Haemophilus ducreyi Hfq contributes to virulence gene regulation as cells enter stationary phase.

UNLABELLED: To adapt to stresses encountered in stationary phase, Gram-negative bacteria utilize the alternative sigma factor RpoS. However, some species lack RpoS; thus, it is unclear how stationary-phase adaptation is regulated in these organisms. Here we defined the growth-phase-dependent transcriptomes of Haemophilus ducreyi, which lacks an RpoS homolog. Compared to mid-log-phase organisms, cells harvested from the stationary phase upregulated genes encoding several virulence determinants and a homolog of hfq. Insertional inactivation of hfq altered the expression of ~16% of the H. ducreyi genes. Importantly, there were a significant overlap and an inverse correlation in the transcript levels of genes differentially expressed in the hfq inactivation mutant relative to its parent and the genes differentially expressed in stationary phase relative to mid-log phase in the parent. Inactivation of hfq downregulated genes in the flp-tad and lspB-lspA2 operons, which encode several virulence determinants. To comply with FDA guidelines for human inoculation experiments, an unmarked hfq deletion mutant was constructed and was fully attenuated for virulence in humans. Inactivation or deletion of hfq downregulated Flp1 and impaired the ability of H. ducreyi to form microcolonies, downregulated DsrA and rendered H. ducreyi serum susceptible, and downregulated LspB and LspA2, which allow H. ducreyi to resist phagocytosis. We propose that, in the absence of an RpoS homolog, Hfq serves as a major contributor of H. ducreyi stationary-phase and virulence gene regulation. The contribution of Hfq to stationary-phase gene regulation may have broad implications for other organisms that lack an RpoS homolog. IMPORTANCE: Pathogenic bacteria encounter a wide range of stresses in their hosts, including nutrient limitation; the ability to sense and respond to such stresses is crucial for bacterial pathogens to successfully establish an infection. Gram-negative bacteria frequently utilize the alternative sigma factor RpoS to adapt to stresses and stationary phase. However, homologs of RpoS are absent in some bacterial pathogens, including Haemophilus ducreyi, which causes chancroid and facilitates the acquisition and transmission of HIV-1. Here, we provide evidence that, in the absence of an RpoS homolog, Hfq serves as a major contributor of stationary-phase gene regulation and that Hfq is required for H. ducreyi to infect humans. To our knowledge, this is the first study describing Hfq as a major contributor of stationary-phase gene regulation in bacteria and the requirement of Hfq for the virulence of a bacterial pathogen in humans.

INTRODUCTION

Upon entry into stationary phase or under conditions of nutrient deprivation, bacteria exhibit global changes in gene expression that result in altered virulence and increased tolerance to stresses (1). In Escherichia coli and many other Gram-negative bacteria, this global change in gene expression is regulated in part by the stationary-phase sigma factor RpoS (2–4). However, several Gram-negative pathogens lack an obvious homolog of RpoS; it is unclear how gene expression is coordinated upon entry into stationary phase in these organisms.

Hfq, first identified as a host factor required for Qβ phage replication in E. coli, is a highly conserved, homohexameric RNA-binding protein (5). By preferentially binding to A/U-rich regions in regulatory small RNAs (sRNAs) and mRNAs, Hfq enhances sRNA-mRNA interactions, resulting in altered mRNA stability or translation (5). Hfq affects mRNA stability and translation by multiple mechanisms: (i) prior to interaction with their mRNA targets, Hfq may directly regulate the stability of sRNAs either by protecting them from degradation or by facilitating their degradation; (ii) Hfq may induce degradation of sRNA-mRNA complexes; (iii) Hfq in association with an sRNA may either repress translation by sequestering the ribosome-binding site or activate translation by exposing the translation initiation region; and (iv) Hfq may directly regulate the stability of mRNA transcripts by stimulating their degradation (5).

Hfq controls a wide variety of pathogenesis-related phenotypes in many bacteria, including motility, quorum sensing, biofilm formation, host cell adherence, invasion and intracellular survival, resistance to antimicrobial peptides, multidrug resistance, persister cell formation, and virulence (6–11). Hfq also affects a number of stress- and stationary-phase-related phenotypes due to its ability to regulate RpoS; mutant strains that lack hfq are defective in RpoS-mediated stress responses and stationary-phase adaptation (10, 12). Even in bacteria that lack RpoS, there is compelling evidence that Hfq contributes to stress- and stationary-phase-related phenotypes. For example, a Brucella abortus hfq mutant is more sensitive to a variety of stresses in stationary phase, and a Francisella novicida hfq mutant exhibits increased cell density at the transition to stationary phase (11, 13). In organisms that lack an RpoS homolog, whether Hfq serves as a major contributor of stationary-phase gene regulation is currently unknown.

Haemophilus ducreyi is a Gram-negative, obligate human pathogen that causes chancroid. Chancroid is a sexually transmitted genital ulcer disease (GUD) that manifests as painful genital ulcers and regional lymphadenopathy. Chancroid is now rare in the United States; cases are generally associated with contact with commercial sex workers in areas of endemicity. Chancroid is a prevalent GUD in the resource-poor countries of Africa, Asia, and Latin America (14). Due to syndromic management of sexually transmitted infections and lack of surveillance programs, the global prevalence of chancroid is now unknown (15). Apart from causing morbidity as a GUD, chancroid also facilitates the acquisition and transmission of human immunodeficiency virus type 1 (HIV-1) (16). In addition to chancroid, H. ducreyi also causes a nonsexually transmitted chronic lower limb ulceration syndrome that is reported from the South Pacific (17–19).

H. ducreyi lacks any recognized environmental or animal reservoir; humans are the only known hosts for this organism. In the human host, H. ducreyi associates with macrophages and neutrophils in an abscess and primarily remains extracellular (20, 21). While the doubling time of H. ducreyi in nutrient-rich broth is approximately 2 h, the estimated minimal doubling time in the human skin is 16.5 h (22). Thus, H. ducreyi encounters a variety of stresses, likely including nutrient limitation in vivo. To combat these stresses and successfully establish infection in the human host, H. ducreyi likely modifies its gene expression, physiological state, and virulence traits. However, H. ducreyi lacks an obvious homolog of RpoS, which controls the general stress response and stationary-phase adaptation in other organisms; how H. ducreyi regulates its gene expression in response to stationary phase is unclear.

We previously reported a comparison of the transcriptome sequencing (RNA-Seq)-based transcriptomes of the parent strain 35000HP, an isogenic cpxA deletion mutant, and an isogenic cpxR deletion mutant grown to the mid-log, transition, and stationary phases of growth (23). In this study, we characterized the gene expression differences at different growth phases in the parent strain 35000HP using the previously reported data and explored the contribution of Hfq to H. ducreyi stationary-phase gene regulation and virulence in humans. We found that cells harvested from stationary phase had increased expression of genes encoding several virulence determinants as well as a homolog of hfq. There was an overlap and an inverse relationship in the expression patterns of genes affected by inactivation of hfq and genes differentially expressed in stationary phase relative to mid-log phase. Finally, an hfq deletion mutant was fully attenuated for virulence in humans; inactivation or deletion of hfq downregulated several H. ducreyi virulence determinants.

