Listeria monocytogenes σ^B Regulates Stress Response and Virulence Functions

Abstract

While the stress-responsive alternative sigma factor σ^B has been identified in different species of Bacillus, Listeria, and Staphylococcus, the σ^B regulon has been extensively characterized only in B. subtilis. We combined biocomputing and microarray-based strategies to identify σ^B-dependent genes in the facultative intracellular pathogen Listeria monocytogenes. Hidden Markov model (HMM)-based searches identified 170 candidate σ^B-dependent promoter sequences in the strain EGD-e genome sequence. These data were used to develop a specialized, 208-gene microarray, which included 166 genes downstream of HMM-predicted σ^B-dependent promoters as well as selected virulence and stress response genes. RNA for the microarray experiments was isolated from both wild-type and ΔsigB null mutant L. monocytogenes cells grown to stationary phase or exposed to osmotic stress (0.5 M KCl). Microarray analyses identified a total of 55 genes with statistically significant σ^B-dependent expression under the conditions used in these experiments, with at least 1.5-fold-higher expression in the wild type over the sigB mutant under either stress condition (51 genes showed at least 2.0-fold-higher expression in the wild type). Of the 55 genes exhibiting σ^B-dependent expression, 54 were preceded by a sequence resembling the σ^B promoter consensus sequence. Rapid amplification of cDNA ends-PCR was used to confirm the σ^B-dependent nature of a subset of eight selected promoter regions. Notably, the σ^B-dependent L. monocytogenes genes identified through this HMM/microarray strategy included both stress response genes (e.g., gadB, ctc, and the glutathione reductase gene lmo1433) and virulence genes (e.g., inlA, inlB, and bsh). Our data demonstrate that, in addition to regulating expression of genes important for survival under environmental stress conditions, σ^B also contributes to regulation of virulence gene expression in L. monocytogenes. These findings strongly suggest that σ^B contributes to L. monocytogenes gene expression during infection.

In several low-GC-content gram-positive bacteria, σ^B has been recognized as a general stress-responsive sigma factor. This alternative sigma factor contributes to the ability of organisms such as Listeria monocytogenes, Bacillus subtilis, and Staphylococcus aureus to survive under environmental and energy stress conditions (4, 10, 18, 19, 65). σ^B also contributes to biofilm formation in S. aureus and Staphylococcus epidermidis (37, 55). Biofilm formation may further enhance the survival of these organisms under conditions of environmental stress.

In L. monocytogenes, a food-borne pathogen capable of causing mild to severe infections in humans, σ^B confers stress resistance (e.g., under acid stress and osmotic stress) and contributes to pathogenesis. To illustrate, an L. monocytogenes σ^B null mutant survives less well than the wild-type parent at low pH (pH 2.5) and in a murine infection model (45, 67). σ^B has been demonstrated to contribute to transcription of prfA, which encodes the central virulence gene regulator PrfA in L. monocytogenes (45).

An increasing body of evidence suggests a broad role for σ^B-dependent genes in the virulence of gram-positive bacteria. For example, σ^B has been linked with regulation of virulence gene expression in S. aureus (6, 12). Specifically, σ^B contributes to transcriptional regulation of sarC in S. aureus. The sar locus partially controls expression of the agr locus; agr and sar are both global regulatory elements that control the synthesis of a variety of extracellular and cell surface proteins involved in the pathogenesis of S. aureus (12, 38). A total of 23 proteins showed increased expression in the presence of σ^B in a two-dimensional gel comparison of proteins isolated from both the wild-type S. aureus strain and its isogenic sigB null mutant (27). σ^B also has been shown to contribute to virulence in Bacillus anthracis. Specifically, a B. anthracis ΔsigB strain is less virulent than its wild-type parent (21). Links between environmental stress responses and virulence in these bacteria suggest a central role for σ^B in contributing to the ability of bacterial pathogens to survive environmental stress conditions, to direct expression of virulence genes, and to cause disease.

Identification of the genes regulated by σ^B is the first step in deciphering specific mechanisms through which this alternative sigma factor confers resistance to multiple environmental stresses and contributes to virulence. Previous efforts at defining the σ^B regulon have focused primarily on B. subtilis, a gram-positive, nonpathogenic model organism. Through a combination of in vitro transcription, reporter fusion transposon mutagenesis, two-dimensional protein gel electrophoresis, and RNA hybridizations (reviewed in reference 52), a total of 75 genes and proteins with σ^B-dependent expression patterns were identified in B. subtilis. Recently, full genome macroarray analyses enabled two research groups to independently define over 120 B. subtilis genes that showed σ^B-dependent expression profiles by comparing gene expression patterns of wild-type strains with those of isogenic sigB null mutants (51, 53). In addition, use of a full-genome B. subtilis array allowed Helmann et al. (30) to identify 44 heat shock-induced genes preceded by previously unidentified potential σ^B-dependent promoters. These transcriptome studies of the B. subtilis stress response illustrate the power of genome-wide, array-based expression studies.

The σ^B regulon in B. subtilis, as determined by Price et al. (53) and Petersohn et al. (51), includes genes with a wide variety of functions. Approximately one quarter of the B. subtilis σ^B-dependent genes encode proteins involved in metabolic functions, including glucose metabolism, protein degradation, and lipid metabolism. In addition, many genes regulated by σ^B in B. subtilis encode transporters, e.g., solute transporters, permeases, and ATP-binding cassette (ABC) transport systems. The broad range of gene functions associated with the σ^B regulon in B. subtilis suggests that, in addition to enhancement of fundamental cellular processes, σ^B-mediated stress resistance mechanisms are also responsible for targeted action against specific stresses.

To better understand the role of σ^B in stress resistance and virulence in L. monocytogenes, we combined promoter searches and DNA microarrays to identify genes directly activated by σ^B from experimentally defined or predicted σ^B-dependent promoter sequences. As some B. subtilis genes have shown apparent σ^B-dependent induction in the absence of well-conserved σ^B-dependent promoter sequences, some σ^B-mediated effects may also occur through indirect regulatory mechanisms (e.g., possibly through σ^B-dependent expression of additional transcriptional regulators [51]).

For the purposes of this study, we chose to focus specifically on defining L. monocytogenes genes that are transcribed from σ^B-dependent promoters. To that end, we first performed a hidden Markov model (HMM) similarity search for putative σ^B-dependent promoters against the published complete L. monocytogenes genome sequence (28). To allow a focused analysis of σ^B-dependent gene expression, a microarray was constructed with the genes identified by the HMM search. To more fully explore the role of σ^B in L. monocytogenes stress response and virulence, our microarray also included previously identified stress response and virulence genes that were not necessarily identified by HMM.

To be classified as being under direct σ^B control, genes had to meet the following two independent criteria: (i) significantly higher expression in the wild-type strain compared to the sigB null mutant under the conditions used for the microarray experiments and (ii) the presence of an identifiable σ^B-dependent promoter-like sequence upstream of the coding region. With these criteria, we identified 54 genes under direct σ^B control. Although we refer to these genes as σ^B dependent, it is important to recognize that the genes identified with this approach may not be exclusively σ^B dependent but may also be regulated by σ^B-independent mechanisms. Interestingly, while the genes that we found to be controlled by σ^B include many that had been identified previously as part of the B. subtilis σ^B regulon, several virulence and virulence-related genes were also identified (e.g., a gene encoding a bile salt hydrolase and genes encoding surface molecules with possible or confirmed roles in host cell attachment). Our data provide further evidence in support of the contributions of σ^B to L. monocytogenes stress resistance. Importantly, identification of multiple σ^B-regulated virulence genes strongly suggests that this alternative sigma factor also contributes to the ability of L. monocytogenes to infect and survive within a host.

