The second iteration of the IMAGE (Integrating Metabolism and Genomics) meeting was characterized by bridges and chasms: between new and old, between computational and experimental, between big and small, between young and old. The goal of the meeting was best described by a question from the audience to Linc Sonenshein: “How do we take that vast storehouse of metabolic intuition and knowledge and get it into a database that is accessible to everyone else?” A dozen years ago, that sort of transfer required either a long apprenticeship with a one-to-one transfer from mentor to disciple or a mind prone to fantasies about artificial intelligence. The IMAGE 2 meeting last May in Montreal, Canada, suggests that the goal may be achieved by the coming generation of scientists. Like the previous IMAGE meeting in 2004, the focus of the meeting was bacteria, but not to the exclusion of other systems.
INTUITIVE AND COUNTERINTUITIVE BEHAVIOR
The meeting opened with a discussion of intuitive and counterintuitive views of metabolism. The “contrarian” view of aerobic respiration presented by Robert Bender (University of Michigan) suggested that the role of the electron transport chain is to protect the interior of the cell from the effects of oxygen rather than to maximize the energy yield from glucose. After all, the cell's cytoplasm is a reducing environment, and many metabolic activities rely on ferrous iron. He also argued that the phosphotransferase system system is dangerous to cells because its product, glucose-6-phosphate, is toxic. Thus, the cell works hard to eliminate this compound by further metabolism to acetate, with concomitant production of ATP and proton motive force, both of which would be dangerous if overproduced. Escherichia coli avoids both of these problems (when a vitamin is provided within the environment) by oxidizing glucose extracellularly, by uncoupling oxidative phosphorylation, and by using acetyl phosphate as a signal to increase biosynthesis. In other words, metabolism is designed to optimize rather than maximize. John Roth (University of California at Davis) continued the counterintuitive view by asking why 13 gene products are needed to code for the two-step catabolism of ethanolamine to acetate in Salmonella enterica. Since the intermediate (acetaldehyde) is toxic, caging the acetaldehyde as it is made might help protect the cell. On the other hand, the volatility of acetaldehyde might be the underlying reason for the protein shells. However, still more puzzling is the apparent “futile cycle” by which the cell derives one ATP at the expense of two ATPs in the terminal catabolism of the compound. Biphasic growth (first on ethanolamine and then on the acetate produced) seemed the most logical explanation. The degradation of ethanolamine requires vitamin B12, so the total number of genes necessary to code for this process requires about 1% of the S. enterica genome. This and the fact that all salmonellae have this capacity argue that the process must be important in the lifestyle of salmonellae in general.
The arguments then became more intuitive as Christoff Schilling (Genomatica) described how modeling and experiments can be integrated. In the first example described, the model derived for a Geobacter species ultimately predicted not only the yields of products but also the labeling pattern of all the amino acids except one, isoleucine. A literature search revealed the existence of an alternative pathway known in a few organisms, but not found in the Geobacter genome as annotated. The model was hand curated to include a potential citromalate pathway to isoleucine, and indeed operation of the citromalate pathway (along with the traditional aspartate pathway) explained the labeling pattern rather well. A second example showed how modeling could operate to direct genetic engineering for product optimization. Using succinate production as a goal, the OptKnock program allowed a simple prediction of which genes could be eliminated in order to make growth dependent on succinate production. The resulting mutant strain produced more succinate than the wild type, but nowhere near the predicted maximum. The mutant was allowed to evolve (in an anaerobic chemostat) until it achieved a rapid growth rate. Since the knockouts were designed such that maximum growth rate would require a high yield of succinate, the evolutionary pressure on the cells led to changes that greatly increased the yield of succinate to levels very near that of the prediction. Thus, the meeting opened with demonstrations of the unpredictable and counterintuitive behaviors of simple metabolic paths, suggesting that there remain many new and exciting discoveries to be made through the use of the new genomic and modeling technologies.
