Genes Differentially Expressed by Haemophilus ducreyi during Anaerobic Growth Significantly Overlap Those Differentially Expressed during Experimental Infection of Human Volunteers

ABSTRACT

Haemophilus ducreyi causes cutaneous ulcers in children and the genital ulcer disease chancroid in adults. In humans, H. ducreyi is found in the anaerobic environment of an abscess; previous studies comparing bacterial gene expression levels in pustules with the inocula (∼4-h aerobic mid-log-phase cultures) identified several upregulated differentially expressed genes (DEGs) that are associated with anaerobic metabolism. To determine how H. ducreyi alters its gene expression in response to anaerobiosis, we performed RNA sequencing (RNA-seq) on both aerobic and anaerobic broth cultures harvested after 4, 8, and 18 h of growth. Principal-coordinate analysis (PCoA) plots showed that anaerobic growth resulted in distinct transcriptional profiles compared to aerobic growth. During anaerobic growth, early-time-point comparisons (4 versus 8 h) identified few DEGs at a 2-fold change in expression and a false discovery rate (FDR) of <0.01. By 18 h, we observed 18 upregulated and 16 downregulated DEGs. DEGs involved in purine metabolism, the uptake and use of alternative carbon sources, toxin production, nitrate reduction, glycine metabolism, and tetrahydrofolate synthesis were upregulated; DEGs involved in electron transport, thiamine biosynthesis, DNA recombination, peptidoglycan synthesis, and riboflavin synthesis or modification were downregulated. To examine whether transcriptional changes that occur during anaerobiosis overlap those that occur during infection of human volunteers, we compared the overlap of DEGs obtained from 4 h of aerobic growth to 18 h of anaerobic growth to those found between the inocula and pustules in previous studies; the DEGs significantly overlapped. Thus, a major component of H. ducreyi gene regulation in vivo involves adaptation to anaerobiosis.

IMPORTANCE In humans, H. ducreyi resides in the anaerobic environment of an abscess and appears to upregulate genes involved in anaerobic metabolism. How anaerobiosis alone affects gene transcription in H. ducreyi is unknown. Using RNA-seq, we investigated how anaerobiosis affects gene transcription over time compared to aerobic growth. Our results suggest that a substantial component of H. ducreyi gene regulation in vivo overlaps the organism’s response to anaerobiosis in vitro. Our data identify potential therapeutic targets that could be inhibited during in vivo growth.

INTRODUCTION

Haemophilus ducreyi causes chancroid, a sexually transmitted genital ulcer disease that facilitates HIV transmission (1, 2). H. ducreyi has more recently been found to be the causative agent of cutaneous ulcers that primarily occur on the lower limbs of children in regions of the tropics where yaws is endemic (1). Although syndromic management of genital ulcers has markedly reduced the prevalence of chancroid, mass drug administration of azithromycin has failed to eliminate H. ducreyi-associated cutaneous ulcers in communities of endemicity (3–5). Thus, understanding how H. ducreyi survives in humans may help to inform novel treatment strategies.

Our laboratory developed an experimental challenge model for H. ducreyi in which human volunteers are infected on the upper arm with 1 to 150 CFU of the bacterium via puncture wounds (6). Papules form within 24 h, and over the course of approximately 1 week, the infection either spontaneously resolves or progresses to form a painful pustule (7). Experimental disease progression mirrors the early stages of natural infection; in both experimental and natural infection, H. ducreyi is found in an abscess surrounded by polymorphonuclear leukocytes and macrophages, which fail to ingest the organism (8, 9). Thus, the model offers an opportunity to understand bacterial metabolism and gene expression in an abscess, which is an anaerobic (10), nutrient-limiting (11), and hostile environment produced by the recruitment and activity of innate immune cells.

We recently used the human challenge model to identify H. ducreyi differentially expressed genes (DEGs) during human infection (12, 13). DEGs in both studies indicated that H. ducreyi gene expression differs between the inoculum—an aerobic mid-log-phase culture—and the abscess. A subset of the DEGs identified in both studies identified upregulated pathways that are associated with anaerobic growth in other organisms, such as nitrate reduction and formate utilization (14), likely reflecting the anaerobic environment of the infected sites. However, the transcriptional response of H. ducreyi to anaerobiosis has not yet been defined and may be uniquely regulated given the lack of intact canonical two-component systems in H. ducreyi that usually respond to anaerobiosis (e.g., ArcAB, NarPQ, and NarXL). A gene for the fumarate and nitrate reduction regulatory (FNR)-like protein (HD_1427) is found in the H. ducreyi genome, but it has not been characterized. Nevertheless, binding sites for ArcA, NarP, and FNR have been bioinformatically predicted upstream of canonical anaerobic metabolism genes found in H. ducreyi (15).