MATERIALS AND METHODS

Bacterial strains and media.

L. monocytogenes strain 10403S (7) and the isogenic sigB null mutant (ΔsigB) FSL A1-254 (67) were used throughout this study. Cells were grown in brain heart infusion broth (BHI; Difco, Detroit, Mich.) at 37°C with shaking unless indicated otherwise.

HMM searches.

HMM searches were performed with HMMER (http://hmmer.wustl.edu) essentially as described by Price et al. (53). The training alignments included 29 B. subtilis σ^B-dependent promoters (described in reference 53) as well as four confirmed (prfA, rsbV, and opuCA) or likely (betL) L. monocytogenes σ^B-dependent promoter sequences. The models were searched against the complete L. monocytogenes strain EGD-e genome sequence (28). The output results were filtered, and only hits within 350 bp upstream of a coding region, as predicted by the ListiList web server (http://genolist.pasteur.fr/ListiList [44]), and with an E value of ≤0.006 were kept.

DNA and RNA isolation and treatment.

RNA was extracted from L. monocytogenes either grown to stationary phase or exposed to osmotic stress (as described below) with the RNeasy Protect Bacteria midi kit (Qiagen, Valencia, Calif.). The enzymatic and mechanical disruption protocol provided by the manufacturer was used, with the exception that cell lysis was performed with sonication at 21W three times for 20 s each (with cells iced between bursts) instead of bead beating. RNA was eluted from the column in RNase-free water and ethanol precipitated overnight at −20°C. Precipitated RNA was centrifuged, washed in 70% ethanol, and resuspended in RNase-free water. DNase treatment was performed with 30 U of RQ1 RNase-free DNase (Promega, Madison, Wis.) in the presence of 400 U of RNase inhibitor (RNasin; Promega) for 1 h at 37°C. The reaction mix was then extracted twice with an equal volume of 50% phenol-50% chloroform, followed by one extraction with an equal volume of 100% chloroform. RNA was ethanol precipitated from the aqueous layer and stored at −20°C in ethanol.

Cultures for RNA isolation were inoculated from single colonies and grown overnight at 37°C. The cultures were then diluted 1:200 in BHI and grown at 37°C to specified growth phases. Specifically, for harvest of stationary-phase cells, L. monocytogenes was grown for one additional hour after reaching an optical density at 600 nm (OD₆₀₀) of 0.8. At this point, a 9-ml aliquot of culture was added to 18 ml of RNA Protect Bacteria reagent (Qiagen), and RNA isolation was performed as described above. For osmotic stress treatment, cells present in 40 ml of a culture grown to an OD₆₀₀ of 0.4 were collected by centrifugation and resuspended in 8.25 ml of warm (37°C) BHI broth, to which 1.25 ml of 4 M KCl was added to yield a final concentration of 0.5 M KCl. After 7 min of incubation at 37°C, corresponding to the approximate peak of σ^B-induced transcriptional response reported in induction experiments in B. subtilis (64), 20 ml of RNA Protect Bacteria reagent was added, and RNA isolation was performed as described above.

Chromosomal DNA used for genomic DNA microarray control spots was isolated as described by Flamm et al. (20) from an overnight culture of L. monocytogenes. Briefly, cell walls were digested with lysozyme in 20% sucrose, followed by cell lysis with sodium dodecyl sulfate and proteinase K. DNA was purified by phenol-chloroform extractions and ethanol precipitated. DNA was then resuspended in spotting buffer consisting of 3× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) plus 0.1% Sarkosyl) and quantified by UV spectroscopy.

Microarray construction.

DNA microarrays were constructed to include 208 L. monocytogenes genes spotted in triplicate, 30 mouse β-actin cDNA spots for nonhybridizing controls, and 60 spots of genomic DNA to aid in data normalization and analysis. The 208 L. monocytogenes genes included 166 of the 170 genes identified by HMM searches as including an upstream predicted σ^B-dependent promoter (PCR amplification failed for four genes) as well as 36 genes previously reported to be involved in virulence or stress response and six negative control genes (i.e., housekeeping genes predicted not to show σ^B-dependent expression). All 208 target genes as well as the PCR primers used for their amplification are detailed in the supplemental material available at http://www.foodscience.cornell.edu/wiedmann/Mark%20supplemental%20table%20 (4-15).doc.

PCR primers were designed with PrimeArray software (56) to amplify the complete open reading frames (ORFs) of the selected genes. For ORFs of >1 kb, primers were selected to amplify a 1-kb fragment at the 5′ end of each ORF. PCR was performed with AmpliTaq Gold (Applied Biosystems, Foster City, Calif.) with an L. monocytogenes 10403S cell lysate (prepared as described in reference 24) as template DNA. Appropriately sized products free of nonspecific amplification or extraneous bands, as determined by agarose gel electrophoresis, were purified by ethanol precipitation and reconstituted in spotting buffer (3× SSC-0.1% Sarkosyl) to a final concentration of 50 to 150 ng/μl. PCR was also used to generate a 1-kb PCR fragment of the mouse β-actin gene with mouse β-actin cDNA (Sigma, St. Louis, Mo.) as the template.

Arrays were spotted with a GMS 417 robotic arrayer (Affymetrix, Santa Clara, Calif.) on GAPS2 glass slides (Corning, Corning, N.Y.). Immediately before use, slides were blocked in a solution of 1-methyl-2-pyrrolidinone containing 1.44% succinic anhydride and 0.02 M boric acid (pH 8.0), as directed by the slide manufacturer.

cDNA labeling and microarray hybridization.

Precipitated, DNase-treated RNA was centrifuged, washed once in 70% ethanol and once in 100% ethanol, and resuspended in RNase-free water. The RNA was quantified on a UV spectrophotometer and checked for quality by A_260/280 ratio and formaldehyde-agarose gel electrophoresis. cDNA was generated from 10 μg of RNA with random hexamers and SuperScript II (Invitrogen, Carlsbad, Calif.) in the presence of indocarbocyanine (Cy3)-dUTP or indodicarbocyanine (Cy5)-dUTP (Amersham Biosciences, Piscataway, N.J.). cDNA synthesis and labeling were performed with a dye-swapping design (sometimes referred to as reverse labeling [62]); each RNA sample was used for two separate labeling reactions, one with Cy3 and one with Cy5. Completed reactions were incubated with 1.5 μl of 1 N NaOH at 65°C for 10 min to inactivate the enzyme and degrade the RNA. Reactions were then neutralized with 1.5 μl of 1 N HCl.

Wild-type and sigB mutant cDNA probes were mixed and purified with the QIAquick PCR purification kit (Qiagen) as described by the manufacturer, with the addition of a 750-μl 35% guanidine-HCl wash after binding. The purified probes were then dried and reconstituted in 20 μl of hybridization buffer (3.5× SSC, 0.25% sodium dodecyl sulfate, and 0.5 μg of salmon sperm DNA/μl; Invitrogen). Probes were boiled for 2 min, applied to arrays, and hybridized overnight in a 60°C waterbath. Before scanning, slides were washed in 2× SSC-0.1% sodium dodecyl sulfate at 60°C for 5 min, followed by room temperature washes with 2× SSC-0.1% sodium dodecyl sulfate, 2× SSC, and 0.2× SSC for 5 min each. Slides were centrifuged to dry them and scanned with a Perkin Elmer Scan Array 5000 confocal laser scanner (Perkin Elmer, Wellesley, Mass.).