RESULTS

Entry into stationary phase upregulates known H. ducreyi virulence determinants and a homolog of hfq.

To identify growth-phase-regulated genes in H. ducreyi, here we compared the transcriptomes of 35000HP at different growth phases using the previously reported RNA-Seq data (23). Four biological replicates were included for each growth phase, summing to a total of 12 samples. To identify genes differentially expressed at different growth phases, we calculated the fold change in the expression of genes in stationary phase compared to mid-log phase, stationary phase compared to transition phase, and transition phase compared to mid-log phase. We used a false-discovery rate (FDR) value of ≤0.1 and a 2-fold change as criteria for differential transcript expression as described previously (23). Comparison of transcriptomes in stationary phase to those in mid-log phase and of those in stationary phase to those in transition phase yielded 288 and 241 differentially expressed genes, respectively; approximately equal numbers of genes were up- and downregulated (Fig. 1). In general, the genes whose expression decreased in stationary phase relative to the mid-log and transition phases encode proteins involved in energy metabolism, biosynthesis, cell envelope homeostasis, transcription, and transport and binding (see Table S1 in the supplemental material). The genes whose expression increased in stationary phase relative to the mid-log and transition phases encode virulence determinants such as those in the flp-tad and lspB-lspA2 operons, a large number of hypothetical proteins (~64% of the upregulated genes), and a homolog of the RNA binding chaperone Hfq (see Table S2). There were only 7 genes differentially expressed between the mid-log and transition phases, suggesting that these two growth phases have similar gene expression patterns.

FIG 1 

Venn diagram showing the overlap in differential gene expression across different growth phases. The up- and downregulated genes are indicated by up (↑) and down (↓) arrows, respectively. The total number of differentially regulated genes in different comparisons is indicated in bold outside the Venn diagram. Stationary/mid-log, genes differentially expressed in stationary phase relative to mid-log phase; Stationary/transition, genes differentially expressed in stationary phase relative to transition phase; Transition/mid-log, genes differentially expressed in transition phase relative to mid-log phase.

Since the most profound differential regulation was noted in stationary phase relative to mid-log phase, we used this data set for quantitative reverse transcriptase PCR (qRT-PCR) validation. As described previously (23), the differentially expressed genes were grouped into 3 categories based on their expression levels (low, medium, and high); genes in each expression level were grouped into up- and downregulated targets, which were then further subgrouped based on their fold change ranges (2.0-fold to 5.0-fold, 5.1-fold to 10.0-fold, 10.1-fold to 15.0-fold, and 15.1-fold to 20.0-fold). Representative genes were selected arbitrarily from each category; a total of 14 genes were selected for qRT-PCR validation. qRT-PCR analysis confirmed the differential expression of 14/14 genes identified by RNA-Seq (Fig. 2). In general, the fold changes derived from RNA-Seq were in good agreement with those obtained from qRT-PCR (Fig. 2) (R2 = 0.82). Thus, qRT-PCR analysis confirmed that genes encoding known virulence determinants such as those in the flp-tad and lspB-lspA2 operons and hfq were upregulated in stationary phase.

FIG 2 

qRT-PCR validation of growth-phase-dependent differences in gene expression derived from RNA-Seq. (A) The fold change in the expression of target genes in stationary phase relative to mid-log phase. The expression levels of target genes were normalized to that of dnaE. The data represent the means ± SD of the results of four independent experiments. (B) Correlation between the fold changes obtained from qRT-PCR and RNA-Seq analysis. The diagonal line represents the power trendline (R2 = 0.82).

Hfq is a major contributor of gene regulation in H. ducreyi.

Since hfq transcripts were upregulated in stationary phase relative to mid-log phase and Hfq contributes to stationary-phase survival and posttranscriptional regulation of gene expression in other bacteria (10, 13), we compared the global expression profile of the 35000HP parent strain with that of the hfq insertional inactivation mutant (35000HPhfq::cat) using DNA microarray analysis. Inactivation of the hfq gene resulted in the differential regulation of 15.8% (289 open reading frames [ORFs]) of the predicted ORFs in the H. ducreyi genome. A total of 191 genes were significantly upregulated, and 98 genes were significantly downregulated (P < 0.05) (see Table S3 in the supplemental material). A list of the top 50 genes that were differentially regulated when hfq was inactivated, not including those ORFs annotated as encoding hypothetical proteins, is shown in Table 1. Genes encoding several different cellular processes were affected by inactivation of hfq (see Fig. S1 in the supplemental material). Hfq positively affected the transcript levels of several virulence factors, including components of the flp-tad operon (24, 25) and the two-partner secretion system encoded by lspB, lspA2, and lspA1 (26–28). qRT-PCR performed on a subset of 15 genes validated the DNA microarray data (Fig. 3) (R2 = 0.876). Taken together, these data suggest that Hfq is a major contributor of H. ducreyi gene regulation.

TABLE 1 

Inverse relationship in the expression patterns between the genes differentially expressed in 35000HPhfq::cat relative to 35000HP and the 35000HP genes differentially expressed in stationary phase relative to mid-log phase

Locus tag[a]	Gene	Description or homolog	Fold change
			35000HP-Δhfq::cat/35000HP[b]	35000HP-stationary/mid-log[c]
HD1435	ompP2B	Outer membrane protein P2 homolog	9.61	−5.0
HD1591		Conserved hypothetical protein	7.61	—[d]
HD1434		Hypothetical protein	6.99	−4.4
HD0384	nqrF	Na+-translocating NADH-ubiquinone oxidoreductase, subunit F	6.74	−4.2
HD1590	deaD	Cold-shock DEAD box protein-A	6.50	−3.1
HD0382	nqrD	Na+-translocating NADH-ubiquinone oxidoreductase, subunit D	6.41	−3.7
HD0647		Conserved hypothetical protein	6.27	−2.4
HD0386	apbE	Thiamine biosynthesis lipoprotein	6.22	−4.1
HD0357		Probable carbon starvation protein A	5.84	−4.2
HD0383	nqrE	Na+-translocating NADH-ubiquinone oxidoreductase, subunit E	5.75	−3.9
HD1512		Acriflavine resistance protein	5.57	−3.4
HD0381	nqrC	Na+-translocating NADH-ubiquinone oxidoreductase, subunit C	5.52	−4.6
HD0045	momP	Major outer membrane protein	5.23	−17.9
HD0710	mutM	Formamidopyrimidine-DNA glycosylase	5.22	−2.6
HD0648	tnaB	Tryptophan-specific transport protein	4.64	−2.5
HD1470	cpxA	Sensor kinase CpxA	4.48	−3.9
HD0564	aspA	Aspartate ammonia-lyase	4.45	−5.6
HD1109		Putative oxalate/formate antiporter	4.43	—
HD0387		Conserved hypothetical protein	4.41	—
HD1622		Conserved hypothetical protein	4.39	—
HD1163	ribAB	Riboflavin biosynthesis protein RibA	4.38	−3.5
HD0195		Hypothetical protein	4.35	−4.1
HD1343		Hypothetical protein	4.29	—
HD0380	nqrB	NADH dehydrogenase	4.26	−4.3
HD1162	ribE	Riboflavin synthase, alpha chain	4.11	−3.7
HD1357		Conserved possible translation initiation factor	4.11	−2.9
HD0282	fimB	Possible fimbrial structural subunit	3.77	—
HD0766	manZ	Mannose-specific phosphotransferase system IID component	3.66	−4.8
HD0646		Conserved hypothetical protein	3.59	—
HD0709	brnQ	Branched-chain amino acid carrier protein	3.59	−2.6
HD1624	aceF	Dihydrolipoamide acetyltransferase	3.56	−2.1
HD1471		Conserved hypothetical protein	3.54	−3.5
HD0889	deoD	Purine nucleoside phosphorylase	3.46	−3.3
HD0767	manY	Mannose-specific phosphotransferase system IIC component	3.40	−3.6
HD0876		Conserved probable RNase	3.37	—
HD1143	sdaC	Serine transporter	3.37	−2.5
HD1356	pyrF	Orotidine 5′-phosphate decarboxylase	3.37	−3.3
HD1312	flp1	flp operon protein Flp1	−3.70	7.0
HD0740	hflX	GTP-binding protein HflX	−3.70	4.8
HD1311	flp2	flp operon protein Flp2	−3.85	7.4
HD1503	guaB	Inosine-5-monophosphate dehydrogenase	−4.17	—
HD1310	flp3	flp operon protein Flp3	−4.35	7.4
HD0997		Hypothetical protein	−4.55	3.0
HD0805		Conserved hypothetical protein	−5.26	—
HD1433	ompP2A	Outer membrane protein P2 homolog	−7.14	2.2
HD0998	uraA	Uracil permease	−10.00	3.6
HD1985		Possible DNA transformation protein	−11.11	2.4
HD0232	arcB1	Ornithine carbamoyltransferase	−12.50	2.4
HD0233	carB	Carbamoyl-phosphate synthase, large subunit	−16.67	5.5
HD0235	carA	Carbamoyl-phosphate synthase, small subunit	−20.00	9.3