Perhaps the most exciting example of the power of changing assumptions came from Jasper Rine's (University of California at Berkeley) search for human genes whose defects might be remediable by small molecules. A paradigm suggested by Bruce Ames argued that substrate excess could overcome an unfavorable Km if that were the cause of the defect. As proof of principle, methyltetrahydrofolate reductase (MTHF reductase) was examined. The known variant (A222V) and a dozen or so previously undiscovered natural human variants were tested in Saccharomyces cerevisiae (where the yeast gene was replaced by the wild-type gene or mutant alleles from humans). About half the variants (all in the catalytic domain) were remediable by excess MTHF. This was easily demonstrated by simple growth tests of the yeast. When other analogs of folate were tested, at least one, which could not be used as a cofactor in MTHF reductase, also remedied the growth defect associated with the mutant alleles. This strongly suggests that small molecules that can bind to the protein are able to promote the folding of the mutant protein, stabilizing the protein during or after its translation and allowing a native structure to develop. If so, then it should be possible, using high-throughput small-molecule screens, to look for “drugs” that might correct defects in human genes.
UNEXPECTED CONNECTIONS
The speakers in the session entitled “Metabolic Integration and Central Metabolism” provided a global perspective on metabolism. They emphasized the far-reaching consequences of altering one aspect of metabolism on everything else. It was also apparent from the presentations that in many cases these connections and interactions were unanticipated. All levels of technology were alluded to in the talks, emphasizing the synergy of approaches that must be used to dissect metabolic integration.
Diana Downs (University of Wisconsin) described classical genetic approaches that uncovered integration of pathways previously considered independent. If the normal pathway shared by thiamine and purine synthesis is disrupted, cells can find ways to reconfigure the metabolic network to provide thiamine. Links to tryptophan, histidine, and isoleucine biosynthetic pathways were uncovered, suggesting that single metabolic pathways can serve multiple functions. A similarly tangled web of interactions was described for the CodY protein of Bacillus. CodY is clearly important in regulating branched-chain amino acid biosynthesis and many degradative pathways normally induced under conditions of nutrient limitation. Linc Sonenshein (Tufts) proposed that it also cooperates with another regulator, CcpA, to control a large number of pathways involved in the generation and utilization of many central metabolites. CodY determines which of the many possible fates of pyruvate will be chosen. In addition, CodY also appears to be a critical regulator of virulence genes in several different pathogens. A complex set of networked interactions, involving the regulatory protein PII was also uncovered, but this time using high-resolution structural data. Mike Merrick (John Innes Center) proposed that PII interacts directly not only with the regulators of glutamine synthetase but also with the ammonia transporter, two nitrogenase regulators in gram-negative bacteria, and other nitrogen regulators in gram-positive species.
These experiments began with a specific, localized region within the metabolic network and led to an understanding that there are global effects that need to be explored. Three speakers showed the converse: that a global analysis can lead to an understanding that there are specific elements in the network that need to be explored. Stephane Aymerich (INRA) showed that traditional analysis of the regulation of the two glyceraldehyde-3-phosphate dehydrogenases of Bacillus subtilis (GapA and GapB, used for glycolysis and gluconeogenesis, respectively) identified the pleiotropic regulator CcpN. Deletion of CcpN leads to a drastic reduction in glycolytic growth. Flux analysis shows that the result is a drastic rerouting of carbon flow and a futile cycle. However, disruption of the futile cycle did not restore normal growth, so there must be more elements to central metabolism to be discovered. Modern genomic technologies have made it possible to study whole communities rather than pure axenic cultures. As Gene Tyson (Massachusetts Institute of Technology) showed, a metagenomic analysis of the DNA sequences found in an acid-mine drainage community showed many expected types of genes (for biofilm formation, oxygen stress response, heavy metal resistance, motility, chaperones, etc.), but it also can allow one to predict interactions among the members of the community, such as who fixes the carbon and who lives off those primary producers. However, it was Elizabeth Skovran (University of Washington) who proposed that, although each approach informs the other, both are needed to give a complete picture. Microarray analysis of the transcripts that are specific to the growth on methanol of Methylobacterium extorquens showed many expected and some unexpected mRNAs, some of which resemble those required for photosynthesis in this nonphotosynthetic organism. A massive genetic screen for mutants defective in growth on methanol yielded some of the same genes but also showed that the two analyses were not completely overlapping; some of the mutated genes did not show regulation in the microarrays, and some of the highly induced genes in the microarrays were not found in the mutational screen.
