if a gene for a trait such as drought tolerance was mapped in species-A, it would be possible to identify the DNA sequence of the gene in the fully characterized species (species B) based on the conserved gene order. This, in turn, would permit the isolation of the useful variant of the gene from species-A by using the DNA from species-B as a hybridization probe. Unfortunately, although this tactic will work for plants closely related to Arabidopsis, such as oilseed rape or brussels sprouts, it gets less useful for those not so closely related—wheat, say, or pine trees. As Jeff Bermetzen of Purdue University said, "I would agree with Chris [Somerville] that Arabidopsis will give us virtually all plant genes. The difficulty is the ties we will be able to make across these organisms." Although there is some colinearity between even distantly related species, there is not nearly enough. A gene might be very close to a particular marker in Arabidopsis but nowhere near the corresponding marker in, say, soybeans. As a result, Bennetzen said, "You are not going to be able to transfer information across species quite that simply based on map position."
For that reason, many workshop participants agreed that one of the goals of any agricultural genome program should be to develop target species other than Arabidopsis. The genomes of these organisms should be fully sequenced so that they too can serve as models for other, related species.
This raised the question, "Which species should be targeted?" Several criteria for choosing were suggested. First, since it would do little good to sequence another plant closely related to Arabidopsis, the target species should be selected from other parts of the plant world. All crop plants—indeed, all flowering plants—belong to one of two broad groups, monocots and dicots, which are distinguished by the number of leaves that first sprout from the seed—one in the case of monocots, two for dicots. Arabidopsis is a dicot, as are peas and beans, potatoes, tomatoes, carrots, strawberries, apples, artichokes, and beets. Monocots include wheat, rice, corn, bananas, pineapples, onions, and asparagus. "Arabidopsis is a really great model for all of the dicots," Somerville said, "but we really should have extensive information on one monocot." Other workshop participants agreed.
There was no consensus, however, on which monocot should be given priority. If the shortness of the genome is a consideration—as it was in the choice of Arabidopsis—then rice is the obvious candidate. Its genome is only 420 million base-pairs, several times larger than that of Arabidopsis, yet smaller than any other commercially important monocot. Several other countries, including Japan and Korea, have already begun sequencing the rice genome, and a U.S. effort could lead to the sort of international genome effort already at work on Arabidopsis. On the other hand, a number of workshop participants thought a corn genome project would make more sense. Although its genome is more than five times that of rice—2.3 billion base-pairs—corn is a much more valuable crop in the United States, thanks mainly to its use in animal feed. (Wheat is also a valuable crop, but the size of its genome—nearly seven times that of corn—rules it out as a target species. That oversized genome is due partly to wheat's evolution as a merger of three separate species and partly to the genome having a great deal of repetitive DNA. Rice, by contrast, has a very efficient genome, with relatively little space on its chromosomes devoted to DNA that serves no function.)
Nor was there any consensus on how many target organisms should be chosen. Some thought one monocot—probably corn or rice—would do. Brian Larkins of the University of Arizona argued for both. "There are some good technical reasons why maize should be done and not simply assume that it is going to be easy to extrapolate the data from rice." Comparisons of the DNA of rice and corn show, he said, that there is much less similarity in the structure of their genomes than once thought. And Bennetzen commented it may be necessary to go even further. "What a lot of us in the field think now is that we are going to need a number of nodal organisms, organisms that will allow you to study a whole series of species that are closely related." Agricultural genome research would be apportioned into groups of related species, each with its own nodal organism, he said.
Because completely sequencing a genome is expensive, there is a limit to how many target species or nodal organisms can be delineated down to the last
base-pair. Just sequencing corn completely, for example, would run into the hundreds of millions of dollars. Entirely sequencing multiple target species would be prohibitively expensive.
Fortunately, most of the information provided by a complete sequence of a genome can be gained by creating detailed maps—both physical and genetic—of the genome, identifying and sequencing most of the genes (or, more accurately, cDNA clones for the genes), and then positioning the genes on the maps. This avoids sequencing the entire genome, which includes many sections that don't contain genes. "You can get virtually all of the information you need without sequencing through all the repetitive sequences and clusters of retrotransposons," said Nina Fedoroff of Pennsylvania State University. "The bottom line is that there are large chunks of clustered sequence that are probably not going to be useful." Since sequencing is a major part of the cost of a genome project—it's now about 30 cents to $1 a base-pair, although the price is dropping rapidly—bypassing this low-value DNA cuts the cost of analyzing a genome sharply.