Bacterial gene expression and metabolism during infection are poorly understood and often differ from those observed under in vitro growth conditions using enriched media (16). In addition, anaerobiosis can trigger virulence factor expression in addition to altering metabolism in other organisms (17). We were therefore interested in determining how well anaerobic growth in vitro could recapitulate in vivo transcriptional profiles. Here, we determined which in vivo DEGs were likely responding to the anaerobic environment alone by comparing the transcriptomes of H. ducreyi grown under aerobic and anaerobic culture conditions at multiple time points. We also examined whether the DEGs identified during human infection were enriched for those induced by anaerobic growth versus aerobic growth.

RESULTS

Growth of H. ducreyi in aerobic versus anaerobic environments.

H. ducreyi is typically grown aerobically in a complex liquid culture medium containing proteose peptone, hemin, fetal calf serum, and IsoVitaleX. To examine whether we could generate an anaerobic environment in this complex medium, we tested a commercial product called Oxyrase, which consists of overexpressed electron transport chain proteins in purified Escherichia coli membranes and the enzymatic substrates (lactate and succinic acid) required to reduce oxygen in the medium. Compared to complex medium supplemented with the substrates, the addition of Oxyrase and the substrates to the medium immediately lowered the oxygen concentration in the medium to 5% and generated an anaerobic environment in ∼2.5 h that was sustained for 24 h (see Fig. S1 in the supplemental material).

We next assessed the growth of H. ducreyi in anaerobiosis compared to that in aerobiosis by measurement of the optical density and CFU over time (Fig. 1). Bacteria from aerobic broth cultures grown overnight were subcultured into medium containing either the Oxyrase buffer substrates or Oxyrase plus the substrates. By both optical density and CFU, the aerobically grown bacteria grew at a higher rate. By 18 h of growth, the aerobic cultures were well into stationary phase and had decreased viable bacterial counts compared to measurements at 8 h. Although the optical density measurements of the anaerobic cultures were approximately half those of the aerobic cultures throughout the time course, the same numbers of viable bacteria were recovered under both growth conditions at 18 h, suggesting that bacterial growth and death were occurring simultaneously in the aerobic cultures by that time point.

FIG 1.

H. ducreyi growth under aerobic versus anaerobic conditions over an 18-h time course. At the indicated time points, growth was measured by the optical density (A) and CFU (B). Symbols represent the geometric means from four independent experiments; error bars represent the standard deviations. Data were analyzed by 2-way analysis of variance (ANOVA) followed by Sidak’s test to correct for multiple comparisons. *, P < 0.05.

To further compare the kinetics of aerobic and anaerobic growth, we extended the growth experiments to 48 h (see Fig. S2 in the supplemental material). By optical density, the anaerobic cultures were still in logarithmic phase at 18 h and entered stationary phase at 24 h; however, there was little change in CFU between 18 h and 24 h, suggesting that bacterial death was also occurring between these time points (Fig. S2). Although the viability of the bacteria grown under anaerobic conditions was maintained for 48 h, the viability of those grown under aerobic conditions was markedly reduced by 36 h and absent by 48 h. Thus, growth under anaerobic conditions fosters the viability of H. ducreyi over time.

The transcriptomes of anaerobically versus aerobically grown H. ducreyi are distinct.

To understand the differences in gene expression between the aerobically and anaerobically grown bacteria, we performed RNA sequencing (RNA-seq) on the cultures used in the initial growth curve experiments, which yielded similar amounts of viable bacteria at 18 h (Fig. 1). We extracted RNA at 4, 8, and 18 h from both aerobic and anaerobic cultures in four independent experiments for a total of 24 samples. Approximately 27 million to 50 million read pairs were acquired for each of the 24 samples (see Table S1 in the supplemental material).

We analyzed gene expression with the edgeR package (18). A principal-coordinate analysis (PCoA) plot of the data indicates that the replicates of each sample grouped with like samples, but the clusters were all distinct from each other by permutational multivariate analysis of variance (PERMANOVA) (see results in Data Set S1 and Fig. 2A). Using Spearman’s correlation coefficient, we found ≥98% correlation between replicates and ≥88% correlation between all samples, indicating that basal gene expression patterns between all conditions were similar (Fig. 2B). By correlation analysis, most samples were highly related (≥94%), except for comparisons to 18-h aerobic cultures, which were the most distinct of the six conditions tested. The gene expression in the 18-h anaerobic samples was more similar to those for the 4- and 8-h anaerobic time points than any of the aerobic samples. The data suggest that anaerobic growth results in a distinct transcriptome compared to aerobic growth.

FIG 2.