Microarray replicates and data analysis.

For each stress condition described above (osmotic stress or stationary phase), two independent RNA isolations (for both wild-type and ΔsigB cells) were performed on separate days to provide true biological replicates. Each set of independent RNA samples was used to perform two microarray hybridizations with a dye-swapping design to correct for differences in Cy3 and Cy5 dye incorporation and to provide experimental duplicates. Briefly, both wild-type and ΔsigB cell RNAs from each of the two independent RNA isolations were labeled separately with either Cy5 or Cy3 as described above. Labeled cDNA was used to perform two independent microarray hybridizations, one with Cy3-labeled wild-type and Cy5-labeled mutant cDNA and one with Cy5-labeled wild-type and Cy3-labeled mutant cDNA. Thus, data for four microarray repetitions were collected for each stress condition.

Raw TIFF images from the scanner were loaded into ScanAlyze (http://rana.lbl.gov/EisenSoftware.htm) for analysis. Spot grids were manually fitted to the microarray images. Spots were flagged and eliminated from analysis only in obvious instances of high background or stray fluorescent signals.

Because our microarray contained a relatively small number of genes, most of which were expected to show differential expression, proper normalization of the raw data was critical. Equalizing the mean intensities of both channels of an array, the most common method used for microarray normalization, would thus have provided severely skewed data on an array, with the majority of spots appearing to be unequally expressed. Therefore, we normalized our data using the mean intensities of spots containing genomic DNA. Data normalization was performed simultaneously on all four array data sets for a given stress condition, as follows. For each channel (Cy3 or Cy5), the mean intensity of all genomic DNA control spots was calculated independently. A correlation factor comparing genomic means among channels was calculated and used to proportionally adjust all spot intensities (experimental and control) so that the means of genomic spot intensities were equal across all arrays. From these normalized values, a floor was calculated as the average intensity plus 2 standard deviations of all β-actin spots in a channel. Spots with values below the floor in both channels were eliminated from analysis; spots below the floor in only one channel had that channel intensity raised to the floor value, and the resulting data were included in the analyses (16).

Floored and normalized channel intensities were analyzed with the Significance Analysis for Microarrays (SAM) program (63). This statistical analysis involves factoring the change in expression of each gene relative to the standard deviation of all replicates for that gene. Therefore, genes with a low change will not be discounted if the ratios are consistent among repeats, effectively reducing false-negatives. False-positives are also avoided when genes have poor reproducibility between replicates. This method of statistical analysis maximizes both the quantity of genes found and the reliability of the results. Spot intensities for all channels were input in SAM as paired, unlogged values. All individual spots were considered repetitions, generating 24 data points for each gene (3 spots per gene × 4 arrays × 2 channels per array). Delta values were chosen according to the lowest false discovery rate. Only genes with expression ratios of >1.5 were considered biologically significant.

RACE-PCR.

Promoter regions were mapped with the 5′ rapid amplification of cDNA ends (RACE) system (Invitrogen) according to the manufacturer's protocol. Briefly, gene-specific first-strand cDNA was tailed with dCTP by using terminal transferase. The products were then amplified with a nested gene-specific primer and a poly(G/I) primer in a touchdown PCR with AmpliTaq Gold (Applied Biosystems). PCR products present in the wild-type L. monocytogenes but absent in the ΔsigB strain were purified with the Qiagen QIAquick PCR purification kit and sequenced with the Big Dye terminator system and an Applied Biosystems 3700 sequencer at the Cornell University Bioresource Center.

RESULTS

HMM search to identify potential σ^B promoters.

An HMM developed with 29 B. subtilis and four L. monocytogenes σ^B-dependent promoter sequences was used to search the L. monocytogenes EGD-e genome (28) to identify genes preceded by a σ^B consensus promoter and thus potentially regulated by σ^B. A total of 170 genes met the following criteria for putative σ^B-dependent genes: (i) location of a predicted σ^B consensus promoter sequence within 350 bp upstream of a predicted open reading frame and (ii) an E value of ≤0.006. The presence of a sequence adequately fitting the promoter consensus was confirmed visually at positions indicated by the HMM. The results of this search were used to create a 208-gene microarray to specifically study the σ^B-mediated stress response in L. monocytogenes. The array consisted of 166 of the 170 HMM-identified genes (PCR amplification failed for four genes), 36 genes involved in stress response or virulence, and six control genes.

Transcriptional analysis of potential σ^B-dependent genes.

Microarray analyses were performed with RNA isolated from wild-type L. monocytogenes and an isogenic ΔsigB strain grown to stationary phase or exposed to osmotic stress, two conditions shown to activate σ^B (4, 19, 22). We identified 55 genes that showed significant expression ratios, with >1.5-fold expression in the wild-type strain over that in the mutant strain in at least one of the two stress conditions tested (Table 1). Fifty-one of these genes displayed >2-fold induction; the highest relative induction was 27-fold.

TABLE 1.

L. monocytogenes genes with significantly higher expression in the wild-type versus the ΔsigB strain under osmotic and stationary-phase stresses

Changea (fold)	No. of genes with statistically significant higher expression under:
	Stationary phase stress	Osmotic stress (0.5 M KCl)	Stationary-phase and osmotic stress	Stationary-phase or osmotic stress
≥1.5	41	51	37	55
≥2.0	37	44	30	51

^aChange was calculated by SAM based on microarray data (see Materials and Methods) and equals the average wild-type spot intensity divided by the average mutant spot intensity for two separate arrays from each of two replicate RNA isolations.

Figure 1 shows scatter plots of wild-type versus sigB mutant spot intensities. As the microarray predominantly targeted genes predicted to be positively regulated by σ^B, the distribution of the majority of the points above the diagonal, which indicates that a large proportion of the genes in the microarray were expressed more strongly in the wild type than in the mutant, was expected. As the presence of points falling far below the diagonal would signify increased gene expression in the absence of σ^B, none to few were expected. For the purpose of this study and in accordance with previous studies (31), we applied a cutoff of 1.5-fold induction to signify biological significance. Two genes included in the microarray fell below this cutoff (showing statistically significant expression ratios between 1.2- and 1.5-fold) and are not discussed further here. As shown in Fig. 1, a few genes showed higher expression in the ΔsigB mutant than in the wild-type strain; expression differences between strains for these genes were determined to be not statistically significant.

FIG. 1.

Scatter plots of normalized microarray spot intensities for wild-type and sigB mutant strains. The center line depicts equal expression in both strains. Outer lines depict 1.5-fold higher expression in either strain. Spot intensities were normalized as described in the text. Data for each gene represent the averages of three replicate spots from each of four hybridizations. RNA for microarray experiments was isolated from (A) stationary-phase cells or (B) osmotically stressed cells.

Overall, 44 of the 169 genes (26%) in our microarray that were either (i) predicted to be σ^B dependent by HMM or (ii) members of operons previously identified as being regulated by a σ^B-dependent promoter also showed σ^B-dependent expression in our microarray experiments. There was no apparent correlation between a promoter's HMM E value and the relative expression level of a corresponding gene, as determined by microarray (r² < 0.01 for correlations between HMM E values and relative expression levels under either osmotic stress or stationary-phase conditions). In addition, 11 of the 36 currently recognized stress response and virulence genes that were included in the array even though they did not show an HMM-predicted σ^B-dependent consensus promoter displayed σ^B-dependent expression patterns. Subsequent visual inspection of the upstream regions for these 11 genes identified putative σ^B consensus promoter sequences for 10 of them (Table 2). Although inlD showed σ^B-dependent expression in our microarray experiments, no putative σ^B consensus promoter sequence was identified for this gene.

