Haemophilus ducreyi Lipooligosaccharide Mutant Defective in Expression of β-1,4-Glucosyltransferase Is Virulent in Humans

Abstract

The lipooligosaccharide (LOS) of Haemophilus ducreyi contains a major glycoform that is immunochemically identical to paragloboside, a glycosphingolipid precursor of major human blood group antigens. We recently identified the gene responsible for the glucosyltransferase activity and constructed an isogenic mutant (35000glu-) deficient in this activity. 35000glu- makes an LOS that consists only of the heptose trisaccharide core and 2-keto-deoxyoctulosonic acid (KDO). For this study, the mutant was reconstructed in the 35000HP (human passaged [HP]) background. Five human subjects were inoculated with 35000HP and 35000HPglu- in a dose-response trial. The pustule formation rates were 40% (95% confidence interval [CI], 13.7 to 72.6%) at 10 sites for 35000HP and 46.7% (95% CI, 24.8 to 69.9%) at 15 sites for 35000HPglu-. The histopathology and recovery rates of H. ducreyi from surface cultures and biopsies obtained from mutant and parent sites were similar. These results indicate that the expression of glycoforms with sugar moieties extending beyond the heptose trisaccharide core is not required for pustule formation by H. ducreyi in humans.

Haemophilus ducreyi causes the genital ulcer disease chancroid. Structural and immunochemical analyses have demonstrated that the principal glycoform of H. ducreyi lipooligosaccharide (LOS) shares common epitopes with the LOS of other mucosal pathogens, such as Neisseria gonorrhoeae, Neisseria meningitidis, and Haemophilus influenzae (12, 20, 26, 27, 36). The major oligosaccharide structure of H. ducreyi LOS (Galβ1–4-GlcNAcβ1–3Galβ1–4Hepα1–6Glcβ1–4Hepα1–5KDO) contains a terminal lactosamine (Galβ1–4-GlcNAc) and is similar in structure to paragloboside (Galβ1–4GlcNAcβ1–3Galβ1–4Glc), a precursor of the major human blood group antigens, I and i (11, 12, 19, 26). The terminal N-acetyllactosamine of the major glycoform of H. ducreyi LOS is modified by sialic acid, much like the mature human I and i antigens (26). The LOS is thought to help H. ducreyi evade the host immune response by mimicking human antigens or by facilitating adherence to and/or invasion of host cells by binding to human cell surface receptors for glycosphingolipids or sialic acid.

Several lines of evidence suggest that H. ducreyi LOS plays a role in the pathogenesis of chancroid. Injection of purified LOS causes intradermal inflammation in experimental animal models (13). Purified LOS induces interleukin-8 (IL-8) expression from keratinocytes in vitro and may stimulate an inflammatory response that leads to lesion formation (46). Mutants whose LOS consisted only of 2-keto-deoxyoctulosonic acid (KDO) or KDO and heptose were attenuated in the temperature-dependent rabbit model of infection, but these mutants also had altered outer membrane protein (OMP) profiles (5, 6). A mutant with a disruption in the d-glycero-d-manno-heptosyltransferase gene exhibited reduced adherence and invasion of human keratinocytes in vitro (19). However, the d-glycero-d-manno-heptosyltransferase mutant and a sialyltransferase mutant were virulent in the human challenge model of infection, indicating that expression of an LOS lacking sialic acid or an LOS consisting of 6Glcβ1–4Hepα1–5KDO is sufficient for pustule formation by H. ducreyi in humans (45).

We recently identified the gene responsible for glucosyltransferase activity (lgtF) and constructed an isogenic mutant (35000glu-) deficient in the expression of LgtF (17). 35000glu- makes an LOS that consists only of the heptose trisaccharide core and KDO and lacks all the terminal sugars found on human glycosphingolipids. Since mammalian cells don't express heptose residues, 35000glu- could be recognized as foreign and eradicated more rapidly than wild-type H. ducreyi or not bind to host receptors for glycosphingolipids.

Here, we tested the hypothesis that an isogenic glucosyltransferase-deficient H. ducreyi mutant is attenuated in the human model of infection. We constructed a new isogenic mutant (35000HPglu-) by insertion of a cat (chloramphenicol acetyltransferase) cassette in the β-1,4-glucosyltransferase gene (lgtF). The virulence of 35000HPglu- was tested in a double-blinded, escalating dose-response study. We compared the papule and pustule formation rates, the cellular infiltrate, and recovery of bacteria from lesions inoculated with the mutant and the parent.