Furthermore, it's not necessary to completely sequence the genes themselves, either. It's possible, Somerville pointed out, to get a lot of the information about a gene by using expressed sequence tags, or ESTs, a version of the gene for which one sequences only a few hundred of the thousands of base-pairs that make up the full gene. (See Box 4: Gene Chips.) "Based on similar studies of human genes you only get five percent more information about the probable function of a gene, based on sequence analysis by fully sequencing the genes," Somerville said. That is, if a researcher has isolated a gene from, say, potatoes and wishes to discover its function by comparing it with a collection of known genes from Arabidopsis or another target species, the chances of success are only about five percent greater if the researcher works with a completely sequenced gene than with ESTs.
The use of ESTs makes it feasible to amass a surprisingly large amount of information about even those plants and animals that are not chosen as target species. "In almost every area of genomics, there is always this battle over what species are going to be done," said Craig Venter of The Institute for Genomic Research. "We find this quite a lot in the microbial world because everybody has their own pet species.... But instead of people fighting over whether it is going to be wheat or corn or cows, fundamental data can be put out in a very short period of time that will rapidly advance all of these areas simultaneously."
Specifically, Venter recommended, "you could generate for $20 million spread over three to five years, 100,000 ESTs from each of forty different species. So there would be over four million ESTs from forty species that could be rapidly put into the public domain." Although there is invariably both redundancy and omission in a set of ESTs—some genes are represented by several or many tags, while other genes are missed altogether—a collection of 100,000 ESTs would pick out most of the important genes in a species. These ESTs could then, Venter said, be placed on physical maps of each of the forty genomes by
|
BOX 4 An agricultural genome project would advance agricultural research in many ways. One of the most intriguing is that it would make possible the use of the so-called "gene chips" for dozens of different agricultural species, opening up an entirely new way of analyzing the inner workings of these plants and animals. A gene chip offers a way of testing which of an organism's genes are active. Of the tens of thousands of genes that make up a plant's or animal's genome, only a portion are turned on—that is, working to produce their particular proteins—at any given time. Perhaps a few dozen genes are very highly active, others are less busy, still others are barely working, and a majority are idle. This pattern of gene activity varies according to cell type—it will be quite different in a root cell than in a leaf cell, for instance—and it changes according to circumstances. An infection will alter the pattern of gene activity, for example, as will such environmental stresses as a lack of water or the presence of pollutants. By analyzing the patterns and how they change, researchers can get clues into how an organism functions—for example, finding genes for disease resistance by seeing which genes become active in the presence of an infection. "We already have gene chips with thousands of genes," said Christopher Somerville of the Carnegie Institution of Washington. "But we envision, in the very near future, small chips in which all of the genes will be placed on these chips and these can be used to measure the expression of all of the genes in the plant in a single experiment. These kinds of experiments are going to qualitatively change how we go about doing plant biology." To date, however, the gene chips can be made for only a small number of species—those plants and animals for which gene libraries already exist. When a gene is turned on in a cell, the cellular machinery uses the long strand of DNA that makes up the gene as a template, creating a complementary strand of messenger RNA, or mRNA. This mRNA in turn is used to direct the assembly of a protein. |
using radiation hybrids—a technique in which chromosomes are broken apart by radiation and the fragments inserted into the cells of other organisms for handling. Furthermore, by using "gene chips," microscopic arrays that contain probes for thousands of different genes, it will be possible to match up the ESTs from the various species against each other and also relate them to genes in Arabidopsis or other target species. "I would argue," Venter said, "that this comparative data is going to be far more valuable to have than to completely sequence any one species."
