Haemophilus ducreyi Seeks Alternative Carbon Sources and Adapts to Nutrient Stress and Anaerobiosis during Experimental Infection of Human Volunteers

Abstract

Haemophilus ducreyi causes the sexually transmitted disease chancroid in adults and cutaneous ulcers in children. In humans, H. ducreyi resides in an abscess and must adapt to a variety of stresses. Previous studies (D. Gangaiah, M. Labandeira-Rey, X. Zhang, K. R. Fortney, S. Ellinger, B. Zwickl, B. Baker, Y. Liu, D. M. Janowicz, B. P. Katz, C. A. Brautigam, R. S. Munson, Jr., E. J. Hansen, and S. M. Spinola, mBio 5:e01081-13, 2014, http://dx.doi.org/10.1128/mBio.01081-13) suggested that H. ducreyi encounters growth conditions in human lesions resembling those found in stationary phase. However, how H. ducreyi transcriptionally responds to stress during human infection is unknown. Here, we determined the H. ducreyi transcriptome in biopsy specimens of human lesions and compared it to the transcriptomes of bacteria grown to mid-log, transition, and stationary phases. Multidimensional scaling showed that the in vivo transcriptome is distinct from those of in vitro growth. Compared to the inoculum (mid-log-phase bacteria), H. ducreyi harvested from pustules differentially expressed ∼93 genes, of which 62 were upregulated. The upregulated genes encode homologs of proteins involved in nutrient transport, alternative carbon pathways (l-ascorbate utilization and metabolism), growth arrest response, heat shock response, DNA recombination, and anaerobiosis. H. ducreyi upregulated few genes (hgbA, flp-tad, and lspB-lspA2) encoding virulence determinants required for human infection. Most genes regulated by CpxRA, RpoE, Hfq, (p)ppGpp, and DksA, which control the expression of virulence determinants and adaptation to a variety of stresses, were not differentially expressed in vivo, suggesting that these systems are cycling on and off during infection. Taken together, these data suggest that the in vivo transcriptome is distinct from those of in vitro growth and that adaptation to nutrient stress and anaerobiosis is crucial for H. ducreyi survival in humans.

INTRODUCTION

The Gram-negative bacterium Haemophilus ducreyi is the causative agent of the sexually transmitted disease chancroid. Chancroid facilitates the acquisition and transmission of human immunodeficiency virus type 1 (1). Although the global prevalence of chancroid has declined due to syndromic management of genital ulcer disease, the disease is still prevalent in several regions of Africa, Latin America, and Asia (2). In addition to causing chancroid, H. ducreyi now is recognized as a leading cause of nonsexually transmitted cutaneous ulcers in children in regions of the South Pacific islands and equatorial Africa where yaws is endemic (3–5). Strains that cause cutaneous ulcers are almost genetically identical to the genital ulcer strain 35000HP and likely evolved from the 35000HP lineage ∼180,000 years ago (6, 7).

The molecular pathogenesis of H. ducreyi 35000HP has been extensively characterized in a human challenge model (8). In this model, H. ducreyi is inoculated into the skin of healthy adults via puncture wounds made by an allergy-testing device. Within 24 h of inoculation, collagen and fibrin are deposited in the wounds; polymorphonuclear leukocytes (PMNs) and macrophages traffic onto collagen and fibrin, forming micropustules (9, 10). By 48 h, the micropustules coalesce to form an abscess, which eventually ulcerates through the epidermis (9, 10). In the abscess, H. ducreyi is found in aggregates and colocalizes with PMNs and macrophages, which fail to ingest the organism. Thus, H. ducreyi must overcome a variety of stresses, including toxic products released by phagocytes, the bactericidal activity of serum that transudates into the wound, the hypoxic and nutrient-limited environment of an abscess, and the presence of host signaling molecules, such as cytokines and chemokines, to establish infection.

Multiple stress response systems usually are employed by Gram-negative pathogens to adapt to environmental changes. Of those systems, the H. ducreyi genome contains homologs of the envelope stress response regulators CpxRA and RpoE, the stringent-response regulators (p)ppGpp and DksA, and the general posttranscriptional regulators Hfq and CsrA (11–16). Although CpxR is dispensable for human infection, the activation of CpxR by the deletion of cpxA renders H. ducreyi avirulent in humans (17, 18). The activation of CpxR primarily downregulates envelope-localized targets, including 7 virulence determinants, each of which is required for human infection, while the activation of RpoE primarily upregulates factors associated with membrane repair and maintenance (11, 12). Thus, CpxRA and RpoE play complementary roles in combating membrane stress in H. ducreyi.

In vivo, pathogenic bacteria sometimes encounter growth conditions similar to those found in stationary phase, which is characterized by growth arrest owing to nutrient limitation and exposure to other stresses (19, 20). Consistent with this idea, the doubling time of H. ducreyi during exponential growth in broth is approximately 2 h, but the estimated minimal doubling time of H. ducreyi in human lesions is 16.5 h (21). Although the rates at which individual H. ducreyi organisms grow and are cleared in vivo are unknown, these data suggest that H. ducreyi spends some portion of its lifetime in vivo in a state similar to that of stationary-phase cells. Several lines of data support this hypothesis. Genes in the flp-tad and lspB-lspA2 operons, which are absolutely required for pustule formation in humans, are upregulated in stationary phase compared to their levels in mid-log phase (13). The RNA-binding protein Hfq is absolutely required for virulence in humans and is a major regulator of stationary-phase gene expression and virulence determinants in H. ducreyi (13). The stringent-response mediators (p)ppGpp and DksA are partially required for pustule formation in humans and have more profound effects on gene expression in stationary-phase cells than in mid-log-phase cells (14, 15). Similarly, the RNA-binding protein CsrA is partially required for virulence in humans and has a more profound effect on resistance to oxidative stress and macrophage killing in stationary-phase cells than in mid-log cells (16).

