Treponema pallidum Macrolide Resistance and Molecular Epidemiology in Southern Africa, 2008 to 2018

ABSTRACT

Treponema pallidum macrolide resistance and clinical treatment failure have emerged rapidly within communities where macrolides have been used as convenient, oral therapeutic alternatives to benzathine penicillin G for syphilis or for other clinical indications. Macrolides are not included in the South African syndromic management guidelines for genital ulcer disease; however, in 2015, a 1-g dose of azithromycin was incorporated into treatment algorithms for genital discharge. We determined the prevalence of 23S rRNA macrolide resistance-associated point mutations in 135 T. pallidum-positive surveillance specimens from Botswana, Zimbabwe, and South Africa between 2008 and 2018. Additionally, we investigated the association between macrolide resistance, T. pallidum strain type, and HIV coinfection. A significant increase in the prevalence of the A2058G macrolide resistance-associated point mutation was observed in specimens collected after 2015. There was a high level of molecular heterogeneity among T. pallidum strains circulating in the study communities, with strain type 14d/f being the most predominant in South Africa. Fourteen novel strain types, derived from three new tpr gene restriction fragment length polymorphism patterns and seven new tp0548 gene sequence types, were identified. There was an association between A2058G-associated macrolide resistance and T. pallidum strain types 14d/f and 14d/g but no association between T. pallidum macrolide resistance and HIV coinfection. The majority of T. pallidum strains, as well as strains containing the A2058G mutation, belonged to the SS14-like clade. This is the first study to extensively detail the molecular epidemiology and emergence of macrolide resistance in T. pallidum in southern Africa.

INTRODUCTION

Syphilis, a multistage sexually transmitted infection (STI) caused by the spirochete Treponema pallidum, is endemic in South Africa (1), with incidence estimates in 2017 calculated at 70,675 cases among adult men and women aged 15 to 49 years (2). It is surpassed only by herpes simplex virus 2 (HSV-2) as a causative agent of genital ulcer disease (GUD) in South Africa (3). Syphilis continues to be a major global health problem, with an estimated 6.3 million new syphilis cases in 2016 (4).

Benzathine penicillin G (BPG) has been the recommended treatment of choice for primary syphilis since the 1950s (5). It is still the first-line treatment for syphilis in South Africa but is primarily reserved for use in pregnant women and children, owing to recent national and global shortages of BPG (6, 7). Doxycycline is listed as the alternative therapeutic option for primary syphilis for men, nonpregnant women, and penicillin-allergic patients in the South African STI treatment guidelines (7, 8). Macrolides, such as azithromycin (AZM), erythromycin, and spiramycin, have been used for the treatment of primary syphilis in Africa (9, 10), Europe (11–13), Asia (14), and the United States (11, 15) in the past. From a clinical and public health perspective, AZM is an attractive and convenient alternative one-off oral therapeutic option to BPG, especially in developing, resource-poor countries. It is simple to administer, with no invasive procedures, has few side effects, and may be used in expedited partner therapy for syphilis (10, 16).

T. pallidum macrolide resistance and clinical treatment failure have emerged rapidly worldwide within communities where macrolides have been used as alternative treatment for primary syphilis (11, 17–20), as well as for other infections within the preceding 12-month period (21). The first documented case of erythromycin-resistant syphilis was reported in 1977 (22), and the first cases of AZM resistance emerged among men who have sex with men (MSM) in the United States between 2002 and 2003 (11). Macrolide resistance occurs as the result of a target site alteration due to either an A2058G or A2059G (based on the Escherichia coli 23S rRNA gene numbering) point mutation in the peptidyl transferase region in domain V of the 23S rRNA component of the 50S rRNA subunit (13, 16, 23, 24).

In 1998, the U.S. Centers for Disease Control and Prevention (CDC) described a molecular typing method for T. pallidum, based on the variable number of 60-bp tandem repeats found in the acidic repeat protein (arp) gene and the restriction fragment length polymorphism (RFLP) pattern of three Treponema pallidum repeat (tpr) genes (tprE, tprG, and tprJ) (25). Later, the CDC typing method was modified when sequence analysis of a short region within the tp0548 gene was added to improve the molecular discrimination of T. pallidum strains circulating in communities (26, 27).