This table includes only the top 50 genes differentially expressed in the hfq inactivation mutant relative to its parent and their inverse relationship in expression patterns to the genes differentially expressed in stationary phase relative to mid-log phase.

35000HPΔhfq::cat/35000HP, fold change in transcript levels in 35000HPΔhfq::cat relative to 35000HP.

35000HP-stationary/mid-log, fold change in transcript levels in stationary phase relative to mid-log phase.

—, no difference in expression was found.

FIG 3 

Relative expression levels of selected H. ducreyi genes in 35000HPhfq::cat. (A) Expression levels of 15 selected genes in 35000HPhfq::cat compared to 35000HP were measured by DNA microarray (black bars) or real-time RT-PCR (white bars) as described in Materials and Methods. These data are from a representative experiment. (B) Correlation between log2 values obtained by DNA microarray analysis and qRT-PCR analysis. The diagonal line represents the power trendline (R2 = 0.8760).

Hfq is a major contributor of H. ducreyi stationary-phase gene regulation.

Given that H. ducreyi hfq was upregulated in stationary phase relative to mid-log phase, we examined whether there was an overlap between the genes altered by inactivation of hfq and the genes differentially expressed in stationary phase relative to mid-log phase. Since transcripts encoding ribosomal proteins were depleted by RNA-Seq, 17 of the 289 genes directly or indirectly regulated by Hfq were excluded from this analysis; since the microarray analysis was done on the hfq inactivation mutant, hfq was excluded from the 288 genes affected by the transition from mid-log to stationary phase. Our results showed that there was a significant overlap (n = 118 genes) in the genes differentially expressed in the hfq inactivation mutant relative to its parent and the genes differentially expressed in stationary phase relative to mid-log phase (chi-square test, P <2.2e-16) (Fig. 4A and Table 1; see also Table S4 in the supplemental material). Comparison of the fold changes of the overlapping genes showed that the fold changes of genes altered by inactivation of hfq inversely correlated with the fold changes of genes differentially expressed in stationary phase relative to mid-log phase (R2 = 0.8) (Fig. 4B and Table 1; see also Table S4). There was also an overlap in the expression patterns of several genes encoding known virulence determinants, including the components of the flp-tad operon (flp1, flp2, flp3, and tadA) and lspB-lspA2 operon (lspB) (Table 1; see also Table S4). These data support the hypothesis that Hfq serves as a major contributor of H. ducreyi stationary-phase and virulence gene regulation.

FIG 4 

Comparison of the expression patterns of the genes altered by inactivation of hfq to those of the genes differentially expressed in stationary phase relative to mid-log phase. (A) Venn diagram showing the overlap in the genes altered by inactivation of hfq and the genes differentially expressed in stationary phase relative to mid-log phase. The total number of genes in each comparison is indicated in bold outside the Venn diagram. Potential Hfq targets, genes differentially expressed in 35000HPΔhfq::cat relative to 35000HP; Stationary/mid-log, genes differentially expressed in stationary phase relative to mid-log phase. The significance of the overlap was tested using the chi-square test (P < 2.2e-16). (B) Inverse correlation between the fold changes of the genes altered by inactivation of hfq and those of the genes differentially expressed in stationary phase relative to mid-log phase. Only genes that overlapped in their expression patterns as shown in panel A were used for the correlation analysis.

Hfq is required for pustule formation in humans.

Inactivation of hfq altered the expression of several genes encoding virulence determinants, suggesting that Hfq might contribute to H. ducreyi infection in humans. Both by microarray and qRT-PCR analyses, the expression of hflX, the gene downstream of hfq, was reduced approximately 3.5-fold in the inactivation mutant. Given that the U.S. Food and Drug Administration prefers use of unmarked mutants for testing in humans and that we did not know if the downregulation of hflX was due to regulation by Hfq or to a polar effect of the insertion, we constructed an unmarked deletion mutant of hfq (35000HPΔhfq) for mutant-parent trials in human volunteers in strain 35000HP using recombineering methodology (29). Sequence analysis confirmed the expected deletion with no secondary mutations in the flanking regions. By qRT-PCR analysis, there was no difference in the expression levels of hflX between the deletion mutant and the parent, suggesting that Hfq does not regulate hflX expression. 35000HP∆hfq grew at the same rate as 35000HP to mid-log phase but lagged slightly in the late log and stationary phases (see Fig. S2 in the supplemental material), suggesting that deletion of hfq affected the growth of H. ducreyi in broth culture only slightly. Except for downregulation of DsrA, the outer membrane proteins (OMPs) and the lipo-oligosaccharide profiles of the 2 strains were identical (data not shown). Thus, the unmarked hfq deletion mutant met the phenotypic criteria required by our protocol for mutant-parent trials in humans.