Gene expression at different time points in the presence and absence of oxygen is unique. (A) A PCoA plot was generated to determine whether like samples clustered together. Note that aerobic and anaerobic cultures cluster separately and that the 18-h aerobic cluster is the most unique of the 6 sample types. (B) The correlation of gene expression between samples was measured using Spearman’s correlation analysis. The numerical results of the PERMANOVA comparisons for all 6 groups and Spearman’s coefficients for all samples can be found in Data Set S1 in the supplemental material.

For all comparisons described below, we defined DEGs as those having a 2-fold level of differential expression with a false discovery rate (FDR) cutoff of <0.01. By quantitative real-time PCR (qRT-PCR), we verified the differential expression of seven genes that were significantly differentially expressed in our data set in at least one time point and that were also differentially expressed by H. ducreyi in vivo in our previous studies (12, 13) (Fig. 3). Except for discordant fold changes in the expression of HD_0673 measured at 18 h under both aerobic and anaerobic conditions, the differential expression of the other six genes (citC, glpF, napD, nrfD, ulaG, and yfeA) followed the same general trends and change in direction of expression whether measured by RNA-seq or by qRT-PCR.

FIG 3.

Validation of select DEGs in RNA-seq data sets by qRT-PCR. The transcription of select genes was assayed by using the same RNA samples as those from the RNA-seq experiment using qRT-PCR. Fold changes for the qRT-PCR data were determined using the ΔΔC_T method, and dnaE served as the housekeeping control gene. Three independent experiments on duplicate samples were done to determine the qRT-PCR fold change values; the final fold change for the 4 biological RNA-seq replicates is also displayed on the graphs. Both the RNA-seq and qRT-PCR fold changes are in comparison to expression levels in the 4-h aerobic samples. Error bars indicate the standard deviations. The fold change determined by qRT-PCR was compared to the value determined by RNA-seq using one-sample t tests. Significant differences between RNA-seq and qRT-PCR fold changes for each sample are indicated either above or below the relevant symbols. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Differentially regulated genes in anaerobiosis versus aerobiosis at specific time points.

We first compared the transcriptomes of anaerobic and aerobic samples after 4 h of growth. At this time point, anaerobiosis should have been achieved in the medium treated with Oxyrase. Thirty-seven genes were differentially expressed under anaerobic versus aerobic conditions at 4 h (Data Set S2). There were 11 upregulated genes, which encompassed all subunits of the nitrate reductase (nrfABCD), a putative lactate dehydrogenase complex (ykgEF), and part of the cytochrome c complex (napBC). The 26 downregulated genes included decreased expression of several protein chaperones (cspD, hscB, and hscA); proteins involved in iron and/or sulfur metabolism (fdx2, iscU, and iscS), nucleotide metabolism (guaAB), and fructose metabolism (fbp); and a virulence factor (hgbA), a hemoglobin binding protein required for human infection (19). The remaining genes all encoded hypothetical proteins.

At the 8-h time point, there were 7 upregulated and 12 downregulated DEGs under anaerobic versus aerobic conditions. The upregulated genes included those encoding ribosomal proteins (rpmE2 and rpmJ1), a tRNA-modifying enzyme (trmD), a cold shock protein (cspC), a conserved protein of unknown function (slyX), and two hypothetical genes. Interestingly, the 12 downregulated DEGs encompassed alternative carbon source utilization (ulaE, ulaR, purT, fdhE, and glpF), cytochrome c proteins (napB, napC, and napG), and the pentose phosphate pathway (tktA). The downregulation of the ula, fdhE, glpF, and nap genes at this time point was somewhat surprising in light of their upregulation in the 4-h anaerobic sample and previous studies showing their upregulation in vivo (12, 13).

At the 18-h time point, there were 228 upregulated and 143 downregulated DEGs. Kyoto Encyclopedia of Genes and Genomes (KEGG) (20) identifiers were used to sort DEGs into categories to make general conclusions about changes in growth during anaerobiosis compared to aerobiosis (Table 1). There was a dramatic shift in gene expression for carbohydrate and energy metabolism. Genes involved in nucleotide metabolism and genetic information processing were also increased in the 18-h anaerobic samples compared to the 18-h aerobic samples. A large driver of the differences for the latter category is that nearly all subunits of the ribosome were upregulated under anaerobic conditions. Transcripts corresponding to membrane transporters were also upregulated in anaerobiosis. These findings may reflect the fact that the bacteria were still replicating under anaerobic conditions but were not replicating under aerobic conditions. Finally, a large proportion of DEGs either had only a general-function prediction or were hypothetical proteins; 22.3% (51/228) of upregulated and 67.8% (97/143) of downregulated DEGs at 18 h have not been characterized.

TABLE 1.

