Evolution of the metabolic and regulatory networks associated with oxygen availability in two phytopathogenic enterobacteria
- 4.4k Downloads
Dickeya dadantii and Pectobacterium atrosepticum are phytopathogenic enterobacteria capable of facultative anaerobic growth in a wide range of O2 concentrations found in plant and natural environments. The transcriptional response to O2 remains under-explored for these and other phytopathogenic enterobacteria although it has been well characterized for animal-associated genera including Escherichia coli and Salmonella enterica. Knowledge of the extent of conservation of the transcriptional response across orthologous genes in more distantly related species is useful to identify rates and patterns of regulon evolution. Evolutionary events such as loss and acquisition of genes by lateral transfer events along each evolutionary branch results in lineage-specific genes, some of which may have been subsequently incorporated into the O2-responsive stimulon. Here we present a comparison of transcriptional profiles measured using densely tiled oligonucleotide arrays for two phytopathogens, Dickeya dadantii 3937 and Pectobacterium atrosepticum SCRI1043, grown to mid-log phase in MOPS minimal medium (0.1% glucose) with and without O2.
More than 7% of the genes of each phytopathogen are differentially expressed with greater than 3-fold changes under anaerobic conditions. In addition to anaerobic metabolism genes, the O2 responsive stimulon includes a variety of virulence and pathogenicity-genes. Few of these genes overlap with orthologous genes in the anaerobic stimulon of E. coli. We define these as the conserved core, in which the transcriptional pattern as well as genetic architecture are well preserved. This conserved core includes previously described anaerobic metabolic pathways such as fermentation. Other components of the anaerobic stimulon show variation in genetic content, genome architecture and regulation. Notably formate metabolism, nitrate/nitrite metabolism, and fermentative butanediol production, differ between E. coli and the phytopathogens. Surprisingly, the overlap of the anaerobic stimulon between the phytopathogens is also relatively small considering that they are closely related, occupy similar niches and employ similar strategies to cause disease. There are cases of interesting divergences in the pattern of transcription of genes between Dickeya and Pectobacterium for virulence-associated subsystems including the type VI secretion system (T6SS), suggesting that fine-tuning of the stimulon impacts interaction with plants or competing microbes.
The small number of genes (an even smaller number if we consider operons) comprising the conserved core transcriptional response to O2 limitation demonstrates the extent of regulatory divergence prevalent in the Enterobacteriaceae. Our orthology-driven comparative transcriptomics approach indicates that the adaptive response in the eneterobacteria is a result of interaction of core (regulators) and lineage-specific (structural and regulatory) genes. Our subsystems based approach reveals that similar phenotypic outcomes are sometimes achieved by each organism using different genes and regulatory strategies.
KeywordsTranscriptional Response TMAO Formate Dehydrogenase Cell Wall Degrading Enzyme Enterobactin
Dickeya dadantii and Pectobacterium atrosepticum cause soft-rot diseases characterized by maceration of plant tissues through the action of multiple secreted plant cell wall degrading enzymes . D. dadantii strain 3937 (D. dadantii) was originally isolated from African violet and is better known by its former name Erwinia chrysanthemi 3937, and P. atrosepticum strain SCRI1043 was isolated from potato [2, 3], but individual strains and these genera as a whole have broad host range, affecting over 50% of angiosperm plant orders . They are a world-wide problem for economically important crops and ornamental plants . Both D. dadantii and P. atrosepticum are relatively well-studied model organisms for understanding the molecular biology of soft-rot pathogenesis [6, 7]. Like most enterobacteria, Dickeya and Pectobacterium are facultative anaerobes that are able to grow with or without O2 by shifting metabolic strategies from aerobic respiration to anaerobic respiration or fermentation . They experience a wide range of O2 concentrations in different plant tissues and natural reservoirs like soil and water . Lack of O2 is thought to be one of the factors that can trigger rapid expansion of latent infections leading to devastating post-harvest destruction of entire crops in storage .
Apart from a small number of important virulence factors, such as pectinases PelA, D and E , little is known about which genes are regulated by O2 availability in these two soft-rot pathogens. In contrast, O2 -regulated genes have been extensively studied in the model animal-associated enterobacteria Escherichia coli and Salmonella enterica, where available data includes genome-scale expression profiling of the anaerobic stimulon of wild-type strains as well as mutants of key regulators FNR, ArcA, NarPQ and NarXL [11, 12, 13, 14, 15, 16]. Most of these regulators, and many of the known target genes associated with anaerobic metabolism are conserved across the enterobacteria and among more distantly related gamma-proteobacteria . Thus, we expect a conserved core transcriptional response to O2 limitation that includes the basic cellular machinery required to generate energy in an anaerobic environment. Yet, some of the O2-regulated genes found in E. coli are simply not present in other genera, and other genes that may be O2 - responsive in the phytopathogens are not shared with animal-associated organisms. A more complete picture of the anaerobic stimulon of plant-pathogenic enterobacteria requires direct experimentation in these organisms.
Here, we characterize the transcript profiles of P. atrosepticum and D. dadantii grown with and without O2 under controlled laboratory conditions. We performed our experiments in defined media to illuminate solely the O2-responsive regulatory network. These conditions are not expected to mirror the complex, dynamic, and largely undefined environment of a plant host. Rather, we seek to identify components of the anaerobic stimulon for follow-up experimentation, provide a framework for identification of characteristics of the response to O2 in more complex datasets, and investigate the conservation and divergence in O2-mediated regulation among the enterobacteria.
Results and discussion
A large number of genes are in the O2-response stimulon
Using a conditional false-discovery rate (cFDR) of 0.01 (permissive criterion for differential expression), EBarrays  detects over 2204 differentially expressed genes in D. dadantii (48.5%), and 599 in P. atrosepticum (13.4%). Particularly in D. dadantii where the extremely good agreement between replicates improves the sensitivity, many of these are genes that show small changes between the aerobic and anaerobic conditions. Requiring the changes (anaerobic/aerobic) be at least 3-fold reduces the numbers to 443 differentially expressed genes in D. dadantii (9.8%) and 320 genes in P. atrosepticum (7.3%). Thus, a substantial fraction of each genome is involved in the anaerobic stimulon even using our stringent criteria (cFDR = 0.01 and fold change > 3), consistent with published reports for E. coli K-12  despite differences in array platforms and analysis methods. The most extreme and conserved transcriptional responses are associated with anaerobic metabolism indicating that these organisms are responding to O2 availability.
Differences in the genetic architecture of the O2-responsive stimulon are illuminated by a biological subsystems approach
Pectobacterium and Dickeya (with Brenneria) form a monophyletic clade of phytopathogens distinct from other genera of enterobacteria, like Escherichia and Salmonella , where the transcriptional response to O2 has been extensively studied. Nevertheless, all free-living enterobacteria with sequenced genomes share a substantial fraction of ancestral genes that are thought to reflect clonal or vertical descent. Gene losses, duplications, and lateral gene transfers lead to content differences among genomes. Little is known about the extent to which these types of events factor into variation in the response of different enterobacteria to O2 availability.
We used OrthoMCL  to cluster protein-coding genes from D. dadantii, P. atrosepticum, and E. coli, and used these ortholog groups to compare transcript profiles across organisms. Genes for some of the well-characterized components of the E. coli anaerobic energy metabolism architecture are entirely missing from one or both of the phytopathogens. Some functional equivalents within and between organisms are carried out by genes that are not orthologous. For these reasons, we find it useful to approach the comparison from a biological subsystem-oriented perspective that accommodates the complexity of the evolutionary history and functional redundancy, and considers related gene products, such as genes associated with a single molecular complex or biological process . In the following sections, we report our findings from a largely statistical perspective. We discuss the possible biological significance in later sections following a subsystems oriented approach that groups orthologous, paralogous and even analogous genes with functionally related products.
Transcriptional response to O2 limitation for genes orthologous in the two phytopathogens
Minimal conserved anaerobic transcriptional response shared by D. dadantii 3937 and P. atrosepticum SCRI1043 and its comparison to E. coli.