Using selective capture of transcribed sequences (SCOTS), our laboratory previously identified H. ducreyi genes preferentially expressed during human infection relative to mid-log-phase cells, which are used to infect human volunteers (22). This study identified several genes involved in heat shock response, DNA damage repair, and nutrient transport as being enriched in vivo (22). However, SCOTS is not quantitative and only identifies enriched transcripts. Thus, the molecular mechanisms that H. ducreyi utilizes to adapt to the in vivo environment are not fully understood.

In the present study, we profiled the genome-wide transcriptome of H. ducreyi during experimental human infection using RNA sequencing (RNA-Seq) and compared it to the transcriptomes of mid-log-, transition-, and stationary-phase cells. The study was undertaken to address three important but unanswered questions. Does the in vivo transcriptome resemble that of any phase of in vitro growth, especially stationary phase? How does H. ducreyi change its transcriptional profile from mid-log-phase growth to adapt to the environment of an abscess? Are the transcriptomes controlled by CpxRA, RpoE, Hfq, (p)ppGpp, and DksA, which regulate the expression of virulence determinants and adaptation to membrane and nutrient stress, differentially expressed in vivo?

MATERIALS AND METHODS

Bacterial strains and culture conditions.

H. ducreyi strain 35000HP was used in the present study. 35000HP was isolated from a volunteer who was experimentally infected on the arm with strain 35000 (23). 35000HP was grown on chocolate agar plates supplemented with 1% IsoVitaleX at 33°C with 5% CO₂ or in gonococcal (GC) broth supplemented with 5% fetal bovine serum (HyClone), 1% IsoVitaleX, and 50 μg/ml of hemin (Aldrich Chemical Co.) at 33°C. For both the human challenge and in vitro transcriptome experiments, the bacteria were grown under good laboratory practice conditions using identical lots of medium components, as required by the FDA (BB-IND 13064).

Biopsy specimen collection.

Biopsy specimens of endpoint pustules infected with 35000HP were obtained from four healthy adult volunteers who participated in csrA and relA spoT mutant-parent trials (15, 16). Demographics of the volunteers, inoculation doses, and duration of infection are described in Table 1. The harvested biopsy specimens were submerged in 2 ml of RNAlater and incubated at room temperature for 30 min. Following incubation, the specimens were either processed for RNA isolation immediately or stored at −80°C for 2 to 3 days before processing.

TABLE 1.

Human sample and RNA-Seq read statistics

Volunteer no.	Gender	Inoculation dosea	No. of days of infection	Total no. of reads	No. of reads mapped to H. ducreyi genome	% reads mapped to H. ducreyi genome	Fold coverage (H. ducreyi genome)
412	Male	146	8	396,873,361	249,924	0.063	4.6
413	Female	146	8	394,572,137	174,172	0.044	3.8
437	Female	74	7	372,203,054	390,494	0.105	12.5
439	Male	49	7	390,661,444	265,759	0.068	8.8

^aEstimated delivered dose in CFU.

RNA isolation and quality assessment.

The biopsy specimens were homogenized by using a mini-beadbeater (Biospec Products). Total RNA was extracted from the homogenized tissue using an RNeasy fibrous tissue minikit according to the manufacturer's instructions. RNA was treated twice with Turbo DNA-free DNase (Ambion). The integrity and the concentration of RNA were determined using an Agilent 2100 Bioanalyzer (Agilent Technologies) and a NanoDrop ND-1000 spectrophotometer (Thermo Scientific), respectively. The efficacy of DNase treatment and the functionality of RNA were confirmed by PCR and reverse transcription-PCR (RT-PCR) analyses, respectively, of dnaE using the QuantiTect SYBR green RT-PCR kit (Qiagen).

rRNA depletion.

The removal of human and bacterial rRNA was performed using the Ribo-Zero gold rRNA removal kit (epidemiology) (Epicentre Biotechnologies) by following the manufacturer's instructions. The removal of rRNA was confirmed using an Agilent 2100 Bioanalyzer.

RNA-Seq analysis.

The RNA samples were subjected to RNA-Seq exactly as described previously (14). To obtain relative gene expression levels, the H. ducreyi gene expression in biopsy specimens was compared to that of H. ducreyi grown in vitro to mid-log, transition, and stationary growth phases. RNA-Seq data can be affected by library preparation and sequencing platforms (24); therefore, we used historical data sets of in vitro transcriptomes in which 35000HP was grown in media and under conditions identical to those used in the human challenge experiments and in which RNA-Seq was done by following a methodology identical to that used for the biopsy specimens (14).

Since the numbers of reads obtained from the broth-grown bacteria were far greater than the bacterial reads in the biopsy specimens, the expression level for each gene was normalized to the total number of reads that mapped to the H. ducreyi genome. To assess the global similarities in transcriptional profiles between different in vivo and in vitro conditions, a multidimensional scaling plot was generated using edgeR (25). To determine if the global transcriptional profiles between different in vivo and in vitro conditions were significantly different from one another, a permutational multivariate analysis of variance (PERMANOVA) was performed using PAST 3.10 (26). Genes that were differentially expressed between in vivo and in vitro conditions were identified using a prespecified false discovery rate (FDR) of ≤0.1 and 2-fold change as the criteria (11). The differentially expressed genes were classified using features from the recently reannotated 35000HP genome (NCBI reference sequence accession number NC_002940.2; http://www.ncbi.nlm.nih.gov/nuccore/NC_002940.2) and the COG database (27). Chi-square analysis was performed to assess if the observed overlap between the genes differentially expressed in vivo and those regulated by CpxR, RpoE, Hfq, (p)ppGpp, and DksA were statistically significant (11–14).

qRT-PCR analysis.