TABLE 2.

Summary of genes with σ^B-dependent expression patterns as identified by microarray analyses

Category	Gene	Function or closest homologuea (E value)	Predicted σB promoter sequenceb	B. subtilis homologuec (E value)	Changed (fold)
					ST	OS
Stress	lmo2230	Arsenate reductase [Staphylococcus aureus] (le-08)	TTTAATGTTTCTAGTAATTTAAAAAGGGTAGATATTACTGT-N108-ATG	arsC (7e-10)	27.0	21.9
	gadB	GadB glutamate decarboxylase (Lmo2434)	TGAACGGTTTGTCTCTGTGGTTTAATGGGTATTGGTGAGAGA-N32-ATG		5.3	4.9
	sepA	Metallo-beta-lactamase superfamily protein [Shewanella oneidensis] (e-173)	CAAAAGGTTTTGAATAATTTTATGGAGGTATAAAAAGAAAGT-N33-ATG		12.0	13.4
	ltrC	Low-temperature-requirement C protein	CCGTCCGTTTTTAGTTCGTGATGGCGGGGACTATGAACTTAT-N260-ATG†	yutG (le-53)‡	4.8	1.9
	lmo1433	Glutathione reductase (NADPH) [Methanosarcina acetivorans] (4e-91)	TTTTTCGTTTGAAAGTGAAATCAGACGGGAAAACAAGCTAAG-N14-GTG		6.8	5.7
	ctc	B. subtilis general stress protein Ctc (le-27)	TCCTTTGTTTTGCTATTTTTCTAAAGGGTAGATATAATGTA-N35-ATG	ctc (le-33)‡	1.5	2.2
	lmo1602	General stress protein [Oceanobacillus iheyensis] (6e-11)	AAGAAAGTTTTAGAGGGGAATACTCAGGGTATAGAAAAAGGA-N18-ATG	ytxG (6e-13)‡	2.7	2.1
Virulence and virulence associated	bsh	Bsh bile salt hydrolase (Lmo2067)	AATTATGTTTTACTCCAAACTCCGAGGGTACTGGTATACAT-N31-ATG		10.8	20.1
	inlA	Internalin A	CCATTAGTGTTATTTTGAACATAAAGGGTAGAGGATAACAT-N437-GTG†	6.2	3.9
	inlB	Internalin B	CCTTTTGTTAGGGTTATTAGCAGTAGGAACTGCAATGGCTC-N113-GTG†	4.6	2.1
	inlC2	Internalin C2	TTTTTTGTTAATTTGGTCTAAAAAAGGGTATCTATTATTAA-N55-ATG†		11.4	7.3
	inlD	Internalin D	None found		3.3	2.3
	inlE	Internalin E	TAAATCGTTAACAAGTCTAATTTTAGTGATTAAACGAAATC-N84-ATG†		3.7	2.4
Transporters	opuC operon		TAAAAAGTTTAAATCTATACTAGTTAGGGAAATTAGTTATCG-N48-GTG
	opuCA	Glycine betaine/carnitine/choline ABC transporter (ATP-binding protein)		opuCA (e-159)	5.3	6.5
	opuCB	Glycine betaine/carnitine/choline ABC transporter (membrane protein)		opuCB (le-74)	3.8	4.7
	opuCC	Glycine betaine/carnitine/choline ABC transporter (osmoprotectant-binding)		opuCC (le-92)	4.1	5.2
	opuCD	Betaine/carnitine/choline ABC transporter (membrane protein)		opuCD (5e-66)	5.0	5.8
	lmo0784	PTS enzyme IIAB, mannose-specific [Yersinia pestis KIM] (5e-34)	AATTGCGTTTTCTGACTAATCTTTTAGGGTAATGTTGTATGT-N195-ATG	levD (le-21)	7.5	5.0
	lmo2602	Cation-transporting ATPase [Oceanobacillus iheyensis] (2e-12)	TTGATTGTTTTGGTTTAATGCCAAAGGGAATATATTAACGT-N16-ATG	sapB (le-09)	5.5	3.8
	lmo0405	PHO4, phosphate transporter family [Bacillus anthracis] (4e-76)	GATTCTTTTTATATTTGTATAAAAGGGGTATAGACAATAAA-N30-ATG		3.9	4.9
	lmo2463	Probable export protein (antibiotic) [Streptomyces coelicolor] (6e-80)	AATGATGTTTGGCATATGTAAAAAAGAGGTATAAATTAGCGTG	ydfJ (4e-30)	3.0	5.2
	lmo1539	Glycerol uptake facilitator [Bacillus halodurans] (8e-66)	AAATGGGTTATAACTCTCGCGAATTGGGGTAAAAGTAAAGAG-N254-ATG	glpF (5e-81)	2.4	2.3
	lmo1421	ABC transporter [Bacillus anthracis A2012] (8e-98)	AATTAGGAATATTTAGGGATGATTTAGGGTAATTGGATGATA-N74-ATG†		2.5	2.1
	lmo0524	Sulfate transporter [Mycobacterium tuberculosis CDC1551] (2e-50)	AGAGATGATTGTTCTATGTCAAAACGGGTAAATAGTATAAG-N65-ATG		2.0	2.9
	lmo0593	Formate/nitrite transporter family protein [Streptococcus agalactiae] (7e-37)	TTTTGTGTTTTAAGAGTTTGAAAACGGGGAAATTAACAACAT-N146-TTG	ywcJ (le-18)	2.9	2.3
Metabolism	lmo0669	Oxidoreductase [Oceanobacillus iheyensis] (3e-86)	TTAAACGTTTTAGCGTAAAACAGGAGGGAAGACATAAAGTA-N139-ATG	yhxD (3e-95)‡	14.2	9.5
	lmo1694	Epimerase, NAD-dependent family [Bacillus anthracis] (2e-62)	AAAATGGTTTTAATACTACTAAAAAGGGAATAAACTAGTTA-N22-TTG	yfhF (5e-59)‡	11.5	11.7
	lmo1883	Chitinase [Serratia marcescens] (e-117)	TTTCTAGTTTTATTTTCACTATGTTGGGTATTTTCTAGTAG-N47-ATG		4.1	2.1
	lmo2695	Dihydroxyacetone kinase [Bacillus halodurans] (e-103)	AACCACGTTTTGACTTTCTAGTAAAGGGAAATTGAGGTAAG-N30-ATG		2.8	4.3
	lmo0956	N-Acetylglucosamine-6-phosphate deacetylase [Bacillus anthracis] (e-110)	TTTTTTGGTTATTTTACTTTTTTTCGGGTAAATAGGAAAAG-N71-ATG	nagA (le-72)	1.8	1.4
	lmo2205	Phosphoglycerate mutase [Neisseria meningitidis] (4e-86)	AAAAAGGTTTGACACTTCACTTGAAAGGGAAAACTTTGGTTG-N44-ATG		1.8	3.2
	pdhA	Pyruvate dehydrogenase (E1 alpha subunit) [Bacillus subtilis] (e-151)	ACCCATGTTTTCAATGAGGAATTTGTGGTAATAACTGGAGAG-N231-ATG	pdhA (e-162)	1.7	1.6
	lmo1933	GTP cyclohydrolase I [Bacillus halodurans] (6e-67)	TTAGTGGTTTTCTGTTTTAGAAATAGGGAATATAATTACGA-N113-ATG	mtrA (7e-73)‡	2.0	2.0
Transcriptional regulation	rsbV operon		TTGAATGTTTTAATTTTATTTGTTAGGGTAAAATCGACAGT-N27-ATG†
	rsbV	Anti-anti-sigma factor (antagonist of RsbW)		rsbV (3e-24)‡	2.1	2.0
	rsbW	Sigma B activity negative regulator RsbW		rsbW (3e-42)‡	2.0	2.4
	sigB	RNA polymerase sigma-37 factor (sigma B)		sigB (7e-97)‡	2.2	2.3
	rsbX	Indirect negative regulation of sigma B (serine phosphatase)		rsbX (le-26)‡	2.0	2.2
	lmo2485	Transcriptional regulator [Methanosarcina mazei Goel] (le-04)	GCAAAAGTAGGAAATGGAATTATCAAGGTTTCTGAACTCGA-N42-ATG		1.8	3.9
Cell wall	lmo2085	Surface-anchored protein (LPXTG motif) [Listeria innocua] (le-18)	CTGGCTGTTTTCTTTTGCTGTTTTATGGGTATTTAATGATGT-N28-ATG		14.7	10.7
	lmo0880	Surface-anchored protein (LPXTG motif) [Listeria innocua] (le-26)	TGTACGGTTTTTAACAAGCAAGTTGTGGGAACCTATAAAGATA-N26-ATG		5.2	2.1
Energy	qoxA	AA3-600 quinol oxidase subunit II	TTTTTTGTTTCGAATTTCACAATCTAGGGAATAGTGAACAGA-N265-GTG	qoxA (3e-91)	1.5	1.9
	lmo2389	Pyridine nucleotide-disulfide oxidoreductase [Bacillus anthracis] (e-123)	GATATAGTTTGTAAGCTGTGAAATAGGGAATATATTGGTAC-N127-ATG	yumB (e-128)‡	1.6	3.4
Translation	lmo1698	Ribosomal protein (S5)-alanine N-acetyltransferase [Bacillus halodurans] (3e-)	TGGAAAGTTTATTTTTTAATAAAATGGGTATATAGAAAAAT-N72-ATG	yjcK (3e-55)	3.1	1.9
	lmo2511	Ribosomal_S30, sigma 54 modulation protein [Bacillus anthracis] (2e-53)	AAAGAGGTTTGCGGAAGCGGTATTAGTGGAATAAAATACTAA-N135-ATG	yvyD (e-162)‡	3.8	1.9
DNA translocase	lmo1606	DNA translocase (spoIIIE) [Bacillus halodurans] (e-168)	TACCATGTTTAACCCCTCTATTACCAAGGTATTATAACGAAT-N97-ATG	ytpT (0.0)	2.4	2.6
Hypothetical	lmo2570	Hypothetical protein YvaZ [Bacillus subtilis] (.023)	CAACAGGTGTTCTAGAATATTTACGGGAACAAAAAGTCGA-N234-ATG	yvaZ (5e-09)	5.3	5.9
	lmo2269	YhzC of Listria monocytogenes (9e-26)	TTTTAAGATTGTAATAAATATGAGTGAAGGAGTGGTAGTCGTG		4.0	5.3
	lmo0911	BH2119-unknown [Bacillus halodurans] (le-06)	CATCTTGTTTTAACTTGCCCTCAGGCGGGTATTTATTATATG-N53-ATG		2.2	1.9
	lmo0794	Conserved hypothetical protein YwnB [Bacillus subtilis] (2e-50)	ATAAATGTTTCCCAGTCCCCTCTTTCGGGAATAATTCTTCTA-N45-ATG	ywnB (5e-54)	5.2	5.8
	lmo1580	Conserved hypothetical ATP-binding domain [Bacillus anthracis] (2e-40)	TTTTATGGTTCTTTTAGGAAAAAGAGGGTAAAATATAAGTA-N35-ATG	yxiE (4e-11)	2.0	2.7
	lmo2673	Conserved hypothetical ATP-binding domain [Bacillus anthracis] (4e-28)	AATCATGCTTCTTTCTTTTATTTATGGGTATTAAGTAAATA-N30-ATG		3.7	6.9
	lmo2386	Conserved hypothetical protein YuiD [Bacillus subtilis] (le-37)	GAATAGGTTTTTAATAAGCTCATTGTGGTAAAATAAGAATAG-N22-ATG	yuiD (le-50)	1.8	2.4
	lmo2399	Hemolysin homolog YhdP [Bacillus subtilis] (2e-89)	CTTGTGGTTTAAACGCTAAGAACGGGTATAGAATGGGGA-N15-ATG		2.0	2.5
	lmo0438	No similarity	ACTAACGAATCGCGTGGATGCGCTTCGGGAATTGTTGAAATA-N157-GTG		2.7	3.4
	lmo0994	No similarity	GAAAGAGGTTATTTTTCACTAAATGGGGTAAAAATACGTTA-N257-ATG		6.1	6.5

