Dysregulated Immune Profiles for Skin and Dendritic Cells Are Associated with Increased Host Susceptibility to Haemophilus ducreyi Infection in Human Volunteers

Bacterial strains and culture conditions. H. ducreyi 35000HP is a human-passaged variant of strain 35000 (4). Actinobacillus actinomycetemcomitans (ATCC 29522) was a gift from Dominique Galli, and nontypeable Haemophilus influenzae 1479 was a gift from Timothy Murphy (53). Haemophilus parainfluenzae (ATCC 33392), Haemophilus parahaemolyticus (ATCC 10014), Haemophilus paraphrohaemolyticus (ATCC 29237), and Haemophilus parainfluenzae (“Haemophilus paraphrophilus”) (ATCC 29242) were purchased from the American Type Culture Collection, Manassas, VA. All bacteria were cultured as described previously (25).

Human inoculation experiments.Eight healthy adult volunteers (seven female and one male; mean age ± standard deviation, 37 ± 12 years), who had been infected twice and either had formed at least one pustule twice or had had the disease resolve at all sites twice, enrolled in the study. Informed consent was obtained from the subjects for human immunodeficiency virus type 1 serology and participation in accordance with the guidelines for human experimentation of the U.S. Department of Health and Human Services and the Institutional Review Board of Indiana University-Purdue University at Indianapolis. One RR subject was excluded due to an underlying illness, one PP subject withdrew from iteration 1 due to a scheduling conflict, and six subjects were infected (Table 1). Procedures for enrollment, exclusion criteria, preparation of the bacteria, inoculation, calculation of the estimated delivered dose, clinical observations, biopsy, and treatment with antibiotics were as described elsewhere (4, 65, 66).

DC-H. ducreyi cocultures.Peripheral blood was obtained from the subjects several months after they were infected a third time. CD14⁺ peripheral blood monocytes were isolated from peripheral blood mononuclear cells (PBMC) by positive-selection magnetic beads (Miltenyi Biotec, Auburn, CA) and grown in the presence of recombinant human IL-4 (1 ng/ml) and recombinant human granulocyte-macrophage colony-stimulating factor (GM-CSF) (0.2 ng/ml) (R&D Systems, Minneapolis, MN). After 7 days of culture, the cells were HLA-DR⁺, CD86⁺ CD40⁺, CD3⁻, CD14⁻, and CD19⁻. DC were incubated with nonopsonized H. ducreyi for 90 min at a multiplicity of infection of 20:1, washed to remove nonassociated bacteria, and incubated for an additional 22.5 h. The experiment was done in pairs so that DC from one RR subject were exposed to the same bacterial inoculum as DC from one PP subject. Cells were collected and processed for RNA. Supernatants were collected and analyzed for cytokines using the human Th1/Th2 II cytometric bead array kit per the manufacturer's instructions (BD Biosciences). Cytokine levels were defined as the cytokine concentration in an infected well minus the concentration in the corresponding uninfected well.

Flow cytometry.Biopsies were processed, and cells were collected and stained, as described previously (32), with the following antibodies conjugated to fluorescein isothiocyanate, allophycocyanin, phycoerythrin, or peridinin chlorophyll a proteins: CD3 (clone SK7), CD14 (MfP9), CD19 (4G7), and CD45 (2D1) (BD Biosciences, San Jose, CA). Monocyte-derived DC were stained with the following antibodies conjugated to fluorescein isothiocyanate, allophycocyanin, phycoerythrin, CyChrome, or peridinin chlorophyll a proteins: CD80 (clone L307.4), CD86 (2331), HLA-DR (646-6), CD40 (5C3), and CD83 (HB15e) (BD Biosciences).

Isolation of RNA.Skin specimens were placed in RNA Later (QIAGEN, Valencia, CA) for 30 min immediately after biopsy. Tissues were homogenized in lysis buffer using a Mini Bead Beater (Research Products International, Mt. Prospect, IL) with 2.4-mm silica beads, and total RNA was isolated using the RNeasy fibrous tissue minikit (QIAGEN). RNA was prepared from DC using the RNeasy kit (QIAGEN). An additional DNase step was performed on all samples using Ambion's DNA-free kit (Ambion, Austin, TX). RNA integrity was assessed by capillary gel electrophoresis on an Agilent Bioanalyzer (Agilent Technologies, Palo Alto, CA).