We determined the prevalence of the A2058G and A2059G 23S rRNA point mutations associated with macrolide resistance in T. pallidum-positive specimens from patients presenting with GUD to primary health care centers (PHCs) in Botswana in 2008, Zimbabwe in 2014, and South Africa between 2010 and 2018. We investigated the level of molecular heterogeneity among the T. pallidum strains circulating in these communities using the modified CDC subtyping methodology and then analyzed the association between the presence of macrolide resistance-associated point mutations and T. pallidum strain type. Additionally, in the South African cohort, an association between macrolide resistance and HIV coinfection was investigated.

MATERIALS AND METHODS

Study population.

Consecutive, consenting, sexually active, adult patients who presented with GUD to the PHC facilities involved in the Botswana (2008), Zimbabwe (2014), and South African (2010 to 2018) STI etiological surveillance programs were enrolled and ulcer swab specimens were collected. Clinical samples were linked to participants using unique survey numbers and delinked from any personal identifiers. Participants provided consent for long-term storage of these specimens. Ulcer specimens were screened for herpes simplex viruses 1 and 2 (HSV-1/2), Haemophilus ducreyi, T. pallidum, and Chlamydia trachomatis serovars L1 to L3, using a validated in-house multiplex real-time PCR for genital ulcer pathogens, as previously described (28). Venous blood specimens (10 ml) were collected from study participants and screened for HIV using two rapid immunoassays: Alere Determine HIV1/2 (Alere Medical Co., Ltd., Chiba, Japan) and Uni-Gold HIV (Trinity Biotech Plc., Wicklow, Ireland).

PHC facilities involved in the South African surveys were based in all nine provinces, i.e., Eastern Cape, Free State, Gauteng, Kwa-Zulu Natal, Limpopo, Mpumalanga, North West, Northern Cape, and Western Cape. Surveillance was conducted annually in Gauteng and periodically in the other provinces. In total, 135 T. pallidum-positive GUD specimens were obtained from all surveys: 99 from South Africa, 32 from Zimbabwe, and 4 from Botswana. These specimens were used for macrolide resistance testing and strain type analysis.

Ethical approval for the study was obtained from the Human Research Ethics Committee (Medical) of the University of the Witwatersrand (clearance certificate no. M051024, M131129, and M1606677) for South African surveillance. The principal investigators of the Botswana and Zimbabwe surveillance programs obtained ethical approval from their respective institutional review boards for samples to be tested.

T. pallidum PCR assays, macrolide resistance testing, and strain typing.

(i) DNA extraction. As per the surveillance testing procedure, 1 ml of Tris-HCl buffer (pH 8.0) was added to the dry GUD swab specimens and vortexed for 20 to 30 s. DNA was extracted from a 200-μl aliquot of the GUD swab suspension on the automated XtractorGene (Qiagen, Hilden, Germany) platform, using the DX reagent kit (Qiagen), and stored at −70°C for molecular testing and analysis. In cases where stored nucleic acid was not of adequate volume to complete testing for macrolide resistance and molecular typing, DNA was reextracted from the original T. pallidum-positive GUD swab specimens using the same platform. Extracted DNA was stored at −70°C and subsequently used in downstream PCR assay applications.

(ii) Resistance testing. Previously published PCR assays, which target the peptidyl transferase region of the 23S rRNA in the 50S ribosomal subunit, followed by restriction digest assays, were used to determine the presence of macrolide resistance-associated mutations in our specimens (11, 13). PCR assays were performed on either the G-STORM GS482 (Vacutec, South Africa) or GeneAmp 9700 (Applied Biosystems, CA) platform. Restriction enzyme digestion of the amplicons was performed using BsaI (New England BioLabs Inc., MA) for the A2059G mutation and MboII (New England BioLabs Inc.) for the A2058G mutation, according the manufacturer’s instructions. All enzymatically treated amplicons were sized on the Agilent 2100 BioAnalyzer (Agilent Technologies Inc., CA). Samples without any of the point mutations would render a single amplicon size of 629 bp. Samples with the A2058G mutation resulted in two restriction enzyme fragments of 449 and 180 bp, while samples with the A2059G mutation yielded 432- and 197-bp fragments.

