Multiple Class I and Class II Haemophilus ducreyi Strains Cause Cutaneous Ulcers in Children on an Endemic Island

Haemophilus ducreyi is a major cause of skin ulcers in the tropics. On an endemic island, multiple strains of H. ducreyi cause infection, coinfections are common, and mass treatment with azithromycin did not exert selection pressure on the organism.

Abstract

Background

Together with Treponema pallidum subspecies pertenue, Haemophilus ducreyi is a major cause of exudative cutaneous ulcers (CUs) in children. For H. ducreyi, both class I and class II strains, asymptomatic colonization, and environmental reservoirs have been found in endemic regions, but the epidemiology of this infection is unknown.

Methods

Based on published whole-genome sequences of H. ducreyi CU strains, a single-locus typing system was developed and applied to H. ducreyi–positive CU samples obtained prior to, 1 year after, and 2 years after the initiation of a mass drug administration campaign to eradicate CU on Lihir Island in Papua New Guinea. DNA from the CU samples was amplified with class I and class II dsrA-specific primers and sequenced; the samples were classified into dsrA types, which were geospatially mapped. Selection pressure analysis was performed on the dsrA sequences.

Results

Thirty-seven samples contained class I sequences, 27 contained class II sequences, and 13 contained both. There were 5 class I and 4 class II types circulating on the island; 3 types accounted for approximately 87% of the strains. The composition and geospatial distribution of the types varied little over time and there was no evidence of selection pressure.

Conclusions

Multiple strains of H. ducreyi cause CU on an endemic island and coinfections are common. In contrast to recent findings with T. pallidum pertenue, strain composition is not affected by antibiotic pressure, consistent with environmental reservoirs of H. ducreyi. Such reservoirs must be addressed to achieve eradication of H. ducreyi.

In endemic regions of Africa and the South Pacific, 5%–15% of children have disfiguring exudative cutaneous ulcers (CUs). Approximately 100 million children are at risk for CU, and 100000 cases are reported annually to the World Health Organization (WHO) [1, 2]. Although CU are usually attributed to Treponema pallidum subspecies pertenue, or yaws, molecular testing shows that Haemophilus ducreyi is also a major cause of CU [3–8].

A single oral dose of azithromycin (Az) is effective in the treatment of yaws [9], raising the possibility that mass drug administration (MDA) of oral Az could eradicate yaws. Because H. ducreyi CU strains are susceptible to Az [10, 11] and Az is effective in the treatment of H. ducreyi–associated CU [12], MDA could also eradicate H. ducreyi. In the WHO yaws eradication program, MDA is followed every 6 months by case finding, testing of ulcer swabs, and treatment of new CU cases and household contacts with Az [3]. Following MDA on Lihir Island in Papua New Guinea, the overall prevalence of CU in children fell from approximately 10% to 1.3% but remained at 1.6% after 42 months of follow-up [13, 14]. The overall number of CU cases with T. pallidum pertenue DNA declined but then significantly increased between 18 and 42 months [14]. Although 8 T. pallidum pertenue genotypes were detected, by 24 months only 1 genotype remained, suggesting that the program reduced strain diversity by interrupting transmission [14, 15]. Throughout the study, the percentage of ulcers due to H. ducreyi was stable [14]. The failure to eradicate H. ducreyi may be due to colonization of the skin of asymptomatic children, flies, and bed linens, which allows escape from antimicrobial pressure [16]. In contrast, asymptomatic colonization of the skin by T. pallidum pertenue is rare [16].

Haemophilus ducreyi also causes chancroid, a genital ulcer (GU) disease that occurs in Africa and Asia [17]. GU strains are phylogenetically differentiated into 2 clades, called class I and class II, which diverged from each other approximately 1.95 million years ago [10, 18–21]. Phylogenetic analysis of the whole-genome sequences of CU strains shows that both class I and class II strains of H. ducreyi cause CU [10, 11, 22, 23]. A multilocus typing system based on dsrA, hgbA, and ncaA recapitulates the phylogenetic tree based on whole-genome sequencing (WGS) [23]. These data suggest that a multilocus typing system could be used for epidemiologic investigations of CU strains, show that both classes of strains cause CU within an endemic country, and raise the possibility that dual infections with both classes of H. ducreyi could occur.

Here we examined which components of the multilocus typing system are required to discriminate between CU strains of H. ducreyi. Using this information, we typed H. ducreyi–positive CU samples obtained prior to, 1 year after, and 2 years after MDA on Lihir Island. We examined whether class I and class II CU strains circulate on the island and whether dual infections occurred. Finally, due to the presence of environmental reservoirs, we tested the hypothesis that the yaws eradication program did not exert selection pressure and/or change the composition or distribution of circulating H. ducreyi strains over time.