Altogether this session highlighted the broad and significant questions in metabolic integration that are being addressed with multiple combined approaches. The talks also served to emphasize the need that exists for increased communication between experimentalists and modelers to make sense of the vast networks that each of us seeks to explore in our individual labs. As was mentioned in comments and questions after this session and throughout the meeting, we are at the limit of being able to retain the networks in our minds. An introduction to one's work now requires not only a description of the components but also of the complete network, its characteristics, flux behavior, and regulation.
BRIDGES AND CHASMS
A session on metabolic engineering and computational approaches highlighted the need to bring distinct research communities together. In looking at nitrosative stress from a bioengineering perspective, J. Liao (University of California at Los Angeles) analyzed transcriptional profiles using network component analysis. This led him to suspect metal-containing proteins as key targets, especially the Fe-S-containing protein IlvD. This enzyme is remarkably sensitive to oxygen, consistent with O. R. Brown's demonstration many years ago that IlvD of E. coli is exquisitely sensitive to hyperbaric oxygen in vivo. This connection illustrates the ongoing need to mine the accumulated wisdom of the earlier microbial physiologists; many of those voices live on only in the memory of their students who are themselves nearing retirement. In a similar vein, a lively and instructive discussion accompanied Greg Stephanopolous' (Massachusetts Institute of Technology) observation that strong expression of mutated rpoD genes in the presence of a copy of rpoD+ led to heterozygotes with increased ethanol tolerance. The differing perspectives of the engineers, modelers, and physiologists led to a spirited discussion that captured the excitement and importance of the IMAGE meeting's success in building the bridges among the various communities. Nikolas Anesiades (University of Toronto) took a different approach to the problem of ethanol toxicity. Using in silico analysis, he designed a quorum-sensing switch that would turn off acetate production and turn on ethanol production. This separated the fermentation into two phases, biomass accumulation and ethanol production. Acetate secretion was well controlled by this toggle switch, and ethanol production was measurable. These experiments were meant as a proof of principle for the design and application of similar genetic “toggle switches” for use in engineered fermentations.
Three speakers in this session emphasized the importance of bridges between engineering, modeling, and physiological communities. Kai-Yu San (Rice) modeled anaerobic succinate formation and found it to be limited by NADH availability. Taking advantage of a long and productive collaboration with George Bennett, San designed and constructed mutant strains to reduce carbon flow to overflow products and increase carbon flow to oxaloacetate. In the end, they found that activation of the glyoxylate cycle, which requires less NADH, resulted in succinate production at a high rate and yield in their mutant strain. Flux analysis indicated that an optimal distribution of flux from pyruvate through either oxaloacetate or citrate was needed. Maxim Durot (CNRS) reported the construction of a complete set of knockout mutations of Acinetobacter baylyi, which was used to determine essential and nonessential genes. In a first pass, growth on succinate was tested and suggested that nearly 100 genes were either misannotated or mismodeled. Hypothesis generation and expert assessment corrected some of these, and a further iteration was conducted on a dozen different media, leading to further inconsistencies that will be reannotated and remodeled. The take-home lesson is that such studies can be automated to a large degree, making it possible to focus hand curation on the true inconsistencies.
With an exponentially increasing amount of data and number of sequenced genomes, the need to take complementary perspectives is even greater, as is the desire to check theory against experimental evidence. Claudia Reich (Argonne) presented a Web-based tool (SEED) that can organize a gene set based upon any reasonable biological grouping and can investigate that group throughout the sequenced microbial genomes. The system is organized computationally, but the subsystems are annotated by experts. Moreover, SEED provides a database of “open hypotheses” that need to be tested. Valerie de Crecy-Legard (University of Florida) has used SEED, as well as programs such as STRING and COG, in her attempt to identify “missing” genes required for the modification of RNAs, particularly tRNAs and rRNAs. By using these bioinformatics platforms, which incorporate information about phylogenetic occurrence, chromosome clustering, protein interactions, protein structure, and coexpression of genes, she and colleagues have been able to predict candidates for missing functions. These predictions were then tested and have led to the identification of nine new families of genes required for tRNA modification, including a new GTP cyclohydrolase I involved in tetrahydrofolate biosynthesis.