To examine whether Hfq is required for virulence in humans, we infected groups of volunteers with the hfq deletion mutant and its parent in escalating dose-ranging studies. In this model, papule formation signifies initiation of infection, while pustule formation signifies disease progression. In one group, two volunteers (420 and 423) formed pustules at 4 of 6 parent sites that were inoculated with 109 CFU of the parent and at 0 of 6 mutant sites that were inoculated with 29, 58, or 116 CFU of the mutant (Table 2). These results suggested that the hfq deletion mutant was attenuated for virulence. In another group, three volunteers (428, 429, and 430) were inoculated at 3 sites with 87 CFU of the parent and at 3 sites with 52, 104, and 209 CFU of the mutant. Pustules formed at 3 of 9 parent sites and at 0 of 9 mutant sites (Table 2). Cumulative results from the two groups of subjects showed that papule formation rate was 93.3% (95% confidence interval [CI], 81.7% to 99.9%) at 15 parent sites and 86.7% (95% CI, 72.4% to 99.9%) at 15 mutant sites (P = 0.55) (Table 2). After 24 h of infection, the mean parent papule size (16.5 ± 10.4 mm2) was significantly larger than mutant papule size (6.8 ± 6.9 mm2) (P = 0.003). The pustule formation rates were 46.7% (95% CI, 7.0% to 86.3%) at 15 parent sites and 0% (95% CI, 0.0% to 0.45%) at 15 mutant sites (P = 0.011) (Table 2). At least one positive surface culture, defined as a culture that yielded at least one colony of H. ducreyi, was obtained during follow-up visits from 6.7% of the parent-inoculated and 0% of the mutant-inoculated sites. Thus, the hfq deletion mutant was fully attenuated for virulence in humans (30).

TABLE 2 

Response to inoculation with live H. ducreyi strains

Volunteer (sex)[a]	Observation period (days)	Strain[b]	Dose(s) (CFU)[c]	No. of initial papules	No. of pustules at endpoint
420 (M)	9	P	109	3	3
		M	29–116	2	0
423 (M)	13	P	109	3	1
		M	29–116	3	0
428 (M)	7	P	87	3	0
		M	52–209	2	0
429 (F)	7	P	87	3	3
		M	52–209	3	0
430 (M)	8	P	87	2	0
		M	52–209	3	0

Volunteers 420 and 423 were inoculated in one group; volunteers 428, 429, and 430 were inoculated in another group. M, male; F, female.

P, parent (35000HP); M, mutant (35000HP∆hfq).

Data represent the doses inoculated at 3 sites, except 29–116 (one dose each of 29, 58, and 116 CFU) and 52–209 (one dose each of 52, 104, and 209 CFU).

Colonies from subcultures of the parent (n = 71) and mutant (n = 72) inocula used to infect volunteers were tested for the presence of hfq and dnaE sequences by colony hybridization. The dnaE probe hybridized to all colonies from both the parent and mutant inocula, while the hfq probe hybridized only to colonies from parent inocula. One biopsy specimen was cultured from a parent site; both probes hybridized to all colonies tested (n = 35). Thus, there was no evidence of cross-contamination between mutant and parent inocula for the five subjects included in the trial.

Hfq contributes to positive regulation of important H. ducreyi virulence determinants.

Given that the hfq deletion mutant was compromised for virulence in vivo (Table 2), we sought to determine if Hfq contributes to the regulation of known H. ducreyi virulence determinants. Inactivation of the hfq gene in H. ducreyi 35000HP resulted in decreased synthesis of LspB, LspA2, DsrA, and Flp1 but not of LspA1 (Fig. 5, lanes 1 and 2); LspA2, DsrA, and Flp1 are absolutely required for pustule formation in humans (30, 31). LspB is involved in the secretion of LspA1 and LspA2 proteins that allow H. ducreyi to avoid phagocytosis (32), whereas DsrA is an autotransporter protein involved in resistance to serum killing (33). The Flp proteins are necessary for H. ducreyi microcolony formation in vitro; the ability of H. ducreyi to form microcolonies in vitro is correlated with virulence in humans (24, 25, 31). The decreased synthesis of these proteins in the hfq inactivation mutant was restored to parental levels by complementation with the hfq gene in trans (Fig. 5, lane 3). These data suggest that Hfq contributes to the positive regulation of several known H. ducreyi virulence determinants.

FIG 5 

Inactivation of the H. ducreyi hfq gene has a positive effect on the synthesis of known virulence determinants. Data represent the results of Western blot analysis of whole-cell lysates from 35000HP (lane 1), 35000HPhfq::cat (lane 2), 35000HPhfq::cat(pML129) (lane 3), and 35000HPhfq::cat(pACYC177) (lane 4) probed with LspA1 MAb 40A4, LspA2 MAb 1H9, an LspB polyclonal antibody, a DsrA polyclonal antibody, or a Flp1 polyclonal antibody. The LspA1 and LspA2 proteins do not have discrete banding patterns in Western blots but instead form smears (27). PAL MAb 3B9 was used to confirm equivalent loading among lanes.

Hfq contributes to H. ducreyi microcolony formation.

Western blot analysis (Fig. 5) indicated that inactivation of hfq downregulated the synthesis of Flp proteins, suggesting that the H. ducreyi hfq inactivation mutant might be deficient in microcolony formation. The hfq inactivation mutant formed bacterial cell aggregates, but these clusters of bacterial cells were much less compact than those formed by the parent strain (Fig. 6). This deficiency in microcolony formation by the hfq inactivation mutant could be corrected by complementation (Fig. 6). These data suggest that Hfq is important for H. ducreyi microcolony formation.

FIG 6 

H. ducreyi microcolony formation assay. The relative abilities of 35000HP, 35000HPhfq::cat, 35000HPhfq::cat(pML129), and 35000HPhfq::cat(pACYC177) to form microcolonies was tested by incubating these strains with Hs27 fibroblasts. The larger pictures were taken at a ×14 magnification. The contents of the white boxes are shown to the right of each picture at a ×40 magnification. Results of a representative experiment are shown. Note that the hfq mutant formed bacterial cell aggregates but that these clusters of bacterial cells were much less compact than those formed by the parent and complemented strains.

Deletion of hfq results in an intermediate serum resistance phenotype.

Similarly to the inactivation mutant, the hfq deletion mutant synthesized less DsrA than 35000HP as assessed by Western blot analysis (Fig. 5) and by OMP profiles (data not shown). Thus, we compared the survival rates of strain 35000HP, strain 35000HP∆hfq, and a previously constructed isogenic dsrA mutant (33) in 50% normal human serum (NHS). In these assays, the hfq deletion mutant survived at significantly lower levels (mean survival ± standard deviation [SD], 46% ± 17%) than did the parent (87% ± 19%) and at significantly higher levels than did the dsrA mutant (7% ± 4%) (Fig. 7A). We next compared the survival rates of strains 35000HP(pACYC177), 35000HP∆hfq(pACYC177), 35000HP∆hfq(pML129), and 35000HP∆dsrA(pACYC177) in 50% NHS. In these experiments, the mean percentages of survival ± the SD were 54% ± 8% for the parent, 40% ± 19% for the hfq deletion mutant, 80% ± 9% for the complemented strain, and 6% ± 2% for the dsrA mutant (Fig. 7B). In these experiments, the complemented strain survived at significantly higher levels than the hfq deletion mutant. All together, these data suggest that Hfq contributes to H. ducreyi serum resistance.