Quantitative RT-PCR (qRT-PCR) was performed using the QuantiTect SYBR green RT-PCR kit (Qiagen) in an ABI Prism 7000 sequence detection system (Applied Biosystems) as described previously (11). Representative genes were selected based on their expression level, up- or downregulation, and fold change (11). Primers used for qRT-PCR analysis are listed in Table S1 in the supplemental material. Expression levels of selected genes were normalized to that of dnaE, which was expressed to the same level in the biopsy specimens as in mid-log-phase organisms.

RNA-Seq data accession numbers.

Data from the RNA-Seq experiments were deposited in the NCBI Gene Expression Omnibus (GEO) database under the accession numbers GSM1946558 (raw data, volunteer 412), GSM1946559 (raw data, volunteer 413), GSM1946560 (raw data, volunteer 437), GSM1946561 (raw data, volunteer 439), and GSE75236 (processed data).

RESULTS AND DISCUSSION

RNA-Seq analysis of experimentally infected pustules.

To examine how H. ducreyi adapts to the environment of an abscess during human infection, we biopsied pustules from four volunteers experimentally infected with H. ducreyi 35000HP, isolated RNA, and determined the bacterial transcriptome using RNA-Seq. Each of the samples generated 372 to 397 million reads, of which 0.04 to 0.1% mapped to the H. ducreyi genome (Table 1). The estimated average genome coverage ranged from 3.8- to 12.5-fold (Table 1). The rank ordering of genes based on the raw read counts showed that nearly 80% of the genes had counts between 10 reads and 3,900 reads (data not shown). The coefficients of determination (R²) of H. ducreyi gene expression between different pustule samples ranged from 0.68 to 0.82 (see Fig. S1 in the supplemental material).

Global gene expression analysis of H. ducreyi during in vitro and in vivo conditions.

Multidimensional scaling showed that the transcriptional profile of H. ducreyi in vivo was distinct from the profiles of H. ducreyi grown in vitro to mid-log, transition, and stationary phases (P = 0.007 by PERMANOVA) (Fig. 1). A pairwise PERMANOVA analysis showed that, except for mid-log-phase versus transition-phase comparisons, the in vivo and in vitro transcriptional profiles significantly differed from each other (see Table S2 in the supplemental material). Taken together, these data suggest that the transcriptome of H. ducreyi in vivo is distinct from those of in vitro-grown cells.

FIG 1.

Multidimensional scaling plot of H. ducreyi transcriptional profiles during in vivo and in vitro growth. The plot was generated from pairwise distances using edgeR. A PERMANOVA analysis was performed to determine if the in vivo and in vitro transcriptomes differed significantly from one another. Each symbol represents one independent sample.

Differential gene expression by H. ducreyi between in vitro and in vivo conditions.

To identify genes differentially expressed in vivo, we calculated the fold change in expression of genes in vivo relative to mid-log, transition, and stationary phases. Using an FDR of ≤0.1 and a 2-fold change as criteria for differential expression, 93, 168, and 385 genes were differentially expressed in vivo compared to mid-log, transition, and stationary growth phases, respectively (Fig. 2). COG classification showed that the differentially expressed genes belonged to a variety of functional categories (Tables 2 and 3; also see Table S3 and S4 in the supplemental material).

FIG 2.

Venn diagram showing the number of genes differentially expressed in vivo compared to those in mid-log, transition, and stationary growth phases. The up- and downregulated genes or operons are indicated by up and down arrows, respectively. The total number of genes differentially expressed is indicated in boldface outside the Venn diagram.

TABLE 2.

Summary of COG functional classification of genes differentially expressed in biopsy specimens relative to samples from mid-log, transition, and stationary phases

COG functional class/category	No. of genes differentially expressed
	Biopsy vs. mid-log		Biopsy vs. transition		Biopsy vs. stationary
	Up	Down	Up	Down	Up	Down
Cellular processes and signaling
Cell cycle control, cell division, and chromosome partitioning	1	1	1	1	4	7
Cell wall/membrane/envelope biogenesis	3	4	4	2	17	11
Defense mechanisms	0	0	0	0	3	0
Intracellular trafficking, secretion, and vesicular transport	17	0	12	1	10	8
Posttranslational modification, protein turnover, and chaperones	3	1	6	1	8	4
Signal transduction mechanisms	1	0	4	0	2	2
Information storage and processing
Phage-derived proteins, transposases, and other mobilome components	1	0	2	1	0	3
Replication, recombination, and repair	5	0	4	3	9	12
RNA processing and modification	0	1	0	1	0	0
Transcription	2	2	4	4	12	14
Translation, ribosomal structure, and biogenesis	2	3	10	5	22	21
Metabolism
Amino acid transport and metabolism	2	1	7	5	12	11
Carbohydrate transport and metabolism	8	0	9	1	17	3
Coenzyme transport and metabolism	2	2	12	3	5	15
Energy production and conversion	6	0	15	1	20	5
Inorganic ion transport and metabolism	7	2	12	3	18	3
Lipid transport and metabolism	1	0	1	1	4	5
Nucleotide transport and metabolism	0	3	2	6	5	4
Secondary metabolites biosynthesis, transport, and catabolism	0	0	0	0	1	3
Poorly characterized
Function unknown	0	6	2	5	12	17
General function prediction only	0	0	1	2	4	7
Unclassified	1	5	1	13	4	41
Total no. of differentially expressed genes	62	31	109	59	189	196

TABLE 3.