Construction of an lgtF mutant.

H. ducreyi 35000HP is a human-passaged variant of 35000, described previously (3, 37). H. ducreyi 35000HPglu- was constructed identically to 35000glu- (17). Briefly, 35000HP was electroporated with pMJFglu and transformants were selected on chloramphenicol- and X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside)- containing plates (10). Chloramphenicol-resistant (Cm^r) transformants that grew normally (large white colonies) in the presence of X-Gal were further characterized by Southern hybridization. A transformant that had undergone allelic replacement in lgtF was designated 35000HPglu-.

LOS and OMPs were prepared from 35000HP and 35000HPglu- as described previously (29, 45). OMPs were subjected to analysis by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) in 12.5% acrylamide gels as described previously (29, 45). LOS was analyzed by SDS-PAGE in 14% gels and silver staining as previously described (45).

Human challenge protocol.

Adult volunteers in good health and over 18 years of age were recruited for the study. In accordance with the human experimentation guidelines of the Institutional Review Board of Indiana University Purdue University Indianapolis and the U.S. Department of Health and Human Services, informed consent was obtained from the subjects for participation in the study and for human immunodeficiency virus serology. The experimental challenge protocol, preparation and inoculation of the bacteria, and clinical observations were done as described previously (3, 37, 38). Although we did not determine the actual delivered dose, the estimated delivered dose (EDD) was calculated based on the CFU loaded on the tines of the inoculation device and the delivery characteristics of the device for antigenic solutions in human skin and bacterial suspensions in swine skin, as previously described (23, 34, 37, 41).

A modification of an escalating dose response study was used to compare the virulence of 35000HP and 35000HPglu- as described previously (2, 18, 41). Each subject was infected at six sites. On one arm, three sites were inoculated with twofold serial dilutions of the mutant. On the other arm, two sites were inoculated with the parent and one site was inoculated with the highest dose of the heat-killed mutant. To blind the study, the six suspensions containing bacteria were placed in random order, given a code number, and inoculated at identical sites on each subject in each iteration. The physicians who evaluated the subjects were unaware of the identity of the suspensions. Subjects were observed until they reached a clinical end point, defined as either development of a painful pustule, resolution of infection at all sites, or 14 days after inoculation. When the end point was achieved, the code was broken and up to two sites with active disease (one inoculated with the parent and one with the mutant), if present, were biopsied. The subjects were then treated with antibiotics as described (2).

Each biopsy was cut into portions. One portion was fixed in formalin and used for immunohistological studies as described elsewhere (28, 37, 38). The slides were coded and evaluated by a dermatopathologist, who was unaware of the code. One portion was semiquantitatively cultured as described (37, 38). Individual colonies from the inocula, surface cultures and biopsies were picked, suspended in freezing medium, and frozen in 96-well plates. Colonies were scored for susceptibility to chloramphenicol on chocolate agar plates.

Characterization of 35000HPglu-.

In Southern blotting, genomic DNA from 35000HP and 35000HPglu- were digested with HindIII and probed with the lgtF open reading frame as well as the cat cassette. The lgtF probe bound to a 6-kb DNA fragment in the parent and a 7-kb DNA fragment in the mutant. The cat probe did not bind to 35000HP DNA but did bind to a 7-kb DNA fragment in 35000HPglu- (data not shown). 35000HP and 35000HPglu- cells had similar growth rates in broth (data not shown). OMPs and LOS prepared from 35000HP and 35000HPglu- were analyzed by SDS-PAGE. As expected, 35000HPglu- LOS migrated more rapidly in SDS-PAGE than 35000HP (data not shown). Both isolates had similar OMP profiles (data not shown).

Evaluation of 35000HPglu- in human subjects.

Four men and three women (one black, six white; age range, 22 to 51 years; age mean ± SD, 33.0 ± 11.7) with no history of chancroid enrolled in the study. Two subjects withdrew on the day of inoculation. Three subjects (176, 177, and 180) were challenged in the first iteration, and two subjects (184 and 185) were challenged in a second iteration (Table 1).

TABLE 1.