^aThe function of the protein encoded is listed if it has been supported by genetic or biochemical data. Otherwise, the nearest homologue of the protein with known function is listed. The E values in parentheses were obtained with Blast 2.0 searches with the BLOSUM62 matrix against the nr database. PTS, phosphotransferase system.

^bThe promoter sequence as predicted by HMM is listed. For genes without an HMM-predicted promoter sequence (marked with †), a putative promoter identified by visual inspection is indicated; “none found” indicates that no suitable promoter could be found by HMM or visual inspection. The expected −35 and −10 regions are in bold, and the last nucleotide triplet is the presumed translational start codon.

^cThe B. subtilis genome was searched for ORFs encoding proteins homologous to the given L. monocytogenes protein by Blast 2.0 on the SubtiList web server (http://genolist.pasteur.fr/SubtiList). The closest homologue is listed, if one was found, along with the E value in parentheses. Homologues that have been determined to be σ^B dependent in B. subtilis (51,53) are marked with ‡.

^dChange was calculated with SAM based on microarray data (see the text) and equals the average wild-type spot intensity divided by the average mutant spot intensity for two separate arrays from each of two replicate RNA isolations. Numbers in italics indicate ratios that were determined by SAM to not be statistically significantly different from a ratio of 1.0. ST, stationary-phase cells; OS, osmotically stressed cells (0.5 M KCl).

Overall, we identified 54 σ^B-dependent L. monocytogenes genes through combined application of HMM and microarray analyses. Based on identification of the predicted proteins, the genes were grouped into 10 functional groups (Table 2). The two most highly represented categories were transport and metabolism proteins, which together comprised 20 of the 54 genes (Table 2). Statistical analysis of microarray data (SAM [63]) showed that, in addition to high expression ratios, the putative oxidoreductase gene lmo0669 also had the highest significance score for differential expression of all genes represented in the microarray (Table 2). The known σ^B-dependent betaine/carnitine transporter operon opuC also showed σ^B-dependent expression, along with eight other solute transport genes. The large proportion of metabolism and transport genes regulated by σ^B underscores the importance of maintaining proper cellular functions during exposure to stress conditions and suggests that σ^B provides protection by enhancing the operation of a specialized subset of basic cellular processes, which may help to produce a more stress-resistant state for the cell.

