A Pilus-Deficient Mutant of Haemophilus ducreyi Is Virulent in the Human Model of Experimental Infection

Abstract

Haemophilus ducreyi expresses fine tangled pili, which are composed predominantly of a major subunit (FtpA). Confocal microscopy showed that an FtpA-specific monoclonal antibody bound to bacteria in biopsy samples obtained from infected human volunteers. To test the role of pili in pathogenesis, an isogenic mutant (35000HP-SMS1) was constructed by insertionally inactivating ftpA. 35000HP-SMS1 did not express FtpA and was nonpiliated but was otherwise identical to its parent, 35000HP. Seven healthy adults were challenged on the upper arm with the isogenic isolates in a double-blinded, escalating dose-response study. Sites inoculated with the mutant produced papules and pustules at rates similar to the rates observed at sites inoculated with the parent. The recovery rate of H. ducreyi from cultures and the histopathology of biopsy samples obtained from pustules inoculated with 35000HP or 35000HP-SMS1 were similar. Although pili are expressed in vivo, FtpA is not required for pustule formation in the human challenge model.

Pili (fimbriae) are filamentous appendages that are present on the surface of most gram-negative bacteria and that serve as adhesins, enabling colonization, infection, or both. Haemophilus ducreyi, the etiologic agent of chancroid, synthesizes filamentous structures called “fine tangled pili,” which are composed predominantly of a major subunit (FtpA) whose apparent molecular mass is 24,000 Daltons [1–3]. FtpA reacts with monoclonal antibody (MAb) 2D8 and has no homology to other known pilus proteins, although it shares homology with proteins that polymerize to form ordered rings [4]. Immunization with purified pili provides partial protection against experimental infection with homologous and heterologous strains of H. ducreyi in the temperature-dependent rabbit model [2]. However, the putative receptor for FtpA binding is unknown [4, 5].

We examined whether H. ducreyi expressed FtpA in vivo. We constructed an isogenic ftpA mutant by insertionally inactivating the ftpA in a human-passaged variant of 35000 (35000HP), which was used to standardize a human infection model [6, 7]. The virulence of the isogenic pair of isolates was tested in a double-blinded, escalating dose-response study.

Materials and Methods

Bacteria and plasmids

H. ducreyi 35000HP-SMS1 was constructed identically to H. ducreyi 35000 ftpA::mTn3(Cm), as described elsewhere [4]. In brief, 35000HP was transformed by elec-troporation with linearized pHD24 ftpA::mTn3(Cm), and the transformants were plated on chocolate agar containing chloram-phenicol (2 μg/mL). Cm^r transformants were screened by use of Western blot analysis for loss of reactivity to MAb 2D8 [4], and the genotype was verified by Southern blot hybridization. A 2D8-nonreactive transformant that had undergone allelic replacement in ftpA was designated 35000HP-SMS1.

Outer membrane, lipo-oligosaccharide, and Southern blot analyses

Lipo-oligosaccharide and outer membranes were prepared from 35000HP and 35000HP-SMS1 and were subjected to SDS-PAGE analysis as described elsewhere [4, 8, 9]. Southern blot analysis was done as described elsewhere [4].

Electron microscopy

To examine whether H. ducreyi cells synthesized pili, 35000HP and 35000HP-SMS1 were grown, harvested, processed, stained with phosphotungstic acid, and examined by electron microscopy as described elsewhere [4].

Human volunteers

Seven healthy women (5 white, 2 black; age range, 22–42 years; mean age ± SD, 28.4 ± 7.8 years) volunteered for the study. For confocal microscopy, tissue specimens were obtained from an additional 3 volunteers (nos. 102, 105, and 110) who participated in the comparison trial of hemoglobin receptor (HgbA)-deficient mutant and parent [10]. Control tissue was obtained from the upper arm of an uninfected healthy white woman (age, 35 years). Enrollment procedures and exclusion criteria were described elsewhere [6, 9, 11].

Human challenge protocol

The experimental human challenge protocol, preparation and inoculation of the bacteria, clinical observations, study design, biopsy techniques, and treatment of subjects were exactly as described elsewhere [6, 9, 11], except that the 5 suspensions containing live bacteria were placed in random order, given a code number, and inoculated at identical sites on each subject. To provide a reference for clinical evaluation, the heat-killed control bacteria were inoculated at a known site on each subject.