KEGG pathway analysis of anaerobic versus aerobic DEGs over time

Pathway	No. of DEGs
	4 h		8 h		18 h
	Up	Down	Up	Down	Up	Down
Metabolism
Carbohydrate	3	2	0	3	19	3
Energy	6	3	0	4	15	0
Lipid	0	0	0	0	7	2
Nucleotide	0	3	0	1	10	2
Amino acid	1	0	0	0	6	4
Other amino acids	0	1	0	0	4	0
Glycan biosynthesis and metabolism	0	0	0	0	7	6
Metabolism of cofactors and vitamins	0	1	0	0	11	1
Metabolism of terpenoids and polyketides	0	0	0	0	2	1
Biosynthesis of other secondary metabolites	0	0	0	0	1	0
Other	0	0	0	0	7	1
Total	10	10	0	8	89	20
Genetic information processing
Transcription	0	2	1	1	9	4
Translation	0	1	3	0	68	5
Folding, sorting, and degradation	0	4	0	1	10	4
Replication and repair	0	0	0	0	9	6
Total	0	7	4	2	96	19
Environmental information processing
Membrane transport	1	3	1	1	17	6
Signal transduction	0	0	0	0	8	1
Total	1	3	1	1	25	7
Signaling and cellular processes
Cellular community	0	0	0	0	3	1
Cell motility	0	1	1	0	0	0
Others	0	0	0	0	1	1
Total	0	1	1	0	6	3
Human diseases
Bacterial toxins	0	0	0	0	0	3
Total	0	0	0	0	0	3
Poorly characterized
General function prediction only	0	1	1	1	9	13
Function unknown	0	7	1	0	42	84
Total	0	8	2	1	51	97

Differential gene expression in the presence and absence of oxygen during exponential growth.

The growth rate of H. ducreyi under anaerobic conditions was lower than that under aerobic conditions and remained exponential throughout the 18-h time course, at which time the aerobically grown bacteria had entered into stationary phase (Fig. 1). To control for the confounding variable of growth phase on gene expression, we compared the change in gene expression over time from 8 h to 18 h in anaerobiosis to the change in gene expression from 4 h to 8 h in aerobiosis, as the bacteria were in exponential phase under both growth conditions (Fig. 1). Using a linear regression comparison strategy, there were 63 upregulated and 36 downregulated genes in anaerobiosis versus aerobiosis over time (Data Set S2). Since linear regression models compare the significance of the rate of change between two time points, the additional DEGs identified in this comparison likely represent either (i) transcripts that do not change from one time point to the next under one condition but do change under the other condition or (ii) transcripts under each condition that change in opposite directions over time.

For this comparison, a large portion of upregulated genes in anaerobiosis was still related to growth. For example, 18 genes encoding ribosomal proteins were upregulated, as were genes involved in purine and pyrimidine metabolism (purT, nrdA, and HD_1029) and transcription (rpoA, rho, deaD, and nusA). Additional upregulated genes in anaerobiosis included sulfur metabolism genes (iscS, cysK, and metK), cofactor biosynthesis genes (moaA, moaC, and HD_0042), and lactate dehydrogenase (lldD), suggesting that the bacteria were expressing proteins requiring anaerobic cofactors, such as molybdopterin (21, 22), and using alternative carbon sources. We also observed the upregulation of dsrA, which encodes an outer membrane protein involved in serum resistance that is required for human infection (23). In this comparison, napC, napG, and fdhE were downregulated, suggesting that nitrate and/or nitrite was limiting. Supporting this, we also observed the downregulation of the anaerobic glycerol-3-phosphate dehydrogenase (glpABC), which uses fumarate or nitrate to oxidize NADH back to NAD⁺ during anaerobiosis (24–26). Genes involved in riboflavin biosynthesis (ribH) and mannose metabolism were also downregulated.

Differentially regulated genes during anaerobic growth.

We next compared gene expression levels after 4, 8, and 18 h of growth under anaerobic conditions (Data Set S2). There was only one gene that was differentially expressed (downregulated) at 8 h compared to 4 h of anaerobic growth, HD_1166, which encodes a 30-residue hypothetical protein whose translation start site overlaps the preceding gene, greB, which encodes a transcription elongation factor.

In comparing gene expression levels between 18 h and 4 h of anaerobiosis, there were 18 upregulated and 16 downregulated genes. The upregulated transcripts included several purine metabolism genes (purM, purT, and purH), the uptake of alternative carbon sources (glpF and dcuB1), toxin production (cvpA), the outer membrane protein encoded by ompP1, members of the nitrate reductase (nrfA and nrfB), glycine metabolism (gcvA), tetrahydrofolate synthesis (fhs), and the putative lactate dehydrogenase subunit (ykgF). The downregulated genes included alterations in the electron transport chain (sucB and tonB), thiamine biosynthesis (thiL), DNA recombination (tnpR), peptidoglycan synthesis (glmU), and riboflavin synthesis or modification (apbE, ribD, ribE, ribAB, and ribH). Both greB and its downstream gene HD_1166 were downregulated in the 18-h to 4-h comparison. Overall, these data support the idea that H. ducreyi shifts to using nitrate as an electron acceptor, imports alternative carbon sources, and downregulates genes involved in replication under anaerobic conditions.