Two other notable functional categories of σ^B-dependent genes are stress response and virulence genes (Table 2). The stress response genes, which have apparent functions in bacterial stress resistance but do not encode transporters or proteins from other recognized categories included in Table 2, represented some of the most highly differentially expressed genes. For example, in the SAM output, which ranks differentially expressed genes based on the statistical significance associated with the expression ratios, four of the seven stress response genes ranked among the 16 genes with the highest significance values for differential expression, for RNA isolated from cells exposed to osmotic stress. The stress response genes identified include gadB, which was ranked as the third or fifth most significantly differentially expressed gene in cells grown to stationary phase or exposed to osmotic stress, respectively.

Three confirmed virulence genes were expressed at a higher level in the wild-type L. monocytogenes strain compared to the ΔsigB mutant (Table 2). bsh exhibited 20.1-fold higher mRNA levels in the wild type compared to the ΔsigB strain under osmotic stress (Table 2); this expression ratio was determined to be highly significant by SAM analysis (ranked as the second most significant ratio for all the genes in the array). The virulence genes inlA and inlB showed significantly higher expression under osmotic and/or stationary-phase stress in the wild type compared to the mutant strain, with expression ratios consistently >2.0 (Table 2). In addition to these confirmed virulence genes, three other internalins (inlC2, inlD, and inlE) also showed significant expression ratios of >2.0 under osmotic and/or stationary-phase stress (Table 2).

PCR mapping showed that the L. monocytogenes strain used in our study (10403S) bears an inlC2DE operon rather than an inlGHE operon, which is present in the L. monocytogenes EGD-e strain (57). Each ORF of the inlC2DE gene cluster is potentially transcribed independently of the others (13). We identified a σ^B promoter consensus sequence upstream of both inlC2 and inlE but not inlD. Only one previously described internalin gene (inlC) that was included in our array and is present in strain 10403S did not show differential expression between the wild-type and the ΔsigB mutant strains.

Microarray identification of operons and known σ^B-dependent genes.

For putative σ^B-regulated genes that were identified only by HMM, only PCR products from the genes most proximal to given HMM-predicted promoters were spotted on the microarray. Therefore, downstream genes that may have been cotranscribed from a given promoter were not tested for σ^B dependence in this array. However, the microarray did include genes comprising two operons (opuC and rsbV) that were known a priori to be σ^B dependent (4, 22). All genes in these two operons were expressed at significantly higher levels in the wild-type compared to the ΔsigB strain. Furthermore, expression ratios for each gene within these operons were approximately equal (Fig. 2A and B). The inlAB operon was not previously known to be regulated by σ^B, but inlA and inlB were among the virulence genes included on the microarray. We found that both genes in this operon were expressed at higher levels in the presence of σ^B and that inlA and inlB displayed similar expression ratios (Fig. 2C).

FIG. 2.

Graphic depiction of three operons represented on the microarray. Numbers above genes indicate wild-type-to-mutant expression ratios in stationary-phase cells. Arrows indicate transcriptional start sites. (A) The four genes of the rsbV operon are transcribed from a σ^B-dependent promoter (7). Transcription can also occur from a σ^A-dependent promoter further upstream, which also transcribes the four upstream genes rsbR, rsbS, rsbT, and rsbU. (B) The four genes of the opuC operon are transcribed from a σ^B-dependent promoter (22). (C) Transcription upstream of inlA can be initiated from one of four promoters (40), one of which is σ^B dependent (as confirmed by RACE). Cotranscription of inlB can occur from at least one of the inlA promoters (40). Transcription of inlB can also occur independently from promoters directly upstream of inlB (i.e., not including inlA), including one promoter that does not have a σ^B consensus sequence (40) and potentially from one with a visually predicted σ^B-dependent promoter.

Confirmation of predicted σ^B-dependent promoters by RACE-PCR.

To confirm that predicted σ^B-dependent promoters were in fact responsible for the differential gene expression seen in the microarrays, we performed RACE-PCR on eight selected genes. These genes were chosen to represent highly differentially expressed genes and genes with high significance scores in SAM as well as known stress response or virulence genes. Gene-specific first-strand cDNA was generated for each gene, and a 3′ poly(dC) tail was added to each cDNA product with terminal transferase. The tailed cDNA product was amplified by touchdown PCR with a poly(G/I) primer (Invitrogen) and a nested gene-specific primer.

PCR bands from wild-type cDNA reactions that met the following conditions were purified and sequenced (Fig. 3): (i) bands that must have been generated by reactions with wild-type cDNA and (ii) equivalent bands that must have been absent in sigB mutant cDNA reactions. For all genes tested, the transcriptional start site determined by RACE was 10 nucleotides (± 2 nucleotides) downstream of the σ^B-dependent promoter −10 region that had been predicted by HMM or by visual inspection (Fig. 4). The genes selected for promoter confirmation by RACE were bsh and inlA, both virulence genes; lmo0669, a putative metabolic gene; lmo1421 and opuCA (which encode putative and known compatible solute transporter proteins, respectively); and lmo2230, lmo1433, and gadB, all putative or proven stress response genes.

FIG. 3.

Determination of σ^B-dependent transcriptional start sites by RACE-PCR. (A) Agarose gel electrophoresis of touchdown-PCR amplified poly(dC)-tailed cDNA. DNA was stained with ethidium bromide and visualized under UV light. Two gene-specific experiments (lmo2230 and gadB) are shown. Lanes 1 and 6, pGEM DNA size marker (Promega). Lanes 2 and 7, negative control PCR of untailed wild-type cDNA. Lanes 3 and 8, PCR of poly(dC)-tailed wild-type cDNA. Lanes 4 and 9, PCR of poly(dC)-tailed ΔsigB cDNA. Lanes 5 and 10, negative control PCR of untailed ΔsigB cDNA. (B) A typical chromatogram from sequencing the wild-type RACE-PCR product (gadB shown). As a reverse-oriented primer was used in sequencing reactions, the reverse complement of the sequence obtained is shown in order to depict the actual transcript sequence. The poly(dC) tail is shown, and the transcript beginning is marked with an arrow.

FIG. 4.

Promoter sequences of genes with σ^B-dependent transcriptional start sites confirmed by RACE-PCR. The −35 and −10 regions are underlined and in bold. Diamonds indicate transcriptional start sites determined by RACE. All promoter sequences displayed were predicted by HMM except for those of inlA and lmo1421, which were identified by visual inspection.

L. monocytogenes σ^B promoter consensus sequence resembles that of B. subtilis.

Successful identification of specific σ^B-dependent genes in L. monocytogenes provided new information needed to facilitate further refinement of the predicted consensus sequence for σ^B-dependent promoters in L. monocytogenes. The 54 σ^B-dependent promoter sequences described above were aligned and used to generate a new consensus recognition site (Fig. 5). The −35 region (GTTT) is separated from the −10 region (GGGWAT) by 13 to 17 nucleotides, most frequently by 15 or 16. The guanidine in the −35 box is almost completely conserved (98%). Overall, the L. monocytogenes σ^B consensus sequence resembles the consensus sequence for B. subtilis σ^B, which was reported by Petersohn et al. (50, 51) as GTTTAA-N_12-15-GGGWAW. This finding is not surprising, as the majority of the sequences used to train the HMM used in this study were obtained from B. subtilis genes. However, the predicted L. monocytogenes σ^B consensus sequence does differ slightly from the B. subtilis consensus promoter sequence. Specifically, the two adenosines at the 3′ end of the B. subtilis −35 region are not conserved in L. monocytogenes. The different consensus sequences suggest that σ^B may vary in its ability to recognize certain promoters in these two species.

FIG. 5.