METHODS

Examination of dsrA, hgbA, and ncaA for Polymorphisms Among Sequenced CU Strains

The sequences of hgbA, ncaA, and dsrA were trimmed from the complete genomes of the class I reference strain 35000HP (GenBank accession number NC_002940.2) and the class II reference strain CIP542 (GenBank accession number CP011229). The Basic Local Alignment Search Tool (https://blast.ncbi.nlm.nih.gov/Blast.cgi) was used to align the gene sequences to each of the 17 previously sequenced CU strain genomes (GenBank accession numbers CP011218–CP011221, CP011227, NZ_LMZZ01000014.1, and CP015424–CP015434). The aligned sequences were analyzed for single-nucleotide polymorphisms (SNPs). For hgbA and ncaA, we compared the entire gene sequences. As there is little homology in dsrA between class 1 and class 2 strains, we examined regions of dsrA that are amplifiable with the class 1– and class 2–specific primers described below. For class I strains, the primers produce an amplicon that is 441 base pairs (bp) in length. Due to the presence of an unstable repeat region at the 3ʹ end of this amplicon, we restricted our analysis to 363 bp (positions 149–511, which correspond to complete codons 50–169) of the 35000HP reference strain [24]. For class II strains, the primers produce a 711 bp amplicon. The analysis was restricted to bp 78–745 and complete codons 26–247 of the CIP542 reference strain. Manual inspection of the alignments was performed to determine which of the alleles distinguished between the CU strains.

Ethics Statement

All participants, or their parents or guardians, provided oral informed consent to be screened and treated during the yaws eradication campaign; written informed consent was obtained from all patients with lesions before enrollment [3, 13]. The protocol was approved by the National Medical Research Advisory Committee of the Papua New Guinea Ministry of Health (MRAC number 12.36) [3, 13].

Samples

Swabs of CU were taken before (April 2013), 1 year after (May 2014), and 2 years after (May 2015) MDA, as described [3]. The swabs were placed in a lysis/transport buffer that stabilizes DNA. Samples obtained in 2013 were sent to the Molecular Diagnostic Unit at Queensland Royal Brisbane and Women’s Hospital (Australia) for PCR detection of H. ducreyi DNA and to the University of Washington (Seattle) for detection of T. pallidum pertenue DNA [3]. Samples obtained in 2014 and 2015 were tested for both H. ducreyi and T. pallidum pertenue DNA at the University of Washington. The samples were classified as being positive for H. ducreyi DNA, for T. pallidum pertenue DNA, for both, or for neither DNA. Of 272 ulcer samples, 123 contained H. ducreyi DNA and 29 contained both H. ducreyi and T. pallidum pertenue DNA. As most of the dual-positive samples were exhausted due other analyses, they were excluded from this study. Of the 123 samples, 117 were recovered and used for H. ducreyi typing (Figure 1). DNA was stored at –80°C until PCR was performed.

Figure 1.

Sample flowchart for Haemophilus ducreyi–positive specimens from Lihir Island and overall results of dsrA typing. The dsrA types are defined in Table 1; the numbers of positive samples for each type are in parentheses. An asterisk (*) denotes a novel type. Abbreviation: MDA, mass drug administration.

DNA Amplification

All specimens underwent PCR with primers specific for class I dsrA, based on the sequence of strain 35000HP, using the primers: sense, 5ʹAGGGTAAATGGACTTGGTCTAATG3 ʹ; antisense 5ʹTGGCTAAACCAGTTTGCAATTC3ʹ. Ten to 15 µL of each specimen was used as a template in the PCR reaction, which was performed with the Roche Fast Start High Fidelity System. The PCR conditions were as follows: initial denaturation at 94°C for 4 minutes, followed by 35 cycles of 94°C denaturation for 1 minute, 53°C annealing for 1 minute, 72°C elongation for 1 minute, and final extension for 7 minutes. To reduce the likelihood of contamination, experiments were conducted under PCR-clean conditions; pipettes, tips, tubes, and buffers were ultraviolet-irradiated for 15 minutes. For each batch of specimens amplified, a negative control with no DNA was processed in parallel. Bidirectional sequencing of each PCR product was performed by Eurofins Genomics. If sequencing yielded null or uninterpretable data, a second round of PCR, using conditions identical to those described above, was performed on the first-round PCR products. Negative controls from the first round were also amplified in the second round. If, after the second round of amplification, the sequencing produced null or ambiguous data, the specimen was declared to be negative for class I H. ducreyi.