Comparison of DEGs identified during human infection to DEGs induced during anaerobic versus aerobic growth.

In previous studies, we examined differential bacterial gene expression between the inoculum used in the human challenge model (mid-log-phase aerobic cultures) and the endpoint pustules biopsied 6 to 8 days after infection by RNA-seq (12, 13). Some DEGs, such as focA and dcuB2, which were upregulated in lesions compared to the inocula, are known to be upregulated during anaerobiosis in other bacteria (27, 28).

As surrogates for the abscess environment and the inocula used to infect human volunteers, we examined genes that were differentially expressed at 18 h of anaerobic growth compared to 4 h of aerobic growth. We compared these DEGs to those identified in two previous studies. In the first study, H. ducreyi gene expression in pustules was compared to historical in vitro growth data sets obtained from mid-log-phase, transitional, and stationary-phase cultures; the study identified 62 upregulated and 31 downregulated H. ducreyi DEGs in pustules compared to mid-log-phase cultures (12). In the second study, bacterial gene expression in the pustule was compared to that in the mid-log-phase cultures used to infect the volunteers; that study identified 113 upregulated and 105 downregulated H. ducreyi DEGs (13). These gene lists were then compared to the 18-h anaerobic versus 4-h aerobic DEGs (called the surrogate comparison here) to identify transcripts that were shared among the three studies (Fig. 4). There were 17 upregulated and 13 downregulated DEGs that overlapped between the DEGs in the surrogate comparison and the DEGs identified in the study by Griesenauer et al. (13), and this was significant (P = 6.4 × 10⁻⁹ and P = 3.9 × 10⁻⁵, respectively). Similarly, there were 11 upregulated and 2 downregulated DEGs that overlapped between the DEGs in the surrogate comparison and those identified in the study by Gangaiah et al. (12); the overlap was significant for the upregulated but not the downregulated DEGs (P = 9.0 × 10⁻⁷ and P = 0.24, respectively).

FIG 4.

Overlap of anaerobic (18 h) versus aerobic (4 h) DEGs with DEGs in pustules versus the inocula (4-h aerobic cultures). (A) Venn diagram showing the number of overlapping upregulated and downregulated genes in common in the three studies. (B) Gene lists of DEGs shared between this and at least one of the other studies (12, 13). PTS, phosphotransferase system.

Ten upregulated genes in the surrogate comparison were upregulated in both of the in vivo studies, including genes involved in ascorbic acid metabolism (ulaBCDEGR), citrate metabolism (citCD), and formate transport (focA). An additional eight upregulated genes were in common with at least one of the in vivo studies, including a member of the nitrate reductase (nrfD), glycerol transport (glpF), a cytochrome (ccmB), and methionine transport (metN). Four downregulated genes in the surrogate comparison were also downregulated in both of the other in vivo studies, including the guaAB operon. An additional 10 downregulated genes were shared with at least one of the in vivo studies, including iron-sulfur metabolism (fdx2 and iscU), protein chaperones (groEL and groES), nucleotide metabolism (thiL), and an outer membrane protein (ompP2B). This suggests that aside from genes that are directly associated with anaerobic metabolism (e.g., focA, glpF, ccmB, and nrfD), the utilization of ascorbic acid and citrate may occur in an anaerobic environment. The downregulated genes are suggestive of slower cellular metabolism overall during anaerobiosis, specifically with regard to DNA replication and protein and cofactor synthesis.

Transcriptional regulation of anaerobic genes.

Transcriptional responses due to a shift from aerobic to anaerobic environments are classically regulated by two different transcription factors, ArcA and FNR, in other gammaproteobacteria. ArcA is the response regulator of the ArcAB two-component system and is phosphorylated by the sensor kinase ArcB, which senses the increased abundance of menaquinone relative to ubiquinone in the quinone pool during anaerobiosis at the cytoplasmic membrane (29). Phospho-ArcA primarily represses aerobic genes, such as those that encode components of the tricarboxylic acid (TCA) cycle and oxidative phosphorylation, but can activate others, presumably by occlusion of repressor binding (30). In H. ducreyi, ArcA is an orphan response regulator in that H. ducreyi lacks an annotated ArcB homolog. FNR is typically regarded as a transcriptional activator, although it can also repress genes (31). FNR senses oxygen availability via the Fe-S cluster to which it binds; a reduced Fe-S cluster is capable of binding to FNR, resulting in FNR dimerization and subsequent DNA binding (32). The only potential E. coli FNR homolog in H. ducreyi is HD_1427. Although arcA expression trended toward significance at 18 h of anaerobic growth compared to 4 h of aerobic growth, neither arcA nor HD_1427 was significantly differentially expressed at any time point under anaerobic versus aerobic conditions (Fig. S3).