Phenotypes of recovered bacteria

Individual colonies were collected from the inocula, surface cultures, and biopsy samples; suspended in medium; and frozen in 96-well plates. The colonies were scored for susceptibility to chloramphenicol, as described elsewhere [10].

Confocal microscopy

Biopsy samples were fixed with 4% para-formaldehyde in PBS for 90–120 min and cryoprotected in 20% sucrose at 4°C overnight. Samples were embedded in optimal-cutting-temperature medium (Miles Laboratories, Elkhart, IN), frozen in liquid N₂, and cut into 10-μm sections with a cryostat. Sections were collected on Superfrost Plus microscope slides (Fisher Scientific, Pittsburgh) and stored at −20°C. Sections were permea-bilized with 0.2% Triton X-100 in PBS for 15 min, washed three times for 2 min in PBS, and blocked with 5% normal goat serum in PBS for 30 min. To detect bacteria, sections were stained with MAb 2D8 [4] and rabbit polyclonal anti-H. ducreyi antiserum [12] for 2 h, blocked with 5% normal goat serum in PBS for 30 min, and incubated with fluorescein isothiocyanate (FITC)-labeled goat anti-mouse IgG and indodicarbocyanine-5 (Cy5)-labeled goat anti-rabbit IgG (Jackson ImmunoResearch Laboratories, West Grove, PA) for 1 h. Sections were washed 3 times for 2 min in PBS after each incubation. Samples were mounted with Vectashield mounting medium (Vector Laboratories, Burlingame, CA) and examined on a confocal laser-scanning microscope (MRC 1024; Bio-Rad Laboratories, Richmond, CA). Images were collected separately for FITC and Cy5 signals, and the images were colorized and combined by use of Metamorph software to demonstrate areas of co-localizing signals. Negative controls included omission of the primary antibody, omission of the secondary antibody, and staining of sections of uninfected upper arm skin.

Results

In vivo expression of FtpA

To determine whether the fine tangled pili of H. ducreyi were expressed in vivo, we examined biopsy samples obtained from 3 subjects who participated in the HgbA-deficient mutant-parent trial [10]. The 3 biopsy samples were obtained from sites inoculated with an estimated delivered dose (EDD) of 48–60 cfu of the parent 35000 and were culture-positive for H. ducreyi and no other bacterial species. To visualize H. ducreyi in these lesions, sections were stained with polyclonal anti-H. ducreyi antiserum and Cy5-labeled secondary antibody and were then examined by confocal microscopy. In each biopsy sample, positively staining structures were present that had a morphology characteristic of H. ducreyi (figure 1), including rod-shaped cells occurring singly or in chains or clusters (data not shown). To determine whether the bacteria expressed FtpA, sections were doubly stained with 2D8 and polyclonal anti-H. ducreyi antiserum followed by FITC-labeled anti-mouse and Cy5-labeled anti-rabbit secondary antibodies. Images of positively staining bacteria were similar with either primary antibody (figure 1), and when the images were combined, the two antibody signals colocalized to the bacterial structures (figure 1C). No staining was observed when the primary antibody was omitted (data not shown). No positive signals were found in uninfected tissue stained with 2D8 or the polyclonal antiserum (data not shown). These data indicate that H. ducreyi expresses FtpA in vivo.

Figure 1

FtpA (major subunit of fine tangled pilus) expression in a biopsy sample taken from subject no. 102 and stained with rabbit polyclonal anti-Haemophilus ducreyi antiserum and monoclonal antibody 2D8. A, Bacteria stained with polyclonal antiserum, detected with indodicarbo-cyanine-5-labeled goat anti-rabbit IgG (red); B, bacteria stained with 2D8, detected with fluorescein isothiocyanate-labeled goat anti-mouse IgG (green). C, Combined image demonstrating colocalization (yellow/orange) of primary antibodies. Bar = 2.5 μm.