After class I dsrA amplification and sequencing had been completed on all samples, the specimens underwent PCR for Class II dsrA. class II dsrA primers, based on the published sequence of strain CIP542, were sense, 5ʹGGCATCAAACGGCTCTTTATC3 ʹ; antisense, 5ʹGCTAACGCACTCTTACCTCTAT3 ʹ. Negative controls, PCR conditions, DNA sequencing, and data interpretation were identical to those described for class I strains.

Nucleotide Alignment and Comparison

We used Sequencher 5.4.6 (Gene Codes Corporation, Ann Arbor, Michigan) to align each ulcer specimen dsrA sequence with its respective dsrA reference sequence. 35000HP and CIP542, respectively, served as the class I and class II reference strains. Differences in dsrA sequences among the reference and sequenced CU strains were used to classify each specimen into a dsrA type. Sequences were submitted to GenBank under accession numbers MG953427–MG953516.

Geospatial Mapping

Lihir Island maps were obtained from the Newcrest Mining Department of Community Relations. Three maps of strain distribution were created for the time points before, 1 year, and 2 years after MDA. Village information for each specimen was recorded during collection. Samples with multiple detected strains were mapped as multiple nodes for each strain present. Nodes were manually placed upon the maps at the village locations for each strain.

Selection Pressure Analysis

Given the large genetic distance between the 2 H. ducreyi classes, we generated multiple alignments for each collection time for each dsrA class type using Sequencher 5.4.6 (Gene Codes Corporation). For example, for the data collected before MDA, we created an alignment including the 7 class I.3 types and the 8 class I.4 types, and another separate alignment including 12 class II.3 types and 1 class II.1 type. Similar alignments were created for the 1-year and 2-year time periods after MDA. Alignments were trimmed to include only complete codons. The class I sequence alignment for the samples collected before MDA included 15 strains and 360 bp (120 codons). The alignments for the class I samples collected 1 year and 2 years after MDA were each 363 bp (121 codons) with 19 and 16 strains, respectively. The class II alignments included 666 bp (222 codons) with 13 samples in the before MDA dataset, 20 samples in the 1 year after MDA dataset, and 7 samples for the 2 years after MDA dataset.

Site-specific selection pressure analysis was performed using 5 algorithms: fixed-effects likelihood (FEL) [25], fast unconstrained Bayesian approximation (FUBAR) [26], internal fixed-effects likelihood (IFEL) [27], mixed-effects model of episodic selection (MEME) [28], and single-likelihood ancestor counting (SLAC) [25] using the HyPhy software package [25] as implemented through the Datamonkey server (www.datamonkey.org). The universal genetic code was used for all datasets. The Model Selection tool was used to identify the best-fit model for each dataset. The significance level was set at P < .05 for the FEL, IFEL, MEME, and SLAC analyses; a posterior probability of >0.9 was used for FUBAR. A site was considered to be under selection if it was identified as such by >1 of the algorithms [29–31]. This was done to account for the uncertainty in whether random-effects models or counting methods are more reliable for detecting selection [32–34].

RESULTS

Development of a Single-Locus Typing System

We had shown that a multilocus typing system based on dsrA, ncaA, and hgbA yielded a phylogenetic tree of class I and class II GU and CU strains that was similar to that obtained by WGS [21, 23]. We examined which of these alleles or portions of these alleles were required to discriminate among the 17 previously sequenced CU strains and the reference strains 35000HP and CIP542.

Within the 14 previously published class I sequenced CU strains, there were 3 SNPs within ncaA and no SNPs within hgbA that differed among the CU strains and 35000HP. Within the 363 bp region of dsrA, 19 SNPs and a variable region corresponding to positions 269–285 of 35000HP were documented (Supplementary Table 1). Within the class I CU strains and 35000HP, variation in dsrA alone segregated the strains into 7 types (Supplementary Table 1). Strain 35000HP was unique and was designated type I.1; GHA8 and GHA9 had identical sequences and were designated type I.2; AUSPNG1 was unique and called I.3; NZS4, NZV1, VAN 1, VAN3, VAN4, VAN5 were identical and designated I.4; NZS2 and NZS3 were identical and called I.5; NZS1 was unique and designated I.6; GHA3 and GHA5 were identical and called I.7. The ncaA SNPs did not add any discrimination to dsrA in segregating these strains (data not shown).