DNA from the macrolide-susceptible T. pallidum Nichols strain (GenBank accession number CP004010.2) was used in all macrolide resistance and subtyping assays as a wild-type control. Positive controls for macrolide resistance analysis included the A2058G mutation-containing SS14 strain and a clinical isolate with the A2059G mutation.

(iii) Molecular subtyping. The modified CDC subtyping method has been well described and is based on three target genes, namely, the arp gene, the tpr gene, and the tp0548 gene (25–27, 29). All PCR assays were performed on either the G-STORM GS482 (Vacutec) or GeneAmp 9700 (Applied Biosystems) platform, and arp gene sizing and tpr RFLP pattern visualization were performed on the Agilent 2100 BioAnalyzer (Agilent Technologies Inc.). The tpr gene results could then be compared with published tpr-type patterns, grouped from a to r (25, 30).

T. pallidumtp0548 gene subtyping was conducted according to the method described by Marra et al. (26). The resultant amplicons were purified with exonuclease 1 (EXOI; 4,000 U) (Thermo Fisher Scientific) and FastAP thermosensitive alkaline phosphatase (FastAP; 1,000 U) (Thermo Fisher Scientific), according to the manufacturer’s instructions. Sequencing of the purified product was performed with the BigDye Terminator v3.1 cycle sequencing kit (Thermo Fisher Scientific) according to the manufacturer’s instructions, followed by sequencing purification with the BigDye Xterminator purification kit (Thermo Fisher Scientific). The samples were sequenced on an ABI 3130xl genetic analyzer (Thermo Fisher Scientific), and the resultant sequences were analyzed on Sequencher software, version 5.4.6 (Gene Codes Corporation), and MEGA version X (31). The sequences were compared to published T. pallidum tp0548 gene sequences (26, 32–34).

Statistical analysis.

Data were exported into STATA 14.2 (Stata Corporation, College Station, TX) for analysis. Piecewise logistic regression of calendar years (with splines placed at 2-year intervals) as the independent variable and proportions of samples with macrolide resistance as the dependent variable (macrolide resistance-associated mutations) was performed to determine trends in macrolide resistance in Gauteng province (2011 to 2018). Univariable binomial logistic regression was used to determine the association between strain type and the presence of the A2058G mutation in all T. pallidum samples, comparing strain types detected in two or more specimens and having the mutation (14b/f, 14d/c, 14d/f, 14d/g, 14e/f, and 15d/f) to the group of strains without this mutation or where individual strain types also had the mutation, used as a baseline. A chi-squared test was used to determine the association between macrolide resistance and HIV coinfection.

Data availability.

We have submitted all new tp0548 gene (83-bp) sequences to the GenBank database under accession numbers MZ475825 to MZ475883.

RESULTS

Demographics of T. pallidum-positive participants.

Of the 99 T. pallidum-positive specimens from South African participants enrolled between 2010 and 2018, 96 (97%) had sufficient material to perform macrolide resistance testing and molecular subtyping. Gender information was available for 94 participants: 70 (74.5%) were male. The median age of participants was 28 years (interquartile range [IQR], 24 to 34). Data on HIV serostatus were available for 87 participants, and of these, 46 (52.9%) were HIV infected. Only one T. pallidum-positive specimen was coinfected with another GUD pathogen, namely, HSV-2.

Of the 32 Zimbabwean participants with T. pallidum infection, 19 (59.4%) were male. The study sites included PHC facilities in Harare (one clinic; n = 5) and Bulawayo (three clinics; n = 27). Thirty (94%) T. pallidum-positive ulcer specimens had sufficient material to perform macrolide resistance testing and molecular typing. No age-related or HIV serostatus data were available for the Zimbabwean cohort and four Botswana survey participants.