ASAP Feature ID
A. Orthologs down-regulated ≥ 3-fold in both D. dadantii and P. atrosepticum
membrane spanning protein in TonB-ExbB-ExbD complex
ribonucleotide reductase of class Ia (aerobic), alpha subunit
ribonucleotide reductase of class Ia (aerobic), beta subunit
iron-binding periplasmic protein
predicted cytochrome b561
predicted ribosomal protein
predicted ribosomal protein
zinc ABC transporter, periplasmic-binding protein ZnuA
high-affinity zinc transport system membrane protein
high-affinity zinc transport system ATP-binding protein
Iron dicitrate-binding protein
putative iron ABC transporter permease protein
putative iron ABC transporter, periplasmic-binding protein
putative iron ABC transporter ATP-binding protein
TonB-dependent ferric achromobactin receptor
putative ABC transporter substrate-binding protein
putative ABC transporter substrate-binding protein
putative transport system permease protein
putative ABC transporter substrate-binding protein
putative ABC transporter substrate-binding protein
putative ABC transporter substrate-binding protein
ABC-type transporter, periplasmic component
ABC transporter, permease protein
ABC transporter, permease protein
ABC transporter ATP-binding protein
ABC transporter substrate-binding protein
B. Orthologs up-regulated ≥ 3-fold in both D. dadantii and P. atrosepticum
iron-dependent alcohol dehydrogenase
alkyl hydroperoxide reductase, C22 subunit
C4-dicarboxylate transporter DcuB
Fe-binding and storage protein
fumarate reductase (anaerobic) NAD/flavoprotein subunit
fumarate reductase (anaerobic), Fe-S subunit
fr d C
fumarate reductase (anaerobic), membrane anchor subunit
fumarate reductase (anaerobic), membrane anchor subunit
glutaredoxin 1,coenzyme for ribonucleotide reductase
high-affinity nickel transport protein
predicted hydrogenase 2 cytochrome b type component
hydrogenase 2-specific chaperone
hydrogenase 2, small subunit
protease involved in processing C-terminal end of HycE
formate dehydrogenase-H, ferredoxin subunit
hydrogenase 4, 4Fe-4S subunit
hydrogenase 4, membrane subunit
hydrogenase 4, membrane subunit
hydrogenase 4, membrane subunit
hydrogenase 4, membrane subunit
hydrogenase 4, membrane subunit
hydrogenase 4, subunit
hydrogenase 4, Fe-S subunit
hydrogenase 4, Fe-S subunit
predicted processing element hydrogenase 4
GTP hydrolase involved in nickel liganding into hydrogenases
[NiFe] hydrogenase metallocenter assembly protein HybG
protein required for maturation of hydrogenases
mannose-specific enzyme IID component of PTS
anaerobic ribonucleoside-triphosphate reductase
anaerobic ribonucleotide reductase activating protein
pyruvate formate lyase I
ribosome modulation factor
pyruvate formate lyase subunit
predicted peptidase (collagenase-like)
predicted oxidoreductase, Zn-dependent and NAD(P)-binding
predicted dethiobiotin synthetase
putative membrane protein
lactoylglutathione lyase-like lyase
formate dehydrogenase, cytochrome B556 subunit
Comparison of divergent and differentially expressed genes in the phytopathogens to E. coli.
ASAP Feature ID
putative membrane protein
putative membrane protein
soluble cytochrome b562
If we include simple ortholog groups where both orthologs are detected as differentially expressed (cFDR = 0.01), but one or the other, or even both have fold changes < 3, then we identify 222 ortholog groups in total that are differentially expressed in a congruent direction, and 51 that are differentially expressed in divergent directions across the two phytopathogens (Additional File 2). This provides a more generous estimate of the conserved core and divergently expressed members of the stimulon. This more permissive congruent set includes operons associated with anaerobiosis in other organisms and closer inspection shows that in these cases, some genes do meet our stringent criteria. This observation provides evidence that our more permissive congruent set includes real members of the anaerobic stimulon. Below, where we detail the genes and biological processes that are implicated in the transcriptional response to O2, we guide inclusion by the stringent set (96 genes), but do not limit discussion to genes meeting the stringent significance criteria.
Comparison of the expression patterns for orthologs shared by D. dadantii, P. atrosepticum and E. coli
D. dadantii and P. atrosepticum are more closely related to each other than to E. coli. Our OrthoMCL analysis identified 2231 groups that include at least one gene from each of the three organisms (totaling 2309 E. coli genes, 2283 D. dadantii genes, and 2263 P. atrosepticum genes). Of these, 2124 ortholog groups include a single gene from each organism.1124 of the 2124 groups have a differentially expressed gene (permissive criteria) in at least one of the three organisms, and 261 groups have at least one gene that shows fold change > 3 (114 genes in P. atrosepticum, 111 in E. coli, and 153 in D. dadantii, see Additional File 3). Only 20 ortholog groups contain genes that show a congruent expression pattern with fold change greater than 3 for all three orthologs (Table 1, bold gene names), suggesting that the conserved response to O2 is small in terms of the number of genes involved, or that the conserved response lies in orthologs with smaller magnitude changes. This set is mainly comprised of genes known to function in cellular metabolism under anaerobic conditions namely, frdABCD, dcuB, adhE, hypC, focA, hybO, yfiD, nrdG, nrdD, beside others such as the collagenase encoding genes yhbUV, a peptidase coding gene pepT, and two other genes ynfK and ycbJ which are all up-regulated. The stringent congruent set also includes four down-regulated genes. These are exbB that encodes a component of the TonB-exbBD complex, yceI that encodes a cytochrome b561 and two other genes (yceJ, yigI), which encode uncharacterized proteins.
Relaxing the analysis stringency to consider all 1-1-1 orthologs that are differentially expressed (no fold change threshold) in all three organisms results in only a small increase to 39 ortholog groups with congruent expression pattern (see Additional File 2, bold gene names). This suggests that the magnitude of the response is not the primary reason the conserved stimulon is so small. Even our most permissive analysis suggests that there are more genes in 1-1 ortholog groups that are differentially expressed in a subset of the organisms than congruent across all three.
Transcriptional response to O2 limitation for genes orthologous in the two phytopathogens and not shared with E. Coli
D. dadantii and P. atrosepticum share lifestyle characteristics, including a plant-host environment, not common to E. coli. They are also more closely related to each other and have acquired genes by lateral transfer events along the shared branch since their divergence from E. coli. We examined the set of orthologs shared by the two phytopathogens, but absent from E. coli to examine which if any of these lineage-specific genes are O2-responsive. A total of 780 OrthoMCL groups include genes from both phytopathogens and none from E. coli, and 716 of these are simple 1-1 ortholog groups. Only 22 genes without E. coli orthologs are differentially expressed with fold changes greater than 3 in both D. dadantii and P. atrosepticum (see Table 1). Of these 5 are up-regulated and encode butanediol dehydrogenase, lactoylglutathione lyase, a nickel transporter (different from the E. coli nikABCDE nickel transport system), a putative membrane protein and a hypothetical protein. All of the 17 down-regulated genes encode proteins that constitute transport systems, and many of them likely transport iron. These are discussed in greater detail in later sections.
Transcriptional response of genes shared with E. Coli and only one of the phytopathogens
Forty P. atrosepticum genes shared with E. coli but not with D. dadantii are O2 responsive in P. atrosepticum in our experiments, and for 26 of them transcript levels change more than 3-fold between the conditions. Similarly, 70 D. dadantii genes that have orthologs in E. coli but not in P. atrosepticum are differentially expressed in D. dadantii, of which 18 show fold changes > 3 (see Additional File 4).
Each phytopathogen shows a distinct response to O2 limitation - differential expression of genes without orthologs in the other organism
D. dadantii and P. atrosepticum each encode a substantial number of genes that are not predicted to have orthologs in the other phytopathogen or in E. coli. There are 1267 D. dadantii protein-coding genes in the ASAP database that were not found in OrthoMCL groups. Of these, 501 are differentially expressed and 142 show fold changes greater than 3. These 142 genes include several which were previously implicated in virulence or growth and survival in plant hosts (Additional File 5), and other recognizable biological processes, but 57 genes encode proteins of unknown function underscoring our incomplete understanding of the response to O2 limitation. In P. atrosepticum, there are 1130 ungrouped protein-coding genes, of which 144 are differentially expressed and 73 show fold changes greater than 3. The 73 genes include the coronofacic acid synthesis genes all of which are > 3- fold up-regulated, three genes that encode putative oxidoreductases, at least ten genes that encode putative exported proteins, and 13 genes of unknown function (Additional File 6).
We attribute the larger number of unique O2-regulated genes from D. dadantii to the smaller variance among replicates. But even using the numbers from P. atrosepticum, the number of genes (144 and 73) in the organism-specific transcriptional response is comparable to the number of genes (81) in the conserved transcriptional response, or greatly exceeds it, if we limit the core to the 20 differentially expressed genes with fold changes greater than 3 that are shared across all three organisms.
The anaerobic stimulon
Global transcriptional regulators associated with anaerobiosis
In E. coli, the key O2 responsive transcriptional regulators include FNR, ArcAB, NarXL, and NarPQ. FNR is an iron-sulfur cluster-containing protein that dimerizes in the absence of O2, and can act as either a transcriptional activator or repressor . The amino acid sequence of FNR from the two phytopathogens is 97.6% identical to the E. coli protein with no differences in important functional domains. Surprisingly, fnr is more down regulated in D. dadantii under anaerobic conditions than either P. atroscepticum or E. coli. If this decreased expression is mediated by FNR as it is in E. coli, it suggests that FNR represses its own synthesis to a much greater extent in D. dadantii than in E. coli.
ArcB is the sensor kinase of the two-component regulatory system, ArcAB, which detects signals emanating from the aerobic respiratory chain, such as the oxidation state of ubiquinones . Under anaerobic conditions ArcB phosphorylates ArcA, activating its site-specific DNA binding activity, which primarily represses genes required for aerobic metabolism . D. dadantii arcB contains a nonsense mutation at codon 383, a change that is predicted to interfere with the multi-step phosphorelay transfer to ArcA . It is possible that ArcA is partnered with a different sensor kinase in D. dadantii, or that this organism does not have a functional ArcA-mediated regulatory system. Curiously, arcA is up-regulated in D. dadantii, while arcA and arcB transcript levels are unaffected in P. atrosepticum and E. coli.