FIG 7 

Serum bactericidal assays. (A) Percent survival of 35000HP, 35000HP∆hfq, and the dsrA mutant in 50% NHS, calculated as follows: (geometric mean CFU in NHS/geometric mean CFU in heat-inactivated NHS) × 100. (B) Percent survival of 35000HP(pACYC177), 35000HP∆hfq(pACYC177), 35000HP∆hfq(pML129), and 35000HP∆dsrA(pACYC177) in 50% NHS, calculated as follows: (geometric mean CFU in NHS/geometric mean CFU in heat-inactivated NHS) × 100. Values are means ± SD of the results of 5 independent experiments.

Structural biology of the H. ducreyi Hfq protein.

Like Hfq from other organisms (10), H. ducreyi Hfq contributed to gene regulation, including the regulation of the genes that encode virulence determinants. Therefore, we sought to determine if H. ducreyi Hfq binds to RNA similarly to Hfq in other organisms. To this end, we performed structural analysis of the H. ducreyi Hfq protein. To gain insights into the structural properties of the H. ducreyi Hfq protein, the Protein Data Bank (PDB) was queried for similar sequences using a hidden-Markov approach (34). Several bacterial proteins with the Sm fold, a highly conserved bipartite sequence motif that mediates RNA binding and protein-protein interactions, were returned; the best match was to the Hfq protein from Herbaspirillum seropedicae (35). This protein shared about 70% amino acid sequence identity with H. ducreyi Hfq, and the reported probability of the match was 100%. These indicators of sequence similarity suggested that the structure of the H. ducreyi Hfq protein would be very similar to that of the H. seropedicae molecule. Consequently, using H. seropedicae Hfq as a template, MODELLER (36) was used to build a hypothetical model of the tertiary structure of the H. ducreyi Hfq protein (see Fig. S3 in the supplemental material). The model predicted that H. ducreyi Hfq likely adopts the classical Hfq fold (5), with an amino-terminal α-helix followed by 5 β-strands. These strands form two β-sheets that are arranged in a “squat barrel.” The second β-strand in such structures adopts a highly curved configuration that allows it to participate in both sheets. The C termini of Hfq proteins tend to vary; therefore, this portion of the protein was removed from the model. The quaternary structure of other Hfq proteins is homohexameric (37), and the H. ducreyi protein very likely conforms to this expectation. This discoid structure is described as having two faces: the “proximal” face, which is close to the amino termini of the monomers, and the “distal” face, which is opposite the proximal face.

Several structures are known for other Hfq proteins bound to RNAs. For the purpose of examining the likely modes of RNA binding to H. ducreyi Hfq, we focused on two RNA-containing structures: E. coli Hfq bound to poly(A) (38) and Hfq bound to AU6A (39). In the E. coli Hfq/poly(A) structure, the RNA is bound to the distal face of the hexamer, with some bases oriented away from the protein and some penetrating into invaginations in its surface. The E. coli Hfq/AU6A structure features RNA bound to the proximal face, with some bases making contacts with amino acid side chains and others facing away from the protein. When three-dimensional alignments of these structures with the model of the H. ducreyi Hfq protein were examined, no conflicts were found that would preclude either of these RNA-binding modes. Thus, the H. ducreyi Hfq protein is likely to bind RNA similarly to E. coli Hfq.

DISCUSSION

To survive in the host environment, H. ducreyi likely senses and responds to stresses by altering its gene expression. Here, we sought to define the global gene expression patterns in H. ducreyi at distinct growth phases, with the goal of understanding how this organism adapts to stationary phase in the absence of RpoS and regulates its virulence determinants in response to alterations in growth. We showed that cells in stationary phase had a broad transcriptional response and upregulation of several virulence determinants and a homolog of hfq. We provided evidence that Hfq likely contributes to regulation of gene expression in stationary phase. We also showed that an hfq mutant exhibits reduced expression of genes encoding several virulence determinants and is attenuated for virulence in human volunteers. Taken together, these data suggest an important role for Hfq in controlling H. ducreyi stationary-phase gene expression and virulence.

In E. coli, entry into stationary phase is accompanied by distinct changes in gene expression that result in altered growth rate and increased resistance to a variety of stresses (1). In general, entry into stationary phase downregulates many genes involved in transcription, translation, biosynthesis of macromolecules, and energy metabolism; this is consistent with the idea that stationary-phase organisms exhibit metabolic downshift that is accompanied by reduced cell division and no net bacterial growth. A large number of genes involved in stress adaptation are induced upon entry into stationary phase; the majority of these genes are regulated by the alternative sigma factor RpoS (40). Stationary-phase H. ducreyi had a distinct transcriptional profile compared to organisms grown to mid-log and transition phases. Similarly to organisms that contain RpoS (41), cells harvested from stationary phase had decreased expression of homologs of genes involved in transcription, translation, biosynthesis of macromolecules, and energy metabolism. Stationary-phase cells had increased expression of homologs of genes involved in arginine biosynthesis, cell envelope homeostasis, and regulation and a large number of genes encoding hypothetical proteins. Thus, H. ducreyi regulates its gene expression in stationary phase similarly to organisms that contain RpoS.

In many Gram negative pathogens, entry into stationary phase is accompanied by upregulation of genes encoding virulence determinants (3, 4). Stationary-phase H. ducreyi had increased expression of genes encoding components of the flp-tad operon and a component of the two-partner secretion system, lspB. The Flp proteins are necessary for H. ducreyi microcolony formation; mutants that do not synthesize Flp proteins are fully attenuated for virulence in humans (24, 25, 31). LspB is required for secretion of LspA1 and LspA2 proteins, which are involved in the inhibition of phagocytosis; an lspA1 lspA2 double mutant is also fully attenuated for virulence in humans (28, 32). Thus, stationary-phase cells had increased expression of genes encoding several known H. ducreyi virulence determinants.

H. ducreyi harvested from stationary phase also had increased expression of a homolog of the RNA-binding chaperone Hfq. The fact that hfq was upregulated in stationary phase in H. ducreyi led us to explore the potential contribution of Hfq to regulation of gene expression in H. ducreyi. To this end, we constructed an insertional inactivation mutant of hfq; sequence analysis indicated the presence of an in-frame insertion in the hfq ORF. Microarray analysis showed that inactivation of hfq led to differential expression of ~16% of the H. ducreyi open reading frames. Despite the presence of an in-frame insertion in the hfq ORF, both microarray and qRT-PCR analyses showed that the downstream gene, hflX, was downregulated in the hfq inactivation mutant. HflX is a GTPase that coprecipitates with the 50S ribosomal subunit and is hypothesized to have a role in translation (42). Given that the function of HflX is unknown, whether downregulation of hflX in the hfq inactivation mutant contributes to altered gene expression in this mutant is unclear. However, the altered synthesis of virulence determinants and their associated phenotypes in the hfq inactivation mutant were restored to parental levels by complementation, suggesting that these phenotypes were due to inactivation of hfq.