Using the MEME algorithm (33), we examined whether there were any common sequence motifs in the 450 bp upstream of the predicted operons of DEGs identified in the 18-h anaerobiosis versus 18-h aerobiosis comparison. In E. coli, both FNR and ArcA are known to bind ∼18-bp-long palindromic sequences (30, 31). We did not identify any palindromic sequence motifs in the upstream sequences. The only short motif identified (see Fig. S4) did not return any likely transcription factor matches when searched in the TomTom database (33). Thus, the transcription factors regulating the response of H. ducreyi to anaerobiosis remain to be identified.

DISCUSSION

H. ducreyi is a facultative anaerobe and grew under anaerobic conditions at a reduced growth rate compared to aerobic conditions, which is typical of other facultative anaerobes (34). Interestingly, the organism remained viable for a longer time under anaerobic conditions; this may contribute to the ability of the organism to survive in the anaerobic environment of an abscess in vivo.

RNA-seq was used to profile the transcriptomes of bacteria that were grown under both aerobic and anaerobic conditions at three different time points (4, 8, and 18 h). The most consistently induced genes included the nitrate reductase operon (nrfABCD), which indicated that the bacteria were transcriptionally responsive to the anaerobic environment and that they experienced a shift in their metabolic state.

One limitation of our study is that we only investigated the transcriptional response of H. ducreyi to anaerobiosis. Posttranscriptional regulatory mechanisms, including small RNAs (sRNAs), are another method used by bacteria to modulate the translation of key carbon metabolism enzymes, such as those in the TCA cycle (35). The posttranscriptional regulation of these genes decreases the response time needed for metabolism to adapt to changes in the environment. We saw an increase in the expression of only two predicted noncoding RNAs (EBG00001101828 and EBG00001101812) when comparing 18 h of anaerobic to 18 h of aerobic growth. The targets and/or functions of these RNAs are unknown. We also analyzed our data using RockHopper (36) to detect any novel differentially expressed sRNAs; however, we did not detect any significantly differentially expressed sRNAs. We did, however, see an upregulation of rnc, which encodes RNase III and is involved in rRNA processing, mRNA degradation, and cleavage of sRNA-mRNA duplexes (37), in anaerobiosis versus aerobiosis at 18 h. In addition, transcripts of rnc were downregulated in late stationary phase (18 h) compared to those at 8 h in aerobiosis. Because stationary phase is characterized by slow growth, we hypothesize that the reduced need for ribosomes would explain the reduction in rnc transcripts later in aerobiosis. However, although anaerobic growth is generally less robust, the increase in rnc expression may reflect an increased role in mRNA and sRNA-mRNA duplex degradation.

We also noted that there was a high proportion of DEGs encoding hypothetical proteins in many of our comparisons across time and between anaerobic and aerobic conditions. Due to their small size (<50 residues), many of these proteins qualify as small open reading frames (sORFs) (38). Investigation of hypothetical small proteins to date has revealed that some of these proteins are able to modulate carbohydrate transporters in the membrane (39). In general, sORF genes were more lowly expressed in aerobic growth and either unchanged or increased in anaerobic growth. We used BLAST to identify potential homologs in other bacteria. Only HD_0157 was highly conserved in other species (Aggregatibacter spp.), but its function is unknown. Under aerobic conditions, sORFs tended to be in or near known integrated phages in the genome, which are known to be composed of multiple sORFs. However, this was not the case for the differentially expressed anaerobic hypothetical sORFs.

Anaerobiosis induced some genes that could alter protein translation. We noted that the Zn²⁺-independent paralog of ribosomal protein L31, rpmE2, was upregulated in both the 8-h and 18-h anaerobic versus aerobic comparisons. This protein has a higher affinity for the ribosome than the Zn²⁺-dependent paralog, rpmE1, and is directly induced by either zinc starvation or anaerobiosis in other organisms (40). How this impacts translation is unknown. The expression of cspC was also upregulated at 8 h and 18 h in anaerobic versus aerobic growth. CspC is a chaperone that binds both single-stranded DNA and RNA to preferentially preserve the transcription and/or translation of certain genes during times of stress as well as anaerobiosis (41).