IRON
Iron plays an important role in a vast number of metabolic activities in the cell. Thus, an entire session of the meeting was devoted to the synthesis, repair, and significance of iron-sulfur clusters in proteins. Dennis Dean's (Virginia Tech) analysis of the Fe-S biogenesis pathways in Azotobacter vinelandii focused on the properties that provide substrate specificity of target proteins. A Nif-specific system functions in building Fe-S clusters for the nitrogenase Fe protein, whereas the Isc pathway functions as the housekeeping system for loading other cellular Fe-S proteins with Fe-S clusters. Dean and colleagues demonstrated that activation of the enzyme aconitase by cluster transfer from the scaffold protein IscU depends on the presence of the 4Fe-4S cluster in IscU. Results from Dean and his collaborator, Mike Johnson, indicate that the Azotobacter Nfu protein functions as an intermediate carrier in Fe-S biogenesis, efficiently shuttling assembled Fe-S clusters from the scaffold protein IscU to apoproteins. Frederic Barras (Laboratoire de Chimie Bacterienne) described another intermediate carrier function in E. coli, similar to the Nfu function described by Dean. He discussed the generic roles of the Suf and Isc Fe-S biogenesis pathways of E. coli and a third pathway characterized by CsdA, CsdE, and a protein encoded by ygdL. He described yet another Isc- and SufA-like protein, YadR. YadR has an essential function under aerobic conditions because of a defect in the synthesis of isoprenylpyrophosphate, a critical intermediate in the essential pathway of isoprenoid biosynthesis, which requires two Fe-S proteins, IspG and IspH. Thus, both specific and general pathways exist for iron sulfur cluster biogenesis, and the role of intermediate carriers is an area of active investigation. Patricia Kiley (University of Wisconsin) continued the discussion of Fe-S cluster biogenesis by focusing on the regulation of the pathways in E. coli. One major transcription factor, IscR, controls transcription of several genes that function in Fe-S biogenesis. IscR is itself a Fe-S protein that negatively regulates the major Isc pathway to maintain sufficient amounts of Fe-S synthesis. While the Fe-S form of IscR was required for repression of the isc operon, the apoform of IscR regulates the Suf pathway in addition to some anaerobic respiratory enzymes. Finally, aerobiosis was shown to upregulate the Fe-S biogenesis pathways, suggesting that these growth conditions increase the demand for Fe-S biogenesis.
Iron metabolism was also implicated in the mechanism of cell death that is mediated by the DNA gyrase inhibitors CcdB and quinoline. Daniel Dwyer (Boston University) used microarrays to show that treatment with these inhibitors led not only to induction of DNA repair genes as expected but also to induction of genes of the SoxS, Fur, and Fe-S synthesis regulons. Iron chelators reduce the formation of hydroxyl radicals under these conditions, suggesting that increased iron uptake contributes to cell death. Reducing Fe-S biogenesis under these conditions also reversed cell death, arguing that Fe-S proteins may be the cause.
Helicobacter pylori has a significant requirement for nickel in the ecological niche of the stomach because it produces hydrogenase and large amounts of urease, both of which are nickel-requiring enzymes. Hilde De Reuse (Institut Pasteur) proposed that the NikR (nickel) regulon was about as large as the Fur regulon in H. pylori and that many genes annotated as “iron genes” are actually regulated either solely by NikR or jointly by Fur and NikR. Nickel uptake required TonB and a novel outer membrane TonB receptor, FrpB4. These data and additional bioinformatics analysis suggest that TonB via interaction with a family of TonB receptors can transport many different small molecules in addition to Fe-containing siderophores and cobalamin.