H. ducreyi genes differentially expressed in the biopsy specimens relative to mid-log-phase samples

Category and operon IDa	Old locus tag	New locus tag	Geneb	Descriptionc	Log counts per million	Fold change	FDR
Upregulated genes
Cellular processes and signaling
Cell cycle control, cell division and chromosome partitioning
Operon_0382	HD0974	HD_RS03990	HD0974	Chromosome partitioning ATPase	5.87	3.3	7.2E−04
Cell wall/membrane/envelope biogenesis
Operon_0515	HD1289	HD_RS05260	lrgB	Deoxyguanosinetriphosphate triphosphohydrolase	7	2.32	3.6E−05
Operon_0711	HD1783	HD_RS07280	HD1783	TolA-TolQ-TolR complex ABC transporter ATPase	8.38	2.53	3.6E−07
Operon_0783	HD1939	HD_RS07950	HD1939	Membrane protein	8.3	2.25	1.4E−04
Intracellular trafficking, secretion and vesicular transport
Operon_0466	HD1155	HD_RS04745	lspB	Large supernatant protein exporter	7.9	6.94	2.3E−13
Operon_0520	HD1298	HD_RS05300	tadG	Tight adherence protein G	10.3	2.12	5.9E−03
Operon_0520	HD1299	HD_RS05305	tadF	Tight adherence protein F	9.06	2.09	5.0E−04
Operon_0520	HD1301	HD_RS05315	tadD	Tight adherence protein D	10.1	2.27	4.6E−03
Operon_0520	HD1302	HD_RS05320	tadC	Tight adherence protein C	9.74	2.15	1.2E−02
Operon_0520	HD1303	HD_RS05325	tadB	Tight adherence protein B	9.62	3.15	1.6E−04
Operon_0520	HD1304	HD_RS05330	tadA	Tight adherence protein A	10.7	3.66	1.8E−05
Operon_0520	HD1305	HD_RS05335	HD1305	Flp operon protein D	10.4	3.46	4.5E−05
Operon_0520	HD1306	HD_RS05340	rcpB	Rough colony protein B	9.16	2.86	2.8E−04
Operon_05207	HD1307	HD_RS05345	rcpA	Rough colony protein A	10.4	2.65	2.5E−03
Operon_0520	HD1308	HD_RS05350	HD1308	Flp operon protein C	9.52	2.35	1.3E−02
Operon_0520	HD1309	HD_RS05355	HD1309	Flp operon protein B	9.18	3.51	6.1E−04
Operon_0520	HD1310	HD_RS05360	flp3	Flp operon protein Flp3	8.9	4.51	7.2E−07
Operon_0520	HD1311	HD_RS05365	flp2	Flp operon protein Flp2	7.9	5	4.2E−08
Operon_0520	HD1312	HD_RS05370	flp1	Flp operon protein Flp1	7.75	4.51	1.0E−05
Operon_0691	HD1736	HD_RS07080	HD1736	Membrane protein	6.99	2.06	2.0E−04
Operon_0714	HD1788	HD_RS07300	secA	Protein translocase subunit SecA	9.23	2.14	1.0E−02
Posttranslational modification, protein turnover and chaperones
Operon_0064	HD0189	HD_RS00740	dnaK	Molecular chaperone DnaK	12.1	2.39	1.4E−02
Operon_0800	HD2006	HD_RS08230	hslV	ATP-dependent protease subunit HslV	8.24	2.3	6.8E−03
Operon_0801	HD2007	HD_RS08235	hslU	ATP-dependent protease ATPase subunit HslU	8.89	2.2	9.6E−03
Signal transduction mechanisms
Operon_0140	HD0357	HD_RS01465	HD0357	Carbon starvation protein A	9.65	2.05	2.7E−02
Information storage and processing
Phage-derived proteins, transposases, and other mobilome components
Operon_0193	HD0517	HD_RS02120	mug-2	Mu-like bacteriophage G protein 2	3.09	3.17	1.7E−02
Replication, recombination, and repair
Operon_0107	HD0290	HD_RS01160	HD0290	Transposase	5.12	2.4	1.0E−03
Operon_0388	HD0986	HD_RS04055	HD0986	dsDNA-mimic protein	6.53	2.07	2.8E−04
Operon_0636	HD1612	HD_RS06560	HD1612	Transposase	4.51	2.03	7.3E−02
Operon_0669	HD1689	HD_RS06895	HD1689	Transposase	5.31	2.17	3.4E−02
Operon_0671	HD1695	HD_RS06920	HD1695	Transposase	4.94	2.02	5.2E−02
Transcription
Operon_0440	HD1096	HD_RS04515	hipB	Transcriptional regulator HipB/DNA-binding protein	6.13	2.55	7.7E−04
Operon_0514	HD1287	HD_RS05250	HD1287	ATP-dependent helicase	8.65	2.04	1.6E−02
Translation, ribosomal structure, and biogenesis
Operon_0586	HD1463	HD_RS05980	HD1463	Ribosome maturation factor RimP	7.69	2.05	1.4E−04
Operon_0706	HD1765	HD_RS07200	rumA	23S rRNA (uracil-1939–C-5)-methyltransferase RlmD	7.16	2.19	6.1E−03
Metabolism
Amino acid transport and metabolism
Operon_0173	HD0444	HD_RS01840	aroF	DAHP synthetase, tyrosine repressible	8.63	2.98	2.2E−05
Operon_0808	HD2035	HD_RS08335	tcyA	Amino acid ABC transporter substrate-binding protein	8.46	2.32	1.4E−02
Carbohydrate transport and metabolism
Operon_0092	HD0237	HD_RS00935	dcuB2	C-4-dicarboxylate ABC transporter	8.12	2	1.0E−02
Operon_0743	HD1857	HD_RS07625	ulaD	3-Keto-l-gulonate-6-phosphate decarboxylase	8	2.14	3.9E−03
Operon_0743	HD1859	HD_RS07630	ulaC	PTS ascorbate transporter subunit IIA	7.75	3.43	5.2E−07
Operon_0744	HD1860	HD_RS07635	ulaAB	PTS ascorbate transporter subunit IIBC	8.91	4.19	4.2E−06
Operon_0745	HD1861	HD_RS07640	ulaG	Ascorbate 6-phosphate lactonase	11.1	5.84	4.4E−08
Operon_0746	HD1863	HD_RS07645	ulaR	Transcriptional regulator	9.31	2.73	7.2E−04
Operon_0746	HD1864	HD_RS07650	ulaE/sgbU	l-Xylulose 5-phosphate 3-epimerase	9.44	2.44	1.9E−02
Operon_0746	HD1866	HD_RS07655	ulaF/araD	l-Ribulose-5-phosphate 4-epimerase	9.28	2.37	1.1E−02
Coenzyme transport and metabolism