NarPQ and NarXL are paralogous two-component regulatory systems associated with nitrate/nitrite regulation under anaerobic conditions in E. coli. P. atrosepticum encodes NarXL and NarPQ of which the latter is missing in D. dadantii. The sensor kinase NarX responds to higher concentrations of nitrate than its paralog NarQ, which also responds to nitrite and aeration [25, 26]. While each sensor kinase can phosphorylate both response regulators, dephosphorylation to the inactive state for NarP is restricted to the cognate partner [27, 28]. Both response regulators activate genes associated with nitrate and nitrite catabolism and repress genes involved in other anaerobic respiratory and fermentative pathways. In P. atrosepticum, narPQ is up-regulated, while its ortholog in E. coli, although not detected as differentially expressed in Kang et al., is up-regulated less than 3- fold.
TCA cycle and glycolysis
Many genes associated with the central metabolic enzymes of glycolysis and the aerobic TCA cycle, have simple 1-1-1 orthologous relationships. While the transcription patterns are largely conserved, however, most are congruent between E. coli and P. atrosepticum but different in D. dadantii (Figure 2). For example, all four genes of the TCA cycle enzyme succinate dehydrogenase (sdhCDAB) are > 3-fold down-regulated in E. coli and P. atrosepticum but in D. dadantii only sdhD is down-regulated. These differences may be related to the arcB mutation found in D. dadantii, which may prevent the anaerobic repression of ArcA regulated genes. In genes that encode isozymes, such as aconitase, all three organisms have at least one isozyme with a conserved response. The fumarase paralogs (fumA, fumB) are the only TCA cycle genes that do not share simple orthology relationships according to OrthoMCL, although the fumC isozymes do. Nevertheless, there is some level of anaerobically induced expression in one of the isozymes across all three organisms (Figure 2).
Conserved genes that participate in fermentation include those associated with the reduction of pyruvate to lactate (ldhA) or the non-oxidative conversion to acetyl-coenzyme-A and formate by pyruvate formate lyase (PFL, pflB), and the subsequent conversion to acetate and ethanol (pta, ackA and adhE). In E. coli, fermentative lactate dehydrogenase (ldhA) is induced under low pH ; however under our growth conditions and in the Kang et al. data, no significant changes in gene expression were detected in any organism. The remaining genes pflB and adhE are up-regulated in all three, however, only pta is upregulated in E. coli and D. dadantii and ackA in P. atrosepticum (see Figure 2).
Expression of fdhF ( including a paralog of fdhF (ABL-0061761) in P. atrosepticum) is up-regulated in both phytopathogens, but remains unaffected in E. coli where FHL expression is known to be dependent on formate and an acidic pH under fermentative conditions (Figure 2). FHL is regulated by the formate-dependent regulator (fhlA) and sigma-54 [30, 31], but are not differentially expressed in E. coli. The E. coli FhlA regulon includes many components (for example, hydrogenases 3, 4) missing in the phytopathogens, and is an obvious example of regulatory divergence where the network is not only smaller in the phytopathogen lineage, but is transcriptionally divergent (Figure 2). Additionally, OrthoMCL groups flhA with the E. coli regulator of hydrogenase 4 (hyfR), together with single genes, which are differentially expressed, in D. dadantii and P. atrosepticum.
Two additional E. coli formate dehydrogenase and hydrogenase isozymes linked to respiration are discussed in the following section. Unlike the phytopathogens, E. coli formate dehydrogenases are the only proteins in E. coli that require selenocysteine for assembly and maturation [32, 33, 34]. The specific requirement for selenocysteine in the oxidation of formate has been shown for formate dehydrogenase-H (fdhF) where replacement with a cysteine residue resulted in a 20-fold less active protein than the wild-type [32, 33]. The phytopathogens are missing genes for tRNASec (selC), selenocysteine synthase, the specialized translation elongation factor and the selenophosphate synthase (selD). The phytopathogens encode cysteine residues into each of the formate dehydrogenases raising the question of whether they have reduced specific activity.
Both D. dadantii and P. atrosepticum have a budAB operon (encoding alpha-acetolactate decarboxylase and acetolactate synthase) adjacent to a divergently transcribed budR-like gene encoding a LysR family transcriptional regulator and budC (encoding 2, 3-butanediol dehydrogenase). In our experiments budC is up-regulated in both phytopathogens. The bud genes enable fermentative butanediol production, a feature that limits the channeling of pyruvate to acid producing pathways and thus counteracts the lethal effects of acidification . It has recently been shown that during soft-rot infection the bud genes play the essential role of increasing the pH of the plant apoplast to facilitate activity of pectate lyases . The pathway is not present in Escherichia or Salmonella but is present in other members of the enterobacteria such as Serratia, Enterobacter, Erwinia, and Klebsiella . In these organisms, butanediol is involved in interactions among plant, animal and insect hosts by acting as a signaling molecule. The mechanism of insect-attraction has been described . It has been shown to produce an anti-inflammatory response in endotoxin-induced lung injury in rats [39, 40].
Aerobic and anaerobic respiration
In the absence of O2, E. coli is able to reduce a variety of alternate electron acceptors, including fumarate, dimethyl sulfoxide (DMSO), trimethylamine N-oxide (TMAO), nitrate and nitrite , to conserve energy through anaerobic respiration using electrons from a variety of donors. The ability to respire fumarate, nitrate and nitrite anaerobically has been reported for D. dadantii and P. atrosepticum  and a subset of the pathways are conserved with E. coli, for example, fumarate reductase (frdABCD) and nitrate reductase (narGHI). Other nitrate/nitrite reductases, which have a complex evolutionary history, are discussed in more detail below. There are no phytopathogen orthologs of the E. coli DMSO reductase genes (dmsABC), the two TMAO reductases (torCAD and torYZ) and their associated regulators (torR, torS and torT).
E. coli has three nitrate reductases and two nitrite reductases, which are part of the NarL and NarP regulons (see Figure 2). These include two membrane-bound proton-translocating nitrate reductases, (narGHJI and narZYWV operons), the periplasmic nitrate reductase (napFDAGHBC operon), a formate-dependent respiratory nitrite reductase (nrfABCDEFG) and the NADH-dependent nitrite reductase (nirBDC). The E. coli narGHJI operon has orthologs in both phytopathogens, but narZYWV does not. The narGHJI operon is up-regulated in both D. dadantii and P. atrosepticum, but is not differentially expressed in E. coli. However, narGHI in E. coli is known to be induced by FNR during anaerobic growth and further induced by NarL [43, 44]. Although there are predicted binding sites for NarL and FNR in the regulatory region of narG in the phytopathogens, the sequence upstream of the conserved FNR binding site, 53 bp from the transcriptional start, has diverged.
We believe nirB gene was erroneously assigned in D. dadantii, because this gene is part of a larger cluster conserved among the two phytopathogens that include a transcriptional regulator (nasR), an ABC transporter (nasF, nasE and nasD), two subunits of an assimilatory nitrate reductase (nasB and nasA), and an uroporphyrin-III C-methyltransferase (nirE). We hesitate to assign orthology of the nas systems between the phytopathogens because the genome context is not conserved beyond the nas subsystem itself, but it is clear that the phytopathogen loci are more structurally conserved (order and content) than either is with the nir system of E. coli (Figure 4).
E. coli has two characterized aerobic terminal reductases that are conserved in the phytopathogens. These are: the cytochrome bo3 type quinol oxidase (encoded by cyoABCD), which has a low affinity for O2, and the high-affinity bd-type cytochrome oxidase (encoded by cydABCD). The cyoABCD operon is down-regulated in E. coli and P. atrosepticum, but shows very little change in D. dadantii. The cydABCD operon is only up-regulated in D. dadantii. The cytochrome oxidase loci are regulated in part by ArcAB in E. coli, which may explain why D. dadantii shows the most divergent expression patterns. Additional Cyt bd type oxidases are encoded in E. coli (appCB) and P. atrosepticum, which may be strain-specific, as suggested by their genomic context and OrthoMCL clustering
E. coli encodes additional cytochromes with predicted roles in the electron transport chain. One of which, yceJ, is > 3-fold down-regulated in all three organisms under anaerobic conditions. The others show divergent patterns of gene expression. The gene cybBD (cytochrome b561) groups with yodB in E. coli, and single orthologs from both phytopathogens, of which only the D. dadantii ortholog is differentially expressed. The E. coli cybC, a pseudogene in strain MG1655, is a soluble cytochrome b562 of unknown function  but has orthologs that are divergently expressed in the phytopathogens (Additional File 8).
In E. coli there are 15 known dehydrogenases that donate electrons to the respiratory chain . Some of these have already been mentioned, but most are conserved in the phytopathogen lineage (sdhCDAB, mqo, glpD, ndh, nuo). Only the operon encoding NADH dehydrogenase I, nuo, is differentially expressed in the phytopathogens while ndh remains unaffected by O2.
There are two systems for the oxidation of hydrogen under anaerobic conditions in E. coli: the Tat dependent periplasmic uptake  hydrogenases hydrogenase 1 and hydrogenase 2 that are involved in hydrogen oxidation coupled to quinone reduction under anaerobic conditions. Like the FHL-type hydrogenases there are differences in the structure and content of these loci between E. coli and the phytopathogens (Figure 3). There are no phytopathogen orthologs for any genes of the E. coli hydrogenase 1 operon, hyaABCDEF. In P. atrosepticum, all hydrogenase related genes are clustered in a single chromosomal locus that OrthoMCL groups with E. coli hydrogenase 2. In E. coli, several genes of the hydrogenase 1 and 2 are up-regulated with others changing in a congruent though not statistically significant way. In D. dadantii and P. atrosepticum, all genes associated with the hydrogenase 2-like system are up-regulated.