T. pallidum macrolide resistance.

The overall prevalence of the A2058G point mutation in the South African T. pallidum-positive specimens was 23% (22/96). The mutation was only detected from 2013 onwards (Fig. 1). The A2058G point mutation was absent in the four Botswana 2008 specimens but present in 6/30 (20%) T. pallidum-positive Zimbabwe 2014 specimens. The A2059G point mutation was not detected in any T. pallidum-positive survey specimens from all three countries.

FIG 1.

Macrolide resistance-associated mutations in T. pallidum by year of surveillance, South Africa, 2010 to 2018 (n = 96).

As Gauteng was the only province in South Africa where a continuous data set was available from 2011 to 2018, a trend analysis for macrolide resistance was done. The prevalence of the A2058G mutation ranged from 0 to 71% between 2011 and 2018 (Fig. 2). A significant increase in macrolide resistance prevalence was observed over the surveillance period, especially between 2017 and 2018 compared to 2015-2016 (odds ratio [OR], 15.75; 95% confidence interval [CI], 1.42 to 174.25; P = 0.025). HIV data were available for 22 South African participants with macrolide-resistant T. pallidum specimens: 9 (41%) were HIV seropositive. There was no association between macrolide-resistant T. pallidum and HIV coinfection (P = 0.189).

FIG 2.

Prevalence of the T. pallidum A2058G macrolide resistance-associated mutation, Gauteng Province, South Africa, 2011 to 2018 (n = 49). Trendline depicts a 2-year moving average. (AZM*, azithromycin).

Molecular epidemiology of T. pallidum strains.

Overall, 58 individual T. pallidum strain types could be identified using all three genes, with 3 additional specimens that could be typed only with the tpr and tp0548 gene targets.

The T. pallidum strain types circulating in Gauteng province between 2011 and 2018 were diverse, with the predominant types being 14d/f (n = 9 [18.8%]) and 15d/f (n = 5 [10.4%]) (Fig. 3). T. pallidum strain types identified in all South African provinces over the entire surveillance period are summarized in Table 1 by year of surveillance. Only one T. pallidum-positive specimen could not be typed using all three genes. Three new tpr gene RFLP patterns, designated s, t, and u, were identified in the South African samples (Fig. 4). Seven new tp0548 gene sequences, which did not align with any of the published sequences to date (30, 32, 33), have been given the designations al, am, an, ao, ap, aq, and ar (see Fig. S1 in the supplemental material). The resultant 14 new strain types identified were from T. pallidum specimens collected in Gauteng (n = 4), Free State (n = 3), Northern Cape (n = 2), Mpumalanga (n = 2), Kwa-Zulu Natal (n = 2), and North West (n = 1) provinces.

FIG 3.

Strain type distribution of T. pallidum in Gauteng province by year of surveillance, 2011 to 2018.

TABLE 1.

T. pallidum strain type distribution in all South African provinces, 2010 to 2018

Strain type(s)	No. (%) of each strain type (n = 96)
6h/aq, 6u/c, 7b/c, 7d/al, 8/d/f, 9d/f, 9s/c, 10d/c, 11d/e, 11d/f, 12d/f, 12e/c, 13j/e, 14a/am, 14a/c, 14b/c, 14d/p, 14e/c, 14j/c, 14k/ao, 14s/an, 14t/f, 14t/i, 15b/c, 15d/e, 15d/g, 16d/al, 16d/f, 18d/ap, 18j/ar, 19b/c, 20b/c, 20e/an, 20s/an, 21a/c, 23d/c, unknown d/f	1 (1)
14b/f	2 (2.1)
13d/f, 14d/c, 15d/c, 18d/e	3 (3.1)
24b/c	4 (4.2)
14d/g, 14e/f, 15d/f	5 (5.2)
14d/f	26 (27.1)

FIG 4.

New South African T. pallidum tpr gene types.