Operon_0695	HD1744	HD_RS07110	HD1744	Sulfur acceptor protein CsdL	6	2.15	5.0E−03
Operon_0306	HD0789	HD_RS03235	ccmC	Heme ABC transporter permease	5.25	2.87	1.6E−04
Energy production and conversion
Operon_0099	HD0264	HD_RS01050	mdh	Malate dehydrogenase	6.68	2.45	7.3E−03
Operon_0499	HD1240	HD_RS05080	citC	Citrate (pro-3S)-lyase ligase	7.75	2.34	4.9E−03
Operon_0499	HD1241	HD_RS05085	citD	Citrate lyase ACP	5.54	4.15	2.5E−08
Operon_0499	HD1242	HD_RS05090	citE	Citrate lyase subunit beta	6.81	3.23	6.0E−06
Operon_0772	HD1915	HD_RS07845	HD1915	Sulfite oxidase	6.8	3.31	1.2E−06
Operon_0772	HD1916	HD_RS07850	HD1916	Sulfoxide reductase catalytic subunit YedY	7.55	6.72	2.0E−20
Inorganic ion transport and metabolism
Operon_0154	HD0388	HD_RS01605	tdhA	TonB-dependent receptor	9.47	6.98	5.6E−16
Operon_0277	HD0721	HD_RS02930	corA	Magnesium transporter CorA	9.81	2.32	1.6E−02
Operon_0390	HD0991	HD_RS04080	focA	Formate transporter FocA	8.67	2.31	5.3E−05
Operon_0410	HD1025	HD_RS04230	yfeC	Membrane protein	7.64	2.18	3.8E−05
Operon_0441	HD1098	HD_RS04520	metN	Methionine import ATP-binding protein MetN	7.88	2.59	8.9E−03
Operon_0728	HD1816	HD_RS07435	yfeA	Iron-binding protein	9.48	2.68	3.1E−05
Operon_0805	HD2025	HD_RS08300	hgbA	Hemoglobin and hemoglobin-haptoglobin-binding protein	12	2.77	1.0E−03
Lipid transport and metabolism
Operon_0299	HD0774	HD_RS03145	fabH	3-Oxoacyl-ACP synthase III	5.74	2.03	5.5E−02
Unclassified
Operon_0359	HD0919	HD_RS03775	HD0919	Hypothetical protein	3.98	2.17	7.4E−02
Downregulated genes
Cellular processes and signaling
Cell cycle control, cell division, and chromosome partitioning
Operon_0396	HD1001	HD_RS04120	HD1001	Cell division protein ZapA	6.73	−2.31	4.2E−04
Cell wall/membrane/envelope biogenesis
Operon_0019	HD0045	HD_RS00210	momp	Membrane protein	14.9	−2.51	9.8E−02
Operon_0071	HD0196	HD_RS00765	HD0196	Membrane protein	8.43	−2.03	1.0E−03
Operon_0100	HD0266	HD_RS01055	lpp	Membrane protein	9	−2.93	1.5E−06
Operon_0608	HD1511	HD_RS06170	glmU	Bifunctional N-acetylglucosamine-1-phosphate uridyltransferase/glucosamine-1-phosphate acetyltransferase	9.77	−2.15	9.4E−02
Posttranslational modification, protein turnover, and chaperones
Operon_0270	HD0700	HD_RS02850	HD0700	Glutathione peroxidase	12.1	−2.44	9.9E−03
Information storage and processing
RNA processing and modification
Operon_0634	HD1606	HD_RS06535	rnc	RNase 3	7.19	−2.02	4.5E−03
Transcription
Operon_0096	HD0254	HD_RS01005	HD0254	Hypothetical protein	9.31	−3.01	1.9E−03
Operon_0258	HD0675	HD_RS02755	metJ	Met repressor	5.55	−2.33	2.4E−02
Translation, ribosomal structure, and biogenesis
Operon_0081	HD0212	HD_RS00830	rimI	Ribosomal-protein-alanine acetyltransferase	6.96	−2.31	4.2E−04
Operon_0488	HD1212	HD_RS04970	smpA	Hypothetical protein	9.55	−2.16	4.0E−02
Operon_0497	HD1238	HD_RS05070	rimK	Alpha-l-glutamate ligase	8.93	−2.1	7.5E−03
Metabolism
Amino acid transport and metabolism
Operon_0238	HD0630	HD_RS02565	dapD	2,3,4,5-Tetrahydropyridine-2,6-dicarboxylate N-succinyltransferase	7.68	−2.82	1.5E−06
Coenzyme transport and metabolism
Operon_0258	HD0671	HD_RS02745	panF	Sodium-pantothenate symporter	9.14	−2.03	8.7E−02
Operon_0415	HD1041	HD_RS04300	HD1041	Nicotinamide riboside transporter PnuC	7.86	−2.07	1.3E−02
Inorganic ion transport and metabolism
Operon_0700	HD1754	HD_RS07150	ftna	Ferritin	8.99	−2.45	9.2E−05
Operon_0700	HD1755	HD_RS07155	ftnB	Ferritin	8.63	−2.6	1.4E−05
Nucleotide transport and metabolism
Operon_0421	HD1053	HD_RS04350	ndk	Nucleoside diphosphate kinase	9.51	−2.03	5.1E−03
Operon_0604	HD1503	HD_RS06145	guaB	Inosine-5-monophosphate dehydrogenase	8.94	−3.5	1.3E−05
Operon_0604	HD1504	HD_RS06150	guaA	GMP synthase (glutamine-hydrolyzing)	8.83	−2.91	1.5E−04
Poorly characterized
Function unknown
Operon_0094	HD0250	HD_RS00990	HD0250	Hypothetical protein	7.47	−2.24	4.0E−03
Operon_0258	HD0673	HD_RS02750	HD0673	Membrane protein	6.91	−3.64	1.3E−05
Operon_0436	HD1089	HD_RS04480	HD1089	Hypothetical protein	6.52	−2.12	5.9E−03
Operon_0535	HD1354	HD_RS05525	HD1354	Hypothetical protein	7.99	−2.21	1.2E−04
Operon_0537	HD1362	HD_RS05560	HD1362	Hypothetical protein	7.48	−2.18	1.8E−04
Operon_0637	HD1614	HD_RS06570	HD1614	Hypothetical protein	5.61	−3.23	2.8E−03
Unclassified
Operon_0179	HD0458	HD_RS01905	HD0458	Hypothetical protein	6.23	−3.12	6.2E−04
Operon_0288	HD0746	HD_RS03030	HD0746	Hypothetical protein	9.8	−2.2	5.6E−03
Operon_0363	HD0926	HD_RS03810	HD0926	Hypothetical protein	5.35	−2.33	3.4E−02
Operon_0611	HD1524	HD_RS06215	HD1524	PerC	1.9	−131.44	9.5E−02
Operon_0762	HD1893	HD_RS07750	HD1893	Hypothetical protein	5.68	−2.44	1.4E−02