INCREASING ROLES FOR SMALL RNAs IN METABOLIC REGULATION
In the past, much of metabolism and bioengineering has focused on the enzymatic and regulatory roles of proteins. However, in recent years, the importance of RNA in both enzymatic and regulatory roles has become increasingly evident. Talks by Eric Massé (University of Sherbrooke) and Terrance Hwa (University of California at San Diego) continued the discussion of iron by focusing on the Fur-dependent small RNA RyhB. Massé showed that the accumulation of 18 transcripts was reduced by RyhB and at least one, encoding shiA, a shikimate permease, was increased. Base pairing of RyhB RNA with shiA mRNA appears to stabilize the mRNA and disrupt a structure that is inhibitory to translation. This allows the cell to accumulate shikimate, a precursor of a siderophore needed for growth under low iron conditions. Hwa used regulation of sodB by RyhB to propose that regulation by the small RNA results in a threshold response rather than the linear response seen with protein regulators. By varying the rate of expression of ryhB or the target gene sodB, the transcriptional rate of the small RNA relative to the target mRNA seemed to be a more critical parameter than the absolute levels of the RNAs, hence the threshold response. Cross talk between target mRNAs was also proposed by suggesting that increased expression of sodB competes with the effect of RyhB on other target gene mRNAs. These studies also identified some features of small RNA regulation that are favorable for the design of synthetic gene circuits, such as noise resistance and fast recovery from removal of the regulatory RNA.
Gisela Storz (NIH) continued the discussion of how small RNAs regulate gene action through limited base pairing interactions with a target mRNA, through extensive base pairing interactions with target mRNAs, and by binding to and modulating the activity of proteins. She described several small RNAs that control the expression of proteins that appear to be toxic at high levels. One example is the SymE protein, which has properties of the MazF family of toxins, but has homology to the MazE family of antitoxins. SymE levels are kept low by the LexA transcriptional repressor, the SymR antisense RNA, and the Lon protease. Another example is the 39-amino-acid TisB protein whose synthesis also is controlled by the LexA repressor as well as by the divergently encoded IstR RNA. Why so many toxins are encoded by the E. coli genome, whether they have roles in normal cellular metabolism, and why the expression of the toxins is controlled by small RNA regulators remain interesting questions for the future.
As mentioned in the opening talk, all sugar phosphates are toxic. Carin Vanderpool (University of Illinois) showed that the SgrS RNA is required for survival under sugar phosphate stress. SgrS expression leads to translational repression and degradation of the ptsG message. PtsG is the glucose-specific sugar transporter whose activity results in accumulation of intracellular glucose phosphate. When glucose-6-phosphate levels increase, the SgrR regulatory protein leads to an increase in SgrS accumulation and the downregulation of PtsG, the sugar transporter and phosphorylator, allowing homeostasis to be achieved. The exact signal sensed by SgrS (and hence the central regulatory mechanism) remains unknown.
RIBOSWITCHES
A discussion of riboswitches, the 5′ leader sequences in nascent RNAs that directly bind and sense effector molecules, began with Tina Henkin's (Ohio State University) description of the general features of the growing number of transcripts found to be regulated in this manner. They all have long leader sequences, are usually regulated as a group, and tend to have a conserved RNA structure. Henkin described two classes of riboswitches in more detail: T-box riboswitches, which monitor the charged/uncharged ratio of a specific tRNA, and S-box riboswitches, which respond to S-adenosylmethionine (SAM). T-box-containing RNAs are abundant in gram-positive bacteria, and a few are also found in gram-negative organisms. Henkin and colleagues were able to characterize the T box at the 5′ end of the B. subtilis glyQS mRNA in vitro and mapped the contacts between the specific tRNAGly and the glyQS mRNA as well as the structural changes induced by the interaction between the RNAs. The tRNA serves as a cross strut using its anticodon and free acceptor ends to interact with the transcript. The specific nature of these interactions can be used to predict the amino acid specificity of gene products encoded by mRNAs preceded by T-box riboswitches. In some organisms, the genes involved in methionine biosynthesis did not have T-box sequences but rather had other conserved leader sequences that corresponded to S-box sequences. In general, induction of S-box-regulated genes correlates with the intracellular levels of SAM, but studies to determine whether there are differences in how individual S boxes respond to SAM are ongoing. Digby Warner (University of Witwatersrand) described a riboswitch upstream of the metE gene of Mycobacterium tuberculosis. The identification of this riboswitch was the outcome of studies of the two methionine synthases in M. tuberculosis, a vitamin B12-dependent enzyme encoded by metH and a B12-independent enzyme encoded by metE. The findings that the growth of a clinical isolate (M. tuberculosis CDC1551) lacking metH is inhibited on solid medium supplemented with vitamin B12, together with the isolation of B12 suppressor mutants that map to the 5′ leader of the metE transcript indicate that transcription of the metE gene is controlled by a B12-controlled riboswitch.