It appears that much of the transcriptional variation between the phytopathogens and E. coli with respect to metabolic responses to O2, is mainly the result of changes in gene content, genome rearrangements of both important regulatory and metabolic components of the anaerobic stimulon, specifically in relation to the loss of the NarP regulon and a mutation in ArcB in D. dadantii and also of genes involved in nitrate/nitrite metabolism. Variation in the components of the respiratory formate dehydrogenases and hydrogenases in both phytopathogens, exhibit more complex evolutionary histories with equivalent functions carried out by paralogous or analogous systems, and many of these exhibit differential responses to O2 across these organisms. The only obvious energy metabolism subsystem present in the phytopathogens that is missing from E. coli involves butanediol fermentation.
Sequences of the arcB locus from D. dadantii strains in our lab confirm the arcB nonsense mutation (data not shown), but we have yet to confirm whether this mutation is found in D. dadantii 3937 strains from other labs, although it is not present in any of the recently sequenced Dickeya species. Whether this mutation is laboratory derived or is a lineage specific event remains to be determined. Regardless, this strain has been successfully used to identify and test various pathogenicity related phenotypes in planta suggesting that it may not affect its ability to macerate host tissues.
Analysis of the expression pattern of genes associated with various types of stress responses reveals interesting similarities and differences among the three organisms. Overall, the patterns are suggestive of a phytopathogen-specific oxidative stress response. For example, ahpC, trxC, sodC and dps encode bona fide virulence factors in bacterial pathogens [49, 50, 51, 52, 53] and they counteract damage due to reactive O2 species. Counterintuitively, these genes are up-regulated in the phytopathogens under anaerobic conditions and as expected, remain unaffected in E. coli. It is possible that prolonged growth in an O2 limited environment simulates a situation that soft-rotting bacteria experience prior to encountering the host oxidative burst. Our data suggest the possibility that both P. atrosepticum and D. dadantii may be able to anticipate and respond to a host induced oxidative environment before its onset, leading to the speculation that in these two phytopathogens the regulatory networks that govern responses to two opposite stressors (anaerobic stress and oxidative stress) may be linked to increase chances of survival in the plant environment. Such anticipatory responses are proposed to occur in organisms living in environments that change in predictable ways . The phytopathogen-lineage specific up-regulation of both the narGHJI genes and the nitrate-dependent formate dehydrogenase genes (fdnGHI) involved in respiratory nitrate reduction, despite the absence of nitrate in our growth media, may also be regarded as an anticipatory response in view of the fact that nitrate is an abundant anion in plant. Similar to E. coli, it is possible that a complex regulatory network that includes PecS , OxyR , FNR, (represses sodC in E. coli), RpoS (induces sodC) as well as regulatory elements of nitrate metabolism may be involved.
Some oxidative stress responsive genes that are down-regulated in the phytopathogens, as expected in an anaerobic environment, are either lineage-specific (ohrR and ohr) or strain-specific (indABC, vfmABCDE) and are not present in E. coli. The expression pattern of several other genes associated with oxidative stress is similar in E. coli and P. atrosepticum, but differs in D. dadantii and for two of these genes, sodA and tpx, ArcA-mediated regulation has been demonstrated for the E. coli ortholog . Orthologs of universal stress proteins and cold shock proteins, which are involved in the response to a variety of environmental stresses in enterobacteria [58, 59, 60, 61, 62], are up-regulated in one or both phytopathogens but remain unaffected in E. coli (see Figure 2)
Metal transport systems
A variety of transition metals including iron, manganese, nickel, zinc and copper are required by bacteria for the activity and stabilty of proteins, including FNR and most of the respiratory enzymes. A large number of genes are devoted to their acquisition, uptake, storage and efflux in order to maintain homeostasis. Since several of the anaerobic respiratory enzymes require a different complement of metals (e.g. nickel) than the aerobic respiratory enzymes, and because unbound iron oxidation states is readily influenced by the O2 status of the growth media, we are not surprised to observe transcriptional changes for associated genes, and some of them are described below.
Each of the three organisms up-regulates at least one nickel uptake system (nikABCDE in E. coli, hoxN in the phytopathogens) indicating an increased requirement for this metal during anaerobiosis. In E. coli the NikABCDE system transports nickel for the NiFe hydrogenases. Genes encoding transport systems for copper and zinc largely show phytopathogen-specific responses. For example, the Cus metal efflux system (cusCFBA operon) is up-regulated in both phytopathogens and is down-regulated in E. coli whereas, the copA gene is down-regulated in both phytopathogens, but not in E. coli. The Cus system and CopA are associated with copper homeostasis under anaerobic and aerobic conditions respectively, in E. coli . Similarly, zinc uptake systems (znuABC) are down-regulated only in the phytopathogens. Additionally, both phytopathogens down-regulate transcripts (> 3-fold) for an ABC transport system predicted to transport zinc (according to the D. dadantii annotations). Genes for this system are not present in E. coli. All three organisms encode a zinc uptake regulator (Zur), the gene for which is down-regulated only in D. dadantii.
All three organisms have a substantial number of genes involved in synthesis and transport of iron chelating siderophores and other iron-containing compounds. Overall, these subsystems are down-regulated in the phytopathogens, and to a lesser extent in E. coli suggesting that the phytopathogens may have a reduced demand for iron during anaerobiosis or may reduce iron levels to avoid damage in the anticipated oxidative environment of the host. Many of these genes are likely regulated by Fur in all three organisms.
Genes that belong to the same ortholog group do not necessarily synthesize the same siderophore although they might be able to transport some of them. For example, OrthoMCL clusters genes for synthesis of the D. dadantii siderophore, chrysobactin, with enterobactin synthesis genes of E. coli, but the siderophores are distinct . D. dadantii does not synthesize enterobactin, although it is capable of uptake and utilization of exogenous enterobactin . The OrthoMCL clusters also include single members from P. atrosepticum, but none of the genes share extended conserved genomic context across organisms, and there is no data on whether P. atrosepticum produces enterobactin, chrysobactin or another siderophore using these genes. It is also established that some bacteria can "steal" siderophores from their neighbors as seen in P. atrosepticum which is unlikely to produce achromobactin, but may be able to uptake and transport it via genes orthologous to the D. dadantii cbrABCD . A detailed comparison of the D. dadantii and P. atrosepticum iron homeostasis systems is found in Franza and Expert .
Genes for several iron transporting ABC transport systems shared only between the phytopathogens are all largely down-regulated in both organisms (for example OrthoMCL groups 3161-3163, yfeABCD, sfuABC ). In E. coli, under anaerobic conditions uptake of ferrous iron is expected to increase relative to oxidized ferric iron. In line with this expectation, the feoAB genes that are involved in transport of ferrous iron are up-regulated in D. dadantii. However, in the Kang et al. experiments, neither the efeBOU genes nor the feoAB genes, which encode ferrous iron transporters were up-regulated. Several genes encoding iron storage proteins are up-regulated in the phytopathogens and remain unaffected or are down-regulated in E. coli (see Additional File 7). These include the bfr gene, encoding a bacterioferritin which contributes differentially to the virulence of D. dadantii depending on the host  and dps which encodes a ferritin-like protein, Dps, with pleiotropic functions. D. dadantii also up-regulates transcripts for a second strain-specific Dps-like protein (ABF-0015905).
Most of the transporters for amino acids are down-regulated in the phytopathogens, and unaffected in E. coli. Only one gene (sulfate transporter; yfbS) shows > 3-fold congruent up-regulation between the phytopathogens and a sodium-serine transporter gene (sst) is up-regulated both in E. coli and P. atrosepticum. Even though OusA has been implicated in anaerobiosis in D. dadantii , we did not detect changes in expression for its gene in our experiments.
In P. atrosepticum and D. dadantii, pathogenicity is largely due to their capacity to depolymerize plant cell wall polymers including cellulose, hemicellulose and pectic substances, as well as other components such as lignin and proteins, through the coordinate production of multiple cell wall degrading enzymes (CWDE). Because of their indispensable role in pathogenicity, expression of CWDE is strictly regulated at the transcriptional level by multiple regulators and further fine tuned by environmental factors including pH, osmolarity and O2 concentrations all of which influence the successful onset of disease symptoms . Anaerobic regulation in the presence of an inducer has been demonstrated for pelA, pelD, pelE and pelL  in D. dadantii. Except for pelD none of these four genes has an ortholog in P. atrosepticum (see Additional File 7). As expected for cells grown in non-inducing conditions, all the reported pectate lyase genes (pelA to E, L, Z, I) are repressed in D. dadantii. This trend is not seen in P. atrosepticum where in fact, one gene, pelB, is up-regulated. The only CWDE-encoding gene that is down-regulated in P. atrosepticum is a pectin lyase, pnl, which is a member of a complex ortholog group. It is interesting to note that a D. dadantii -specific gene, xynA (ABF-0019026), encoding a putative endoxylanase, characterized in a related corn pathogen , is > 3-fold up-regulated.