Responses to inoculation of live H. ducreyi strains^a

Subject no.	Days of observation	Isolate	No. of initial papules	Final outcome of initial papule
				No. of papules	No. of pustules	No. resolved
176	13	35000HP	2	0	1	1
		35000HPglu-	3	0	2	1
177	7	35000HP	2	0	2	0
		35000HPglu-	3	0	2	1
180	13	35000HP	2	0	0	2
		35000HPglu-	2	0	1	1
184	13	35000HP	2	0	0	2
		35000HPglu-	3	0	1	2
185	14	35000HP	1	0	1	0
		35000HPglu-	1	0	1	0

^aEach volunteer was inoculated at two sites with the parent (35000HP) and at three sites with the glucosyltransferase mutant (35000HPglu-). 

The EDDs in the first iteration were 77 CFU for 35000HP and 34, 69 and 137 CFU for 35000HPglu-. Papules developed at six of six sites inoculated with the parent and at eight of nine sites inoculated with the mutant. At the end point, pustules were present at three of six parent sites and five of nine mutant sites.

Since inoculation of both the mutant and the parent caused pustules at similar rates, we continued the experiment with similar target doses. In the second iteration, two subjects were inoculated with an EDD of 77 CFU of 35000HP and 18, 36, and 72 CFU of 35000HPglu-. Papules developed at three of four sites inoculated with the parent and at four of six sites inoculated with the mutant. At the end point, one of four parent sites and two of six mutant sites contained pustules.

Overall, the pustule formation rates were 40% (exact binomial 95% confidence interval [CI], 13.7 to 72.6%) at 10 sites for 35000HP and 46.7% (exact binomial 95% CI, 24.8 to 69.9%) at 15 sites for 35000HPglu-. Thus, expression of glucosyltransferase was not required for pustule formation.

Surface cultures were obtained from all inoculation sites at each follow-up visit. No bacteria were recovered from sites inoculated with the heat-killed control. H. ducreyi was recovered intermittently from parent and mutant sites in two of the subjects. Overall, the recovery rate was 8% from sites inoculated with the parent (n = 49) and 1% from the mutant (n = 83) (two-sided Fisher's exact test, P = 0.063). Bacteria were recovered from three of three parent sites and four of five mutant sites that were biopsied. The yields of 35000HP and 35000HPglu- from biopsy cultures that were positive ranged from 2.4 × 10⁵ to 1.4 × 10⁷ CFU/g of tissue and 1.4 × 10⁴ to 1.6 × 10⁶ CFU/g of tissue, respectively. Thus, the numbers of bacteria recovered from mutant and parent biopsies were similar.

We examined the cellular infiltrate in three parent and four mutant sites that were present at the end point. The histopathology for both mutant and parent biopsy specimens were indistinguishable. In biopsy specimens obtained from both the parent and mutant sites, the dermis contained a perivascular and perifollicular infiltrate of mononuclear cells and some polymorphonuclear leukocytes (PMNs) and the venules were lined with reactive endothelial cells. The majority of the mononuclear cells were stained with a CD3 marker and were predominantly in the mid-reticular dermis in a perivascular location (data not shown). Micropustules with PMNs were present in the epidermis.

To confirm that the inocula were correct and that we had inoculated the sites as intended, individual colonies from the cultures used to prepare the inocula, surface cultures, and biopsy cultures were analyzed for antibiotic susceptibility as described previously (45). All colonies tested from the inocula (n = 80 parent and 80 mutant), surface cultures (n = 160 parent and 3 mutant), and biopsies (n = 128 parent and 178 mutant) were correct.

Conclusions.

We had previously evaluated two LOS mutants in the human model of H. ducreyi infection. 35000HP-RSM203 is incapable of sialylating its LOS, while 35000HP-RSM2 has a major glycoform that terminates in a single glucose attached to a heptose trisaccharide core and KDO. Surprisingly, both mutants formed pustules at the same rate as 35000HP in humans (45). In this study, we showed that a third isogenic LOS mutant, 35000HPglu-, caused papules and pustules at rates similar to those of its parent. Thus, expression by H. ducreyi of an LOS that consists only of a heptose trisaccharide core and KDO and lacks any structures homologous to those found in human paragloboside is sufficient for virulence in humans.