Within the 3 previously published class II CU strains, there were 9 SNPs in dsrA (Supplementary Table 2), 2 SNPs in ncaA, and 2 SNPs in hgbA that distinguished the class II CU strains from CIP542. Strain CIP542 had a unique dsrA sequence and was designated as type II.1; GHA1 and GHA2 had the same dsrA sequences and were called II.2; VAN2 was unique and called II.3. The ncaA SNPs did distinguish between GHA1 and GHA2, but the hgbA SNPs did not add any discrimination to dsrA in classifying these strains (data not shown). Given that variations in ncaA and hgbA added little to variations in dsrA in distinguishing among the CU strains, we used dsrA as a cost-efficient single-locus typing system for the H. ducreyi–positive samples from Lihir island.

Sample Typing

Of the 117 H. ducreyi–positive samples, 77 yielded 90 dsrA sequences that could be unambiguously aligned with class I, class II, or both classes of CU strains (Figure 1). Of the 77 specimens, 24 were obtained prior to MDA; 33 were obtained 1 year and 20 were obtained 2 years after MDA. Thirty-seven samples yielded class I amplicons, 27 yielded class II amplicons and 13 yielded both (Figure 1). Polymorphisms found between bp 149 and bp 511 were used to type the class I samples; polymorphisms found between bp 79 to bp 744 were used to type the class II samples (Table 1). Overall, there were 50 class I strains comprised of 5 types including 1 new type (I.8) not present in the previously sequenced CU strains (Figure 1 and Table 1). Within the 40 strains identified as class II, there were 4 types, 2 of which were not present in the previously sequenced strains (Figure 1 and Table 1).

Table 1.

Polymorphisms in dsrA Used to Type the Haemophilus ducreyi–Positive Clinical Samples

Class I
Type	152	170	175	193	196	209	215	226	227	266	269	271	271.1	272	272.1	272.2	273	274	275	277
I.1	G	T	G	A	G	A	C	G	A	C	A	G	–	C	–	–	T	C	C	G
I.2	A	C	●	●	●	●	T	A	●	●	●	●	●	●	●	●	●	●	●	●
I.3	A	C	●	●	●	●	T	A	●	●	C	●	●	A	●	●	C	G	●	C
I.4	A	C	A	●	●	●	T	A	●	●	C	●	●	A	●	●	C	G	T	C
I.5	A	C	A	●	●	●	T	A	●	●	C	●	A	●	T	T	●	●	●	C
I.6	A	C	A	●	●	●	T	A	●	●	C	●	●	–	●	●	A	●	G	–
I.7	●	C	●	●	A	G	●	A	G	A	●	C	●	●	●	●	●	T	●	–
I.8*	A	C	A	G	●	●	T	A	●	●	C	●	●	A	●	●	C	G	T	C
Type	278	279	280	281	282	283	284	285	290	292	299	313	405	411	427	460	469	476	491
I.1	G	C	G	T	T	T	C	T	G	C	T	T	A	T	A	T	G	G	A
I.2	A	●	●	C	●	C	●	●	A	T	●	●	T	●	●	●	●	●	G
I.3	C	T	C	C	●	C	●	●	●	T	●	●	T	●	●	●	●	●	G
I.4	C	T	C	C	●	C	●	●	●	T	●	●	T	●	●	C	●	●	G
I.5	C	T	C	C	●	C	●	●	●	T	●	●	T	●	●	C	●	●	G
I.6	T	●	–	C	●	C	●	●	●	T	●	●	T	●	●	C	●	●	G
I.7	●	●	–	–	A	A	A	A	●	A	G	C	T	●	C	●	A	A	G
I.8*	C	T	C	C	●	C	●	●	●	T	●	●	T	C	●	C	●	●	G
Class II
Type	125	214	216	273	319	344	346	358	359	380	383
II.1	G	G	T	T	C	T	A	T	A	G	G
II.2	●	●	G	A	A	●	G	C	G	●	T
II.3	●	A	●	A	A	●	G	C	●	A	T
II.4*	●	●	●	●	●	C	●	●	●	●	●
II.5*	A	A	●	A	A	●	G	C	●	A	T

Data represent single-nucleotide polymorphisms in the dsrA amplicons that were used to type the clinical samples. For the class I types, the numbers correspond to the bp of the dsrA allele of 35000HP (type I.1); for the class II types, the numbers correspond to the bp of the dsrA allele of CIP542 (type II.1). A dot (● ) indicates a match to the reference sequence and a dash (–) indicates that that position does not exist in the sample. An asterisk (*) denotes a novel type.