MACROMOLECULAR MACHINES: PROTEASOMES AND POLYMERASES
An unexpected link between central metabolism and proteolysis in E. coli was revealed by Eliora Ron (Tel Aviv University), who found that knocking out the ability to make acetyl phosphate results in a reduced rate of ATP-dependent proteolysis, especially by Lon, a major cellular protease responsible for degrading unstable polypeptides, and activates the heat shock response. Whether acetyl phosphate phosphorylates a factor that influences Lon activity or provides a high-energy phosphate needed for proteolysis remains to be determined, but as was suggested in the opening talk, acetyl phosphate seems to play a central role in many regulatory phenomena.
The roles of the pleiotropic metabolite ppGpp in transcriptional specificity and chromosome replication were explained by Rick Gourse (University of Wisconsin) and Jade Wang (Baylor College of Medicine), respectively. Gourse showed that ppGpp interacts directly with E. coli RNA polymerase, decreasing the half-life of RNA polymerase-promoter complexes. Promoters with short half-lives, such as those for rRNA genes, are particularly sensitive to ppGpp. This effect of ppGpp is greatly enhanced by DksA, a protein that binds to RNA polymerase. Thus, the mutually interacting effects of ppGpp, the concentration of initiating nucleotide, and the DksA protein all serve to coordinate transcription of growth rate-dependent promoters in E. coli. Wang proposed that ppGpp mediates a protective arrest in replication elongation by interacting with DNA primase directly when B. subtilis cells are nutritionally limited. She also proposed that B. subtilis has evolved to have 75% of its transcription units in the same orientation as the movement of replication forks to limit potential head butting of DNA and RNA polymerases. When a large segment of the chromosome was inverted, replication of that segment slowed down considerably.
Wayne Nicholson (University of Florida) reported that some mutations to rifampin resistance in rpoB, the gene for the beta subunit of RNA polymerase, have unexpected consequences in B. subtilis in addition to the previously seen defects in sporulation. Surprisingly, utilization of multiple carbon sources is strongly affected, including novel acquisition of the ability to utilize some carbon sources and to produce fermentation products that are not typical of wild-type cells.
The function of Sda, a B. subtilis protein that prevents sporulation in cells that have damaged or incompletely replicated chromosomes, was proposed by Bill Burkholder (Stanford University). He has been able to dissociate the signals that induce initiation of sporulation (nutritional limitation and population crowding) from those that activate Sda, suggesting that Sda intervenes in response to several kinds of replication stress and acts in growing cells as well as in cells that are initiating sporulation. Stephanie Gon (CNRS, Marseille, France) described how an exploration of intracellular mechanisms of sulfhydryl group reduction in E. coli led her to conclude that the ATP-bound form of the replication initiation protein DnaA is a repressor of nrdAB, the genes that encode the cell's primary ribonucleotide reductase. Since DnaN, the beta-clamp protein of DNA polymerase III, stimulates conversion of ATP-DnaA to ADP-DnaA when DnaN joins the replication complex, DnaN can be considered an inducer of nrdAB expression, whose activity causes a burst of deoxynucleoside triphosphate (dNTP) synthesis. A danger of overexpression of ribonucleotide reductase is an increase in the spontaneous mutation rate. NrdAB usually guards against such overproduction by its responsiveness to feedback inhibition by dNTPs, but a second ribonucleotide reductase (NrdEF) that is induced by DNA damage lacks the feedback control, leading to higher dNTP pools and higher rates of mutagenesis concomitant with DNA repair.