Given the value of obtaining complete genomes for target species and the possibility of using ESTs to skim the cream off several dozen genomes for a relatively small investment, the researchers at the workshop thought the best approach to an agricultural genome project would be a multi-tiered one. At the top tier would be a few target species, such as Arabidopsis and rice or corn, whose genomes are sequenced completely. At the lowest tier would be dozens of
|
Thus the amount and type of mRNA in a cell is a direct measure of the pattern of gene activity in that cell, and researchers have taken advantage of this to create gene libraries based on the expression of genes in cells. Scientists isolate all the mRNA from a cell, make DNA strands that are complementary to each mRNA strand, and then make many copies of each of these complementary DNA, or cDNA strands. The resulting cDNA samples, each consisting of many copies of a single cDNA strand corresponding to one particular gene, make up a gene library, and scientists have assembled libraries with tens of thousands of these samples for certain plants and animals. Since each of these cDNA strands consists of tens of thousands of base-pairs—the chemical units that make up DNA—researchers generally determine only a partial sequence of a few hundred base-pairs, enough to identify the strand. These partially sequenced strands of cDNA are called expressed sequence tags, or ESTs. The gene chips consist of thousands of these samples of cDNA from one organism—up to 10,000 samples in a square centimeter—on a glass slide. To determine gene activity, a researcher isolates the mRNA from the cells of interest, uses the mRNA to make complementary strands of cDNA that are labeled with fluorescent molecules, and then lets that cDNA react chemically with the cDNA samples on the gene chip. Because the labeled cDNA sticks to complementary bits of cDNA on the gene chip, a researcher can then scan over the gene chip and tell by the spots of fluorescence which cDNA from the library corresponds to genes that are active in the cell. Another use envisioned for gene chips is to detect polymorphisms in genomic DNA. By developing gene chips that can assay the allelic composition of hundreds of genes at a time it will be much more feasible for plant breeders to use marker-assisted breeding. "This is the technology that will take over plant breeding at some point in the future," Somerville said, but not until extensive amounts of DNA sequence information is available for the major agricultural species. And that, he said, should be one result of an agricultural genome project. |
plants and animals whose genomes are mapped out in some detail and most of whose genes are identified by expressed sequence tags. In the middle would be nodal organisms with genomes delineated to an intermediate amount.
The project should include not just crop plants, but also livestock, crop trees, and even microorganisms, each group with its own target species. The microorganisms are often overlooked in discussions about an agricultural genome project, noted Jim Cook of USDA's Agricultural Research Service at Washington State University, but they are of crucial importance to agriculture. Many are pathogens, of course, which farmers would like to rid their crops of, but some are beneficial, such as bacteria that live symbiotically with the roots of crop plants and help the plant absorb nutrients from the soil.
In addition to this multi-tiered approach, workshop participants had several other bits of scientific advice. One key to the success of an agricultural genome project will be keeping abreast of the most up-to-date mapping and sequencing
techniques. "There are new technologies emerging, the EST sequencing and the radiation hybrid maps, that make this a lot cheaper than it used to be," said Neal Copeland at the National Cancer Institute in Frederick, Maryland. "Looking back at the mouse, if we knew about this technology we could have done it a lot cheaper and a lot faster."
Another key will be coordination. Since only a few genomes will be sequenced completely, comparison of genetic information across species will play a major role in an agricultural genome project, and the maps will have to be made with that in mind. Much of the mapping done today does not match up from one organism to another, Bermetzen said. "They are maps that are specific to a given species. That is a big problem." If the multi-tiered approach is to be a success, "mapping is going to need to be done in each species in a transferable mode."
Finally, it is important to keep in mind that all of this—the ESTs and radiation hybrids, the sequences and maps—represents just the first phase of what will be a very long, though ultimately very profitable, process. "When you go back to think about it, this is the easy and cheap phase," Copeland said. "We're talking about a lot of dollars, but this is cheap compared to what it's going to take to figure out what all these genes are doing." After the mapping and sequencing are done, researchers will face the task of understanding the structure and function of the various genes, how they are controlled, and how they interact with one another—and also uncovering how the variations in genes from species to species alter these details. It will be these discoveries that make possible the coming revolution in agriculture. "What we are really doing here [with an agricultural genome project], Copeland said, "is generating the basic infrastructure for the next millennium."
Organizing the Project
Any genome project will involve two very different types of science. Generating ESTs, creating maps, sequencing DNA—these are mostly repetitive procedures which involve doing the same thing over and over again hundreds, thousands, sometimes millions of times. To a large degree, they can be automated. On the other hand, working with an individual gene to understand its function and how that function is linked to the gene's structure is a different challenge with each new gene, demanding the attention of and intellectual contributions from individual researchers. These two types of science will be most effective under two very different types of organizational structure.