Strain Composition and Geospatial Mapping

Prior to MDA, 2 class I dsrA types (I.3 and I.4) and 1 class II type (II.3) were predominant on the island (Figure 2). These types remained predominant throughout the 24-month period and accounted for approximately 18%, 29%, and 40% of the 90 types detected, respectively (Figure 2). Another class II type (II.1) was detected in 1 sample prior to and 1 year after MDA but was not detected thereafter (Figure 2). In a small number of samples, types I.1, I.8, and II.4 were detected only 1 year after MDA, whereas types I.5 and II.5 were detected only 2 years after MDA. Geospatial mapping showed that the 3 predominant types (I.3, I.4, and II.3) were dispersed over the island throughout the study period (Figure 3).

Figure 2.

Percentage of Haemophilus ducreyi dsrA types detected on Lihir Island prior to, 1 year after, and 2 years after mass drug administration (MDA). For dual infections, both types were included.

Figure 3.

Geospatial mapping of Haemophilus ducreyi strains on Lihir Island before mass drug administration (MDA) (A), 1 year (B), and 2 years after MDA (C). If a sample contained >1 dsrA type, both types were mapped to the same village.

Selection Pressure Analysis

F81 was found to be the best model for all datasets except the class I samples collected 1 year after MDA; HKY85 was most appropriate for this dataset. Selection was detected in 2 of the 6 datasets by the FUBAR algorithm only; thus, no sites were considered to be under selection pressure.

DISCUSSION

We found that the single-locus typing scheme is sufficient to discriminate among CU strains, that at least 9 different types of H. ducreyi have been present on Lihir Island, and that coinfections with both classes are common, occurring in approximately 17% of 77 cases. Direct WGS of 21 CU samples also shows that class I and II strains circulate in Ghana and the Solomon Islands, but coinfections were not detected in this smaller study [35].

Prior to MDA, 3 types (I.3, I.4, II.3) were predominant on the island. These strains persisted throughout the observation period. Geospatial mapping showed that the predominant strains were dispersed over the entire island throughout the 2-year time period. One type (II.1) was present prior to MDA and 1 year later, while 5 types (I.1, I.5, I.8, II.4, and II.5) were detected at a single time point after MDA. As only a convenience sample (90 of 690 CU cases) was tested at baseline [3, 13] and samples that were positive for both H. ducreyi and T. pallidum pertenue DNA were excluded from this study, the types that were detected only after MDA could have been present in low levels prior to MDA. Alternatively, the types that were detected only after MDA could have been introduced onto Lihir Island by travel of persons from other endemic islands that were not part of the clinical trial. Overall, approximately 87% of the types detected corresponded to strains that were isolated from the South Pacific previously; no types corresponding to Ghanaian strains were detected.

Limitations of the study include that fact that dsrA sequences were found in only 66% of the 117 H. ducreyi–positive samples; repeated freeze thawing of the samples likely limited our ability to recover unambiguous sequences. As we defined strain types by variations in a single gene, there may be more variability in the circulating strains than what we detected.

Selection pressure analyses indicate that MDA of Az and subsequent treatment of CU cases with Az every 6 months exerted slight or no changes in the composition of the H. ducreyi strains during the 2-year observation period. These negative results can be attributed to several reasons: 2 years is a short time frame for selection pressure to become evident in a slowly evolving organism; biannual treatment of patients with CU and their contacts with Az results in very limited pressure on the organism; for selection pressure analysis, our data set was relatively small. Azithromycin is predominantly concentrated intracellularly in fibroblasts [36]; colonizing strains that are on the surface of the skin are likely to escape Az pressure and cause new infections. Finally, given its mechanism of action, Az may not exert pressure on the dsrA gene, whose gene product confers serum resistance and binding to fibrinogen, vitronectin, and fibronectin to the organism [37, 38].

In summary, multiple strains of H. ducreyi cause CU in an isolated endemic community and coinfections are common. In contrast to the reduction of T. pallidum pertenue diversity [14]. MDA did not exert selection pressure and/or change the composition or distribution of circulating H. ducreyi strains over a 2-year observation period, consistent with asymptomatic colonization and other environmental reservoirs, which must be addressed to achieve eradication of the H. ducreyi component of CU.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. We thank Sheila Lukehart and Jennifer Robson for providing us with DNA from the clinical samples.

Financial support. This work was supported in part by Newcrest Mining Company and by the National Institutes of Health (grant number T35 HL110854 to J. C. G.).

Potential conflicts of interest. All authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.