Construction of an isogenic ftpA mutant in the 35000HP background

The ftpA open-reading frame was insertionally inactivated in 35000HP exactly as reported elsewhere [4]. H. ducreyi 35000HP was transformed with pHD24 ftpA::mTn3(Cm) by electroporation. One transformant, 35000HP-SMS1, no longer bound the MAb 2D8 in Western blot analysis (data not shown). Southern blots of H. ducreyi 35000HP and 35000HP-SMS1 chromosomal DNA were probed with a 450-bp fragment of ftpA and with the cat cassette of pACYC184. The ftpA probe hybridized to a 6.5-kb AvaI fragment in the parent DNA and to an 8.1-kb AvaI fragment in the mutant DNA, as reported elsewhere [4]. The cat cassette probe did not hybridize to 35000HP DNA but did hybridize to an 8.1-kb AvaI fragment in the mutant (35000HP-SMS1) DNA, confirming that allele exchange had occurred in ftp A (data not shown). Whole cells of 35000HP and 35000HP-SMS1 were examined by use of electron microscopy. Many piliated cells of 35000HP were easily found, whereas no piliated organisms were seen in samples prepared from 35000HP-SMS1 (data not shown). The growth rates of 35000HP and 35000HP-SMS1 in broth were identical, as were the outer-membrane protein profile and the lipo-oligo-saccharide profiles (data not shown). To ensure that the mutation was stable, we passed the pilus mutant on nonselective chocolate agar plates 20 times during ∼7 weeks. Southern blot analysis showed that the cat insertion was stable in the absence of antibiotic selection.

Human inoculation experiments

An escalating dose-response study was used to compare the virulence of the mutant and the parent. In the first iteration, the EDD for 35000HP was 40 cfu, and it was 20, 40, and 80 cfu for 35000HP-SMS1. No lesions developed at sites inoculated with heat-killed bacteria. Papules developed at all 6 sites inoculated with the parent bacteria and at all 9 sites inoculated with the mutant bacteria. Skin lesions resolved at 1 parent site and 6 mutant sites (table 1). Pustules developed at 5 (83%) of 6 sites inoculated with the parent bacteria and at 3 (33%) of 9 sites inoculated with the mutant. One subject developed pustules at all sites inoculated with live bacteria. In 2 subjects, lesions resolved at all sites inoculated with mutant bacteria.

Table 1

Response to inoculation of live Haemophilus ducreyi 35000HP and 35000HP-SMS1 (mutant with inactivated gene for fine tangled pili).

The result of the first iteration suggested that the mutant's ability to cause pustules was partially impaired. A power analysis (80% at a 5% significance level) showed that the difference in the pustule formation rates (50%) would achieve significance if the rates remained the same when we inoculated a total of 12 sites with the parent and 19 sites with the mutant. We therefore infected 4 more subjects in the second iteration, and each subject was inoculated with an EDD of 20 cfu of 35000HP and with 20, 40, and 80 cfu of 35000HP-SMS1. No lesions developed at sites inoculated with the heat-killed control bacilli. Seven of 8 parent sites and 11 of 12 mutant sites developed papules (table 1). Lesions at 5 parent sites and 6 mutant sites resolved, and 1 papule remained at 1 mutant site in 1 subject at the end of the observation period. Pustules developed at 2 of 8 parent sites and at 4 of 12 mutant sites.

The cumulative results for both iterations showed that papules developed at 92.8% (95% confidence interval [CI], 66.1%– 99.8%) of sites inoculated with 35000HP and at 95% (95% CI, 76.2%–99.9%) of sites inoculated with 35000HP-SMS1 (P = 1.0). Pustules formed at 50% (95% CI, 23%–77%) of sites inoculated with 35000HP compared with 33% (95% CI, 14.6%–57%) of sites inoculated with 35000HP-SMS1 (P = .16). The days to pustule formation in sites inoculated with the parent or the mutant were similar (P = .408, log-rank test). Thus, no significant difference in pustule formation was observed when mutant and parent sites were compared.

Cellular infiltrate of lesions induced by mutant and parent isolates

A similar histologic pattern was present in biopsy samples obtained from the parent and mutant sites. Micropustules with polymorphonuclear leukocytes were present in the epidermis. The dermis contained a perivascular infiltrate of mono-nuclear cells and some polymorphonuclear leukocytes, and the venules were lined with reactive endothelial cells (data not shown). Most of the mononuclear cells were CD3-positive in both mutant and parent pustules (data not shown).