Microarray analysis also showed that inactivation of hfq altered the expression of genes encoding several virulence determinants (30, 31). Therefore, we hypothesized that Hfq would be required for virulence in humans. Given that inactivation of hfq decreased the expression of hflX and that the FDA prefers the use of unmarked mutants in human volunteers, we generated an unmarked deletion mutant of hfq for the human inoculation experiments; hflX expression was unchanged in the deletion mutant relative to the parent. The hfq deletion mutant caused papules that were significantly smaller than those caused by its parent and formed no pustules. Similarly to the insertion mutant, the deletion mutant had decreased synthesis of DsrA; deletion of hfq decreased the ability of H. ducreyi to resist human serum. Inactivation of hfq reduced the ability of H. ducreyi to form microcolonies in vitro; Flp1, a determinant of microcolony formation, was also decreased in the hfq inactivation mutant. In addition, inactivation of hfq decreased the synthesis of LspB and LspA2, which are involved in resisting phagocytosis. Although we did not evaluate either hfq mutant for resistance to phagocytosis, it is highly likely that both mutants would also be impaired in this regard. Thus, the attenuation of the hfq deletion mutant in the human challenge model is likely due to downregulation of several virulence determinants.

Inactivation of hfq affected the expression levels of several genes that are involved in gene regulation. For example, the transcript levels of cpxRA were upregulated in the hfq inactivation mutant. Activation of the CpxRA system is dependent on the levels of phosphorylated CpxR, and activation of CpxR renders H. ducreyi totally avirulent in humans (29). Inactivation of hfq also increased the expression of pta, which is involved in the synthesis of acetyl phosphate (AcP), but did not affect the expression of ackA, which is involved in the degradation of AcP. CpxR can accept phosphoryl groups from AcP, causing the Cpx system to be activated (43). Whether the increase in cpxRA and pta transcript levels results in accumulation of phosphorylated CpxR in the hfq mutant is unknown. However, comparing the genes altered by inactivation of hfq to those regulated by activation of CpxR in the cpxA deletion mutant (23), approximately 20% of the genes altered by inactivation of hfq overlapped and were positively correlated with the genes regulated by activated CpxR (data not shown; chi-square test, P = 9.6e-08). Given that activation of the Cpx system attenuates the virulence of H. ducreyi in humans, these data suggest that the Cpx system may be activated in the hfq mutant and that such activation may in part be responsible for the attenuation of the hfq mutant.

In E. coli and many other Gram-negative bacteria, Hfq positively regulates RpoS; mutant strains that lack hfq are compromised in RpoS-dependent stationary-phase adaptation (44). However, several Gram-negative pathogens lack an obvious homolog of rpoS but contain a homolog of hfq, including Brucella abortus, Neisseria gonorrhoeae, Francisella novicida, Actinobacillus pleuropneumoniae, Haemophilus influenzae, Pasteurella multocida, Leptospira interrogans, Bartonella bacilliformis, and Bordetella pertussis (11, 13, 45–47). A Brucella abortus hfq mutant is compromised in its ability to survive under conditions of stresses in stationary phase, and a Francisella novicida hfq mutant shows increased cell density at the transition to stationary phase (11, 13). Similarly, the H. ducreyi hfq mutant slightly lagged in growth in the late log phase and stationary phase. There was a significant inverse relationship in the expression patterns of the genes altered by inactivation of hfq and the genes differentially expressed in stationary phase relative to mid-log phase. Thus, in the absence of an RpoS homolog, Hfq likely serves as a major contributor of stationary-phase gene regulation in H. ducreyi.

Hfq regulated only a subset (41%) of the H. ducreyi genes differentially expressed in stationary phase relative to mid-log phase. It is unknown how the remainder (59%) of the differentially expressed genes are regulated. Carbon storage regulator A (CsrA) and guanosine tetraphosphate (ppGpp) are important posttranscriptional and transcriptional regulators of stationary-phase gene expression, respectively (1, 48). An H. ducreyi csrA mutant is partially attenuated for virulence in humans and exhibits more pronounced stress survival phenotypes in stationary phase (49). Preliminary studies in our laboratory indicate that H. ducreyi synthesizes ppGpp, and a ppGpp null mutant is attenuated for virulence in humans (C. Holley, W. Li, K. R. Fortney, D. M. Janowicz, S. Ellinger, B. Zwickl, B. P. Katz, and S. M. Spinola, unpublished data). Thus, in addition to Hfq, CsrA and ppGpp likely play a role in stationary-phase gene regulation and survival in H. ducreyi.

In E. coli and other bacteria, Hfq serves as a chaperone that facilitates the interaction of sRNAs with their mRNA targets (5). Structure analysis of the H. ducreyi Hfq protein revealed that Hfq might bind RNA similarly to E. coli Hfq. By RNA-Seq analysis, we identified 10 putative sRNAs, 7 of which are homologous to sRNAs in other bacteria (transfer-mRNA [tmRNA], 6S RNA, GcvB, bacterial small RNA signal recognition particle [SRP], RNase P_bact_a, flavin mononucleotide [FMN] riboswitch, and a lysine riboswitch) and 3 of which appear to be unique to H. ducreyi (data not shown). However, the Hfq dependence of these putative sRNAs and their contribution to H. ducreyi pathogenesis are currently unknown.

By facilitating interaction of sRNAs with their mRNA targets, Hfq affects either the stability or translation initiation of its targets. Inactivation of hfq altered both the transcript and protein levels of genes encoded by the flp-tad and the lspB-lspA2 operons; these data suggest that Hfq likely affects the stability of the transcripts of these genes. Despite the decreased expression of DsrA detected in Western blot analysis and OMP profiles and the decreased resistance to human serum, the transcript levels of dsrA were unchanged in the hfq mutant compared to its parent; these data suggest that the regulation of DsrA by Hfq is likely at the translational level. Thus, similarly to other organisms, H. ducreyi Hfq likely contributes to the regulation of its targets at both the transcript and translational levels.

In summary, we show that Hfq likely serves as a major contributor of virulence and stationary-phase gene regulation in H. ducreyi. Future studies will focus on identifying Hfq-dependent proteins and sRNAs and the potential contribution of Hfq-dependent sRNAs to H. ducreyi pathogenesis.

MATERIALS AND METHODS

Bacterial strains, plasmids, and culture conditions.

The bacterial strains and plasmids used in this study are listed in Table 3. H. ducreyi strains were grown on chocolate agar supplemented with 1% IsoVitalex at 33° C with 5% CO2. Alternatively, H. ducreyi strains were grown in gonococcal (GC) broth or Columbia broth supplemented with 2.5% or 5% fetal bovine serum (Hyclone), respectively, 1% IsoVitalex, and 50 µg/ml of hemin (Aldrich Chemical Co.) at 33° C. For RNA-Seq experiments, strain 35000HP was grown to the mid-log phase (OD660 = 0.2), transition phase (OD660 = 0.31), or early stationary phase (referred to here as stationary phase; OD660 = 0.35) in supplemented GC broth. E. coli strains were grown in Luria-Bertani media at 37° C except for strain DY380, which was maintained in L-broth or agar and grown at 32° C or 42° C for induction of the λ red recombinase. Where necessary, the media was supplemented with kanamycin (20 µg/ml for H. ducreyi; 50 µg/ml for E. coli), spectinomycin (200 µg/ml for H. ducreyi; 50 µg/ml for E. coli), or chloramphenicol (1 µg/ml for H. ducreyi; 30 µg/ml for E. coli).