Microarrays.The same amount (100 ng) of total RNA was used from all samples, which were processed in parallel using premixes to minimize experimental variation. Samples were amplified using the Affymetrix two-cycle protocol. First- and second-strand cDNA syntheses were done using the Affymetrix GeneChip Expression 3′ Amplification two-cycle cDNA synthesis kit (Affymetrix, Santa Clara, CA). The first in vitro transcription used the Ambion MEGAscript T7 high-yield transcription kit (Ambion); the second used the Affymetrix GeneChip Expression 3′ Amplification IVT labeling kit. Biotinylated cRNA was fragmented following standard procedures. Each sample was hybridized to a separate human genome U133 Plus 2 array, which contains approximately 55,000 probe sets corresponding to approximately 39,500 well-substantiated genes, as described previously (46). The arrays were scanned on the Affymetrix GeneChip Scanner 3000 controlled by GCOS software, and data were extracted using the Microarray Software Suite 5 (MAS5) algorithm. Each sample was scaled to a target intensity of 1,000. Probe sets that were absent from more than 66% of the samples in all four groups (PP infected [PPI], PP uninfected [PPU], RR infected [RRI], and RR uninfected [RRU]) were removed before further analysis (45).

After applying the 66% present/absent filter, genome-wide expression data were analyzed using singular value decomposition (SVD), which is also known as principal-component analysis (5). SVD is a mathematical transformation that filters noise and redundant information in expression data, searches for values called eigenvectors that explain the variation of the high-dimensional data, and examines whether there is any separation or clustering in the data. Array samples were plotted in a two-dimensional scatter plot, in which the x or y coordinates represented their projections to two eigenvectors, representing infected versus uninfected samples and PP versus RR subjects.

Array-to-array variation was normalized within either the PP or the RR group, using nonlinear normalization techniques (12). A log-normal distribution was assumed for gene expression data. As both infected and uninfected samples were from either a PP or RR subject, a two-way analysis of variance model that accommodated correlated samples was employed to compare transcript levels under multiple conditions, and P values were calculated with t tests [lm() in R, www.r-project.org ]. False-discovery rates (FDR) were calculated for different levels of significance using the following formula: minimum [1, (P value) × (number of transcripts present)/(number of transcripts detected at that P value)].

To examine the biological functions of these transcripts, we utilized websites such as Biocarta (http://www.biocarta.com/ ) and OMIM (Online Mendelian Inheritance in Man; NCBI). Transcripts were reviewed for polarization of expression of signature genes for different blood cell types (www.affymetrix.com/support/technical/technotes/bloodappendix_technote.pdf ), apoptosis pathways, Toll-like receptor (TLR) pathways, cell surface activation markers, cytokines, cytokine receptors, chemokines, and chemokine receptors that may account for different immune responses in the two groups. The two lists were further classified for function using the Gene Ontology categories (http://www.geneontology.org/index.shtml ).

The supplemental material includes tables summarizing the immunologically related genes at a ≥2-fold change and a P value of <0.05 for the tissue and DC microarrays. For genes that were detected with multiple probe sets, all values were examined to ensure consistent results, and the probe set with the highest fold change was reported. In the case of discrepant results, we used the results from probe sets corresponding to the 3′ end of the transcript, which are more accurate than those from the 5′ end when RNA is amplified (46). The raw data are available at http://www.ncbi.nlm.nih.gov/geo/ under accession number GSE5574.

