Journal of Bacteriology, March 2004, p . 1448-1461, Vol . 186, No . 5

Transcriptional Profiling of Caulobacter crescentus during Growth on Complex and Minimal Media

Alison K . Hottes,^1,2 Maliwan Meewan,² Desiree Yang,^3, Naomi Arana,^3, Pedro Romero,⁴ Harley H . McAdams,² and Craig Stephens^3*

Departments of Electrical Engineering,¹ Developmental Biology, Stanford University, Stanford, California 94305,² Biology Department, Santa Clara University, Santa Clara, California 95053,³ Bioinformatics Research Group, SRI International, Menlo Park, California 94025⁴

Received 14 August 2003/ Accepted 12 November 2003

 
Microarray analysis was used to examine gene expression in the freshwater oligotrophic bacterium Caulobacter crescentus during growth on three standard laboratory media, including peptone-yeast extract medium [PYE] and minimal salts medium with glucose or xylose as the carbon source . Nearly 400 genes [approximately10% of the genome] varied significantly in expression betweenat least two of these media . The differentially expressed genesincluded many encoding transport systems, most notably diverseTonB-dependent outer membrane channels of unknown substratespecificity . Amino acid degradation pathways constituted thelargest class of genes induced in PYE . In contrast, many ofthe genes upregulated in minimal media encoded enzymes for synthesisof amino acids, including incorporation of ammonia and sulfateinto glutamate and cysteine . Glucose availability induced expressionof genes encoding enzymes of the Entner-Doudoroff pathway, whichwas demonstrated here through mutational analysis to be essentialin C . crescentus for growth on glucose . Xylose induced expressionof genes encoding several hydrolytic exoenzymes as well as anoperon that may encode a novel pathway for xylose catabolism.A conserved DNA motif upstream of many xylose-induced geneswas identified and shown to confer xylose-specific expression.Xylose is an abundant component of xylan in plant cell walls,and the microarray data suggest that in addition to servingas a carbon source for growth of C . crescentus, this pentosemay be interpreted as a signal to produce enzymes associatedwith plant polymer degradation.

 
Caulobacter species are ubiquitous inhabitants of freshwater, marine, and subsurface environments and are sufficiently adaptableto low nutrient conditions that tapwater and even distilledwater are ready sources for their isolation [19, 48, 49] . Thephysiological properties that enable Caulobacter and other oligotrophicspecies to survive and reproduce in such environments are notwell understood . In this work, we used whole-genome transcriptionalprofiling and genetic analysis to characterize metabolic pathwaysand transport systems that are differentially expressed by Caulobactercrescentus during growth in three standard laboratory media:a defined "minimal" inorganic salts medium supplemented witheither glucose or xylose as the sole carbon source and a complexpeptone-yeast extract medium [PYE], in which amino acids serveas the primary carbon source . These observations provide importantbaseline data on the genetic coordination of metabolism in thismodel organism.

The dimorphic life cycle of C . crescentus has served as a model for understanding prokaryotic development, cell cycle regulation, and asymmetric cell division [24] . At each division, C . crescentusdivides into two morphologically and physiologically distinctcells: motile swarmer progeny with a single polar flagellum,and sessile stalked progeny displaying an adhesive stalk . Theswarmer progeny, in which DNA replication is initially inhibited,is specialized for dispersal and can seek new nutrient sourcesby chemotactic sensing . After about a third of their cell cycle,swarmer progeny cells eject their flagellum and grow a stalk from the same pole . The stalk enhances nutrient acquisitionby increasing the effective membrane area available for transport systems without substantially increasing the cytoplasmic volume[20] . Stalked cells initiate chromosome replication, grow anew flagellum, and divide, yielding new swarmer and stalkedcells.

Genome-scale transcriptional profiling has shown that the expression of over 500 C . crescentus genes varies significantly over the course of the cell cycle [32] . Among these are many genes involvedin energy metabolism, nucleotide and amino acid biosynthesis,and protein synthesis . As shown here, a substantial fractionof these genes are also regulated in response to nutrients, raising intriguing questions as to how regulatory networks controlling metabolism and cell cycle progression might interact.

Caulobacter is obligately aerobic and heterotrophic and can utilize a variety of carbon sources . Previous studies have demonstrated the activity of Entner-Doudoroff pathway enzymes [14] duringCaulobacter growth on glucose [48, 52], but there has been noquantitative analysis of glucose flow into the Entner-Doudoroffpathway versus alternatives such as the Embden-Meyerhof-Parnasglycolytic pathway . Related -proteobacteria, including Rhizobiumand Agrobacterium strains, use both the Entner-Doudoroff andEmbden-Meyerhof-Parnas pathways, sometimes simultaneously [3,42] . The C . crescentus genome, however, lacks a clear homologof phosphofructokinase, a key enzyme of the Embden-Meyerhof-Parnas pathway [45] . Without this enzyme, any glucose-6-phosphate producedcould serve as a substrate for glucose-6-phosphate dehydrogenase,directing carbon from glucose into the Entner-Doudoroff pathway.We show here that the Entner-Doudoroff pathway is essentialin C . crescentus for growth with glucose as the sole carbonsource.

Genome sequence analysis unexpectedly revealed that C . crescentus has the potential to degrade plant-derived biopolymers through the production of exoenzymes, including cellulases [endo-ß-1,4-glucanases and ß-glucosidases], xylanases and xylosidases, and polysaccharide deacetylases [45] . In freshwater ponds, lakes, and rivers, Caulobacter species could use live or decaying aquatic vegetation as surfaces for biofilm formation . The major monomeric products released by degradative exoenzymes acting on lignocellulose would be glucose [from cellulose] and xylose [from xylan] . The mechanism of xylose dissimilation in C . crescentus is unknown. The best-known bacterial pathway for xylose catabolism, isomerization to xylulose followed by phosphorylation to xylulose-5-phosphateand entry into the pentose phosphate pathway [23], is probably not operative in C . crescentus, given the absence of homologs for xylose isomerase and xylulose kinase in the genome . An inducible NADP-dependent xylose dehydrogenase activity has been reported in C . crescentus [48], but neither the gene encoding this activitynor the catabolic pathway in which it might participate is known.We identify here genes whose expression increases in the presenceof xylose . These data suggest that C . crescentus uses xyloseas an indicator of the availability of plant-derived polymers. We also identify a DNA sequence motif associated with many xylose-induciblegenes and show that the motif is important for regulation.

The abundance of putative secreted peptidases in the C . crescentus genome suggests that extracellular degradation of proteins may also be important in natural habitats [45] . Environmental aminoacids support rapid growth of C . crescentus, leading us to comparegene expression in a complex amino acid-containing medium [PYE]to gene expression in defined minimal media . As expected, severalamino acid degradation pathways are upregulated during growth in PYE, in contrast to induction of amino acid biosynthetic pathways [along with ammonia and sulfate assimilation systems]in minimal media . Interestingly, however, expression of manyother genes encoding enzymes of key biosynthetic pathways doesnot change significantly between PYE and minimal media.

As a gram-negative bacterium, C . crescentus must acquire nutrients via transport across both outer and inner membranes, and high-affinity nutrient uptake is presumably necessary to support an oligotrophic lifestyle . The C . crescentus genome seems to lack homologs of the OmpF-type outer membrane porins characteristic of many gram-negative species [45] but is remarkably rich in TonB-dependent receptors,with at least 65 gene products containing the TonB box . As analternative transport system, TonB-dependent receptors are thought to harness energy from the inner membrane proton motive force to the transport of macromolecular complexes, including vitaminB₁₂ and siderophores, across the outer membrane [28, 41] . Withrespect to transport across the inner membrane, over 80 ABC[ATP binding cassette] transporter components [61] are encodedin the Caulobacter genome . We show here that numerous TonB-dependentreceptors and some ABC transporters are differentially expressedin each medium . Whether the deployment of such transportersis a key factor for Caulobacter and other oligotrophs to efficientlyacquire nutrients in diverse nutrient-limited environments clearlywarrants further investigation.

 
Bacterial strains, media, and growth. Cultures were grown in PYE [peptone-yeast extract] broth containing0.2% Bacto Peptone [Difco] [0.1% yeast extract [Difco], 1 mMMgSO₄, and 0.5 mM CaCl₂] in M2 minimal salts medium [6.1 mMNa₂HPO₄, 3.9 mM KH₂PO_4, 9.3 mM NH₄Cl, 0.5 mM MgSO₄, 10 µMFeSO₄ [EDTA chelate; Sigma Chemical Co.], 0.5 mM CaCl₂] with0.2% glucose as the sole carbon source [referred to subsequentlyas M2G] or in M2 medium with 0.2% xylose as the sole carbonsource [referred to subsequently as M2X] [12] . Liquid cultureswere inoculated with C . crescentus colonies randomly pickedoff PYE agar plates [12] . C . crescentus strain CB15N [NA1000]was used for microarray studies . The mutations and transcriptionalfusions generated in this study were analyzed in the CB15N strainbackground . hex mutant strains [provided by Bert Ely, Universityof South Carolina] were derived from the CB15 strain background.Cultures from which RNA was extracted were grown in 125-ml flasksand shaken continuously at 275 rpm and incubated at 30°C.