TABLE 3 

Bacterial strains and plasmids used in this study

Strain or plasmid	Description	Source or reference
Strains
E. coli DH5α + HB101	Strains used for general cloning procedures	Invitrogen
E. coli DY380	DH10B derivative containing a defective λ prophage in which the red, bet, and gam genes are controlled by the temperature-sensitive λcI857 repressor	57
H. ducreyi 35000HP	Human passaged variant of strain 35000	58
H. ducreyi 35000HPhfq::cat	35000HP with a chloramphenicol resistance cassette inserted into the hfq gene	This study
H. ducreyi 35000HPΔhfq	35000HPhfq unmarked, in-frame deletion mutant	This study
H. ducreyi FX517	35000dsrA::cat insertion mutant	33
Plasmids
pCR2.1	Cloning vector	Invitrogen
pML104	pCR2.1 carrying the wild-type 35000HP hfq gene and flanking DNA	This study
pML121	pCR2.1 carrying the cat promoter from ΔEcat and the cat cassette from pSL1 flanked with SmaI sites	This study
pML120	pML104 with a chloramphenicol resistance cassette inserted into the hfq gene	This study
pKF10	pCR-XL-TOPO containing the hfq-coding region along with 0.5-kb flanking regions	This study
pRSM2832	Plasmid containing spectinomycin resistance cassette flanked by the FRT sites	29
pKF11	hfq replaced with spectinomycin resistance cassette in pCR-XL-TOPO	This study
pRSM2072	H. ducreyi suicide vector	59
pKF12	hfq replaced with spectinomycin resistance cassette in pRSM2072	This study
pACYC177	Cloning vector	New England Biolabs
pML129	pACYC177 carrying the wild-type 35000HP hfq gene	This study

RNA-Seq analysis of growth-phase-dependent differences in gene expression.

We previously reported a comparison of the transcriptomes of strain 35000HP, a cpxA deletion mutant, and a cpxR deletion mutant grown to the mid-log, transition, and stationary phases of growth (23). The growth-phase-dependent differences in gene expression were identified by comparing the transcriptomes of 35000HP at different growth phases using previously reported data (23). Given that transcripts encoding ribosomal proteins are depleted from the total transcripts, RNA-Seq experiments in general suffer from the caveat that genes encoding ribosomal proteins cannot be quantified using this technique. Therefore, as described in Results, genes encoding ribosomal proteins were not included in these analyses. As described previously (23), a prespecified FDR of ≤0.1 and a 2-fold change were used as criteria for differential transcript expression analysis. The differentially expressed genes were grouped into functional categories using the role classification available in the comprehensive microbial resource database (50).

Validation of RNA-Seq data by qRT-PCR.

qRT-PCR was performed using a QuantiTect SYBR green RT-PCR kit (Qiagen) and an ABI Prism 7000 sequence detector (Applied Biosystems) as described previously (23). The primer pairs were designed to amplify internal gene-specific fragments ranging from 70 to 200 bp. The amplification efficiency was determined for each primer pair (P1 to P16; see Table S5 in the supplemental material); all primer pairs had >95% efficiency. The expression levels of target genes were normalized to that of dnaE, which was amplified using primer pair P17 and P18 (see Table S5).

DNA microarray analysis and qRT-PCR validation of microarray results.

Total RNA was isolated from broth-grown H. ducreyi cultures as previously described (51). The H. ducreyi custom spotted DNA microarrays used in this study have been previously described (51). For each experiment, 5 µg of total RNA extracted from cells grown to mid-log phase (8 h) was used for first-strand cDNA synthesis as previously described (51). To avoid gene-specific dye bias, each sample was subjected to reverse labeling (dye swap). Differential expression was defined as a minimum of a 2-fold change in expression in the 35000HPhfq::cat strain relative to 35000HP. The data were further analyzed to include only expression profiles that had a P ≤ 0.05 after a one-sample t test analysis. The differentially expressed genes that achieved statistical significance were grouped into functional categories based on the role classification available in the comprehensive microbial resource database (50).

Fifteen genes were randomly selected for further confirmation of their relative transcript levels by two-step qRT-PCR. Primers (P19 to P36) used in this study are listed in Table S5 in the supplemental material. The reverse transcriptase reaction was performed as described previously (51). Assays were performed on two independent biological replicates, using HD1643 (gyrB) to normalize the amount of cDNA per sample. The fold change of each gene was calculated using the 2−ΔΔCT method.

Construction and complementation of an hfq insertional inactivation mutant and an unmarked, nonpolar hfq deletion mutant.

To construct an hfq insertional inactivation mutant, a ~3-kb fragment containing the hfq ORF as well as ~1 kb of both upstream and downstream flanking DNAs was introduced into pCR2.1 (Invitrogen) to obtain plasmid pML104. A cat cartridge from pSL1 (52), modified to contain its native promoter (51) and flanked with SmaI sites, was introduced into pCR2.1 to obtain plasmid pML121. The hfq ORF was interrupted by digesting pML104 with SwaI (native restriction site located 35 nucleotides [nt] inside the ORF) and inserting the cat cassette, which had been excised from pML121 using SmaI. The resultant construct was designated pML120. Primer pair P37 and P38 (see Table S5 in the supplemental material) was used to amplify a ~3-kb fragment from pML120 which was subsequently subjected to digestion with DpnI, gel purified, and used to electroporate H. ducreyi 35000HP as previously described (51). An hfq insertional inactivation mutant (35000HPhfq::cat) was selected on chocolate agar plates containing chloramphenicol; nucleotide sequence analysis confirmed the presence of the in-frame insertion within the hfq ORF.

An unmarked, in-frame deletion mutant of H. ducreyi hfq was made using “recombineering” methodology exactly as described previously (29, 31, 49). All primer pairs used in the construction of the deletion mutant are listed in Table S5 in the supplemental material. Briefly, hfq and its flanking sequences were amplified using primer pair P39 and P40 and cloned into pCR-XL-TOPO to generate pKF10, which was electroporated into DY380, a strain of E. coli that expresses λ red recombinase. A pair of 70-bp primers was used to amplify a spectinomycin (spec) resistance cassette flanked by the flippase recognition target (FRT) sites employing pRSM2832 as the template (29). P41 included 47 bp upstream of hfq, its start codon, and 20 bp homologous to the 5′ end of the spec cassette; P42 included 21 bp at the 3′ end of hfq, 29 bp of the downstream region, and 20 bp corresponding to the 3′ end of the spec cassette. The mutagenic amplicon was electroporated into DY380(pKF10) for recombination, generating pKF11. A SpeI-digested fragment containing the hfq flanking regions and the spec cassette was cloned into the suicide vector pRSM2072, generating pKF12, which was electroporated into 35000HP. After allelic exchange was confirmed by PCR, FLP recombinase was used to excise the spec cassette exactly as previously described (29). This resulted in replacement of the hfq gene by a short ORF that includes the hfq start codon, 81 bp encoding a flippase (FLP) scar peptide, and the last 21 bp of hfq, including its stop codon. By employing this design, the downstream gene, hflX, should be transcribed and translated normally. The hfq deletion was confirmed by sequence analysis; the final mutant was designated 35000HPΔhfq. Primer pair P43 and P44 (see Table S5), which binds to a region within the downstream gene hflX, was used to confirm that the deletion in 35000HPΔhfq did not affect the expression of hflX by qRT-PCR.