Real-time PCR.IL-10, IL-12, and migration inhibition factor (MIF) were quantified relative to the level of mRNA for RPL35A, which had little sample-to-sample variation in the DC microarrays. The primers and TaqMan probes were designed using Roche's Universal Probe Library Assay Design Center (see Table VI at https://www.roche-applied-science.com ). cDNA was made from the same RNA samples used for the microarray experiments with the Roche Transcriptor first-strand cDNA synthesis kit per the manufacturer's instructions. The reaction mixtures consisted of 500 nM of each primer, 5 mM MgCl₂, and 200 nM of each probe (LC Fast Start DNA master mix; Roche). The samples were amplified on a Roche LightCycler (94°C for 5 min; 45 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 1 min; and a final step at 72°C for 7 min). Differences in transcript levels between the groups were compared using the Student t test and the Wilcoxon rank sum test.

Proliferation assays.PBMC were isolated by Ficoll-Hypaque density gradient centrifugation, frozen, and stored in fetal calf serum with 10% dimethyl sulfoxide. Proliferation assays were performed using the Vybrant CFDA SE cell tracer kit (Molecular Probes) (44, 81). PBMC were labeled with carboxyfluorescein diacetate succinimidyl ester (CFSE) and incubated with medium alone, 5 μg/ml phytohemagglutinin (PHA), or 5 μg/ml heat-inactivated bacterial cells (25) at 37°C with 5% CO₂. Cells were harvested, washed, and stained with propidium iodide and anti-CD4-allophycocyanin (BD Biosciences). Live CD4⁺ cells were analyzed for expression of CFSE using a FACSCalibur flow cytometer and Cell Quest software (BD Biosciences). The cell division index (CDI) was calculated as (number of CD4⁺ CFSE^dim cells with antigen)/(number of CD4⁺ CFSE^dim cells without antigen). A CDI of greater than 2 is considered a positive response in this assay (44).

Reinfection of PP and RR subjects.To examine whether subjects who had different clinical outcomes had distinct immune responses at the transcript level to H. ducreyi in tissue, we inoculated PP and RR subjects at three sites with live H. ducreyi 35000HP and at one site with a phosphate-buffered saline control and biopsied all sites 48 h later. We reasoned that by 48 h we could still capture transcripts in RR lesions and compare them to those found in PP lesions and that the infected sites should reflect the response to the bacteria and wounds, while the control site should reflect the response to the wounds. We hypothesized that the PP and RR groups would have different gene expression profiles at their infected sites in response to H. ducreyi.

We infected three subjects in iteration 1 and two subjects in iteration 2 (Table 1). Papules formed at all sites inoculated with live bacteria. We biopsied four sites per subject 48 h after inoculation. RNA was made from the infected site that had the largest diameter of erythema and from the phosphate-buffered saline control site. The biopsies of the remaining two infected sites were pooled and processed for flow cytometry. The mean numbers (PP, 110,000; RR, 92,000) of leukocytes per biopsy obtained from the two groups were not statistically different. The mean percentages of major cell types in the PP and RR lesions were similar for PMN (8.6 versus 10.2), CD14 cells (27.1 versus 36.2), and CD3 cells (63.7 versus 52.7), suggesting that differences in transcript profiles may not merely be due to differences in cell numbers but may reflect qualitative differences in the host response.

Several months after the first five subjects were infected, a sixth subject (no. 243) was infected and formed only one papule. She underwent two biopsies (from the infected site and the wounded site). These biopsies were processed only for RNA.

Differential transcript responses to H. ducreyi in skin are associated with the PP and RR phenotypes.The 12 tissue samples contained intact RNA and were hybridized to separate Affymetrix Human Genome U133 Plus 2 Arrays. SVD showed that the arrays were separated into four distinct groups by two eigenvectors, representing infected versus uninfected samples and PP versus RR subjects (Fig. 1). This analysis suggested that there were distinct global transcriptional responses in infected and uninfected sites and in PP and RR subjects.

• Open in new tab
• Download powerpoint

FIG. 1.

Plot of SVD of the 12 tissue arrays. The x axis represents the projections of 12 samples onto the eigenvector that depicts the major variation between infected and uninfected samples. The arrow on the x axis points toward higher expression levels in infected samples than in uninfected samples. The y axis represents the projections of 12 samples onto the other eigenvector, which depicts the major variation between PP and RR subjects. The arrow on the y axis points toward higher expression levels in PP samples than in RR samples. Thus, the analysis separates the arrays into four groups that reflect the biology of the PP and RR phenotypes and infected and uninfected sites. Closed circles, RRU; open circles, RRI; closed squares, PPU; open squares, PPI.