RNA isolation. For the microarrays, cells were harvested from a total of eightPYE cultures, eight M2G cultures, and five M2X cultures . Thecultures were in exponential growth at the time of harvest,with mean densities as follows: PYE, OD₆₀₀ = 0.37 ± 0.12;M2G, OD₆₀₀ = 0.34 ± 0.13; and M2X, OD₆₀₀ = 0.27 ±0.05 . [The cultures were not synchronized, and thus containedcells from all stages of the cell cycle.] Cells were isolatedby centrifugation in a Beckman J2-21 M centrifuge at 7,740 xg at room temperature for 2 min . [In control experiments, approximately85% of the cells were recovered in the pellet, and all stagesof the cell cycle seemed to be recovered with similar efficiencybased on phase contrast microscopy.]

The complete RNA extraction and purification protocol is described in the supplementary material [http://www.stanford.edu/group/caulobacter/metabolism] and described briefly here . Each cell pellet was resuspendedin Trizol [Invitrogen] and incubated at 65°C for 10 min.Chloroform was added, and the samples were centrifuged in amicrocentrifuge at 16,000 x g at 4°C for 15 min . The aqueouslayer was transferred to a new tube, mixed with isopropanol, and frozen overnight at -80°C . The next day, RNA was pelleted by centrifugation at 16,000 x g at 4°C for 30 min . Supernatantwas poured off, and the pellets were washed in 70% ethanol andresuspended in diethyl pyrocarbonate-treated water . ContaminatingDNA was digested with DNase I [Ambion] . RNA was extracted withacid phenol-chloroform, precipitated with ethanol and sodiumacetate, and then resuspended in RNase-free water . RNA qualitywas visualized on agarose gels, and RNA concentration was determinedspectrophotometrically.

PCR microarrays. PCR products were spotted onto poly-L-lysine-coated glass microarrayslides with an OmniGrid printer [GeneMachines] with 32 Telechempins . The slides were postprocessed as described by Laub etal . [32] . Primer sequences are available in the supplementarymaterial and correspond largely to those used in Laub et al.[31] . Slides from three printings were used.

Labeled indodicarbocyanine-dCTP [Cy3] and indocarbocyanine-dCTP [Cy5] cDNA probes were generated from 20 µg of total RNAby reverse transcription as described by Laub et al . [32], except that the amount of random hexamers was increased from 500 ngto 1 µg . Samples were hybridized competitively under coverslipsto the microarray slides overnight at 65°C, then washedas described by Laub et al . [32] . Hybridized arrays were scannedwith a GenePix 4000B scanner [Axon Instruments], and scannedspots were converted to ratios [red/green] with GenePix Pro4.0 software . Ratios and statistics on each spot were storedin an SQL Server database . The signal for each channel [redand green] was taken to be the mean of the foreground signalminus the median of the local background signal . Some spotswere excluded by the criteria described in the online supplementarymaterial . Array-wide normalization procedures are also describedin the supplementary material.

Identifying genes differentially expressed between pairs of media. Thirty-four microarrays were used to assay relative RNA abundance. Eighteen compared M2G and PYE, nine compared PYE and M2X, andseven compared M2X and M2G . To reduce dye bias, samples fromeach medium were labeled with Cy3 about half the time and withCy5 about half of the time . We used least-squares techniques[59] to simultaneously estimate log₂[M2G/M2X], log₂[M2X/PYE], and log₂[M2G/PYE] expression ratios for each gene . We used a correlated error model to account for dependencies introduced by hybridizing some RNA samples on multiple arrays . The varianceof each estimate was modeled by assuming a normal error distribution.

For each gene, Student's t test was used to test the hypothesis that each log ratio, i.e., log₂[M2G/M2X], log₂[M2G/PYE], andlog₂[M2X/PYE], was zero . A two-sided P value, which accountedfor the different number of samples available for each gene,was assigned to each ratio . For most genes, more arrays comparingM2G and PYE were done than were done for any other combinationof media . Thus, the confidence regions for log₂[M2G/PYE] tendedto be smaller than those for log₂[M2G/M2X] or for log₂[M2X/PYE].To roughly equalize the number of false-negatives, significancelevels were set so that, under the noise assumptions, 10 false-positiveswere expected for each of the log₂[M2G/M2X] and the log₂[M2X/PYE]comparisons and five false-positives were expected for the log₂[M2G/PYE] comparisons . To minimize the effect of possible biases, to be considered significant, gene expression also needed to changeat least 1.4-fold . Additionally, average changes of at leastthreefold were considered significant . Applying the above proceduresto a randomized version of the data set yielded 18 false-positives[all comparisons combined] . For more detail on the analysisprocedure, see the supplementary material.

To aid in identifying patterns of coregulation, differentially expressed genes were sorted into several groups . For example,genes were assigned to the "up-in- PYE" set [equivalent to the"down-in-M2" set] if the increase in expression between PYEand both M2G and M2X was statistically significant or if theincrease in expression between PYE and at least one of the M2media was statistically significant, while the difference inexpression between PYE and the other M2 medium was at least1.75-fold . The down-in-PYE/up-in-M2, up-in- M2G, down-in-M2G,up-in-M2X, and down-in-M2X sets were constructed analogously.

Metabolic pathway analysis and identification of differentially regulated potential isozymes. Functional analysis of microarray data was aided by CauloCyc[http://biocyc.org], a computationally derived pathway/genomedatabase of C . crescentus . A pathway/genome database describesthe genome of an organism, the product of each gene, the biochemicalreaction[s], if any, catalyzed by each gene product, the substratesof each reaction, and the organization of those reactions intopathways . CauloCyc was computationally derived from the annotatedgenome of C . crescentus by the PathoLogic software, part ofthe Pathway Tools software suite [26, 54].

Each metabolic pathway in CauloCyc consists of a set of enzymatic functions [biochemical reactions] and the gene[s] that codefor each function . When CauloCyc contained multiple genes fora given reaction, microarray data were used to construct 95%confidence intervals for each gene for each pairwise comparisonwith the noise model described above . This allowed pairs ofgenes whose products are annotated as catalyzing the same reaction,but which for one or more pairwise comparisons averaged at leasta twofold difference in expression and had disjoint confidenceintervals to be identified . Typically, we required at leasttwo comparisons to yield distinct confidence intervals . Forexample, in a case where one gene in the pair was up in PYEand the other was unchanged, the confidence intervals for bothPYE/M2G and PYE/M2X would need to be distinct . Also, if thePYE/M2G and PYE/M2X ratios for the first gene were 0.9 and 1.2,respectively, then the average PYE/M2G and PYE/M2X ratios forthe second gene would need to be at least 1.8 and 2.4, respectively.

Gene annotations and categorization. Most C . crescentus open reading frame [ORF] annotations weretaken from GenBank accession number AE005673 [45] . For genesannotated in GenBank as hypothetical or conserved hypotheticalproteins, clusters of orthologous gene [COG] descriptions [57, 58] were used if they were more detailed . Gene names shown in the tables came from TIGR and generally refer to the homologous Escherichia coli gene [see www.tigr.org for more information].Annotations for CC2086 and CC1000 were obtained on the strengthof BLAST matches to the NCBI NR database [2] . BLAST was alsoused to identify homology between C . crescentus gene productsand the gene products of other organisms . Classifications weretaken from COG classifications when available; genes withoutCOG classifications were categorized based on their GenBankannotations.

In order to create disjoint groups, genes assigned to multipleCOG categories were placed into a single category [see supplementary materials] . To facilitate comparisons, gene products involvedin transport [based on either their GenBank or their COG annotation] were grouped together . Some genes coding for amino acid metabolism proteins were further classified as biosynthetic or degradativebased on the metabolic pathways in CauloCyc . Nontransport genesresponsible for assimilating sulfate into cysteine were classifiedin the amino acid metabolism and biosynthesis groups . Uncharacterizedgenes annotated simply as hypothetical proteins which did nothave a COG number were placed in the hypothetical category;other uncharacterized genes were placed in the conserved hypothetical category.