It is becoming increasingly clear that the multiprotein machines that produce and degrade the cell's macromolecules have a more intimate relationship with and response to metabolic signals than had been previously recognized.
INTEGRATING METABOLISM AND GENOMICS: THE CHALLENGES OF SCALE AND SCALABILITY
As the -omics era continues to explode, our ability to generate data is far surpassing our ability to integrate those data either computationally or intuitively. Michael Ellison (University of Alberta) has approached this issue by modeling the cell as a collection of discrete automata, where each molecule is a particle that follows simple rules and can be tracked in four dimensions. Although this allows an almost seamless transition from stochastic to continuous, the challenges of scalability intrude. Biological time ranges from picoseconds to hours or more; size ranges from nanometers to meters. His current models deal with cells of about a million molecules with times up to about 100 milliseconds. However, his model is scalable and limited only by computing power of the CPU. The challenge is to build a friendly interface where “small science” of the bench can inform the discrete elements of the model and vice versa.
Alexei Savchenko (University of Toronto) pointed out that about half of all sequenced genes are of unknown (or poorly characterized) function. He described an effort to purify milligram quantities of thousands of proteins and to derive structures where possible and to perform high-throughput assays for enzymatic activities using relaxed, general substrates on the first pass and then more specific ones when possible. So far, about a quarter of the purified proteins have yielded at least a general description of the activity. The same issue was raised by Eran Pichersky (University of Michigan) who used the vast diversity of secondary metabolism in plants to argue that general characteristics and computer predictions are insufficient to describe the reality of the metabolic potential of any given plant or pathway. In the end, biochemical characterization will always be essential to understand the individual organism. Thus, both small and large science will always be necessary complements to each other. Barry Wanner (Purdue University) described the work of his colleague Hirotada Mori (Nara Institute of Science and Technology) on developing an integrated database to describe central metabolism in E. coli. First, Mori's group purified all the enzymes of glycolysis and the tricarboxylic acid cycle, measured Km and Vmax, and raised antibodies against them. They then grew batch cultures of E. coli and sampled at 10-minute intervals during growth and at 1-hour intervals during stationary phase. From each sample, they determined the mRNA, the protein, and the metabolites of these pathways. This takes advantage of the various collections of tagged genes that Mori and his collaborators have developed and will provide an integrated picture of the progress of the cells through their growth phases.
Bob LaRossa (DuPont) undertook a combined genomic and genetic approach to Zymomonas mobilis metabolism, an organism best known for its production of ethanol via the Entner-Doudoroff pathway. Surprisingly, the amino acid biosynthetic genes seem to be unregulated by nutritional status, so he undertook a large-scale transposon screen and retrieved very few auxotrophs, consistent with the paucity of transporters in the genome of Z. mobilis. Those few transporters identified in the annotated genome corresponded well with the auxotrophs identified in the genetic screen.
Paul Straight (Harvard) moved the discussion away from pure cultures with an observation that a single colony of B. subtilis creates a bald, colorless zone in a lawn of Streptomyces coelicolor. Mutagenesis of the B. subtilis yielded a mutant that made a red zone instead of a colorless one and identified a large gene cluster (2% of the genome!) encoding a previously unknown secreted product, bacillaene. The interactions of this compound, several other compounds secreted by the B. subtilis, and the metabolites of S. coelicolor show clearly that microbes not only sense metabolites in their environment but also contribute to them in highly interactive ways.
The goal of the IMAGE 2 meeting was and remains to promote integration of metabolism, genomics, technological advances, and modeling. There has been much progress toward that goal since IMAGE 1, and it is now clear that the pace toward that goal will continue to accelerate as we get ready for IMAGE 3 in a few years. Chasms remain, but the bridges are being built. We may still have to shout to (at?) each other across those bridges, but we are now learning to speak the same language, and the cross-cultural fertilization is moving bacterial metabolism forward toward a level of understanding that was not even dreamed of a scant few decades ago.
Footnotes
Published ahead of print on 7 September 2007.