Phytopathogen secretion systems
Both D. dadantii and P. atrosepticum encode a diverse collection of secretion systems, several of which are known to play key roles in interaction with plant hosts. The T6SS shows one of the most dramatic divergent expression patterns observed in our experiments. The T6SS mediates secretion of proteins encoded within repetitive clusters of genes found distributed throughout the genome, often, though not reliably, annotated as hcp and vgrG. OrthoMCL clusters paralogs of each type into two groups. The hcp cluster includes three members from P. atrosepticum and two from D. dadantii. The vgrG cluster includes three members from D. dadantii and five from P. atrosepticum. The D. dadantii T6SS is down-regulated in the absence of O2 and the P. atrosepticum T6SS is up-regulated. A previous report demonstrated that the P. atrosepticum T6SS cluster and proteins secreted via the T6SS (four hcp genes and three vgrG genes) are induced by plant host extracts [73, 74]. Mutants of two genes, believed to correspond to a structural component of the secretion apparatus (vasK) and a sigma-54 dependent regulator (vasH), showed increased virulence relative to wild-type, a phenotype attributed to increased growth (higher density) and associated increases in pectic enzyme production in the mutants. In our analyses, vasK and vasH are up-regulated, and the corresponding orthologs in D. dadantii are both down-regulated, typical of the T6SS clusters as a whole. All members of both hcp and vgrG groups are up-regulated and exhibit an expression pattern congruent with the T6SS in each organism.
Under the anaerobic conditions used in our experiments, genes for type I secreted proteases (PrtA, PrtB, PrtC, PrtG) and for their accessory proteins (PrtD, PrtE and PrtF) show a similar trend as the CWDE; they are down regulated in D. dadantii, and remain unaffected in P. atrosepticum. The Type II secretion system (T2SS) is responsible for secretion of CWDE as well as several other targets . In addition, it has been linked to iron homeostasis in D. dadantii with interactions between inner membrane components of the T2SS and the machinery for achromobactin synthesis . E. coli has an orthologous secretion system that is not expressed in wild-type E. coli strains but is functional in hns mutant strains , and the corresponding genes are unaffected in E. coli. Genes associated with the T2SS are nearly all down-regulated in D. dadantii and unaffected in P. atrosepticum, where absolute expression levels remain high regardless of O2 availability. D. dadantii also encodes a second locus similar to genes of a T2SS, which is associated with targeting proteins to the outer membrane . Several genes from this system in D. dadantii are also down-regulated. Both phytopathogens encode a Type III secretion system (T3SS), a syringe-like apparatus employed by numerous Gram-negative pathogens to inject bacterial proteins into host cells. Although the T3SS is required for full virulence in D. dadantii  and P. atrosepticum , far fewer secreted effector proteins have been identified in soft-rot associated pathogens than many other bacteria, and some pathogenic Pectobacterium lack a T3SS altogether . The D. dadantii T3SS has also been implicated in multicellular behavior and biofilm formation . Genes associated with the T3SS are largely unaffected in both D. dadantii and P. atrosepticum, with low absolute expression levels with and without O2. However, in D. dadantii several related genes are down-regulated including, hrpS, which encodes a σ54-enhancer binding regulatory protein, two secreted harpin genes, hrpN and hrpW and dspE encoding a T3 secreted effector. None of these genes show a similar response in P. atrosepticum. Genes that encode a putative two partner secreted adhesin and its associated activator/transporter constitute a complex OrthoMCL group that has two D. dadantii genes and three P. atrosepticum genes (includes hecA/B). Only the P. atrosepticum orthologs are up-regulated under anaerobiosis. In E. chrysanthemi strain EC16 a role for HecA in early pathogenesis has been suggested .
Methyl-accepting chemotaxis proteins (MCPs) transduce environmental and cellular signals to the flagella . The D. dadantii genome has 45 genes that encode proteins whose products are annotated as MCPs, while there are 36 such genes in P. atrosepticum. The C-terminal signal transduction domain of MCPs is highly conserved across all members of the family, while the N-terminal sensory domain varies extensively. This complicates reliable prediction of orthology. In our OrthoMCL analysis, 12 out of 15 D. dadantii-specific MCPs and one of 5 P. atrosepticum-specific MCPs are differentially expressed. Of the MCPs shared between the phytopathogens 17 are differentially expressed in D. dadantii and 9 in P. atrosepticum. Setting aside the potential errors with prediction of orthology, there is clearly an O2-availability regulated motility response in both D. dadantii and P. atrosepticum.
Though not an MCP, the E. coli Aer has been associated with aerotaxis via sensing of cellular redox potential using an FAD cofactor . The phytopathogens each have three homologs that show similarity to Aer throughout the entire length of the alignment (aer1, aer2 and ABL-0063893 in P. atrosepticum and aer1, ABF-0014726 and ABF-0014843 in D. dadantii). Both phytopathogens include at least one putative aerotaxis receptor up-regulated and one down-regulated under anaerobic conditions. Transcripts of the E. coli aer gene decrease, though not statistically significantly, during anaerobiosis.
Flagella, motility and attachment
In the soft-rot pathogens, the contribution of flagella to motility is important for virulence [86, 87, 88]. Genes encoding regulatory elements of flagellar assembly as well as some of the genes of the flagellar apparatus are affected under anaerobic conditions in one or both phytopathogens. None of the differentially expressed genes show a similar trend in both phytopathogens (see Additional File 7) suggesting that these genes may be regulated differently during anaerobiosis in these two phytopathogens.
Genes encoding enzymes that produce a range of polysaccharides such as lipopolysaccharides (rfa and waa genes), exopolysaccharides/O-antigens (wza and rfb genes), enterobacterial common antigen (wec and rff genes) and membrane derived oligosaccharides such as periplasmic glucans (opg genes) play important roles in virulence, adhesion, resistance to host-derived compounds and are considered virulence factors in P. atrosepticum and/or D. dadantii [89, 90, 91, 92, 93]. In our experiments, the expression of these genes is mostly unaffected in P. atrosepticum and some of them are down-regulated in D. dadantii.
The P. atrosepticum genome contains a cluster of 9 cfa and cfl genes that are > 3- fold up-regulated in the absence of O2. They encode a putative polyketide biosynthesis system predicted to synthesize a compound similar to coronafacic acid, a component of the coronatine phytotoxin produced by P. syringae , and mutations in the P. atrosepticum genes dramatically reduces virulence on potato. A similar gene cluster was recently characterized from a phytopathogenic Streptomyces, with mutants in the polyketide synthesis system showing reduced virulence on tobacco .
D. dadantii is pathogenic to pea aphids under laboratory conditions and this trait appears relatively widely distributed among species of enterobacteria . Deleting a cluster of four genes encoding proteins similar to cytolytic delta-endotoxins from the gram-positive entomopathogen Bacillus thuringiensis significantly reduced virulence of D. dadantii on aphids. These four genes (typically expressed as a single transcriptional unit) are up-regulated in our experiments. Interestingly these genes were regulated by many of the same regulators that control expression of virulence factors for the plant host but in opposite directions .
Pathogenicity-associated transcriptional regulators (TR)
In P. atrosepticum and D. dadantii, several regulators coordinate expression of virulence factors in response to environmental or physiological conditions [98, 99, 100, 101, 102, 103]. O2 dependent modulation has been demonstrated for very few virulence factors or associated regulators.
In our experiments, D. dadantii and P. atrosepticum each appear to have at least one O2-responsive strain-specific global regulatory gene whose expression is influenced by anaerobiosis. They are a gene for PecM [103, 104, 105, 106, 107] in D. dadantii and that for RdgB in P. atrosepticum [108, 109]. Beside these, the PecS repressed expI gene  coding for the LuxI homolog in D. dadantii for the production of AHL, the expR gene that activates PecS and which encodes the AHL receptor are down-regulated only in D. dadantii, and their orthologs are unaffected in P. atrosepticum. Interestingly, many regulators shared between the phytopathogens are differentially expressed in a lineage-specific pattern that may indicate regulatory divergence for these loci. It is possible that at least some of these regulators respond to host-derived signals under the O2-limiting and inducing conditions encountered within the plant.
Most of the other global regulators of virulence including KdgR [111, 112], Crp [99, 101] and Fur (fur shows statistically significant up-regulation but only minimal (~1.7) fold-change), that are highly conserved between the phytopathogens are unaffected in either organism. They all show moderate levels of expression regardless of O2 availability. Although transcription of several KdgR target genes is affected in our experiments, the involvement of KdgR is unlikely to account for the change in expression for these targets since the inducer 2-keto-3-deoxygluconate (KDG), a pectin degradation compound was not present in our medium. Furthermore other than CWDE genes mentioned above, which are members of complex overlapping regulons, none of the co-regulated transporter genes such as KdgMN and togMNAB are affected in our experiments.