^aGenes that are in a putative operon are indicated by identical operon identifiers (IDs).

^bGenes that were differentially expressed in the biopsy samples compared to expression in all three growth phases are indicated in boldface.

^cdsDNA, double-stranded DNA; PTS, phosphotransferase system; ACP, acyl carrier protein.

Thirty-nine genes were commonly differentially expressed in vivo compared to expression in all three growth phases; of these, 32 were upregulated and 7 were downregulated (Fig. 2). The 32 upregulated genes encode homologs of genes (hereafter referred to as genes for simplicity) involved in l-ascorbate utilization and metabolism, amino acid and transition metal transport, heat shock and growth arrest response, and DNA replication and repair (boldfaced in Table 3). Genes in the flp-tad operon, which are required for infection in humans, also were part of the 32 commonly upregulated genes (Table 3). The 7 downregulated genes were scattered across a variety of functional categories (Table 3). The upregulation of 32 genes in vivo compared to all three growth phases suggests that these genes play important roles during human infection.

Comparison of H. ducreyi gene transcription in mid-log-phase cells (the inoculum) and the abscess.

In the human challenge model, volunteers are inoculated with mid-log-phase organisms. Therefore, we next asked how H. ducreyi gene transcription changes from that in mid-log phase to that in the abscess and focused on these transcriptional changes for the remainder of this study.

We first performed qRT-PCR to confirm the differential expression of select genes in vivo. Using dnaE as a reference gene, qRT-PCR confirmed the differential expression of 8/9 genes identified by RNA-Seq (Fig. 3). While HD0673 was repressed 3.63-fold by RNA-Seq, it was not differentially expressed by qRT-PCR (1.01-fold) (Fig. 3). The fold changes derived from RNA-Seq moderately correlated with the fold changes derived from qRT-PCR with a coefficient of determination of 0.6.

FIG 3.

qRT-PCR validation of the RNA-Seq data. The fold change in expression was calculated by comparing the gene expression in H. ducreyi harvested from pustules to those harvested from mid-log phase. The expression levels of target genes were normalized to that of dnaE. The data represent the mean ratio of transcript expression in four biopsy specimens divided by expression from four broth cultures used for the RNA-Seq study. As the samples were not paired, no standard deviations were calculated.

Differential expression of genes involved in nutrient stress adaptation.

Prominent among the upregulated genes were those that encode proteins involved in l-ascorbate utilization and metabolism (ulaABCD, ulaG, ulaR, ulaE, and ulaF) (Table 3). Escherichia coli can utilize l-ascorbate as an alternative carbon source using the ula pathway (28). In the presence of glucose, UlaR inhibits the ula pathway; under glucose limitation and in the presence of l-ascorbate, UlaR activates the ula pathway (28, 29). H. ducreyi in pustules also had increased the expression of a gene encoding the carbon starvation protein A (cstA), which is induced under glucose starvation in E. coli (Table 3) (30). Pustules typically are anaerobic, and H. ducreyi upregulated two genes involved in anaerobic respiration and fermentation in vivo (dcuB2 and focA) (Table 3) (31–34). Taken together, these data suggest that the in vivo environment is glucose limited and anaerobic and that H. ducreyi utilizes l-ascorbate as an alternative carbon source, perhaps to synthesize sugars for cell wall biogenesis and ribose-5-phosphate for nucleotide biosynthesis under these conditions.

If this is the case, what is the source of l-ascorbate in the abscess? When neutrophils encounter bacterial pathogens, they take up and concentrate l-ascorbate intracellularly up to 30-fold (35). H. ducreyi-induced abscesses contain live and dead neutrophils. If neutrophils concentrate l-ascorbate in response to H. ducreyi, dying and necrotic neutrophils may release l-ascorbate into the environment, providing this carbon source to H. ducreyi. To our knowledge, the potential contribution of l-ascorbate to the survival of an abscess-forming organism has not been reported; this pathway could be exploited for therapeutics.