L. monocytogenes consensus sequence logo generated with the GENIO/logo RNA/DNA and Amino Acid Sequence Logos web server (http://genio.informatik.uni-stuttgart.de/GENIO/logo). Predicted promoter sequences of the 54 σ^B-dependent genes were aligned manually and entered into the program. The vertical axis is information content in bits. The height of a nucleotide represents its frequency at that location. Letters displayed upside down indicate a nucleotide frequency of less than 25%. Numbers below selected residues indicate nucleotide frequencies at that position. Conserved −35 and −10 regions are indicated by bold numbers.

Effect of induction conditions on gene expression patterns.

As σ^B-dependent gene expression patterns were determined under two different stress conditions, we also analyzed whether the magnitude of σ^B dependence for individual genes varied with the condition. Comparison of the microarray data generated under different stress conditions indicated that the relative magnitude of σ^B-dependent expression was similar under the two conditions for the genes tested. The scatter plot in Fig. 6 shows the wild-type-to-ΔsigB mutant gene expression ratios obtained for stationary-phase cells plotted against the gene expression ratios for osmotically stressed cells.

FIG. 6.

Comparison of expression ratios for two stress conditions. Wild-type-to-sigB mutant gene expression ratios for stationary-phase and osmotic stress are plotted on the two axes. Genes that are not expressed at a significantly higher rate in the wild type (×) cluster near the origin. Significantly σ^B-dependent genes (▪) lie near the diagonal, indicating similar expression ratios under both conditions. Outer lines indicate twofold differences in expression ratios among conditions. The point labeled A represents lmo0880, which showed the largest difference in expression ratio.

The majority (67%) of the σ^B-dependent genes defined as described above showed significantly higher mRNA levels, with expression ratios of >1.5 in the wild-type L. monocytogenes strain under both stress conditions (Table 1). For five genes, the expression ratios between the two stress conditions varied from each other by more than twofold; the largest difference was 2.5-fold (lmo0880, point A in Fig. 6). However, the majority of the genes (39 out of 55) had expression ratios that differed by less than 1.5-fold under the two stress conditions. These results suggest that the majority of genes comprising the σ^B regulon are induced to a similar extent in relation to each other regardless of the specific environmental stress. A subset of σ^B-dependent genes (e.g., lmo0880) may also be subject to additional transcriptional regulation by other stress-specific mechanisms. This is not to say, however, that all stress conditions activate σ^B to the same degree. For example, while our specific conditions resulted in similar induction levels among the targeted genes (slope of regression = 0.89), primer extension data suggest that induction of transcription from the σ^B-dependent L. monocytogenes rsbV promoter may vary under the different stress conditions (4).

DISCUSSION

To improve our understanding of the specific roles of the alternative sigma factor σ^B in regulating gene expression and bacterial stress response, we used a combination of whole-genome HMM similarity searches and microarray-based strategies to identify members of the σ^B regulon in the facultative intracellular pathogen L. monocytogenes. A 208-gene microarray, which included 166 HMM-identified L. monocytogenes genes located downstream of putative σ^B-dependent promoter sequences as well as selected additional virulence and stress response genes, allowed us to identify 54 genes as being directly regulated by L. monocytogenes σ^B. Transcriptional start site mapping on selected genes confirmed the functionality of putative σ^B-dependent promoters at the locations predicted by the HMM or by visual inspection.

Similar to the σ^B regulon previously defined for B. subtilis, a gram-positive soil bacterium closely related to the genus Listeria, members of the L. monocytogenes σ^B regulon were shown to encode a variety of protein functions involved in metabolic pathways, transport, and other fundamental cellular functions. As expected, many genes identified in this study encode proteins directly associated with stress resistance. Interestingly, several genes identified as σ^B dependent represent putative or known virulence genes (e.g., inlE and bsh, respectively) or stress response genes that have been shown to contribute to the ability of L. monocytogenes to survive within an infected host (e.g., opuC [60, 66]) or under conditions similar to those encountered in an infected host (e.g., gadB [11]). These results suggest that σ^B, in addition to enhancing bacterial survival under stress conditions, also contributes to virulence in L. monocytogenes.

Role of σ^B in L. monocytogenes stress resistance.

Phenotypic characterization of L. monocytogenes strains lacking sigB has shown that σ^B plays an important role in resistance to osmotic, oxidative, and acid stresses (4, 18, 19, 22, 67). Similar studies with B. subtilis ΔsigB mutants also provided evidence for the importance of a functional σ^B for survival under ethanol, oxidative, osmotic, and acid stresses (2, 25, 65). The σ^B-dependent L. monocytogenes genes identified here provide additional insight into the functional bases of the reduced stress resistance in L. monocytogenes ΔsigB mutants. For example, gadB, which encodes a glutamate decarboxylase important for acid stress survival in L. monocytogenes (11), was shown in this work to be part of the L. monocytogenes σ^B regulon.

Interestingly, both the L. monocytogenes and the B. subtilis σ^B regulons include multiple genes that have been shown experimentally to contribute to protection against osmotic stress, including genes encoding several solute transporters. The ABC transporter OpuC contributes to osmotic stress resistance by facilitating uptake of the osmoprotectants choline and glycine betaine in both L. monocytogenes and B. subtilis (1, 35, 36, 60), and expression of the encoding operon is σ^B dependent in L. monocytogenes. Similarly, lmo1421 and the B. subtilis homologue opuB, each of which is a member of the σ^B regulon in its species, both encode a putative choline transporter which may also contribute to osmotic stress resistance (22, 36). In addition, ctc, which shows σ^B-dependent expression in both L. monocytogenes (this study) and B. subtilis (29, 33), was recently shown to contribute to osmotolerance in L. monocytogenes (26).

Finally, the σ^B regulon in both L. monocytogenes and B. subtilis also includes genes that appear to contribute to oxidative stress resistance, consistent with the observation that an L. monocytogenes ΔsigB mutant shows increased susceptibility to oxidative stress (18). From our data, we propose that σ^B-dependent resistance to oxidative damage in L. monocytogenes is likely due, at least in part, to glutathione reductase, which is encoded by lmo1433. Glutathione reductase is an enzyme that provides protection from oxidative stress by reducing glutathione disulfide to glutathione (9). L. monocytogenes has been shown to accumulate glutathione, supporting the possible activity of this enzyme during oxidative stress (47). While no glutathione reductase homologue was found in B. subtilis, oxidative stress protection in B. subtilis is dependent on the DNA-binding protein Dps, as well as on the trxA-encoded thioredoxin, both of which are σ^B dependent (59). Our HMM searches did not identify L. monocytogenes trxA as bearing a σ^B-dependent promoter.

Role of σ^B in virulence gene expression.

Characterization of L. monocytogenes ΔsigB mutants in both tissue culture and murine models of infection has provided initial evidence that σ^B contributes to the ability of L. monocytogenes to cause infection. While it has been shown that transcription from one of three prfA promoters (specifically the P2 promoter) is abolished in a ΔsigB mutant, no additional evidence for other functional contributions of σ^B to virulence have been reported. Our results provide clear evidence that σ^B in L. monocytogenes contributes to transcriptional activation of genes directly associated with virulence as well as those that likely contribute indirectly to virulence (58).

Virulence genes defined as being part of the σ^B regulon include bsh (encoding a bile salt hydrolase) as well as two genes from the internalin family (inlA and inlB). Although σ^B-dependent transcription of inlA and bsh has also been independently verified by reverse transcription-PCR (D. Sue and M. Wiedmann, unpublished data), and while RACE-PCR confirmed the presence of functional σ^B promoters for these genes, it is important to note that both genes also appear to be transcribed from σ^B-independent promoters (14, 15).