Mapping of hex mutants. C . crescentus mutants unable to use glucose as their sole carbonsource [hex phenotype] were isolated by Ely and colleagues [13, 25] . Mutations in eight such strains were mapped by Cr30 phage-mediatedtransduction, with a set of kanamycin-resistant marker strains[60] . Mutations were further localized to single genes by complementationanalysis . Candidate sets of genes [e.g., those expected to beinvolved in sugar metabolism] in the indicated genomic regionswere amplified by PCR with AccuTaq Long-read Taq polymerase[Sigma-Aldrich] and the supplied reaction buffer plus 2% dimethylsulfoxide . The products were cloned into plasmid pCR2.1 withthe Topo Cloning System [Invitrogen], then subcloned into broad-host-rangeplasmid pMR20 in the E . coli S17.1 host . These plasmids wereintroduced into the hex mutant strains by conjugation, withselection on PYE containing tetracycline and nalidixic acid.Complementation of the hex phenotype was assessed by testingfor colony formation on M2G agar plates, in which glucose servesas the sole carbon source.

Construction of mutant strains. A nonpolar deletion removing nearly half of the coding regionof CC2054, encoding glucokinase, was generated through a PCR-basedstrategy . Flanking primers were designed that annealed approximately300 bp upstream and downstream of the CC2054 coding region.Complementary chimeric internal primers were designed that fusedsequences from the 5' and 3' portions of the coding region,deleting codons 114 through 250 [out of 332 total] but maintaininga continuous reading frame . [The sequences of all primers usedin this work are available on request.] Each of the chimeric internal primers was used in individual PCRs with the appropriate flanking primer . The products were isolated and combined ina subsequent PCR, with only the flanking primers for amplification.

Fusion of the initial PCR products in the second PCR was mediated by complementarity of the terminal chimeric regions, resultingin a single final product with much of the CC2054 coding regionremoved . This construct was cloned into plasmid pCR2.1 withthe Topo Cloning System [Invitrogen], then subcloned into thekanamycin-resistant pNPTS138 vector for integration into theC . crescentus genome and sacB-mediated sucrose counterselection[60] . Sucrose-resistant, kanamycin-sensitive isolates lackingthe integrated plasmid were screened by colony PCR to determinewhether the wild-type or mutant gene remained on the chromosome.Deletion mutants were streaked on M2G, M2X, and PYE to determinewhether CC2054 is required for utilization of glucose as thesole carbon source.

lacZ reporter fusions to confirm microarray data. The promoter regions of several genes were fused to a lacZ reporter to provide an alternative method of assessing regulation of expression of these promoters during growth on various media.The upstream regions of the CC0505, CC0823, and CC2057 geneswere amplified by PCR, generally including approximately 300bases upstream of the start codon for the indicated genes andup to several hundred bases downstream into the coding regions.PCR products were cloned into plasmid pCR2.1 with the Topo CloningSystem [Invitrogen], then subcloned as EcoRI fragments intothe pRKlac290 vector to generate a transcriptional fusion tothe lacZ gene [56] . Orientation in the pRKlac290 vector wasverified by PCR with an insert-specific primer and a primercomplementary to the coding sequence of the lacZ gene . pRK290constructs containing the various promoters were delivered byconjugation from E . coli S17.1 into C . crescentus, with selectionon PYE containing tetracycline and nalidixic acid . ß-Galactosidaseactivity was assayed in log-phase cells as described previously[56].

Motif identification. The MEME software [4] was used to identify matrix models describingDNA sequence motifs present upstream of genes whose expressionis elevated in M2X or [separately] in M2G . The 475 bases upstreamand the 125 bases downstream [600 bases total] of the startof each predicted C . crescentus open reading frame were searched.Bases interior to the predicted coding region were includedin the search set to provide robustness against possibly misannotatedstart codons . Each possible motif location was scored with themotif matrix as well as with a third-order Markov model of C.crescentus intergenic regions [35] . A location's score was thelog of the probability that the motif was present minus thelog of the probability that the sequence came from the background.Both strands of each sequence were searched . The cutoff scorefor the presence of the motif was taken as the score of thelowest scoring sequence used in generating the motif, exceptfor glucose motif 2, where the second lowest score was used.

In Table 4, when overlapping instances of the xylose motif appearedon both DNA strands, only the highest scoring instance is shown.Both as a control and to estimate the number of false-positives,an equal number of 600-bp sequences were generated from thebackground distribution model and scored . One random sequencescored above the cutoff for the xylose motif, three were abovethe cutoff for glucose motif 1, and 22 were above the cutoff for glucose motif 2 . Matrices describing the motifs are available in the supplementary material.

 
When searching for the xylose motif upstream of Xanthomonas campestris and Xanthomonas citri genes, background models of X . campestris and X . citri intergenic regions were used . Genepredictions and coordinates for X . campestris and X . citri camefrom GenBank accession numbers AE008922 and AE008923, respectively[10] . Genome sequences were downloaded from GenBank [ftp://ftp.ncbi.nlm.nih.gov/] and processed locally with the Genome-tools software suite [33]. Genes were predicted to be in the same transcription unit [operon] if they are next to each other on the same chromosome strand and if their intergenic spacing [predicted stop codon to predicted start codon] is less than or equal to 70 bp.

To test the functionality of the motif associated with xylose-regulated genes, the sequences upstream of the CC0823 and CC0505 genes were altered by site-directed mutagenesis [60] . The templates for mutagenesis were the same promoter-containing fragments used to assay regulation of the wild-type promoters . The mutated promoter regions, with identical flanking sequences, were similarly subcloned into the pRKlac290 lacZ transcriptional fusion vectors and introduced into C . crescentus for assay.

 
Growth of C . crescentus on various carbon sources. C . crescentus can use diverse carbon sources for growth, including various carbohydrates [48], fatty acids [46], amino acids [15,48], and aromatic compounds [6] . We examined the growth of C. crescentus strain CB15, the strain used to determine the genome sequence, as well as the CB15N [NA1000] derivative used routinely for Caulobacter cell cycle synchronization experiments, on M2 minimal salts agar [12] with various carbon sources . Carbonsources tested included hexoses and hexose derivatives [glucose,fructose, galactose, mannose, glucosamine, gluconate, glucuronicacid, and mannitol], pentoses [xylose, arabinose, and ribose],disaccharides [sucrose, lactose, and trehalose], and other carboncompounds [acetate, glycerol, and glutamate].

Only glucose, xylose, and glutamate produced colonies of atleast 1 mm in diameter within 72 h, although colony growth wasonly slightly slower on sucrose . Trehalose generated 1-mm colonieswithin 5 days, while galactose, mannose, mannitol, arabinose,and ribose supported growth of 1-mm colonies within 6 to 10days at 30°C . Growth on M2 medium containing 0.2% fructose,acetate, glycerol, gluconate, glucuronic acid, or glucosaminewas barely detectable after 10 days at 30°C . No significantdifferences were observed between CB15 and CB15N during growthon these carbon sources.

To more quantitatively assess growth rates, CB15N was grownin liquid batch culture on PYE [an amino acid-rich medium typicallyused for laboratory culture of C . crescentus], M2 with glucose [M2G], and M2 with xylose [M2X] . During the exponential growthphase of cultures incubated aerobically at 30°C, CB15N culturesexhibited a doubling time of 89 min on PYE, 114 min on M2G,and 127 min on M2X.

Transcriptional profiling of C . crescentus grown on PYE and minimal media. To assess gene expression in cultures grown in PYE, M2G, andM2X, RNA was prepared from log-phase C . crescentus CB15N cultures.RNA samples were fluorescently labeled and competitively hybridizedon spotted arrays displaying PCR product probes representing3,528 predicted C . crescentus genes [93% of the 3,761 open readingframes predicted in the C . crescentus genome] [31, 32] . Analysisof the data produced M2X/M2G expression ratios for 3,074 genes,PYE/M2G ratios for 3,179 genes, and M2X/PYE ratios for 3,089genes . [Due to space constraints, only selected data are presentedand discussed here . The entire data set, as well as all tablesnumbered with an S prefix, are presented in the supplementarymaterials, which are available online at http://www.stanford.edu/group/caulobacter/metabolism.Additional information on probes, data collection, data processing,and statistical analysis can be found at the same site.] Expressionof 390 genes was found to vary significantly and reproducibly.Genes that were found to be differentially expressed betweeneach pair of media are shown by functional categories in Fig.1.

 
Sets of genes that appear to be coordinately regulated in eachmedium were identified, perhaps revealing information as totheir function . For example, there were 88 up-in-PYE genes,whose expression is elevated in PYE compared to both M2G andM2X . [Up-in-PYE genes can also be said to be down in M2.] Similarly,we found 119 up-in-M2/down-in-PYE genes, 26 up-in-glucose [M2G]genes, eight down-in-glucose genes, 51 up-in-xylose [M2X] genes,and five down-in-xylose genes . Not surprisingly, given the differencesin carbon, nitrogen, sulfur, and phosphate sources, the greatest differences in gene expression were observed between PYE andM2 media, irrespective of the carbon source in the M2 cultures. Examination of those with known or predicted functions revealedsome trends . For example, many genes encoding transporters orcomponents of transport systems were differentially regulatedin the three media . Also, the expression of genes encoding componentsof amino acid degradation pathways tended to be elevated inPYE, whereas biosynthetic pathways were frequently elevatedin M2 medium [Fig. 1C and D and Fig . 2] . These observations are discussed in more detail in the following sections.