H. ducreyi also upregulated several genes involved in citrate metabolism in vivo (Table 3). Under anaerobic conditions, many bacteria metabolize citrate via the citrate fermentation pathway, which involves the uptake of citrate by a citrate transporter, the breakdown of citrate by citrate lyase into acetate and oxaloacetate, and the decarboxylation of oxaloacetate by oxaloacetate decarboxylase to pyruvate (36–38). Pyruvate generated from the decarboxylation of oxaloacetate subsequently is metabolized to acetyl-coenzyme A, which ultimately is converted by pyruvate formate lyase into ATP and formate (36–38). The accumulation of formate results in significant reduction in intracellular pH, necessitating the excretion of formate out of the cell (39). H. ducreyi upregulated three genes encoding different components of citrate lyase (citC, citD, and citE) and a gene encoding the formate transporter (focA) in vivo (Table 3). In an alternative pathway, oxaloacetate is converted by malate dehydrogenase to l-malate and l-malate is converted by fumarate hydratase to fumarate, which is metabolized by fumarate reductase to succinate during fumarate respiration (31, 32, 40). H. ducreyi also upregulated a gene encoding malate dehydrogenase (mdh) (Table 3) in vivo. Thus, H. ducreyi may utilize citrate as an alternative carbon and energy source in vivo.

Another hallmark of the transcriptional response of H. ducreyi to the in vivo environment is the upregulation of genes involved in nutrient transport. Genes involved in the transport of transition metals such as iron and manganese (yfeA and yfeC) and magnesium and cobalt (corA) were elevated in vivo (Table 3). Genes involved in the transport of amino acids (metN and tcyA) and peptides (cstA) also were elevated (Table 3). These findings are consistent with the concept of nutritional immunity, in which vertebrate hosts sequester transition metals and nutrients as a defense against invading pathogens (41). Collectively, these data suggest that the environment of an abscess contains a limited supply of transition metals and amino acids and that the scavenging of these nutrients helps H. ducreyi survival in this niche.

In addition to nutrient scavenging, genes involved in heat shock response, growth arrest/dormancy response, and DNA recombination also were induced in vivo. Specifically, the expression of genes encoding the heat shock protein DnaK (dnaK) and the cytoplasmic chaperones HslV and HslU (hslV and hslU) was elevated (Table 3). Genes involved in growth arrest/dormancy response, such as the bacterial apoptosis protein LrgB (lrgB) and the bacterial high persistency antitoxin HipB (hipB), also were induced in vivo (Table 3). In E. coli, the induction of genes involved in heat shock and growth arrest/dormancy responses is a hallmark of a nutrient stress response (42, 43). Furthermore, transposase homologs were induced in vivo (HD0290, HD1612, HD1689, and HD1695) (Table 3). The transposition of mobile elements is increased by nutrient stress in some other organisms (44, 45). Thus, the upregulation of heat shock response, growth arrest response, and DNA transposition genes could be a secondary effect of nutrient stress.

Differential expression of genes required for virulence in humans.

All 18 genes or operons known to be required for the full virulence of H. ducreyi in humans were expressed in vivo (13–17, 46). However, H. ducreyi harvested from pustules only had relatively increased expression of hgbA and genes belonging to the flp-tad and lspB-lspA2 operons (Table 3). hgbA encodes a protein required for hemoglobin uptake, the flp-tad locus encodes proteins involved in microcolony formation, and the lspB-lspA2 locus encodes proteins involved in the evasion of phagocytosis (47–51). It was somewhat surprising that only 3 genes encoding virulence determinants were upregulated in vivo. One explanation is that the expression of hgbA and genes of the flp-tad and lspB-lspA2 operons is regulated by environmental cues and that the expression of the remaining essential genes is constitutive. In accordance with this hypothesis, hgbA expression is increased under heme-limiting conditions, which may be found in vivo. The expression of the flp-tad and lspB-lspA2 operons is increased in stationary phase, suggesting that the in vivo induction of these genes is in response to nutrient stress (13). However, our methods only identified H. ducreyi genes that were differentially expressed at one time point from the entire population of bacteria in a lesion. The expression of other virulence determinants may be differentially regulated at other time points or in bacteria that occupy distinct microniches in the lesions. Surprisingly, H. ducreyi also upregulated a gene encoding the TonB-dependent heme receptor TdhA, which is not required for H. ducreyi infection in humans (8).

Differential expression of putative sRNAs.

Hfq is required for H. ducreyi virulence in humans (13). In other organisms, Hfq functions by facilitating pairing of small RNAs (sRNAs) with mRNA targets, affecting their stability and/or translation (52). Therefore, we asked if any of the putative sRNAs were differentially expressed during human infection. By examining the intergenic regions of the H. ducreyi genome, we identified 10 putative sRNAs; seven were homologous to known sRNAs, and three were unique to H. ducreyi (Table 4). Compared to mid-log phase, the bacterial small RNA SRP/Ffs homolog and HDsRNA06 were induced whereas the GcvB homolog and HDsRNA10 were repressed during human infection; the other sRNAs were not differentially regulated (Table 4). In E. coli, GcvB is known to repress genes involved in amino acid transport and biosynthesis. Thus, the downregulation of a homolog of GcvB in H. ducreyi is consistent with the upregulation of genes involved in amino acid and peptide transport (53).

TABLE 4.

Differential expression of putative H. ducreyi sRNAs identified by RNA-Seq in biopsy specimens relative to mid-log-phase samples

Name	Position		Length (nt)	Flanking gene		Rfam annotationb	Family or descriptionc	Fold changea	FDR
	Start	End		Left	Right
HDsRNA01	70637	70995	359	HD0084	HD0085	RF00023	tmRNA/SsrA	NDE
HDsRNA02	754996	755134	139	HD0953	HD0954	NA	Unique to H. ducreyi	NDE
HDsRNA03	795930	796094	165	HD1000	HD1001	RF00013	6S RNA/SsrS	NDE
HDsRNA04	817927	818100	174	HD1027	HD1028	RF00022	GcvB	−5.30	1.28E−06
HDsRNA05	943036	943169	134	glpC	ribD	RF00050	FMN riboswitch	NDE
HDsRNA06	952062	952286	223	HD1171	HD1172	NA	Unique to H. ducreyi	2.25	5.07E−02
HDsRNA07	992920	993113	194	HD1226	HD1227	RF00168	Lysine riboswitch	NDE
HDsRNA08	1086275	1086371	97	hhdA	HD1328	RF00169	Bacterial small RNA SRP/Ffs	2.64	1.90E−02
HDsRNA09	1339675	1340084	410	HD1622	HD1623	RF00010	RNaseP_bact_a	NDE
HDsRNA10	1477923	1478075	183	HD1765	HD1766	NA	Unique to H. ducreyi	−3.89	1.97E−06

^aMean fold change in expression in the biopsy specimens relative to that in mid-log phase. NDE, not differentially expressed.