Interestingly, both bsh and inlA are regulated by PrfA (14, 15), a positive transcriptional regulator of virulence gene expression. While PrfA binding has been shown to activate transcription by binding to the −35 promoter region, none of the PrfA binding boxes upstream of the σ^B-dependent virulence genes described here overlap the σ^B consensus promoter −35 region. However, the σ^B-dependent P2 promoter of prfA contains a sequence resembling a PrfA box (23), and Milohanic et al. (42) reported a gene, lmo0596, with a putative PrfA box located at the −35 region of a predicted σ^B-dependent promoter. We conclude that transcription of a subset of L. monocytogenes virulence genes is regulated by a complex regulatory network that can include activation by both PrfA and σ^B. We hypothesize that this PrfA/σ^B regulatory mechanism coordinates transcriptional activation of subsets of L. monocytogenes virulence genes in specific host environments, e.g., bsh and inlA have been shown to be critical for listerial pathogenesis in the intestine (15, 39).

While bsh, inlA, and inlB represent experimentally verified virulence genes which were identified as being σ^B dependent, we also defined as members of the σ^B regulon additional putative L. monocytogenes virulence genes, including additional internalins. The L. monocytogenes internalins represent a diverse group of surface proteins with confirmed or putative virulence functions (39, 49, 57). While internalin B does not display an LPXTG motif, the other internalins include this cell wall anchor domain. Proteins with LPXTG motifs are common among gram-positive bacteria, and their functions are broad in range and frequently affect virulence (reviewed in reference 46). In addition to inlA, we identified four other L. monocytogenes genes encoding putative cell wall-anchored proteins displaying an LPXTG motif (including the internalin genes inlC2 and inlE) as being σ^B dependent. While inlC2 and inlE, which are part of the inlC2DE operon in L. monocytogenes 10403S, have not been shown experimentally to contribute to virulence, members of the homologous inlGHE operon present in other L. monocytogenes strains (e.g., EGD-e) have been shown to contribute to host cell internalization (5, 57). In addition to two other σ^B-dependent genes encoding proteins with LPXTG motifs (lmo2085 and lmo0880), we also identified one σ^B-dependent gene (lmo0994) that is unique to L. monocytogenes (i.e., absent from Listeria innocua and B. subtilis). Further studies with appropriate null mutants will be necessary to determine the specific functions of these putative σ^B-dependent virulence genes.

In addition to the established or putative virulence genes discussed above, we also identified σ^B-dependent stress response genes that had been shown previously to contribute to L. monocytogenes virulence, infection, and intrahost survival. Examples include the σ^B-dependent opuC operon; an L. monocytogenes LO28 opuC mutant showed reduced colonization of the mouse upper small intestine following peroral inoculation (60) and reduced numbers of bacteria in the spleens and livers of infected mice (66), although the phenotype appears to be strain specific (60). The σ^B-dependent gadB was previously shown to contribute to L. monocytogenes acid survival, including survival in synthetic gastric and ex vivo porcine stomach fluids (11). In conjunction with our findings described above, these data support a broad role for σ^B-dependent genes in L. monocytogenes virulence and intrahost survival. The majority of L. monocytogenes virulence studies thus far have used the mouse model of infection, but the mouse lacks the appropriate E-cadherin receptor, which is particularly critical for gastrointestinal invasion by L. monocytogenes (67). Use of more appropriate animal models to study L. monocytogenes pathogenesis (e.g., guinea pigs [39]) may help us to more clearly define the in vivo contributions of σ^B to L. monocytogenes virulence.

Studies in other gram-positive pathogens have provided further evidence that σ^B contributes to directing gene expression during host infection. In S. aureus, the virulence determinant SarA is expressed from a promoter that is strictly σ^B dependent (6, 41). SarA specifically activates transcription of agr, which encodes a positive regulator of extracellular virulence gene expression (43, 48). In both S. aureus and S. epidermidis, σ^B is also required for biofilm formation, a probable prerequisite for establishing infections (37, 54, 55). σ^B contributes to virulence in Bacillus anthracis; a sigB mutation in B. anthracis severely attenuates virulence in a murine model of infection (21). In addition to σ^B, stress-responsive sigma factors may contribute broadly to bacterial virulence. For example, the gram-negative stress-responsive alternative sigma factor RpoS has also been shown to contribute to virulence in a variety of pathogens, including Yersinia spp., Salmonella spp., Pseudomonas aeruginosa, and Legionella pneumophila (3, 17, 34, 61).

Comparative genomics of the σ^B-dependent stress response.

The definition of 54 σ^B-dependent genes in L. monocytogenes provides a unique opportunity for a comparative evaluation of the predicted functions of the σ^B-dependent stress response among low-GC-content, gram-positive bacteria. Overall, the range of protein functions associated with the L. monocytogenes σ^B-dependent genes defined in this work appears to be similar to the range of functions encoded by the genes described for the B. subtilis σ^B regulon (51, 53). Through similarity searches, we determined that 31 of the 54 σ^B-dependent genes in L. monocytogenes have homologous gene sequences in B. subtilis and 12 of these 31 genes show σ^B-dependent expression in both L. monocytogenes (this work) and B. subtilis (51, 53). These 12 genes include the stress response genes ctc, ltrC, and lmo1602, the rsbV operon, and metabolic genes (see Table 2). These genome-scale comparisons of the L. monocytogenes and the B. subtilis σ^B regulons thus suggest that the σ^B-dependent stress response system has adapted in L. monocytogenes to facilitate pathogen-host interactions.

The analysis of the S. aureus σ^B regulon performed by Gertz et al. (27) identified predominantly uncharacterized proteins. While 20 of the 27 σ^B-dependent S. aureus proteins identified had homologues in B. subtilis, only 7 of those homologues were σ^B dependent in B. subtilis. This finding parallels our own in that only a portion of the B. subtilis genes that are homologous to the σ^B-dependent genes identified in L. monocytogenes are also σ^B dependent in B. subtilis. Taken together, these observations suggest that the σ^B regulon has evolved to serve different roles among these bacteria.

Towards defining the complete L. monocytogenes σ^B regulon.

Stress-responsive alternative sigma factors, including RpoS in gram-negative bacteria (32) and σ^B in low-GC-content, gram-positive bacteria, contribute to regulation of large sets of genes, both directly and indirectly (51). As commercial, full-genome microarrays become available, more genes than those found here can be identified as σ^B dependent, including those not identified by the original HMM search. Still, defining all genes regulated by an alternative sigma factor represents a significant challenge, even with the availability of technologies such as two-dimensional gel electrophoresis and full-genome microarrays.

While our identification of 54 σ^B-dependent genes likely represents about one-third of the L. monocytogenes σ^B regulon, which is estimated to contain around 150 genes, the functional diversity represented by the proteins encoded by these genes provides valuable new insight into the specific functions of the L. monocytogenes σ^B regulon. In addition to the 54 genes reported here, many additional members of the L. monocytogenes σ^B regulon may have been correctly predicted by our HMM search, but their detection may have been masked by redundant regulation of transcription. Additional approaches such as in vitro transcription analyses with a purified L. monocytogenes σ^B-RNAP complex (8) or transcriptional profiling of an L. monocytogenes strain with an inducible sigB (53) will likely be necessary to identify and confirm additional members of the σ^B regulon. We have demonstrated, however, that a combination of HMM-based similarity searches and construction of a microarray as described here represents an efficient and economical approach for defining genes regulated by a specific mechanism, such as an alternative sigma factor.