We identified transcripts that were up- or downregulated in the infected versus uninfected tissue within each group of subjects with a mixed-model analysis of variance to account for the pairing of an experimental sample and a control sample within each subject. As expected, many transcripts were different, reflecting the infiltration of immune cells at infected sites versus wounded sites. For the comparison between PPI and PPU tissue, there were 9,604 probe sets representing 6,095 unique genes that were differentially regulated at a P value of ≤0.05 with an FDR of 0.14. For the comparison between RRI and RRU tissue, there were 9,284 probe sets representing 5,650 unique genes that were differentially regulated at a P value of ≤0.05 with an FDR of 0.14.

We examined whether the transcripts identified as up- or downregulated in the PPI versus PPU tissues overlapped with those identified in the RRI versus RRU tissues. The tissue RNA was used in its entirety to perform the arrays, and the level of transcript expression could not be validated by quantitative PCR. Thus, we limited our analysis to transcripts with at least twofold difference in expression level at a P value of ≤0.05. Of 4,268 transcripts corresponding to unique genes, 1,543 were found in both groups, while 1,565 were unique to the tissue from the PP group and 1,160 were unique to the tissue from the RR group.

Of the 1,543 common transcripts, we identified 268 that were relevant to immune function and that were differentially regulated in both the PPI versus PPU comparison and the RRI versus RRU comparison (see Table SI in the supplemental material). For the common transcripts, the directions (up- or downregulated) and magnitudes of change were similar in the two groups. These transcripts reflected the infiltration of leukocytes into the infected sites and indicated that the two groups have a core response to H. ducreyi. Upregulated transcripts included markers of antigen presentation and processing, chemokine and chemokine receptors, complement components, cytokines and cytokine receptors, adhesion molecules, IFN-inducible molecules, TLRs, and tumor necrosis factor (TNF)-induced transcripts.

Of the 1,565 transcripts unique to the comparison of the PPI and PPU groups, we identified 52 that were relevant to immune function (see Table SI in the supplemental material). Of the 1,160 transcripts unique to the comparison of the RRI and RRU groups, 58 were relevant to immune function (see Table SI in the supplemental material). Several adhesion molecules, complement and complement-related transcripts, and markers of differentiation, maturation, and activation were upregulated in the RRI sites, while such molecules were downregulated in the PPI sites (Table 2). Several more chemokine and chemokine receptors were upregulated in the RRI sites than in the PPI sites (Table 2). The RRI sites contained downregulated transcripts for markers of tolerance (TGFΒ2 and TGFB1I4) and several prostaglandin synthetases and had upregulated transcripts for CD163 and IL1RN, while the PP sites had upregulated transcripts for several cytokines and cytokine receptors, suggesting that inflammation was resolving at the RR sites and not at the PP sites (21, 51) (Table 2). Taken together, the data are reflective of an effective immune response at the RR sites.

An example of qualitative differences between PP and RR sites is shown in Table 3. Nuclear factor-erythroid 2-related factor 2 (Nrf2) is a transcription factor that regulates the expression of antioxidant genes, whose function results in posttranslational modification of many kinases downstream of TLR4 signaling. Deficiency of Nrf2 leads to a decreased redox state and increased activity of both the MyD88-dependent and MyD88-independent TLR4 pathways (38, 72). In addition, Nrf2 deficiency leads to higher levels of transcript expression in both TLR4 pathways. In models of sepsis and inflammation, Nrf2-deficient mice have higher mortality rates and expression levels of proinflammatory cytokines than wild-type mice (72). We examined the arrays to see if antioxidant synthesis genes and the TLR4 signaling pathways were differentially regulated in the groups. The PPI tissue contained 13 downregulated and 5 upregulated transcripts for glutathione synthesis, while the RRI tissue had 5 downregulated and 4 upregulated transcripts (Table 3). The PPI tissue also contained eight upregulated transcripts and one downregulated transcript in the TLR4 signaling pathways, while the RRI tissue had three upregulated and two downregulated transcripts (Table 3). Taken together, this transcript pattern suggests that the PP response in skin represents a dysregulated, hyperinflammatory state.