^bNA, not available.

^ctmRNA, transfer-messenger RNA; FMN, flavin mononucleotide.

Comparison of genes differentially expressed in vivo to those regulated by CpxR, Hfq, RpoE, (p)ppGpp, and DksA.

We recently defined the genes differentially regulated by CpxR, RpoE, Hfq, (p)ppGpp, and DksA in H. ducreyi. These regulons were determined by comparing a CpxR-activating mutant to a cpxR deletion mutant by the overexpression of RpoE relative to the wild type or by comparing hfq, relA spoT, and dksA mutants to the wild type (11–14). The hfq, relA spoT, and dksA deletion mutants and the CpxRA-activating mutant all downregulated multiple virulence determinants and were attenuated for virulence in humans (13–15, 17). Comparison of 93 genes differentially expressed between mid-log-phase bacteria and the abscess to the 119 CpxR-dependent genes showed an overlap of 6 genes, to the 180 RpoE-dependent genes showed an overlap of 12 genes, to the 282 Hfq-dependent genes showed an overlap of 10 genes, to the 148 (p)ppGpp-dependent genes showed an overlap of 15 genes, and to the 58 DksA-dependent genes showed no overlap. None of these comparisons achieved statistical significance by chi-square analysis. The lack of overlap may be because these regulons were determined by comparing differential gene expression in strains where the regulator was either present or induced to that of a strain where the regulator was either absent or not induced. Together, these data suggest that these systems, which are required for virulence in humans, switch on and off during human infection.

Comparison of H. ducreyi transcriptional profile to those of other bacterial pathogens during human infection.

Transcriptional profiles during human infection have been previously defined for several bacterial pathogens, including Staphylococcus aureus, Neisseria gonorrhoeae, uropathogenic E. coli, Vibrio cholerae, Vibrio vulnificus, group A Streptococcus, and Mycobacterium tuberculosis (54–60). Several of these studies reported the upregulation of genes involved in the transport of transition metals, peptides, and amino acids during human infection. Taken together with our study, these data suggest that nutrient limitation is a general phenomenon during human infection, consistent with the concept of nutritional immunity (41).

All of the above-described studies were performed on samples collected from naturally infected patients. Such studies have limitations, such as person-to-person variability in the infecting bacterial strains, in host immune status, and in the stage of disease and the lack of control samples that allow the determination of differential gene expression in vivo. With the exception of S. aureus and M. tuberculosis studies, the transcriptomes for all of the above-described pathogens were determined using lavage, urine, stool, or swab samples instead of infected tissue; these samples may not recapitulate gene expression at the site of infection. In the S. aureus study, the in vivo transcriptomes were not compared to the in vitro transcriptome of the infecting strain isolated from each patient but were compared to the in vitro transcriptome of a reference strain belonging to the USA300 clone, which displays intraclonal genetic variability (61). In the M. tuberculosis study, the tissue samples were obtained from patients who had received antituberculosis chemotherapy; as discussed by the authors, the antibiotics may have influenced the in vivo gene expression profile.

Our model overcame several of these limitations, in that we infected healthy adults with a single bacterial strain to a defined stage of disease and determined baseline gene transcription from the infecting strain. Although we did not determine the transcriptomes of the actual inocula used to infect these volunteers, we did determine the in vitro transcriptome of 35000HP grown under conditions identical to those of the human inoculation experiments. As stated previously, limitations of our study are that we determined the transcriptome at one time point from an entire excised lesion and we did not capture changes in transcript expression over time or from bacteria that occupy different niches within the lesion.

Genes differentially expressed in biopsy specimens compared to stationary phase.

Compared to that in stationary phase, 385 genes were differentially expressed in biopsy specimens; of these, 189 genes were upregulated and 196 genes were downregulated (Fig. 2). Approximately four times more genes were differentially expressed when the bacteria from biopsy specimens were compared to stationary-phase bacteria than when the bacteria from biopsy specimens were compared to mid-log-phase bacteria; only 43 differentially expressed genes were common to two comparisons (Fig. 2). These differences likely are due to differences in nutrient availability under the two in vitro growth conditions. Similar to the biopsy specimen versus mid-log-phase comparison, several of the genes upregulated in bacteria from the biopsy specimens compared to that with stationary-phase bacteria encode proteins involved in l-ascorbate utilization, nutrient transport, DNA transposition, stress response, anaerobiosis, and respiration (including lactate dehydrogenase), as well as HgbA and the Flp-Tad locus proteins, which are required for human infection (see Table S4 in the supplemental material) (8). Unlike results of the comparison between in vivo and mid-log-phase-grown bacteria, lspB was not differentially expressed, and dsrA, which is required for human infection and encodes the H. ducreyi serum resistance determinant, was upregulated in vivo compared to stationary-phase bacteria (see Table S4).

Conclusions.

In summary, our data demonstrate that H. ducreyi gene expression in vivo does not resemble that of in vitro growth and that the adaptation of H. ducreyi to the pustular environment primarily involves the activation of a genetic program geared toward the acquisition of alternative carbon sources, such as ascorbic acid, and adaptation to nutrient stress and anaerobiosis. Future studies will focus on characterizing the genes that promote adaptation to the host environment and their potential contributions to H. ducreyi pathogenesis and on studying the interaction between the bacterial transcriptome and that of the host by dual RNA-Seq (62).