Non-protein coding genes in the anaerobic stimulon
List of small RNAs that are O2-responsive in at least one of the two phytopathogens D. dadantii and P. atrosepticum
ASAP Feature ID
Several other small RNA genes show fairly compelling evidence of differential transcriptional regulation between the two phytopathogens including spf (Spot 42), rygA/omrA and ryeA, and possibly glmY, glmZ, and rsmB. In D. dadantii, one up-regulated small RNA corresponds to the spf gene, an ortholog of the E. coli Spot 42 small RNA, which plays a role in anti-sense mediated down-regulation of the third gene (galK) of the galactose operon [116, 135], and whose expression is known to be affected by carbon source available in the media and is cAMP-CRP responsive . The corresponding gene in P. atrosepticum does not change expression, rather levels are intermediate in both conditions, similar to the E. coli ortholog in the Kang experiments. The rygA gene is > 3-fold up-regulated in D. dadantii. The P. atrosepticum ortholog is not differentially expressed, and trends in the opposite direction. Targets of the two E. coli orthologs, neither of which have previously been implicated in anaerobiosis, include both transcriptionally up and down regulated genes many of which are involved with cell surface structures or functions. They also negatively regulate fepA and fecA, two genes associated with iron homeostasis, and fimbrial genes associated with adhesion and biofilm formation . The ryeB RNA is up-regulated in D. dadantii and down-regulated in P. atrosepticum. In E. coli, RyeB interacts with RyeA, encoded on the opposite strand, to mediate RNAse III-dependent cleavage  and is known to be pH-responsive in O2-limited conditions . We detect ryeA as differentially expressed in D. dadantii, although not in P. atrosepticum, where it trends in the same direction as ryeB. GlmY, a small RNA implicated in amino-sugar metabolism, is detected as up-regulated in D. dadantii and unaffected in P. atrosepticum. Amino sugars are important precursors of the peptidoglycan and lipopolysaccharide components of the cell wall. The rsmB gene is highly expressed under both aerobic and anaerobic conditions in D. dadantii and P. atrosepticum. It is up-regulated in the absence of O2 only in D. dadantii. This gene, like several others, was also not present on the Kang et al. arrays. RsmB has been linked to production of extracellular enzymes, quorum sensing and T3SS in Dickeya as well as in a related species [124, 125, 126]. This collection of lineage-specific expression patterns suggests that altering regulation of small RNAs may be a particularly labile mechanism of regulatory diversification.
We investigated the transcriptional response to O2 under simple controlled laboratory conditions for two soft rot-associated phytopathogenic enterobacteria to begin to enumerate the regulatory and metabolic networks associated with a key environmental parameter that impacts the interaction of these organisms with plant hosts, and to explore the extent of regulatory divergence that occurs among enterobacteria. We analyzed data from D. dadantii and P. atrosepticum individually, and compared them to each other, as well as the model organism E. coli K12, using predicted gene-by-gene orthology and by grouping related genes into subsystems. The latter approach provides insights into larger scale patterns of conserved and lineage-specific biological processes regulated by O2 availability.
The O2-responsive stimulon for each organism is large, and includes genes conserved across the family enterobacteria, as well as lineage-specific and organism-specific genes that were likely acquired through lateral gene transfer events. Some conserved genes show a conserved response to O2, but others vary across organisms in the magnitude or even direction of response. D. dadantii and P. atrosepticum are more closely related to each other than to E. coli K12, and overall, their gene expression profiles are more conserved in terms of total number of orthologous genes (including those not shared with E. coli) responding in a congruent way, and in the proportion of genes shared across all three organisms responding in a congruent way.
A subset of genes show a different trend, with the expression profile more similar between E. coli and P. atrosepticum, with D. dadantii behaving differently. We attribute many of these to the likely deficiency of ArcAB regulation in D. dadantii where the sensor kinase of this two-component regulatory system is a pseudogene; however, it is not possible from these experiments to rule out that ArcA may be partnered with a different signal transducer, opening up the possibility that existing binding sites for ArcA could be coopted to respond to an entirely different signal.
A relatively small number of conserved genes are > 3-fold up-regulated or down-regulated in all three organisms. While these include some known components of the well characterized E. coli anaerobic stimulon, other important components are missing from this conserved core. Detailed investigation of the orthology relationships and the subsystem-based approach reveal a broader group of processes implicated in a net conserved response to O2, but with variation across the organisms in the number of functionally redundant paralogs and/or non-paralogous isofunctional subsystems, which we refer to as changes in subsystem architecture. For example, hydrogenases are highly up-regulated under anaerobic conditions in all three organisms, but the genes involved show complex homology relationships, involving duplications, and or deletions, as well as genome rearrangements, that obscure the common response. We detailed these types of relationships and the associated expression patterns for subsystems involved with regulation, metabolism and a variety of processes associated with interactions of the phytopathogens with plant hosts.
Our results indicate that the O2-responsive gene network includes a variety of virulence and pathogenicity-relevant processes including secretion, response to environmental stress, metal homeostasis, and taxis. Further experiments aimed at investigating the specific role of O2 regulation of these biological processes may be fruitful. Several virulence-associated subsystems exhibit strikingly divergent behavior in the two phytopathogens, most notably the T6SS. More subtle differences, like down-regulation of the complete complement of pectate lyases in D. dadantii, which are largely unaffected in P. atrosepticum, may lead to insights into the differences in virulence of these two phytopathogens under O2-limited conditions, but this requires exploration under a broader number of experimental conditions. The experimental conditions explored here are very limited, and the number of genes in the plant-pathogen anaerobic stimulons will only increase as additional variables (carbon sources, nitrate availability, intermediate O2 levels, time series of shifting O2 availability, etc.) are investigated.
The genes differentially expressed in one or both phytopathogens include known targets of regulators associated with quorum sensing, oxidative stress, iron homeostasis, nitrate/nitrite, and carbohydrate availability, as well as the established key regulators of anaerobiosis under the experimental conditions we chose, FNR and ArcAB. The underlying regulatory network behind the O2-responsive stimulon we have described is complex, involving a larger number of regulators, and it clearly differs between the two phytopathogens. Further dissection of this network bioinformatically will certainly require simultaneous consideration of a large number of regulators. We do not attempt a comprehensive dissection in this paper. We observe that at least one known key regulator of virulence genes, PecM, is among the genes differentially affected between D. dadantii and P. atrosepticum.
Several small regulatory RNAs are also differentially expressed under aerobic and anaerobic conditions, including ones that are found in E. coli, but have not previously been characterized as O2-responsive. These expression patterns should be experimentally validated using additional techniques, both because of their novelty, and because small genes may be particularly subject to measurement errors using arrays.
Many of the organism-specific and phytopathogen-specific genes in the anaerobic stimulon were likely acquired through lateral gene transfer events. These genes may have become a part of the stimulon in a variety of ways. A lateral transfer can include regulatory elements from the donor that also function in the recipient or a laterally acquired gene could be integrated into the recipient genome in a way that makes use of native O2-responsive regulatory elements. Alternately, evolutionary events (point mutations, rearrangements, further lateral transfers in the same region) subsequent to acquisition of a gene could render it O2-responsive. We expect there to be examples of all of these, but further examination awaits comparison with additional representatives of each genus from ongoing genome projects that will permit more precise definition of the boundaries of laterally acquired elements and reconstruction of the ancestral regulatory states. Finally, these experiments addressed only a single representative of each species, and further investigation will be required to determine whether the expression patterns we observed here are typical of each species, and which aspects of the anaerobic stimulon vary within each species.
Bacterial growth and RNA extraction
We grew three replicates each of Dickeya dadantii 3937 and Pectobacterium atrosepticum SCRI1043 in MOPS minimal medium (purchased from Teknova, Inc.) supplemented with 0.1% glucose at 30°C and 23°C respectively. Overnight cultures of each bacterium were diluted to an O.D.600 of 0.05 in fresh medium and cultured in a gas sparging system  apparatus described in Kang et al.) that permits precise control over the mixtures of O2, N2 and CO2. Cultures were grown to early log phase under aerobic (70% N2, 25% O2 and 5% CO2) and anaerobic (95% N2 and 5% CO2) conditions. 20 ml samples were collected in tubes containing 2 ml phenol-ethanol. RNA was extracted using the hot-phenol method . Quality of RNA samples was assessed using the Agilent Bioanalyzer 2100 nanochip system (Agilent Technologies).
Microarray design and hybridization
334647 genome-specific probes for the D. dadantii and 344859 probes for the P. atrosepticum genomes were selected using chipD (target melting temperature 78°C, target probe length 40 to 70 -mers, interval size 12)  and oligonucleotide arrays were synthesized by Nimblegen Inc. Procedures described in the Nimblegen Arrays User's guide (http://www.nimblegen.com/products/lit/05434505001_NG_Expression_UGuide_v6p0.pdf) were followed for cDNA synthesis, labeling and hybridization. Arrays were scanned at 532 nm and signals were extracted using NimbleScan software (NimbleGen, Inc.).
Analysis of gene expression data
Signals from the 3 replicates of the hybridization experiments from each organism were normalized using RMA  implemented in the NimbleScan software and imported into a custom MicroSoft Access database. Normalized signal intensities for each set of three replicates are in good agreement (R2 ranging from 0.92-0.99). The median of signals from multiple probes (from coding and non-coding strand) for each individual gene was calculated using R, from which log2 expression values were derived. To determine directional changes in gene expression, log2 ratios were determined by calculating the difference between the log2 median anaerobic signal and log2 median aerobic signal.