 
Nutrient transport. The differential gene expression data suggest that C . crescentusmodulates the transporters it synthesizes in response to thegrowth medium [Fig . 1F and G; Table S1 in supplementary materials].Twenty-five TonB-dependent receptors showed differential regulationin at least one pairwise medium comparison, with the largestcoregulated set being the nine specifically induced by xylose[i.e., elevated in M2X relative to both M2G and PYE] [Tables1, 2, and 3, Table S1] . Six ABC transport components were repressedin PYE, including three predicted to be involved in sulfatetransport; five were induced in PYE, including components similarto those used to transport cysteine, sugar[s], and phosphonatesin other bacteria . No ABC transport components responded togrowth on xylose, and only one was induced by glucose.

 
Previous examination of the membrane proteome of C . crescentus found 12 outer membrane proteins [11 of which are TonB-dependent receptors] to be more abundant in cells grown in M2G comparedto PYE [47] . The microarray data presented here are generally consistent with that analysis, suggesting that differences in transporter transcript levels are positively correlated with differences in the respective protein levels . Genes encodingthree of the TonB-dependent receptors identified in the membraneproteome analysis [CC1517, CC0214, and CC0028] showed elevatedexpression in both M2 media in the microarray analysis, withnotably higher expression in M2G than M2X [Table 2] . CC1666showed a fourfold increase in expression in both M2 media relativeto PYE . Expression of CC1750 also increased significantly betweenM2G and PYE, though little change was observed in M2X . Therewere, however, some discrepancies between the microarray andproteome data . Expression of CC3461 was strongly induced inM2X [Table 3], but the difference in expression in M2G versusPYE was not statistically significant . Expression of CC1915and CC1970 did not change significantly in M2 media comparedto PYE, and expression of CC1099 actually decreased in M2G relativeto PYE, contrary to the proteomic data . Microarray data comparingexpression in M2G and PYE were not obtained for CC1781, CC2194,and CC3500.

Degradation pathways induced in PYE. PYE supplies amino acids that can be employed in protein synthesis,used to fuel energy production, or transformed into other necessarymetabolites . The expression of a number of genes encoding componentsof amino acid degradation pathways was elevated in PYE [Fig.1D and 2A] . Particularly notable are components of pathways involved in degrading arginine, histidine, alanine, glycine, glutamate, and proline, which are abundant in peptone and inyeast extract, the main components of PYE [BD Diag-notic Systems, http://www.bd.com/diagnostics/microservices/regulatory/typicalanalysis/index.asp] [Table 1] . Arginine catabolism appears to occur via the argininesuccinyltransferase pathway, with enzymes from two potential transcription units [CC0581-0584 and CC1606-1608], yieldingCO₂, NH₃, NADH, glutamate, and succinate.

Histidine degradation in C . crescentus [15] has been shown toproceed through the initial steps of the classical Hut [histidineutilization] pathway defined in Aerobacter [Klebsiella] aerogenes[36] . That pathway yields glutamate and formamide, but C . crescentusproduces formate rather than formamide, suggesting that theterminal steps of the pathway resemble that of Pseudomonas putida[34] . The known C . crescentus Hut enzymes are products of asingle PYE-induced locus [CC0957-0960] . CC0958 is annotatedin GenBank as formiminoglutamase, which catalyzes the terminalstep of the K . aerogenes Hut pathway, yielding formamide . Theannotation may be erroneous in that the CC0958 polypeptide sequencealigns closely [BLAST E value, 3 x 10^-48] with the P . putidaN-formylglutamate amidohydrolase [GenBank accession number AAB86969],which catalyzes the terminal step of the pathway version yieldingformate . Conservation of this gene product in other -proteobacterial genomes [Agrobacterium tumefaciens, Sinorhizobium meliloti, Mesorhizobium loti, Brucella melitensis, and Rhodospirillumrubrum] suggests that this pathway for histidine utilizationis standard for this group.

The induced pathways suggest mechanisms by which C . crescentus generates energy from amino acid degradation during growth in PYE . For example, induction of alanine dehydrogenase [CC3574]allows the conversion of alanine to pyruvate and generates NADH.The pyruvate is presumably fed into the tricarboxylic acid cycle,while NADH feeds respiratory electron transport . Similarly,a bifunctional proline dehyrogenase/pyrroline-carboxylate dehydrogenase[CC0804, homologous to E . coli putA] is also induced, whichgenerates FADH₂ and NADH while producing glutamate from proline.

Glutamate can serve as an excellent source of carbon and nitrogen for C . crescentus [13] . Expression of CC0088, which encodesa member of a recently identified class of glutamate dehydrogenases[40], increased threefold during growth in PYE relative to M2media . Glutamate dehydrogenase carries out oxidative deaminationof glutamate, yielding -ketoglutarate, ammonia, and NADH . Theactivity of the related glutamate dehydrogenase in Streptomycesclavuligerus increases significantly in cultures grown withglutamate as the sole nitrogen and carbon source compared tocultures grown with ammonia and glycerol [40] . Ely et al . [13] previously reported that C . crescentus has no detectable glutamate dehydrogenase activity, but their assay used NADP⁺ as the electronacceptor, whereas Minambres et al . [40] found the S . clavuligerushomolog to be NAD⁺ specific.

The yeast extract in PYE may contain significant levels of fatty acids, which C . crescentus utilizes as an energy source, based on the observation that growth on PYE induced expression ofseveral genes encoding enzymes in the ß-oxidationpathway for fatty acids . These include acyl-coenzyme A dehydrogenase[CC1350: PYE/M2G = 3.55, PYE/M2X = 3.55], enoyl-coenzyme A hydratase/isomerase[CC1352: PYE/M2G = 2.33, PYE/M2X = 2.34; CC2169: PYE/M2G = 3.54,PYE/M2X = 3.65], and acetyl/propionyl-coenzyme A carboxylase[CC2168: PYE/M2G = 2.17, PYE/M2X = 2.10] . The fatty acid oxidationgenes induced in PYE are a subset of a large family of similarlyannotated genes in the C . crescentus genome [Table S2] . Whetherthe induced gene products are functionally distinct from theuninduced ones is unknown, as is the mechanism of their differentialregulation.

Assimilation and biosynthesis in minimal media. During growth in M2 medium, C . crescentus must synthesize eachamino acid . As expected, this leads to elevated expression ofgenes encoding components of several amino acid biosyntheticpathways, including those for leucine, isoleucine, valine, serine,threonine, cysteine, methionine, glutamine, and glutamate [Fig.2B and Table 1] . Interestingly, several pathways showed no apparentdifference in expression compared to growth on PYE, and in someonly a subset of the genes showed increased expression . For example, among the many enzymes included in the pathways for synthesis of the aromatic amino acids tryptophan, tyrosine,and phenylalanine, only CC2300 [encoding 3-deoxy-D-arabino-heptulosonate-7-phosphate synthase, the first enzyme of the chorismate pathway] and CC2222, encoding chorismate mutase, showed significant induction inM2 . The observed lack of induction of other aromatic amino acid biosynthesis genes in M2 is consistent with previous S1 nuclease mapping data [55], which found that trpF [CC3545], trpB [CC3544],and trpA [CC3543] mRNA levels did not change between cells grownon PYE and M2-sucrose . In contrast, microarray studies foundthat mRNA levels for genes in the E . coli trpEDCBA operon increasedtwo- to fivefold in minimal media compared to the same mediasupplemented with tryptophan [27].

In M2 medium, sulfate supplies sulfur to the cells, and genes encoding components of the sulfate assimilation pathway arestrongly induced relative to PYE [Table 1] . These include a sulfate-specific ABC transport system [encoded by CC1596-1598,whose products are homologous to the E . coli CysTWA sulfate transporter], a periplasmic binding protein that probably delivers sulfate to the ABC transport system [CC0286], and the enzymesthat incorporate sulfate-derived sulfur into cysteine: sulfateadenylate transferase/adenylylsulfate kinase [CC1482-1483], phospho-adenylylsulfate reductase [CC1121], sulfite reductase [CC1119], and cysteine synthase [CC1426 and CC3625] . Expressionof CC3063 [cysJ, the subunit of sulfite reductase] did notchange significantly.