Differential DC responses to H. ducreyi are associated with the PP and RR phenotypes.Myeloid DC are enriched relative to plasmacytoid DC in pustules of subjects infected once (5a). Thus, myeloid DC could orchestrate the differential immune response to the organism in infected tissue. To examine whether DC have differential responses to the bacteria, we derived myeloid DC from PBMC of the PP and RR subjects. After 7 days, DC were incubated with nonopsonized H. ducreyi or in medium alone for 24 h. DC were stained with CD40, CD80, CD83, CD86, and HLA-DR. There was no difference in the percentage or in the mean fluorescence intensity of infected DC that expressed these surface markers compared to uninfected DC that were allowed to mature in the wells for 24 h (data not shown), consistent with previous observations (5a). No differences in expression of these markers were noted between the PP and RR groups.

The 12 DC samples containing intact RNA were hybridized to Affymetrix arrays and analyzed exactly as described for the tissue arrays. At a P value of ≤0.05, for the comparison of RRI and RRU DC, there were 3,317 probe sets representing 2,433 unique genes that were up- or downregulated with an FDR of 0.28. For the comparison of PPI and PPU DC, 3,134 probe sets representing 2,347 unique genes were differentially regulated with an FDR of 0.30.

We examined whether the subset of transcripts that were at least twofold differentially regulated in the RRI versus RRU DC overlapped with the same subset identified in the PPI versus PPU DC. A total of 2,021 transcripts corresponding to unique genes were twofold differentially regulated in both groups, with 464 transcripts unique to the RRI versus RRU DC, 1,107 unique to the PPI versus PPU DC, and 451 common to the two groups.

There were 81 transcripts known to be relevant to immune responses that were common to the infected DC from both groups (see Table SII in the supplemental material). The direction of change was the same for all the common transcripts in the two groups. Key transcripts (Table 4) associated with DC maturation or polarization of T cells that were upregulated in both groups included CD80, ICAM-1, IL-1α, IL-1β, IL-6, IL-8, IL-12p40, IL-15, IL-2Rα, TNF-α, several TNF- and IFN-induced proteins, Jagged-1, and G-CSF. Except for upregulation of programmed cell death ligand 1, the common response to H. ducreyi was generally consistent with a DC1 profile (18).

There were 74 differentially regulated transcripts unique to infected DC from the PP group (see Table SII in the supplemental material). Key transcripts (Table 5) included many markers of DC1, such as CCR7, CD38, CD58, IL-7, IL-18, LAMP3, and GM-CSF (9, 18, 49, 85). Caspase 1, which processes pro-IL-18, was also upregulated (18). Many IFN-responsive genes were also upregulated. The marked upregulation of GM-CSF and other proinflammatory cytokines suggests that the PP DC could sponsor a hyperinflammatory state. However, the PP DC differentially regulated many transcripts that were markers of DCreg, such as IL-10, CD200, and Notch homolog 2 (17, 18, 37, 48) (Table 5). INDO, encoding IDO, which leads to tryptophan depletion and T-cell death, was upregulated 173-fold. IL-16, which fosters DC maturation, was downregulated, as were CD1D and CD84. SOCS2 and SOCS3, which interfere with TLR 4 signaling and DC function (69), were upregulated. Major histocompatibility complex (MHC) class II DQα, HLA-B, and HLA-F were upregulated, but MHC class II DMβ, MHC class II DMα, and HM13 were downregulated, which could lead to defects in antigen processing and presentation.

There were 28 differentially regulated transcripts unique to the infected DC from the RR group (see Table SII in the supplemental material). Key transcripts (Table 5) included only additional markers of DC1, such as upregulated IL-23αp19, IL-1F9, IRF1, IRF4, and MIF (18, 49, 85). LILRB1 (CD85j), an inhibitory receptor whose activation reduces production of IL-12 and polarizes towards Th2 responses (71), was downregulated.