To identify differentially expressed genes between aerobic and anaerobic growth, an empirical Bayesian analysis, EBArrays  was executed within the free statistical analysis software package R  and Bioconductor v2.1 . The posterior probability for each pattern was calculated using a hierarchical log-normal normal expression model with the conditional false discovery rate (cFDR) at 0.01 to determine the appropriate threshold (cFD(τ)). The E. coli WT data set for aerobic and anaerobic conditions derived from Kang et al. was reanalyzed to identify differentially expressed genes as described above. The critical thresholds for the three data sets are as follows:
D. dadantii : 0. 8734342, P. atrosepticum: 0.9067187, E. coli: 0.9289846. Throughout the manuscript, we have used the term "differentially expressed" to denote genes that qualify our permissive criterion namely the probability for observed differential expression is higher than the critical threshold, for the dataset, when the conditional false discovery rate is set at 0.01 and the prefix "highly" is used to denote differentially expressed genes whose transcripts show more than a 3-fold change between the conditions (stringent criteria). Genes whose transcript abundance is higher under anaerobic conditions are referrred to as "up-regulated" and those with lower transcript abundance under anaerobic conditions are referred to as "down-regulated". In some instances, fold change values for poly-cistronic operons that are conserved across all three organisms are averaged across genes to simplify our analyses. Additional File 9, Additional File 10, and Additional File 11 contain the complete datasets for D. dadantii, P. atrosepticum and E. coli, respectively.
Analysis of expression data for non protein coding RNA (small RNA)
Our oligonucleotide arrays included probes tiled across the entire genomes for both D. dadantii and P. atrosepticum. Thus, we are also able to analyze the O2-response for genes typically not included in gene expression arrays, like those that encode small RNAs (sRNAs), even if they were not present in the original genome annotations. OrthoMCL considers only protein-coding genes, and there are not comprehensive predictions of orthology for non-coding RNAs (ncRNAs) or tRNAs in the ASAP database (the sRNAbase  has predictions for 34 small RNAs in P. atrosepticum) so we do not include them in the cross-species analyses above. Instead, we used BLASTn to find orthologs in both phytopathogens on a case-by-case basis. Most small RNA genes discussed in the results and discussion are missing or incorrectly annotated in at least one of the two phytopathogen GenBank genome sequences. We have corrected them in the ASAP database. These RNAs are short, and consequently, inferences about expression patterns are based on a relatively small number (typically around 10) of probes. We manually investigated probe behavior consistency in most cases, and found the expression patterns persuasive. All small RNA genes detected as differentially expressed in either organism are shown in Table 3.
Sequences and annotations for predicted protein-coding genes for D. dadantii 3937, P. atrosepticum SCRI1043 and E. coli were obtained from the ASAP database  and clustered using OrthoMCL  using default parameters. OrthoMCL clustered genes from each of the three organisms into simple or complex ortholog groups depending on whether one or more than one ortholog is present for the gene in the organsims being compared as described using the following example. E. coli encodes paralogous genes for fumarase, fumA and fumB that are homologous to the gene identified as fumA in the phytopathogens. All of these four genes are clustered together in a single orthologous group (complex group) by OrthoMCL. The third E. coli fumarase isozyme, encoded by fumC has a simple 1-1-1 relationship with fumC of D. dadantii and P. atrosepticum and these three genes are clustered together as a different group (simple group). Strain-specific genes do not belong to any OrthoMCL group. Additional File 8 lists all protein-coding genes from all three organisms, OrthoMCL group identifiers, and counts of the number of members of the group from each organism along with a short form of the experimental data results. Operon structures and known and predicted regulator binding sites for E. coli were obtained from EcoCyc  unless otherwise noted.
JA is supported by the NHGRI training grant to the “Genomic Sciences Training Program” (5T32HG002760) and the NLM training grant to the “Computation and Informatics in Biology and Medicine Training Program” (NLM 5T15LM007359). Both LB and JA are also supported by the “Molecular Evolution of Microbial Pathogen Genomes”, NIH R01-GM62994, as were PL and JDG (P.I. NTP; Co-P.I.s JDG, PJK). LB was also supported in part by a Vilas Life Cycle Award to NTP. We would like to thank the Gene Expression Center at the University of Wisconsin-Madison for their support in providing technical support, facilities and equipment for use in our research.
- 1.Perombelon MCM: Potato diseases caused by soft rot erwinias: an overview of pathogenesis. Plant Pathol. 2002, 51 (1): 1-12. 10.1046/j.0032-0862.2001.Short title.doc.x.Google Scholar
- 2.Lemattre P: 1972, Chimie horticole, Nouvelle éd. edn. Paris,: J.-B. BaillièreGoogle Scholar
- 3.Harrison MD, Nielsen LW: Blackleg and bacterial soft rot. Compendium of Potato Diseases. Edited by: Hooker WJ, St. Paul MN. 1981, American Phytopathological Society, 27-29.Google Scholar
- 5.Perombelon MCM, Kelman A: Ecology of the Soft Rot Erwinias. Annu Rev Phytopathol. 1980, 18 (1): 361-387. 10.1146/annurev.py.18.090180.002045.Google Scholar
- 6.Kotoujansky A: Molecular Genetics of Pathogenesis by Soft-Rot Erwinias. Annu Rev Phytopathol. 1987, 25 (1): 405-430. 10.1146/annurev.py.25.090187.002201.Google Scholar
- 7.Robert-Baudouy J: Molecular biology of Erwinia: from soft-rot to antileukaemics. Trends Biotechnol. 1991, 9 (1): 325-329. 10.1016/0167-7799(91)90103-O.Google Scholar
- 15.Constantinidou C, Hobman JL, Griffiths L, Patel MD, Penn CW, Cole JA, Overton TW: A reassessment of the FNR regulon and transcriptomic analysis of the effects of nitrate, nitrite, NarXL, and NarQP as Escherichia coli K12 adapts from aerobic to anaerobic growth. J Biol Chem. 2006, 281 (8): 4802-4815.PubMedGoogle Scholar
- 16.Overton TW, Griffiths L, Patel MD, Hobman JL, Penn CW, Cole JA, Constantinidou C: Microarray analysis of gene regulation by oxygen, nitrate, nitrite, FNR, NarL and NarP during anaerobic growth of Escherichia coli : new insights into microbial physiology. Biochem Soc Trans. 2006, 34 (Pt 1): 104-107.PubMedGoogle Scholar
- 17.Ravcheev DA, Gerasimova AV, Mironov AA, Gelfand MS: Comparative genomic analysis of regulation of anaerobic respiration in ten genomes from three families of gamma-proteobacteria (Enterobacteriaceae, Pasteurellaceae, Vibrionaceae). BMC Genomics. 2007, 8: 54-10.1186/1471-2164-8-54.PubMedCentralPubMedGoogle Scholar
- 20.Overbeek R, Begley T, Butler RM, Choudhuri JV, Chuang HY, Cohoon M, de Crecy-Lagard V, Diaz N, Disz T, Edwards R: The subsystems approach to genome annotation and its use in the project to annotate 1000 genomes. Nucleic Acids Res. 2005, 33 (17): 5691-5702. 10.1093/nar/gki866.PubMedCentralPubMedGoogle Scholar
- 27.Noriega CE, Lin HY, Chen LL, Williams SB, Stewart V: Asymmetric cross-regulation between the nitrate-responsive NarX-NarL and NarQ-NarP two-component regulatory systems from Escherichia coli K-12. Mol Microbiol. 2010, 75 (2): 394-412. 10.1111/j.1365-2958.2009.06987.x.PubMedCentralPubMedGoogle Scholar
- 34.Sawers RG, Blokesch M, Böck A: Anaerobic Formate and Hydrogen Metabolism. EcoSal-Escherichia coli and Salmonella: Cellular and Molecular Biology. Edited by: Böck A, Curtiss R III, Kaper JB, Karp PD, Neidhardt FC, Nyström T, Slauch JM, Squires CL, Ussery DA. 2004, Washington: ASM PressGoogle Scholar
- 42.Dickey RS: Erwinia chrysanthemi: A Comparative Study of Phenotypic Properties of Strains from Several Hosts and Other Erwinia Species. Phytopathology. 1979, 69: 324-329. 10.1094/Phyto-69-324.Google Scholar