During growth in M2, ammonia serves as the sole nitrogen source. Ely et al . [13] showed that ammonia assimilation in C . crescentusoccurs via the combined action of glutamine synthetase and glutamatesynthase . They observed fivefold induction of glutamate synthaseactivity and fourfold induction of glutamine synthetase activityin M2G versus PYE medium . Similarly, our microarray data showelevated expression in M2 media [compared to PYE] for the genes encoding glutamine synthetase [CC1969: M2G/PYE = 2.78, M2X/PYE = 1.93] and glutamate synthase subunits [CC3606: M2G/PYE = 2.54, M2X/PYE = 3.07; CC3607: M2G/PYE = 4.06, M2X/PYE = 4.75].

Very few genes encoding enzymes involved in nucleotide biosynthesis showed significant changes in expression in M2 versus PYE . Some enzymes involved in the synthesis of other relevant metabolitesand cofactors, however, showed significant induction in M2.De novo purine synthesis and some reactions in amino acid synthesisdepend on formyl-tetrahydrofolate as a donor of one-carbon units.The formyl-tetrahydrofolate synthesis pathway in C . crescentushas not been clearly elucidated, but expression of genes encodingseveral enzymes that could be relevant, including CC2138 [5-methyl-tetrahydrofolate-homocysteine methyltransferase], CC2140[5,10-methylenetetrahydrofolate reductase], and CC3630[formyltetrahydrofolate deformylase],is induced during growth in M2 medium . With respect to one-carbonunit metabolism, expression of S-adenosylmethionine synthetase [CC0050], which produces the key methyl donor S-adenosylmethionine, was also increased in M2.

Glucose catabolism. C . crescentus expresses enzymes of the Entner-Doudoroff pathwayduring growth on glucose, and it has been proposed that thisis the primary route of glucose catabolism in this organism[52] . Our microarray results show that the expression of CC1495,which codes for 2-dehydro-3-deoxyphosphogluconate aldolase [commonlyreferred to as KDPG aldolase], which carries out the last stepin the Entner-Doudoroff pathway, was increased in both M2G andM2X compared to PYE but did not vary significantly between M2Gand M2X . The microarray data also show that the genes encodingthe first four enzymes in the Entner-Doudoroff pathway [CC2054to CC2057] are more highly expressed in M2G than in either M2Xor PYE, although there is some elevation of expression in M2Xrelative to PYE.

The CC2057 to CC2054 coding regions are closely spaced [3, 7,and 15 bp of separation, respectively] and probably form anoperon . To independently verify the regulation of this operon,the CC2057 promoter region [including 287 bases upstream ofthe predicted start codon] and coding sequence were cloned intoa low-copy-number lacZ fusion vector . Expression of the CC2057promoter fusion in strain CB15 grown in M2G [1,460 ±445 Miller units] and in PYE supplemented with glucose [1,460± 430 units] was higher than in M2X [917 ± 90 units] or PYE [693 ± 185 units] . The glucose-dependentincrease in expression was lower than observed in the microarrayexperiments [Table 2] but qualitatively consistent.

C . crescentus hex [hexose] mutant strains that are unable to use glucose as their sole carbon source have been isolated [25]. We examined eight hex mutant strains to further characterize the phenotype . All of the strains grew normally on PYE and M2X agar plates . Most of these strains appeared completely incapableof growth on M2G, the exceptions being two strains that formedcolonies very slowly on M2G [see below] . To identify the basisfor this inability to utilize glucose, the hex mutations weremapped by transductional linkage to kanamycin resistance markers[60] . The specific genes in which the mutations resided werethen identified by complementation, based on restoration ofthe ability to grow on M2G . Three strains [SC333, SC337, andSC414] contained mutations that mapped to CC2055, a homologof the E . coli edd gene encoding phosphogluconate dehydratase,which catalyzes the Entner-Doudoroff pathway reaction in whichphosphogluconate is converted to 2-keto-3-deoxy-6-phosphogluconate.Another mutation [SC144] was mapped to CC1495, a homolog ofthe E . coli eda gene encoding KDPG aldolase, which cleaves 2-keto-3-deoxy-6-phosphogluconate to yield glyceraldehyde-3-phosphate and pyruvate.

Four other hex mutant strains were examined, including two [SC114 and SC481] that showed no growth on M2 with glucose or galactose as the carbon source and two [SC155 and SC162] that grew very slowly on glucose or galactose . All four mutations mapped toCC1493 [ppc], which encodes phosphoenolpyruvate carboxylase,an enzyme that catalyzes the addition of CO₂ to phosphoenolpyruvateto generate oxaloacetic acid . Expression of CC1493 was elevatedin M2G compared to both M2X and PYE . During growth on minimalmedia such as M2G, tricarboxylic acid cycle intermediates aresiphoned off for use in various biosynthetic pathways . In orderfor the cycle to continue, oxaloacetate levels must be replenished[17] . The difficulty experienced by ppc mutants in growing onM2 with glucose or galactose as the carbon source demonstratesthe critical role of this replenishment process . We suspectthat the mutations in SC155 and SC162 compromise the functionof the phosphoenolpyruvate carboxylase, reducing [but not eliminating]oxaloacetate production.

Entry of glucose into the Entner-Doudoroff pathway typically requires the sugar to be phosphorylated, though exceptions insome archaea have been noted [9] . In E . coli, glucose is phosphorylatedas it enters the cell through a phosphoenolpyruvate-dependentphosphotransferase transporter system [50] . Interestingly, thoughthe C . crescentus genome encodes several putative phosphotransferasetransporter system-type transport systems, none were inducedn response to glucose in our data . However, the putative operonencoding several other components of the Entner-Doudoroff pathwayalso includes a glucokinase homolog [CC2054], which could accountfor production of glucose-6-phosphate . Since no mutations withthe hex phenotype were identified in this gene, a knockout ofCC2054 was constructed . The mutant strain showed no detectablegrowth with glucose as the sole carbon source, a phenotype consistentwith its being required for phosphorylation of glucose priorto entry into the Entner-Doudoroff pathway . This suggests thatin C . crescentus, glucose is not transported through or phosphorylatedby a phosphotransferase transporter system-type system . Basedon the microarray data, one candidate for transporting glucoseacross the inner membrane is encoded by CC1103, whose productexhibits significant similarity to the E . coli fucose permease.

The MEME software [4] was used to identify conserved DNA motifsassociated with glucose-induced genes . Two palindromic motifs were found, associated with nonoverlapping sets of genes [Tables S3 and S4, supplementary materials] . The first motif appearsin 13 intergenic regions in the C . crescentus genome [TableS3] . Genes with the first motif upstream include CC2057 [thefirst gene of the operon encoding enzymes of the Entner-Doudoroffpathway], four genes encoding transporters [CC0970, CC1103,CC1500, and CC1754], two genes coding for 1,4-ß-D-glucanglucohydrolase homologs [CC0797 and CC2052], and a gene fora putative glucokinase family protein [CC3167] . The second motif,albeit less statistically significant [see above], appears in14 intergenic regions, including probable promoter regions for11 genes showing elevated expression in glucose [Table S4].These include CC1493 [encoding phosphoenolpyruvate carboxylase]and four genes encoding transporters [CC0214, CC1517, CC1518,and CC2928] . Though intriguing, the significance of these twomotifs with respect to expression and regulation of these genesremains to be investigated.

Xylose regulon. A substantial fraction of the genes whose expression increasedin xylose encode enzymes predicted to contribute to the degradationof extracellular polymers [Table 3] . This class includes twoxylosidase/arabinosidases [CC2802, CC3054], an -glucouronidase [CC2811], a putative ß-galactosidase [CC0788], a putative polysaccharide deacetylase [CC2153], and a gene annotated asencoding a secreted protein [CC2152] . Expression of CC0989 andCC2357, which are annotated as encoding xylosidases, did notincrease significantly in M2X in the microarray data . Expressionof a fructose-bisphosphate aldolase [CC3250], which could actin degradation or in gluconeogenesis, was also increased inxylose.

With SignalP [43, 44], four additional nontransport, xylose-inducedgenes encoding proteins with probable signal peptides for exportwere identified . Based on homology, probable functions couldbe identified for only one of these [CC0979, peptidyl-prolylcis-trans isomerase A] . Another one [CC0925] is annotated asan OmpA-related protein, suggesting that it may be targetedto the outer membrane . Also targeted to the outer membrane area set of nine TonB-dependent receptors whose expression is inducedby xylose . As noted previously, these proteins could support transport across the outer membrane . Xylose also induced expression of a gene [CC0814] encoding a member of the major facilitator superfamily of inner membrane proteins . The CC0814 product is strikingly similar to the E . coli XylE xylose/H⁺ symporter [BLASTPE value, 2 x 10^-70], suggesting a role in xylose import intothe cytoplasm . Given that E . coli, as an example, utilizes atleast one other xylose transport system in addition to XylE,it is possible that other inner membrane transporters aid inacquiring xylose in C . crescentus as well.