Recovery of bacteria from lesions

The recovery rate of H. ducreyi from cultures obtained from all sites inoculated with the live parent (n = 114) was 6% and with the live mutant (n = 171) was 6%. All biopsy samples were cultured semiquanti-tatively. Bacteria were recovered from 3 of 4 parent biopsy samples and from 4 of 4 mutant biopsy samples. The yields from parent biopsy samples ranged from 1.5 × 10³ to 1.2 × 10⁶ cfu/g of tissue (geometric mean = 2.2 × 10⁴), and the yields from mutant biopsy samples ranged from 1.7 × 10⁴ to 2.6 × 101.15⁵ cfu/g of tissue (geometric mean = 1.1 × 10⁵). The data suggest that mutant and parent bacteria replicated in lesions.

Confirmation of the phenotype of the recovered bacteria

To confirm that the inocula were correct and that no phenotypic changes occurred during infection, individual colonies from each of the 4 broth cultures used to prepare the inocula, from surface cultures, and from biopsy specimens were scored for chloramphenicol susceptibility. For the broth cultures used to prepare the inocula, all 72 parent colonies and 77 mutant colonies tested were phenotypically correct (mutant, Cm^r; parent, Cm^s). Sixty-one colonies obtained from surface cultures of parent sites and 127 colonies obtained from mutant sites had the expected phenotypes. Of biopsy samples that were culture-positive, all 50 parent colonies and 132 mutant colonies tested were correct. Thus, all colonies tested from the inocula, surface cultures, and biopsy samples had the expected phenotype.

Complications

Two subjects developed hypertrophic scars at the biopsy sites and were treated with intralesional injections of triamcinolone. The scars became flat after treatment.

Discussion

We demonstrated that H. ducreyi expresses FtpA in vivo. To examine the role of FtpA in the early stages of infection, we constructed a pilus mutant of H. ducreyi 35000HP. One transformant, 35000HP-SMS1, no longer bound MAb 2D8 in Western blot analysis. Southern blot analysis of H. ducreyi 35000HP and 35000HP-SMS1 DNA confirmed the insertion of the cat cassette in ftpA in 35000HP-SMS1. Electron microscopic examination did not reveal any piliated organisms in samples prepared from 35000HP-SMS1. The pilus mutant caused pustule formation at a rate similar to that of the isogenic parent in the human challenge model of chancroid.

H. ducreyi clumps both in vivo and in vitro, and it would be technically difficult to perform a mutant-parent comparison with a single dose of bacteria. To achieve statistical power, such comparisons would also require large numbers of volunteers. Therefore, we evaluated the ftpA mutant by means of an escalating dose-response study, as described elsewhere [9]. In the first iteration, pustules developed at 5 of 6 parent sites and at 3 of 9 mutant sites. The results suggested that the mutant's ability to cause pustules was partially impaired. A power analysis showed that the difference in the pustule-formation rates would be significant if the rates remained the same in an additional 4 subjects. In the second iteration, pustules developed at 2 of 8 parent sites and at 4 of 12 mutant sites. Cumulative results indicated that the pustule formation rate at mutant sites (33%) was similar to that at parent sites (50%) (P = .16). For the entire trial, the mean (± SD) EDD for the mutant (46 ± 25 cfu) was similar to that of the parent (28 ±10 cfu). The recovery rate of H. ducreyi from surface cultures and biopsy samples, and the cellular infiltrate in the lesions caused by the mutant and parent, were similar. Thus, no major differences were observed in the ability of the parent and the mutant to cause disease.

Although the clinical course of experimental infection mimics natural infection, the route of inoculation is artificial. The bacteria are mechanically introduced into the epidermis and dermis by the tines of the allergy-testing device. Although the model can discriminate between the ability of a parent and that of a mutant to form pustules [10], the route of inoculation may mask the potential role of an adhesin in pathogenesis. Similarly, in the temperature-dependent rabbit model, 10⁴–10⁵ cfu are injected intradermally into the back of a rabbit. In the rabbit model, the ftpA mutant causes ulcers at a rate similar to that of the parent [13]. Thus, although FtpA is expressed in vivo in humans, FtpA is not required for pustule formation in animal or human models of H. ducreyi infection.

Acknowledgments

We thank Ruben Sandoval for expert technical advice with the bacterial staining and confocal microscopy.

Informed consent was obtained from the subjects for participation and for HIV serology, in accordance with the human experimentation guidelines of the US Department of Health and Human Services and the institutional review board of Indiana University-Purdue University at Indianapolis.

Grant support: NIH (AI-27863, AI-31494, AI-75329, RR-00750; AI-09971 to M.E.B.).