- 46.Unden G, Dünnwald P: The Aerobic and Anaerobic Respiratory Chain of Escherichia coli and Salmonella enterica: Enzymes and Energetics. EcoSal-Escherichia coli and Salmonella: Cellular and Molecular Biology. Edited by: Böck A, III RC, Kaper JB, Karp PD, Neidhardt FC, Nyström T, Slauch JM, Squires CL, Ussery DA. 2004, Washington: ASM PressGoogle Scholar
- 50.Battistoni A, Pacello F, Folcarelli S, Ajello M, Donnarumma G, Greco R, Ammendolia MG, Touati D, Rotilio G, Valenti P: Increased expression of periplasmic Cu, Zn superoxide dismutase enhances survival of Escherichia coli invasive strains within nonphagocytic cells. Infect Immun. 2000, 68 (1): 30-37. 10.1128/IAI.68.1.30-37.2000.PubMedCentralPubMedGoogle Scholar
- 51.Uzzau S, Bossi L, Figueroa-Bossi N: Differential accumulation of Salmonella[Cu, Zn] superoxide dismutases SodCI and SodCII in intracellular bacteria: correlation with their relative contribution to pathogenicity. Mol Microbiol. 2002, 46 (1): 147-156. 10.1046/j.1365-2958.2002.03145.x.PubMedGoogle Scholar
- 53.Piddington DL, Fang FC, Laessig T, Cooper AM, Orme IM, Buchmeier NA: Cu, Zn superoxide dismutase of Mycobacterium tuberculosis contributes to survival in activated macrophages that are generating an oxidative burst. Infect Immun. 2001, 69 (8): 4980-4987. 10.1128/IAI.69.8.4980-4987.2001.PubMedCentralPubMedGoogle Scholar
- 55.Hommais F, Oger-Desfeux C, Van Gijsegem F, Castang S, Ligori S, Expert D, Nasser W, Reverchon S: PecS is a global regulator of the symptomatic phase in the phytopathogenic bacterium Erwinia chrysanthemi 3937. J Bacteriol. 2008, 190 (22): 7508-7522. 10.1128/JB.00553-08.PubMedCentralPubMedGoogle Scholar
- 64.Franza T, Expert D: Iron Uptake in Soft Rot Erwinia. Iron Uptake and Homeostasis in Microorganisms. Edited by: Cornelis P, Andrews SC. 2010, Norfolk, UK: Caister Academic Press, 292-Google Scholar
- 66.Mahe B, Masclaux C, Rauscher L, Enard C, Expert D: Differential expression of two siderophore-dependent iron-acquisition pathways in Erwinia chrysanthemi 3937: characterization of a novel ferrisiderophore permease of the ABC transporter family. Mol Microbiol. 1995, 18 (1): 33-43. 10.1111/j.1365-2958.1995.mmi_18010033.x.PubMedGoogle Scholar
- 68.Boughammoura A, Matzanke BF, Bottger L, Reverchon S, Lesuisse E, Expert D, Franza T: Differential role of ferritins in iron metabolism and virulence of the plant-pathogenic bacterium Erwinia chrysanthemi 3937. J Bacteriol. 2008, 190 (5): 1518-1530. 10.1128/JB.01640-07.PubMedCentralPubMedGoogle Scholar
- 73.Mattinen L, Somervuo P, Nykyri J, Nissinen R, Kouvonen P, Corthals G, Auvinen P, Aittamaa M, Valkonen JP, Pirhonen M: Microarray profiling of host-extract-induced genes and characterization of the type VI secretion cluster in the potato pathogen Pectobacterium atrosepticum. Microbiology. 2008, 154 (Pt 8): 2387-2396.PubMedGoogle Scholar
- 78.Liu H, Coulthurst SJ, Pritchard L, Hedley PE, Ravensdale M, Humphris S, Burr T, Takle G, Brurberg MB, Birch PR: Quorum sensing coordinates brute force and stealth modes of infection in the plant pathogen Pectobacterium atrosepticum. PLoS Pathog. 2008, 4 (6): e1000093-10.1371/journal.ppat.1000093.PubMedCentralPubMedGoogle Scholar
- 80.Bell KS, Sebaihia M, Pritchard L, Holden MT, Hyman LJ, Holeva MC, Thomson NR, Bentley SD, Churcher LJ, Mungall K: Genome sequence of the enterobacterial phytopathogen Erwinia carotovora subsp. atroseptica and characterization of virulence factors. Proc Natl Acad Sci USA. 2004, 101 (30): 11105-11110. 10.1073/pnas.0402424101.PubMedCentralPubMedGoogle Scholar
- 83.Rojas CM, Ham JH, Deng WL, Doyle JJ, Collmer A: HecA, a member of a class of adhesins produced by diverse pathogenic bacteria, contributes to the attachment, aggregation, epidermal cell killing, and virulence phenotypes of Erwinia chrysanthemi EC16 on Nicotiana clevelandii seedlings. Proc Natl Acad Sci USA. 2002, 99 (20): 13142-13147. 10.1073/pnas.202358699.PubMedCentralPubMedGoogle Scholar
- 87.Mulholland V, Hinton JC, Sidebotham J, Toth IK, Hyman LJ, Perombelon MC, Reeves PJ, Salmond GP: A pleiotropic reduced virulence (Rvi-) mutant of Erwinia carotovora subspecies atroseptica is defective in flagella assembly proteins that are conserved in plant and animal bacterial pathogens. Mol Microbiol. 1993, 10 (5): 1154-10.1111/j.1365-2958.1993.tb00986.x.PubMedGoogle Scholar
- 89.Bouchart F, Boussemart G, Prouvost AF, Cogez V, Madec E, Vidal O, Delrue B, Bohin JP, Lacroix JM: The virulence of a Dickeya dadantii 3937 mutant devoid of osmoregulated periplasmic glucans is restored by inactivation of the RcsCD-RcsB phosphorelay. J Bacteriol. 2010, 192 (13): 3484-3490. 10.1128/JB.00143-10.PubMedCentralPubMedGoogle Scholar
- 91.Toth IK, Thorpe CJ, Bentley SD, Mulholland V, Hyman LJ, Perombelon MC, Salmond GP: Mutation in a gene required for lipopolysaccharide and enterobacterial common antigen biosynthesis affects virulence in the plant pathogen Erwinia carotovora subsp. atroseptica. Mol Plant Microbe Interact. 1999, 12 (6): 499-507. 10.1094/MPMI.19184.108.40.2069.PubMedGoogle Scholar
- 103.Lebeau A, Reverchon S, Gaubert S, Kraepiel Y, Simond-Cote E, Nasser W, Van Gijsegem F: The GacA global regulator is required for the appropriate expression of Erwinia chrysanthemi 3937 pathogenicity genes during plant infection. Environ Microbiol. 2008, 10 (3): 545-559. 10.1111/j.1462-2920.2007.01473.x.PubMedGoogle Scholar
- 104.Praillet T, Reverchon S, Robert-Baudouy J, Nasser W: The PecM protein is necessary for the DNA-binding capacity of the PecS repressor, one of the regulators of virulence-factor synthesis in Erwinia chrysanthemi. FEMS Microbiol Lett. 1997, 154 (2): 265-270. 10.1111/j.1574-6968.1997.tb12654.x.PubMedGoogle Scholar
- 108.Liu Y, Chatterjee A, Chatterjee AK: Nucleotide sequence, organization and expression of rdgA and rdgB genes that regulate pectin lyase production in the plant pathogenic bacterium Erwinia carotovora subsp. carotovora in response to DNA-damaging agents. Mol Microbiol. 1994, 14 (5): 999-1010. 10.1111/j.1365-2958.1994.tb01334.x.PubMedGoogle Scholar
- 110.Reverchon S, Chantegrel B, Deshayes C, Doutheau A, Cotte-Pattat N: New synthetic analogues of N-acyl homoserine lactones as agonists or antagonists of transcriptional regulators involved in bacterial quorum sensing. Bioorg Med Chem Lett. 2002, 12 (8): 1153-1157. 10.1016/S0960-894X(02)00124-5.PubMedGoogle Scholar
- 122.Reichenbach B, Maes A, Kalamorz F, Hajnsdorf E, Gorke B: The small RNA GlmY acts upstream of the sRNA GlmZ in the activation of glmS expression and is subject to regulation by polyadenylation in Escherichia coli. Nucleic Acids Res. 2008, 36 (8): 2570-2580. 10.1093/nar/gkn091.PubMedCentralPubMedGoogle Scholar
- 125.Mukherjee A, Cui Y, Ma W, Liu Y, Chatterjee AK: hexA of Erwinia carotovora ssp. carotovora strain Ecc71 negatively regulates production of RpoS and rsmB RNA, a global regulator of extracellular proteins, plant virulence and the quorum-sensing signal, N-(3-oxohexanoyl)-L-homoserine lactone. 215. 2000, 2 (2): 203-Google Scholar
- 126.Yang S, Peng Q, Zhang Q, Yi X, Choi CJ, Reedy RM, Charkowski AO, Yang CH: Dynamic regulation of GacA in type III secretion, pectinase gene expression, pellicle formation, and pathogenicity of Dickeya dadantii(Erwinia chrysanthemi 3937). Mol Plant Microbe Interact. 2008, 21 (1): 133-142. 10.1094/MPMI-21-1-0133.PubMedGoogle Scholar
- 128.Hayes ET, Wilks JC, Sanfilippo P, Yohannes E, Tate DP, Jones BD, Radmacher MD, BonDurant SS, Slonczewski JL: Oxygen limitation modulates pH regulation of catabolism and hydrogenases, multidrug transporters, and envelope composition in Escherichia coli K-12. BMC Microbiol. 2006, 6: 89-10.1186/1471-2180-6-89.PubMedCentralPubMedGoogle Scholar
- 131.Adhya S: Suboperonic regulatory signals. Sci STKE 2003. 2003, pe22-185Google Scholar
- 140.Team RDC: R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing. 2011Google Scholar
- 142.Sathesh-Kumar S, Sridhar J, Rafi ZA: sRNAbase: A web resource for small RNAs in Enterobacteriaceae. Silico Biol. 2009, 35-9Google Scholar
- 144.Keseler IM, Collado-Vides J, Gama-Castro S, Ingraham J, Paley S, Paulsen IT, Peralta-Gil M, Karp PD: EcoCyc: a comprehensive database resource for Escherichia coli. Nucleic Acids Res. 2005, D334-D337. 33 DatabaseGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.