The CC0823 through CC0819 genes probably form an operon, basedon the close spacing of the coding regions and the coordinatedinduction by xylose . The extent of induction observed for thesegenes ranged from a low of 2.8-fold [M2X/M2G] for CC0820 toa high of 11.6-fold for CC0822 . [No expression data were obtainedfor CC0821.] A chromosomal Tn5-lacZ insertion previously isolatedin the CC0823 coding region showed strong induction of ß-galactosidase activity by xylose [40-fold upon addition of xylose to PYE][39] . Plasmid-borne transcriptional fusions of CC0823 to lacZ displayed 10-fold-higher activity in M2X than in M2G [39], similarto the sevenfold induction observed for CC0823 mRNA in M2X versusM2G in the microarray data . The Tn5-lacZ insertion in CC0823prevented growth with xylose as the sole carbon source [39],implying that CC0823 or a downstream gene in this operon iscritical for metabolism of xylose.

Regions upstream of the genes in the C . crescentus xylose regulon were scanned with MEME [4] to identify conserved sequences potentiallyrepresenting transcription factor binding sites . A conserved20-bp palindromic motif strongly associated with xylose-regulatedgenes was identified [Table 4] . The motif appeared 25 timesin intergenic regions in the C . crescentus genome . Fifteen ofthese sites were adjacent to at least one gene whose expressionis significantly increased in xylose [Table 4], including CC0823,where it overlapped the predicted -10 region of the promoter[Fig . 3] . Transcription factors controlling gene expressionin response to xylose have been identified in several bacterialspecies, and binding sites for such factors are sometimes conservedbetween species [38] . However, the motif identified here differssubstantially from the consensus binding site of the XylR repressorfound in the Bacillus/Clostridium group [53], and it does not resemble the recognition sequence for the XylR activator of some -proteobacteria [30].

 
The only xylose-regulated gene whose transcription start sitehas been identified is CC0823 [39] . Overlap of the conserved motif with the -10 region [Fig . 3] of this promoter is consistentwith a repressor binding site designed to inhibit transcriptioninitiation by interfering with RNA polymerase binding . It shouldbe noted that this motif is distinct from the dyad symmetricelement suggested by Meisenzahl et al . [39] as a potential regulatorysite for the CC0823 promoter.

To experimentally examine the role of this motif in xylose-dependent regulation, the motifs associated with CC0823 and CC0505 were mutationally altered [Fig . 3] . CC0505 encodes a signal peptide-containingprotein, possibly secreted, whose function is unknown . CC0505is not necessary for xylose utilization [data not shown], butits expression showed substantial xylose induction in microarrayexperiments [M2X/M2G = 5.8, M2X/PYE = 6.2] . Wild-type and mutantpromoters were fused to the lacZ reporter gene on a low-copy-numberplasmid, and ß-galactosidase expression was assayed in the same media used in the microarray experiments as wellas PYE supplemented with xylose [Fig . 3] . There were clear differencesbetween the two promoters—the CC0823 promoter yielded notably higher expression under all conditions—but the consequences of altering the conserved motif were similar . Inboth cases, the mutant promoters showed strong constitutiveexpression in media lacking xylose . These data support a modelin which the conserved motif functions as a binding site fora repressor protein that obstructs these promoters in the absenceof xylose and releases in the presence of xylose, allowing accessto RNA polymerase . Definition of this regulatory motif is particularlysignificant in that the CC0823 promoter is widely used as aheterologous promoter directing inducible gene expression inC . crescentus [for example, see references 11 and 39]; definingthe target site for regulation should aid in identifying the regulator itself and in manipulating the system to improve its utility.

Possible regulatory proteins. A number of putative transcription factor or proteins involvedin signal transduction pathways were differentially expressedin the media examined here . Regulators induced in M2 media relativeto PYE included a homolog of ArsR [CC2141: M2G/PYE = 2.23, M2X/PYE= 2.15], a response regulator [CC3477: M2G/PYE = 2.81, M2X/PYE= 2.40], a histidine kinase [CC0285: M2G/PYE = 1.88, M2X/PYE= 1.82] a hybrid histidine kinase/response regulator [CC0026:M2G/PYE = 2.09, M2X/PYE = 1.99], a LuxR-related protein [CC0782:M2G/PYE = 2.53, M2X/PYE = 3.08], and extracytoplasmic functionsigma factors CC2883 [M2G/PYE = 2.38, M2X/PYE = 2.07] and CC3475[M2G/PYE = 2.03, M2X/PYE = 1.89] . Two putative transcriptionfactors, both members of the LacI family [CC2053 and CC2355],appeared to be specifically induced by glucose [expression significantlygreater on M2G than M2X or PYE, Table 2] . Expression of a responseregulator [CC0247] and a sensor histidine kinase [CC1294] wasinduced during growth on xylose [Table 3] . A few potential transcriptional regulators showed relatively minor [less than twofold] but statistically significant increases in expression in PYE relative to the M2 media, including a homolog of the PupR anti-sigma factor [CC0982],a response regulator [CC1610], a putative cyclic AMP bindingprotein [CC2171], and an AsnC family member [CC3573] [see supplementary materials].

Cell cycle regulation of metabolic genes. The expression of several C . crescentus genes associated withroutine metabolic functions has been shown to be cell cycleregulated during growth in M2G medium [32] . A subset of thosecell cycle-regulated genes were differentially regulated inthe growth media tested here [supplementary material] . Twelvecell cycle-regulated genes were up and one was down in xylose,two were down and six were up in glucose, and 37 were down and21 were up in PYE . The genes regulated by both the growth mediaand the cell cycle did not seem to have a common time of peakexpression during the cell cycle . Cell cycle-regulated genescould appear to be medium regulated in microarray experimentswith mixed cultures if the cultures grown in various media haddifferent proportions of swarmer and stalk cells, but we sawno obvious evidence for that to be the case here.

The faster growth rate of C . crescentus in PYE compared to the M2 media likely explains why some cell cycle-regulated genes, including dnaA [CC0008] and recA [CC1087], appear to be increased in PYE . Some of the cell cycle-regulated genes whose expression was elevated in PYE, including phnC, phnD, and phnE [CC0361 to CC0363], are involved in transport, while others, such as gcvT [CC3335] and gcvH [CC3334], are involved in degradation. Conversely, many of the cell cycle-regulated genes whose expression was down in PYE are involved in biosynthesis . For promoters regulated by both the cell cycle and nutrient composition, itremains to be determined whether the same transcription factorsare used in both forms of regulation.

 
Past studies of bacterial metabolism have primarily focusedon cells adapted to high nutrient conditions . In this work,in contrast, we analyzed the metabolism of an oligotrophic organism,Caulobacter crescentus . Oligotrophs thrive where nutrients areperpetually scarce, the usual situation in most aquatic andterrestrial habitats on earth . An improved understanding ofoligotrophs' metabolic activities is essential to understandingbiogeochemical processes that have major impact on the environment[20] . We describe here a comparison of gene expression duringgrowth of C . crescentus on common laboratory media . These dataprovide insights into C . crescentus metabolic pathways and regulatory strategies and establish a baseline for future experiments withthese media . We assessed expression patterns for over 80% ofthe known genes in the C . crescentus genome in each pairwisemedium comparison and found that over 12% of the genes examinedshowed significant differences in expression in at least onemedium comparison . Not surprisingly, most of the differencesoccurred between the complex PYE medium and the M2 minimal saltsmedia regardless of the carbon source used in the M2 medium.Fewer differences were seen between the two M2 media, but thedifferences that were observed may identify transporters, enzymes,and regulators specifically relevant to glucose and xylose utilization.

The Entner-Doudoroff pathway has been shown to be active inC . crescentus during growth on glucose [48, 52] . In the datapresented here, all of the genes encoding enzymes of the Entner-Doudoroffpathway were found to be significantly induced during growthon glucose-containing media compared to PYE . Mutations in threegenes [CC1495, CC2054, and CC2055] encoding enzymes of the Entner-Doudoroffpathway rendered C . crescentus incapable of growth with glucoseas the sole carbon source . The Entner-Doudoroff pathway is thusessential for glucose utilization in C . crescentus, a situationpreviously seen only in Pseudomonas aeruginosa [5] and Burkholderia cepacia [1] . In contrast, in several free-living -proteobacteria in which glucose metabolism has been examined, both the Embden-Meyerhof-Parnasand Entner-Doudoroff pathways are used [3, 8, 42].