For the mapping and sequencing part of the project, the Arabidopsis genome project may be a good model. It is a very narrowly focused, carefully coordinated program funded by three agencies working in tandem: the National Science Foundation, the Department of Energy, and the Department of Agriculture. The project is being carried out at more than half a dozen laboratories organized into three
teams, and these teams are in turn coordinating their efforts with Arabidopsis genome programs in Europe and Japan.
This complex division of labor demands careful planning and constant communication, said Christopher Somerville, a co-investigator on one of the teams. "One of the successes of the Arabidopsis group," he said, "has been the development of coordinating committees at both the national scale and an international scale to make sure that we do not engage in any wasteful duplication of effort and to resolve issues that range all of the way from nomenclature to allocating work to the different groups, organizing meetings and [running] databases."
Neal Copeland of the National Cancer Institute, who has worked on the mouse genome, echoed the importance of such communication. "The genetic map [for the mouse] was done in five or six large mapping centers throughout the world. Early on, we recognized the need to interface all this data being generated by these different groups. So what we did was modeled after the Human Genome Project. We set up chromosome committees. In the mouse we now have a chromosome committee for each chromosome. We have an international meeting once a year, usually one year in Europe, one year in the United States, where everybody gets together, all the chromosome committees and anybody else who is interested, and they go through and they revise the maps, and then they get published. They used to get published in print form, but now basically they are being published in electronic form."
Of course, an agricultural genome project will be much broader and bigger than the Arabidopsis or the mouse program, so communication and coordination will be that much more important—and that much more difficult to maintain. Not only will researchers working on the same species need to keep in touch and integrate their efforts, but scientists studying different species, different orders, perhaps ever different phyla and different kingdoms, will also need to develop lines of communication and to harmonize their mapping techniques, database formats, and so on. A further complication arises with the need to coordinate research done in the private sector with that performed in the public sector (see Box 5: Public and Private Genomes).
Economies of scale also become more important with a large genome project, Copeland noted. In the mouse genome, for example, most of the microsatellites—a type of gene marker consisting of short sequences of base-pairs repeated many times—were mapped at one location, the MIT Genome Center. "I think we wouldn't be nearly as far along in the development of the microsatellite map had it not been done in one central facility," Copeland said. Similarly, ESTs for the mouse are generated at just one place, Washington University. "This is not a mouse group," he said, "but the reason they are doing it is, again, economy of scale. These people are experts in the field, and it wouldn't make any sense for a mouse lab that didn't have this technology to do this. So in thinking about doing EST maps for the agricultural species, it makes sense to be doing them in facilities that have the technology in hand and that can do them more cheaply."
|
BOX 5 Because of the importance of genome information, there is often a tension between the public and private sectors over access to it. Private companies, seeing competitive advantage in having genome data that their competitors do not, may seek to keep some of that data to themselves. On the other hand, scientists in universities and government labs often wish to get their hands on as much genome information as possible in order to maximize the quality of their research. Participants in the workshop described two cases in which this tension has been apparent, even in the early days of agricultural genome programs. In Europe, Christopher Somerville said, commercial considerations have delayed the dissemination of Arabidopsis genome data that was paid for partly with public funds. The seventeen labs in the consortium, he said, ''have completed about two megabases [two million base-pairs] of DNA but have imposed a long delay on the release of that sequence because the Europeans decided to organize their genomic sequencing in collaboration with industry.'' As part of the consortium agreement, industry partners were given the right to the first look at the Arabidopsis sequence as it was finished, Somerville explained, and "for whatever reason, that sequence has not yet been released." Meanwhile, in the United States, the vast majority of expressed sequence tags, or ESTs, for the corn genome are in private hands. According to Tony Cavalieri of Pioneer Hi-Bred International, his company has today about 80,000 ESTs from corn, representing an estimated 45,000 different genes. And although the company has shared some of that information with some public-sector scientists, he said, it has no plans to release the data and let its competitors take advantage of research paid for by Pioneer. "That information is important," he explained. "It allows you competitive advantage" by helping researchers find the entire gene and patent it. In general, the workshop participants were not happy with such situations. "I use a lot of Pioneer seed, and I am thankful for the work that Pioneer is doing in |
Such considerations as economies of scale and the desirability of concentrating expertise and the most advanced technology in a few places may argue for a very different type of organization than biologists are used to, said Bob Cook-Deegan from the Institute of Medicine. "I don't think that the [Human] Genome Project is the right model for the technologically intensive part [of an agricultural genome program]. The right policy model might actually be how the Department of Defense, particularly DARPA, has handled chip design, interactive computing, and information processing techniques. They channel a lot of resources into a few centers, rather than NSF's or NIH's more distributed system."