IL-12p40 was upregulated in DC from both groups, but IL-23αp19 was upregulated only in the RR DC. STAT4 was upregulated in the PP DC, while STAT1 was upregulated in RR DC. This coordinate regulation may reflect programming for subsequent signaling through IL-12/STAT4 and IL-23/STAT1 (6). IL-1β and IL-6, which promote Th17 differentiation in humans, were upregulated in both groups. Since IL-23 also promotes human Th17 development or expansion, RR DC could promote a more profound Th17 response than PP DC. Taken together, the data indicate that while DC from the PP and RR groups share elements of a common response to H. ducreyi, other elements of the DC response are divergent. DC from the PP group express unique transcripts that should promote both Th1 and Treg responses, while DC from the RR group express transcripts should promote a Th1 response and promote expansion of Th17 cells by upregulation of IL-23.

Validation of the DC microarrays by quantitative PCR and cytokine analysis.RPL35A (a ribosomal protein transcript) was not differentially regulated in either the PP or RR DC. MIF was upregulated in the infected RR DC, IL-10 was upregulated in the infected PP DC, and IL-12p40 was upregulated in both groups. By quantitative PCR, the levels of RPL35A transcripts were not statistically different when infected DC were compared to uninfected DC in both groups. After normalization to RPL35A transcript levels, MIF transcripts were significantly upregulated in the RR DC, IL-10 was significantly upregulated in the PP DC, and IL-12 p40 was significantly upregulated in both groups (all P < 0.01). The directions of the levels of expression of these transcripts were similar to those determined in the array (Table 6), although the magnitude of the fold change was generally higher in real-time PCR than in the array.

Cytokines were analyzed in the 24-h cell culture supernatants obtained from infected and uninfected DC. Although our sample size lacked sufficient power to determine whether the differences in cytokine levels were significant, the levels of expression of IL-10, TNF-α, and IL-6 (Fig. 2) paralleled the levels of transcript expression for these cytokines expressed by the two groups (Tables 4 and 5). Little IFN-γ was made by both groups, and transcripts for IFN-γ were not differentially regulated in either group. Thus, the cytokines levels produced by the DC were consistent with the relative levels of transcript expression in the microarrays.

• Open in new tab
• Download powerpoint

FIG. 2.

Cytokine production as measured by cytometric bead array. The culture supernatants from the DC cocultures used for microarray analysis were used for determination of TNF-α and IL-6 (A) and for IFN-γ and IL-10 (B). Values represent the means + standard deviations (SD) of cytokine concentrations in infected wells minus those in uninfected wells from the three persons in each group. White bars, PP group; black bars, RR group.

Proliferative responses to H. ducreyi are not associated with the PP and RR phenotypes.An alternative hypothesis for differential host susceptibility is that the RR group had cross-reactive memory cells that were primed by commensals and provided Th1 responses to H. ducreyi that facilitated clearance, while the PP group lacked such cells. For the five subjects who were inoculated in the first two iterations, PBMCs were obtained prior to the third challenge, 48 h and 7 to 10 days after infection. For all subjects, CD4 cells proliferated in response to PHA, but the mean CDIs for both groups were <1.5 at all three time points for all concentrations (1 to 20 μg/ml) of heat-inactivated H. ducreyi whole cells tested (data not shown). All six subjects were recalled approximately 6 months after the sixth subject was inoculated so that we could measure blastogenic responses to H. ducreyi and related species of the Pasteurellaceae. CD4 cells from the RR and PP groups did not proliferate in response to H. ducreyi (CDI, <1.0), although they did proliferate to various degrees in response to the other bacterial species (Fig. 3). The CDI for PHA was >9 for all samples tested. The concentration of IFN-γ in culture supernatants made in response to H. ducreyi did not differ between the two groups (data not shown). We repeated this experiment for blastogenesis and IFN-γ production in response to H. ducreyi, and the results were identical.

• Open in new tab
• Download powerpoint

FIG. 3.

Proliferation of CD4 cells in response to various species of the Pasteurellaceae. The data represent the mean CDIs + standard deviations (SD) from the three subjects in each group. White bars, PP group; black bars, RR group.