The apparent absence of the enzyme phosphofructokinase in C. crescentus would preclude use of the upper branch of the Embden-Meyerhof-Parnaspathway . Thus, instead of generating glyceraldehyde-3-phosphateand dihydroxyacetone phosphate, glucose is processed into glyceraldehyde-3-phosphateand pyruvate . There appears to be little transcriptional regulationof the genes encoding enzymes for the latter steps of glycolysis[taking glyceraldehyde-3-phosphate to pyruvate], as mRNA levelsshowed minimal differences between media . In contrast, in E.coli, expression of all genes in the lower branch of the Embden-Meyerhof-Parnaspathway except pykA and pykF [pyruvate kinases] increases duringfermentation on glucose compared to fermentation on xylose [16]. The E . coli gene expression data agree with metabolic flux estimates in that all steps in the lower part of the glycolytic pathway are expected to have higher flux during fermentation of glucose compared to fermentation of xylose except for the conversionof phosphoenolpyruvate to pyruvate [16] . Whether similar expressionof glycolytic genes during growth of C . crescentus on xyloseand glucose likewise reflects comparable carbon flux throughthe lower part of the glycolytic pathway during growth on glucoseand xylose is uncertain, given that the route of xylose catabolismis unknown.

There are many genes whose products have the same predicted enzymatic function in the current C . crescentus genome annotation but which exhibited distinct expression patterns in the microarray data [Table S2] . Some differentially regulated isozymes may participate in distinct pathways that are differentially induced depending on the medium . For example, in an oligotrophic Spirillum sp., the relative activities of three forms of lactate dehydrogenase vary as a function of the lactate concentration in the growth medium [37] . The exact roles of many of the C . crescentus isozymesremain to be determined . In such cases, differential regulationmay provide clues to distinct functions.

One example is particularly relevant to glucose catabolism.Two genes [CC0784 and CC1495] are annotated as encoding KDPGaldolase, which catalyzes the final step in the Entner-Doudoroffpathway . Expression of CC1495 is elevated in M2 minimal mediawith either xylose or glucose as the carbon source comparedto PYE, whereas expression of CC0784 is unchanged . Subsequentmutational results demonstrated that CC1495 is essential forgrowth on glucose, implying that CC0784 [assuming that it isexpressed] is not capable of functionally substituting for CC1495in glucose catabolism . On the other hand, the CC1495 mutantstrains could still form colonies with galactose as the carbonsource . The CC0784 polypeptide sequence resembles gene productsannotated as 2-dehydro-3-deoxyphosphogalactonate aldolases fromseveral organisms, including Ralstonia solanacearum, Bradyrhizobiumjaponicum, and Brucella species; this resemblance is considerably stronger than that shown by the CC1495 polypeptide . It has been reported previously that C . crescentus may use some versionof the Entner-Doudoroff pathway for galactose catabolism [29]. Our results suggest that the CC0784 product, rather than participating in the glucose Entner-Doudoroff pathway, may be involved ina version of the Entner-Doudoroff pathway dedicated to galactose.

The pathway of xylose degradation in C . crescentus may be distinct from known pathways in other microbes, as genes encoding key enzymes for those pathways [particularly xylose isomerase and xylulose kinase] are not apparent in the C . crescentus genome. An inducible xylose dehydrogenase activity has been demonstratedin C . crescentus, which may serve as the initial step in xylose catabolism [48] . The only known C . crescentus mutant strainthat is unable to grow on xylose resulted from a Tn5-lacZ insertionin the xylX [CC0823] gene [39] . This insertion showed stronglyxylose-inducible ß-galactosidase activity . Microarraydata presented here confirm xylose induction of CC0823, whichturns out to be the first gene in a potential five-gene operon[CC0823 to CC0819], the largest transcription unit whose expressionresponds to xylose.

The CC0823 product does not resemble any gene of known function, but it is homologous to genes in other genomes, including Mesorhizobium loti, Streptomyces coelicolor, Agrobacterium tumefaciens, Pseudomonasputida, Bradyrhizobium japonicum, and Sinorhizobium meliloti.The xyl mutant phenotype of the Tn5 insertion could reflecta critical role for the CC0823 product in xylose metabolism[39]; alternatively, the xyl mutant phenotype could result frompolar effects of the insertion on expression of critical downstreamgenes in the operon . Two putative dehydrogenases [CC0822 andCC0821] of unknown substrate specificity are encoded in thisoperon, either of which could potentially be responsible forinducible xylose dehydrogenase activity.

The ability to replenish intermediates of the tricarboxylicacid cycle can be critical for growth of microbes on a singlecarbon source [17] . Under such conditions, tricarboxylic acid cycle intermediates are continually siphoned off for various biosynthetic processes, compromising continued flux throughthe pathway . Expression data and genetic analysis suggest thatcarbon flux through the tricarboxylic acid cycle is manageddifferently during growth on glucose and xylose . The expressionof phosphoenolpyruvate carboxylase is induced during growthon glucose but not xylose, and genetic evidence is presentedabove that supplementation of the tricarboxylic acid cycle withoxalacetate via phosphoenolpyruvate carboxylase is criticalfor growth on glucose but not xylose.

In contrast, expression of isocitrate lyase [CC1764] is specifically elevated during growth on xylose . Isocitrate lyase catalyzes the conversion of isocitrate to glyoxylate and succinate, initiating the glyoxylate bypass . The succinate generated by this processcan be drawn off for other biosynthetic pathways . Glyoxylate,meanwhile, can be combined with acetyl-coenzyme A by malatesynthase to produce malate, allowing a modified tricarboxylicacid cycle to continue . Interestingly, expression of malatesynthase, encoded by the gene [CC1765] adjacent to that encodingisocitrate lyase, was not observed to increase coordinatelyin response to xylose in the microarrays . Future studies arewarranted to examine the role of the glyoxylate bypass duringgrowth on xylose and other carbon sources.

The ability to simultaneously import a variety of growth substrates could be critical for survival in oligotrophic environments[21] . As such, it seems unlikely that all of the putative transportersand exoenzymes whose expression is induced by xylose are utilizedsolely, or are even necessary, for the import and utilizationof this sugar . The CC0505 gene, for example, is strongly inducedby xylose, but a CC0505 knockout strain shows no defects ingrowth on xylose [data not shown] . We hypothesize that C . crescentususes xylose as an indicator that metabolites derived from thebreakdown of plant cell walls are available . Other microorganisms,particularly soil microbes such as Streptomyces species, areknown to use small molecules resulting from extracellular polymerbreakdown as a positive feedback signal to induce degradativeenzymes [22] . Similarly, growth of Thermotoga maritima on xyloseinduces genes for two xylanases and an -glucosidase, and growthof T . maritima on xylan additionally increases expression ofgenes for an -glucuronidase and a ß-D-galactosidase[7].

Since xylose in natural habitats would presumably be acquiredby breakdown of xylan, xylose may be used to trigger the expressionof extracellular hydrolytic enzymes relevant to breakdown of lignicellulose [of which xylan is a major constituent] as wellas the enzymes needed for the catabolism of xylose itself, asin the fungus Aspergillus niger [18] . Environments rich in xylan should contain cellulose as well; interestingly, however, xylose does not induce expression of endoglucanases or ß-glucosidases that could [at least theoretically] be used by C . crescentus for cellulose degradation.

The abundance of TonB-dependent receptors induced by xylosecould provide cells with the ability to import products of extracellular lignocellulose degradation, but it remains to be determinedwhat metabolites each transporter is capable of bringing intothe periplasm . In an effort to further characterize the transporters induced by xylose, we used BLASTP to compare xylose-inducedC . crescentus transporters to the NCBI database . Somewhat surprisingly, many of the strongest homologies were to proteins encoded by Xanthomonas campestris, a plant pathogen that secretes exoenzymes to attack plant cell walls . The virulence of at least one Xanthomonas species [X . oryzae, which causes rice blast disease] is dependent on the ability to secrete xylanase [51] . The X . campestris genomecontains bidirectional best hits to four of the xylose-inducedTonB-dependent receptors [CC0442, CC0999, CC2832, and CC3336]as well as the putative inner membrane xylose transporter [CC0814]and the SapC peptide permease homolog [CC1000].

Although Caulobacter and Xanthomonas are not close phylogenetically,their genomic contents bear intriguing similarities . In additionto having sets of putative exoenzymes similar to C . crescentus,the sequenced genomes of the Xanthomonas species X . campestrisand X . citri also display an abundance [nearly 50] of TonB-dependentreceptors [10] . There are also hints that regulation by xylosemay be similar in X . campestris and C . crescentus . The X . campestris genome has several close matches to the DNA sequence motif associated with xylose regulation in C . crescentus [Table 4], with threeof the four strongest instances of this motif in the X . campestrisgenome located upstream of putative transcription units containingxylose-associated genes [xylose isomerase, xylosidase, and xylanase][Table S5] . Thus, although C . crescentus has not been reportedto act as a plant pathogen, it may have mechanisms for acquiringplant polymer-derived metabolites similar to those found inXanthomonas species.