The major limitations of the human challenge model are the artificial route of inoculation and that we are permitted to infect subjects until they develop painful pustules or 14 days have passed. In subjects who achieve the clinical end point in 6 to 7 days, approximately 2 × 10⁵ CFU are present in an entire pustule, suggesting that the bacteria replicate for a minimum of 10 to 12 generations in the model (42). Despite these constraints, isogenic mutants with mutations in hgbA, pal, and dsrA are impaired in their ability to form pustules even with inocula that are 10-fold higher than those of the parent (2, 9, 18). Thus, the model is an appropriate test of virulence. However, we cannot examine the role of virulence determinants in the pathogenesis of ulcers or lymphadenitis and cannot exclude the possibility that expression of parental oligosaccharide is important in the later stages of disease.

In vitro, H. ducreyi LOS induces IL-8 expression in keratinocytes and facilitates adherence to and invasion of keratinocytes (19, 46). Lipid A probably induces IL-8 expression (31), while the oligosaccharides are responsible for attachment and invasion (19). In the human challenge model, the bacteria are deposited in puncture wounds made by the tines of the applicator. Although the majority of the dose is delivered to the dermis, bacteria are delivered to the epidermis and should be able to interact with keratinocytes (7). By confocal microscopy, micropustules are present in the epidermis of papules within 24 h of infection (7). Bacteria are not seen at 24 h, probably due to the low EDD. At 48 h, the bacteria are seen in the epidermal micropustules and in the dermis (7). The bacteria do not attach to or invade keratinocytes throughout the papular and pustular stages (7, 8). Taken together, the data suggest that H. ducreyi is rapidly surrounded by PMNs and sequestered from physically interacting with keratinocytes within 24 h. The principal function of H. ducreyi LOS during the initial stages of infection may be induction of IL-8 expression and pustule formation rather than attachment or invasion. Truncations in the oligosaccharides of the LOS should not affect the ability of the organism to recruit PMNs and therefore may not affect lesion formation.

The fact that truncations in LOS do not affect the virulence of H. ducreyi was extremely surprising. Other pathogenic Haemophilus and Neisseria sp. strains express LOS structures that are similar to those expressed by H. ducreyi (reviewed in reference 31), and mutations or variations in their LOS frequently affect their virulence in animal or human models. For example, LOS mutants or variants of H. influenzae type b or H. influenzae biogroup aegyptius are less virulent in rat bacteremia models (14, 25, 33, 43). In a chinchilla model of otitis media, a LOS mutant of nontypeable H. influenzae was cleared from the middle ear even at doses 4 logs higher than that of the parent (15). Signature-tagged mutagenesis of N. meningitidis shows that expression of several LOS biosynthesis genes responsible for assembly of paragloboside-like structures is required for virulence in a rat model of bacteremia (40). In the human challenge model of N. gonorrhoeae infection in humans, inoculation of a variant that produces a truncated LOS reverts to expression of full-length LOS in vivo (35). In natural gonococcal infection, LOS is frequently sialylated (4). In contrast, H. ducreyi does not seem to need to express an LOS beyond the heptose trisaccharide core to cause infection, suggesting that its biology is distinct from that of these other pathogens.

Two factors that are important in the pathogenesis of H. ducreyi infection in the human challenge model are its ability to evade phagocytosis by macrophages and PMNs and its ability to escape complement-mediated killing (7, 9). For some strains of N. gonorrhoeae, expression of sialylated LOS confers resistance to complement-mediated killing by immune and nonimmune sera (21, 30, 31, 44). However, the LOS of H. ducreyi plays a minimal role in serum resistance, which appears to be mediated by the OMP DsrA (16, 22, 39). Sialylation of N. gonorrhoeae LOS is associated with resistance to opsonophagocytosis by PMNs (21, 24, 31, 32) and a decreased ability to stimulate an oxidative burst (32). The factors responsible for evasion of phagocytosis and phagocytic killing by H. ducreyi are not known. It has been suggested that H. ducreyi produces a capsular polysaccharide (1), and several putative glycosyltransferases have been identified by inspection of the genome; however, no alternative surface polysaccharides have been identified (unpublished observations). The results of this trial suggest that truncations in LOS do not affect the ability of the organism to evade the phagocytic response. We speculate that the ability of H. ducreyi to resist phagocytosis and phagocytic killing resides in OMPs and/or secreted products.

In summary, we have conclusively shown that expression of H. ducreyi LOS beyond the triheptose core is not required for pustule formation in human volunteers. Future studies will focus on examination of the role of lipid A in PMN recruitment and pustule formation and identification of factors responsible for the evasion of the host response.