Gene expression during anaerobiosis has not been previously studied in H. ducreyi, but during human infection, H. ducreyi is found in an abscess, which is an anaerobic environment (10). We therefore examined genes that were differentially expressed at 18 h of anaerobic growth compared to 4 h of aerobic growth as surrogates for the lesional environment and the inocula used to infect human volunteers. In the 18-h anaerobic versus 4-h aerobic comparison, ∼34% (18/53) of the upregulated DEGs and ∼22% (14/63) of the downregulated DEGs overlapped the abscess versus inoculum comparisons in at least one of our previous studies (12, 13). An attempt was made to determine whether the eight hypothetical genes that appear both here and in in vivo studies had any conserved domains that might point toward their function in anaerobiosis. Most of these genes encoded putative small proteins. Small proteins are emerging as a novel regulatory mechanism of both cytosolic and transmembrane proteins, including transporters. Although further characterization of these genes is required, they may represent another possible mechanism for H. ducreyi to regulate metabolism outside sRNAs.

We had previously shown that of 18 genes or operons encoding virulence determinants known to be partially or fully required for pustule formation in humans, only 5 (dsrA, flp-tad, hgbA, lspB-lspA2, and sapA) were upregulated in human lesions (12, 13). Using a linear regression comparison to identify DEGs in anaerobiosis versus aerobiosis during exponential growth, we found that dsrA was upregulated. The expression of hgbA was increased at 4 h of anaerobic versus aerobic growth. None of the five known virulence factors were differentially regulated in our 18-h anaerobic versus 4-h aerobic comparison. Thus, anaerobiosis does not appear to play a major role in the expression of virulence determinants in H. ducreyi.

We hypothesize that anaerobiosis may help to prolong the logarithmic phase of growth and foster bacterial viability in human lesions. The transcriptional regulators for the response to anaerobiosis in H. ducreyi are not currently known. Several key homologs related to the detection of anaerobic environments via the lack of oxygen (e.g., the sensor kinase ArcB) and/or the presence of nitrate or nitrite (e.g., the sensor kinases NarQ and NarX) are missing from the genome based on current annotations. Although the H. ducreyi genome contains homologs of the response regulators ArcA and NarP or NarL and the transcription factor FNR, we did not identify any consensus binding sequences for these regulators upstream of DEGs induced by anaerobiosis.

This study has also helped to further our understanding of which metabolic pathways are likely to be functional in anaerobiosis. For example, the export of formate for use as an electron donor was suggested by the upregulation of focA and pflB, which produce and export formate, respectively, and fdhE, which encodes a component of the formate dehydrogenase complex. However, the H. ducreyi formate dehydrogenase complex is incomplete compared to complexes in other bacteria. Thus, although the role of formate in the response to anaerobiosis is unclear, our data suggest that these partial complexes may still be functional.

A limitation of this study was that a rich, complex medium was used to grow H. ducreyi. Since genes induced by anaerobiosis can be regulated by catabolite repression, minimal, defined media are typically used to elucidate the anaerobic transcriptome. Unfortunately, there is no known minimal medium that supports the growth of H. ducreyi. However, we were still able to identify DEGs induced by anaerobiosis under glucose-replete conditions that were also shown to be similarly differentially expressed in pustules. A better understanding of the H. ducreyi response to anaerobiosis may be assisted by better characterization of the many hypothetical genes identified in our study. Definition of regulatory mechanisms involved in the response to anaerobiosis will be required to fully understand how H. ducreyi adapts to the anaerobic environment.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

Haemophilus ducreyi 35000HP was grown overnight on chocolate agar plates at 33°C in 5% CO₂. Plate-grown bacteria were used to inoculate a liquid culture of complete gonococcal (GC) broth (50 μg/mL hemin, 5% heat-inactivated fetal bovine serum [FBS], and 1% IsoVitaleX) and grown aerobically overnight. The broth culture grown overnight was subcultured into complete GC broth containing either 10% Oxyrase (Oxyrase, Inc.) in substrate buffer or 10% Oxyrase buffer alone (15 mM potassium phosphate [pH 7.0], 650 mM dl-lactate, and 650 mM succinic acid) at an optical density at 660 nm (OD₆₆₀) of approximately 0.06. A total of 7.5 mL of the cultures was aliquoted into 15-mL glass tubes with rubber-lined screw caps. For anaerobic (with Oxyrase) cultures, the caps were tightly screwed shut; for aerobic cultures (buffer substrates), the caps were put on loosely. The cultures were then shaken at 33°C.

Preliminary experiments using complete GC medium containing the buffer substrates with or without Oxyrase were used to determine when anaerobiosis was reached using a Microx 4 with O₂ probe (PreSens) under the same conditions in Chien-Chi Lin’s laboratory at Indiana University-Purdue University Indianapolis. Unfortunately, biosafety regulations in this laboratory precluded us from measuring O₂ concentrations in the bacterial cultures.

Growth curves.