Environmentally responsive signal transduction pathways in C. crescentus have not been extensively explored . The collectionof large microarray data sets, as presented here, provides someinitial insight into regulatory responses to nutrient levelsin this organism . A set of genes whose expression responds tothe availability of xylose have been identified, along witha cis-acting site that appears to represent the target for axylose-responsive repressor . Two potential regulatory siteshave also been identified for glucose-responsive genes; theseneed to be tested, and the trans-acting factors controllingxylose and glucose regulation will be pursued in future work.Coregulation of many genes and potential operons encoding aminoacid synthesis or degradation functions has been observed.

Future work will continue to characterize the traits that allow C . crescentus to be a successful oligotroph by exploring this organism's metabolism in more dilute media, in mixtures of substrates, and in continuous cultures . Finally, the stage is set for exploring the interface between environmental and cell cycle regulation. What environmental cues influence cell cycle progression? Arethere cell cycle checkpoints that determine if a cell has sufficient resources to proceed with division? How are individual promoters influenced by both the environmental and cell cycle state? Asjust one example, expression of the glycine cleavage system[gcv] genes [CC3355 to CC3352] is strongly induced in PYE mediumversus M2, presumably due to relatively high levels of exogenousglycine . During the cell cycle, expression of the gcv genespeaks in swarmer cells and then is repressed dramatically inpredivisional cells [32] . Is the same transcription factor usedin both forms of regulation? What controls cell cycle regulationof this promoter, and why? Further application of molecularand classical genetic approaches with microarray analysis canprovide insight into such questions.

 
A.K.H., M.M., P.R., and H.H.M . were supported by grant DE-FG03-01ER63219-A001 from the Department of Energy . A.K.H., M.M., and H.H.M . were supported by grant MDA972-00-1-0032-P00005 from the DefenseAdvanced Research Projects Agency Defense Sciences Office . A.K.H.was also supported by NIH HG00044 from the National Institutesof Health . P.R . was also supported by grant N66001-01-C-8011from the Defense Advanced Research Projects Agency . The workof C.S., D.Y., and N.A . was supported by a grant from SantaClara University and by the Community of Science Scholars Initiativeprogram at SCU, made possible by a grant from the Howard HughesMedical Institute to support undergraduate research . C.S . isalso supported by grant MCB-0317037 from the National ScienceFoundation.

We thank Bert Ely [University of South Carolina], who kindly provided us with several hex mutant strains for analysis, and Lisandra West, for support in generating the microarray dataset.

 
* Corresponding author . Mailing address: Biology Department, Santa Clara University, 500 El Camino Real, Santa Clara, CA 95053 . Phone: [408] 551-1898 . Fax: [408] 554-2710 . E-mail: cstephens@scu.edu .

Present address: Department of Microbiology and Immunology, University of California-San Francisco, San Francisco, CA 94143.

Present address: Department of Developmental Biology, Stanford University, Stanford, CA 94305.

1. Allenza, P., and T . G . Lessie. 1982 . Pseudomonas cepacia mutants blocked in the Entner-Doudoroff pathway . J . Bacteriol . 150:1340-1347.
2. Altschul, S . F., W . Gish, W . Miller, E . W . Myers, and D . J . Lipman. 1990 . Basic local alignment search tool . J . Mol . Biol . 215:403-410.
3. Arthur, L . O., L . K . Nakamura, G . Julian, and L . A . Bulla, Jr. 1975 . Carbohydrate catabolism of selected strains in the genus Agrobacterium. Appl . Microbiol . 30:731-737.
4. Bailey, T . L., and C . Elkan. 1994 . Fitting a mixture model by expectation maximization to discover motifs in biopolymers . Proc . Int . Conf . Intell . Syst . Mol . Biol . 2:28-36.
5. Blevins, W . T., T . W . Feary, and P . V . Phibbs, Jr. 1975 . 6-Phosphogluconate dehydratase deficiency in pleiotropic carbohydrate-negative mutant strains of Pseudomonas aeruginosa. J . Bacteriol . 121:942-949.
6. Chatterjee, D . K., and A . W . Bourquin. 1987 . Metabolism of aromatic compounds by Caulobacter crescentus. J . Bacteriol . 169:1993-1996.
7. Chhabra, S . R., K . R . Shockley, S . B . Conners, K . L . Scott, R . D . Wolfinger, and R . M . Kelly. 2003 . Carbohydrate-induced differential gene expression patterns in the hyperthermophilic bacterium Thermotoga maritima. J . Biol . Chem . 278:7540-7552 .
8. Conrad, R., and H . G . Schlegel. 1977 . Different degradation pathways for glucose and fructose in Rhodopseudomonas capsulata. Arch . Microbiol . 112:39-48.
9. Dandekar, T., S . Schuster, B . Snel, M . Huynen, and P . Bork. 1999 . Pathway alignment: application to the comparative analysis of glycolytic enzymes . Biochem . J . 343:115-124.
10. da Silva, A . C., J . A . Ferro, F . C . Reinach, C . S . Farah, L . R . Furlan, R . B . Quaggio, C . B . Monteiro-Vitorello, M . A . Van Sluys, N . F . Almeida, L . M . Alves, A . M . do Amaral, M . C . Bertolini, L . E . Camargo, G . Camarotte, F . Cannavan, J . Cardozo, F . Chambergo, L . P . Ciapina, R . M . Cicarelli, L . L . Coutinho, J . R . Cursino-Santos, H . El-Dorry, J . B . Faria, A . J . Ferreira, R . C . Ferreira, M . I . Ferro, E . F . Formighieri, M . C . Franco, C . C . Greggio, A . Gruber, A . M . Katsuyama, L . T . Kishi, R . P . Leite, E . G . Lemos, M . V . Lemos, E . C . Locali, M . A . Machado, A . M . Madeira, N . M . Martinez-Rossi, E . C . Martins, J . Meidanis, C . F . Menck, C . Y . Miyaki, D . H . Moon, L . M . Moreira, M . T . Novo, V . K . Okura, M . C . Oliveira, V . R . Oliveira, H . A . Pereira, A . Rossi, J . A . Sena, C . Silva, R . F . de Souza, L . A . Spinola, M . A . Takita, R . E . Tamura, E . C . Teixeira, R . I . Tezza, M . Trindade dos Santos, D . Truffi, S . M . Tsai, F . F . White, J . C . Setubal, and J . P . Kitajima. 2002 . Comparison of the genomes of two Xanthomonas pathogens with differing host specificities . Nature 417:459-463.
11. Domian, I . J., K . C . Quon, and L . Shapiro. 1997 . Cell type-specific phosphorylation and proteolysis of a transcriptional regulator controls the G₁-to-S transition in a bacterial cell cycle . Cell 90:415-424.
12. Ely, B. 1991 . Genetics of Caulobacter crescentus. Methods Enzymol . 204:372-384.
13. Ely, B., A . B . Amarasinghe, and R . A . Bender. 1978 . Ammonia assimilation and glutamate formation in Caulobacter crescentus. J . Bacteriol . 133:225-230.
14. Entner, N., and M . Duodoroff. 1952 . Glucose and gluconic acid oxidation of Pseudomonas saccharophila. J . Biol . Chem . 196:853-862.
15. Ferber, D . M., F . Khambaty, and B . Ely. 1988 . Utilization of histidine by Caulobacter crescentus. J . Gen . Microbiol. 134:2149-2154.
16. Gonzalez, R., H . Tao, K . T . Shanmugam, S . W . York, and L . O . Ingram. 2002 . Global gene expression differences associated with changes in glycolytic flux and growth rate in Escherichia coli during the fermentation of glucose and xylose . Biotechnol . Prog . 18:6-20.
17. Gottschalk, G. 1978 . Bacterial metabolism . Springer-Verlag, New York, N.Y.
18. Hasper, A . A., J . Visser, and L . H . de Graaff. 2000 . The Aspergillus niger transcriptional activator XlnR, which is involved in the degradation of the polysaccharides xylan and cellulose, also regulates D-xylose reductase gene expression . Mol . Microbiol . 36:193-200.
19. Henrici, A . T., and D . E . Johnson. 1935 . Studies on freshwater bacteria . II . Stalked bacteria, a new order of schizomycetes. J . Bacteriol . 30:61-93.
20. Hirsch, P. 1986 . Microbial life at extremely low nutrient levels . Adv . Space Res . 6:287-298.
21. Hirsch, P., M . Bernhard, S . S . Cohen, J . C . Ensign, H . W . Jannasch, A