To complement both hfq mutants, the wild-type 35000HP hfq gene, together with ~500 bp 5′ from the ATG translational start codon and ~200 bp 3′ from the translational stop codon, was amplified from chromosomal DNA using primer pair P45 and P46 (see Table S5 in the supplemental material). The amplicon was digested with SacII and ligated to SacII-digested pACYC177 (New England Biolabs) to obtain pML129. After the result was confirmed by PCR and sequence analysis, pML129 DNA was used to transform strains 35000HPhfq::cat and 35000HPΔhfq to obtain kanamycin-resistant strains 35000HPhfq::cat(pML129) and 35000HPΔhfq(pML129), respectively. The 35000HP wild-type, 35000HPΔhfq, and 35000HPhfq::cat strains were also transformed with pACYC177, and the resulting strains were designated 35000HP(pACYC177), 35000HPΔhfq(pACYC177), and 35000HPhfq::cat(pACYC177), respectively.

Phenotypic comparisons.

Lipo-oligosaccharides and OMPs were isolated from 35000HP and 35000HPΔhfq and analyzed by polyacrylamide gel electrophoresis as described previously (53–55).

Human inoculation experiments.

Eleven healthy adult volunteers (7 men and 4 women; 3 white and 8 black; mean age ± SD, 48.9 ± 9.5 years) initially enrolled in the study. The volunteers gave informed consent for participation and for HIV serology, in compliance with the guidelines of the U.S. Department of Health and Human Services and the Institutional Review Board of Indiana University. One volunteer was excluded due to an underlying medical condition; 1 withdrew consent prior to inoculation; 9 participated in the study.

Stocks of 35000HP and 35000HPΔhfq were prepared according to FDA guidelines under BB–IND 13064. Human inoculation experiments were performed exactly as described in detail elsewhere (30, 56). Comparisons of papule and pustule formation rates were performed using a logistic regression model with generalized estimating equations (GEE) (25); the GEE sandwich estimate for the standard errors was used to calculate 95% confidence intervals (95% CI) for the rates. For a rate of zero, the GEE estimate does not exist; in this instance, we calculated the exact binomial confidence intervals based on the number of subjects rather than sites, as described previously (25).

To confirm the identity of the bacteria in the inocula, surface cultures, and cultures of biopsies, colony hybridization was performed using probes specific for dnaE and the deleted region of hfq generated by primer pair P17 and P18 and primer pair P13 and P14 (see Table S5 in the supplemental material), respectively, using methods described previously (49).

Three groups of volunteers were inoculated in this study; however, the data from the second group of volunteers were excluded due to an experimental error. We intended to inoculate a group of four participants (424, 425, 426, and 427) in the second iteration with the parent and escalating doses of the mutant. Surprisingly, pustules formed at 4 of 9 parent sites and 12 of 12 mutant sites. However, colony PCR performed on colonies isolated from the inocula used to infect the subjects showed that the participants in the second group were mistakenly infected only with the parent (106 CFU at 3 sites on one arm; 244, 488, and 976 CFU at 3 sites on the other arm). All the participants in the second group achieved clinical endpoint (development of a painful pustule) in 6 or 7 days, which is typical for the model; the participants were not harmed by the error. After reporting a protocol violation, we amended our procedures so that PCR is performed on the broth cultures prior to inoculation of the volunteers to confirm that the parent cultures contain the gene of interest and that the mutant cultures lack the gene of interest.

Western blot analysis.

Whole-cell lysates (5 × 107 CFU/well) from broth-grown H. ducreyi strains were resolved by SDS-PAGE in 4% to 20% polyacrylamide gels and electroblotted onto polyvinylidene difluoride (PVDF) membranes as described previously (51). After transfer, membranes were incubated in StartingBlock (phosphate-buffered saline [PBS]) Blocking Buffer (Thermo Scientific) containing 5% normal goat serum for 1 h at room temperature or overnight at 4° C, followed by incubation in primary antibody (Ab) either for 4 h at room temperature or overnight at 4° C. Membranes were subsequently incubated for 1 h at room temperature in a 1:20,000 dilution of either goat anti-mouse IgG-horseradish peroxidase (IgG-HRP) or goat anti-rabbit IgG-HRP (Bio-Rad). The following primary antibodies used in this study have been previously described: LspA1-specific monoclonal antibody (MAb) 40A4 (26), LspA2-specific MAb 1H9 (26), mouse polyclonal LspB antibody (27), PAL-specific MAb 3B9 (53), mouse polyclonal DsrA antibody (33), and rabbit polyclonal Flp1 antibody (24).

Microcolony formation assay.

Microcolony assays were performed as previously described (24). Briefly, 24-well tissue culture plates (Costar) were seeded with ~105 Hs27 human foreskin fibroblasts (CRL-1634; American Type Culture Collection) per well and incubated until they achieved confluence. H. ducreyi cells grown overnight in 5 ml of Columbia broth were collected by centrifugation and resuspended in tissue culture medium to an OD600 = 0.1. Portions (5 µl) of the bacterial suspension were added in triplicate to individual wells, and the bacterial cells were centrifuged onto the monolayers for 5 min at 1,000 × g at room temperature, after which the plates were incubated at 33° C and 5% CO2. After incubation for 24 h, each well was washed three times with PBS (pH 7.4) and stained with crystal violet. Images were taken using an FSX100 Bio Imaging Navigator microscope system (Olympus) at ×14 and ×40 magnifications.

Serum bactericidal assays.

Serum bactericidal assays were performed exactly as described previously (29).

Microarray data accession number.

The data from these DNA microarray experiments were deposited at the NCBI Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE44535.

Genes downregulated in stationary phase compared to mid-log and transition phases.

Table S1, DOCX file, 0.2 MB.

Genes upregulated in stationary phase compared to mid-log and transition phases.

Table S2, DOCX file, 0.2 MB.

Complete list of genes whose expression was most affected by inactivation of hfq as measured by DNA microarray analysisa

Table S3, DOCX file, 0.2 MB.

Comparison of the genes differentially expressed in 35000HPΔhfq::cat compared to 35000HP and the genes differentially expressed in stationary phase compared to mid-log phase.

Table S4, DOCX file, 0.3 MB.

Primers used in this study.

Table S5, DOCX file, 0.2 MB.

Functional grouping of H. ducreyi genes whose expression is affected by the inactivation of hfq. Black and white bars represent genes whose expression is upregulated and downregulated, respectively. Download

Figure S1, TIF file, 3.1 MB

Growth kinetics of strain 35000HP and 35000HPΔhfq in GC broth. Growth kinetics was determined by measuring the optical density at 660 nm (OD660) at different time points following inoculation from overnight cultures. A, B, and C correspond to the mid-log, transition, and stationary growth phases, respectively. Download

Figure S2, TIF file, 3 MB

Structural model of H. ducreyi Hfq. A ribbon representation of the modeled tertiary structure of this protein is shown. The model is shown in green, except for those parts that likely come into contact with RNA on the proximal side (purple) and on the distal side (blue). Parts of the model that are outside the known structure of the template (PDB code 3SB2) or that belong to the highly flexible C terminus have been omitted. Download

Figure S3, TIF file, 6.4 MB