Cultures were grown as described above in duplicate. At the indicated times, the OD₆₆₀ was measured on a spectrophotometer. For measurement of CFU, a portion of one culture was removed and serially diluted at the indicated times. Serial dilutions were plated in triplicate onto chocolate agar plates, and plates were incubated at 33°C in 5% CO₂. Colonies were counted and averaged using geometric means. The growth curves were determined in 4 independent experiments.

RNA extraction.

Five milliliters each of the aerobic or anaerobic culture used to determine the CFU at a given time point was removed from the culture, placed into a 15-mL conical vial while trying to avoid aeration, and pelleted at 3,000 rpm for 10 min in an HN-SII centrifuge (International Equipment Company). Medium was removed, and the pellet was suspended in 1 mL of TRIzol and frozen at −80°C until RNA purification.

RNA was purified using a Zymo Directzol RNA purification kit according to the manufacturer’s instructions. Contaminating DNA was removed both by on-column DNase digestion supplied with the kit and after elution by Turbo DNase (Invitrogen) according to the manufacturer’s instructions. qRT-PCR of the dnaE gene confirmed that contaminating DNA had been removed prior to the construction of RNA-seq libraries.

RNA-seq library construction and sequencing.

Library preparation and sequencing were performed by the Indiana University School of Medicine Genomics Core. RNA quality was measured on an Agilent Bioanalyzer. Prior to library construction, bacterial rRNA was depleted using a QIAseq FastSelect-5S/16S/23S kit (Qiagen). RNA-seq libraries were made using the Kapa RNA kit (Roche) and run on one lane of a NovaSeq S2 flow cell (Illumina). One-hundred-base-pair paired-end reads were obtained.

Sequence alignment and differential gene expression analysis.

Raw reads were mapped to the H. ducreyi 35000HP genome (GenBank accession number AE017143.1) using Bowtie2. The featureCounts package in Rsubread was used to determine the number of high-quality transcripts that mapped to each gene and generate a gene count table. edgeR was used to determine the differentially expressed genes. Lowly expressed genes in all samples were filtered out prior to model building and statistical analysis. A multidimensional linear regression model was built to compare DEGs across both conditions and time. Differentially expressed genes were annotated using the Seq.IO package in Python 3.7. Grouping by gene function(s) was performed using the online KEGG database.

Verification of DEGs by qRT-PCR.

qRT-PCR using primers corresponding to a set of DEGs that were analyzed in previous studies (12, 13) were used to verify the RNA-seq data (Table 2). Additional primers were synthesized to examine arcA and HD_1427 (FNR) expression (Table 2). One nanogram of the RNA used in the RNA-seq experiments was used in each reaction mixture containing one-step QuantiTect Sybr green master mix (Qiagen) with the appropriate primer sets. Reactions were run in a Thermo Fisher QuantStudio3 thermocycler for 30 min at 50°C, 15 min at 95°C, and 40 cycles of 95°C for 15 s, 59°C for 30 s, and 72°C for 30 s. Threshold cycle (C_T) numbers were determined in QuantStudio. ΔΔC_T values were calculated in Microsoft Excel based on primer efficiencies.

TABLE 2.

qRT-PCR primers used in this study

Gene	Forward primer	Reverse primer
arcA	CTG GCG CAG CAA ACA ATA AA	TCA TTG TCA CGT CCC GTT AAA
citC	TGG CAT TGC TGA TTT ATC CA	CTT CCG TTC CGA TAA AAC GA
dnaE	AAC GTT ACC TTC AGC AAG GGG TTC	GGC GTT TGG GAT CGT CGA GTG TAT
glpF	CAA TTA GCC GGT GCA ACA TTA G	CGA TCG TTA CTG CGG GAT TTA
napD	GCC AAA TTA CCC CAA GTG AA	CGC CGT TGA TCT CTT TTA GC
nrfD	CGC CAA ATC TTG TCT GGG ATA G	GCA CTA CGA ATC ACC CAG TTT
ulaG	AGA ACA AGT ACC GGC CAA TAA	GCC TCG ACC ACA CCA TAA A
yfeA	AGG TCG TCA CCA CAT TTA CAG	ATC TTT GGG CGT AGG TTG ATA G
HD_0673	CTA AAT GGG CTT TCT GGC TAT CT	GCA AGA GAG TTC AAA CCA AAG TG
HD_1427	TCA CAA GCG TAT GCT GAG AG	GAC CAA GCA AAC GGC TAA TG

Statistical analysis.

Statistics for all RNA-seq data were performed in R. For all other data, statistical analysis was performed in GraphPad Prism version 9. For statistical analysis of CFU assays, raw CFU measurements were first normalized using a log transformation before testing for significance. Tests performed are indicated in the figure legends, and the reported P values have been corrected for multiple comparisons.

Data availability.

The data from these RNA-seq experiments were deposited at the NCBI Gene Expression Omnibus (GEO) database under accession number GSE193250.