But whether the Human Genome Project or the highly concentrated programs that DARPA runs serve as the model, many participants believed that one tier of an agricultural genome project should be a directed, technologically intensive effort aimed at mapping and sequencing as much of as many different
|
genome work," said Wallie Hardie, National Corn Growers Association, "but I am really not very comfortable with just one company doing this work in genome mapping. I would like to see the power of this technology in the public sector so a lot of the companies could get these differentiated products to me quicker." Neal Copeland of the National Cancer Institute argued that basic genome information, such as ESTs, maps, and sequences, should be thought of as scientific tools and, as such, should be made widely available in order to keep science moving forward as quickly as possible. "These building blocks," he said, "are going to drive all of the science that goes on in the next decade or the next millennium, and this all needs to be in the public domain." Even Pioneer's Cavalieri agreed: ''Our preference would be that the EST information was available, that the tools for figuring out function were available, and that we would compete on the basis of product development using the genes that were known and available to the world at large." Pioneer decided to go after the ESTs itself, he said, only when it seemed that nothing was being done in the public sector and the company became worried that its competitors would move ahead of them in this area of basic science. Thus most researchers at the meeting called for a government project that would make agricultural genome information publicly available. When work is done in collaboration with industry, the basic information—ESTs, maps, and so on—should be put in the public domain, although companies may be given intellectual property rights to other discoveries, such as gene function.1 It may even be useful in some cases for the government to buy data that the public sector currently owns. ''One way of saving a lot of time and work and money," suggested Brian Larkins of the University of Arizona, "would be simply to purchase Pioneer's EST database and make it available." That might be possible, said Cavalieri of Pioneer. "I do think there would be some willingness to talk about what the private and public initiatives could be to solve some of these questions."
|
agricultural genomes as possible, as quickly as is feasible. Much of it should be carried out in places specializing in a particular technique or techniques, rather than having expertise and interest in a particular species. Large portions of the work will not be intellectually challenging, but they will be technically demanding.
The second tier will be quite different. The maps and sequences are of little use without details about what the various genes do and how they do it, and this information cannot be produced in the same assembly-line fashion that the maps and sequences can. Instead, the best approach is to fund individual investigators who are pursuing their own research interests.
To date, much of the individual-investigator funding for agricultural genome research has come from the USDA's National Research Initiative (NRI). Last year, said Ron Phillips, NRI's chief scientist, the National Research Initiative
awarded 72 grants totaling $9 million in the plant genome area. "Since [the plant genome program] started in 1991," he said, "there has been money put toward about 50 different species." NRI also funds research into animal genomes. But whether the support comes from NRI or as part of a broader agricultural genome project, workshop participants agreed that investigator-initiated research will be the most effective way to pursue such specialized investigations as how a gene's structure is related to its function.
Finally, many researchers at the workshop stated that things are changing so quickly in the area of genomic research that any genome plan should be flexible enough to take advantage of advances and new understandings as they come along. "I wonder if it wouldn't make sense to suggest that you emulate the five-year planning process that the Human Genome Project undertook in 1989, 1990," said Daniel Drell, a biologist with the Department of Energy. "It basically involves getting affected scientists, those who can contribute, to work up a draft five-year plan and then circulate it and see what reaction it elicits." And then, of course, Copeland said, keep in mind that the five-year plans will certainly be obsolete long before the end of five years. ''In going back and looking at the 1991, 1995 projections for what would happen in the mouse and the human genome projects, what we have learned is that all these projections have been way underestimated. We have gone much beyond where anybody could have even dreamed that we would be right now.''