The T. pallidum strain type distribution from the Zimbabwe surveillance differs from the pattern seen in South Africa, with the predominant strain types being 14d/c (n = 7 [23.3%]) and 14d/f (n = 3 [10.0%]) (Fig. 5). Only 2/30 (6.6%) Zimbabwe T. pallidum strains could not be typed using all three genes. The four specimens from Botswana surveillance had diverse T. pallidum strain types, namely, 10d/c, 14d/d, 14 d/g, and 17b/c.

FIG 5.

T. pallidum strain type distribution in Zimbabwe, 2014 (n = 30).

There was an association between the A2058G mutation and strain types 14b/f (OR, 5.92; 95% CI, 1.21 to 28.82; P = 0.028), 14d/f (OR, 5.07; 95% CI, 2.11 to 12.19; P < 0.001), and 14d/g (OR, 9.86; 95% CI, 4.24 to 22.96; P < 0.001) (Table 2).

TABLE 2.

Univariable binomial logistic regression to compare associations between A2058G macrolide resistance-associated point mutation and T. pallidum strain types in southern Africa (n = 128)

Strain type detected	No. of specimens	A2058G mutation prevalence by strain type, no. (%)	OR associated with A2058G mutation (95% CI)	P value
14b/f	2	1 (50)	5.92 (1.21–28.82)	0.028
14d/c	9	1 (11)	1.31 (0.18–9.71)	0.789
14d/f	28	12 (43)	5.07 (2.11–12.19)	<0.001
14d/g	6	5 (83)	9.86 (4.24–22.96)	<0.001
14e/f	7	2 (29)	3.38 (0.83–13.70)	0.088
15d/f	5	1 (20)	2.37 (0.35–16.03)	0.377
All other strains	71	6 (8.5)	1.00

A neighbor-joining phylogenetic tree was constructed using a Kimura 2 model with gamma distribution rates to categorize the 83-bp T. pallidum tp0548 gene sequences into the Nichols-like and SS14-like clades (Fig. S2 and Table S1) (31). The majority (65/96 [67.7%]) of South African strains belonged to the SS14 clade, while the Zimbabwean strains clustered evenly into the two clades. Only one of the Botswana strains belonged to the SS14 clade, with the remaining three grouping into the Nichols-like clade. One Botswana specimen in the Nichols-like clade clustered with the Treponema pallidum subsp. endemicum Bosnia A (GenBank accession number CP007548.1) and Cuban (GenBank accession number MG002640.1) strains included for comparison. The new sequence type aq, identified in the South African cohort, was closely related to the Treponema pallidum subsp. pertenue Samoa D reference sequence (GenBank accession number CP002374.1). The majority (24/28 [85.7%]) of strains that were macrolide resistant grouped into the SS14-like clade.

DISCUSSION

This is the first study to extensively detail the molecular epidemiology of T. pallidum in southern Africa and investigate the association between macrolide resistance, T. pallidum strain type, and HIV coinfection. We also report the first findings of T. pallidum macrolide resistance in South Africa (A2058G mutation), as well as several novel T. pallidum strain types. Overall, the A2058G mutation was present in over one-fifth of the T. pallidum-positive specimens collected from South Africa and Zimbabwe. Our surveillance reveals that there was a significant temporal increase in the number of macrolide-resistant symptomatic primary syphilis cases in South Africa after 2016.

Due to widespread emergence of macrolide resistance in T. pallidum and associated treatment failure, the use of high-dose (2 g) AZM is recommended as an alternative treatment of primary syphilis only in special circumstances, e.g., when other recommended treatment options are unavailable and where macrolide susceptibility is likely, based on local epidemiology (35, 36). Macrolides are not recommended for use in pregnant women, HIV-infected persons, or MSM (36). Although macrolides are not included in the South African national syndromic management guidelines for GUD, in 2015, a 1-g dose of AZM was incorporated into the treatment algorithms for male and female genital discharge syndromes (5, 8, 10, 15, 16, 37). AZM has a relatively long half-life (approximately 68 h) and therefore a prolonged therapeutic tail, resulting in subinhibitory concentrations for up to 4 weeks in both intra- and extracellular compartments (38). This may lead to increased exposure to subtherapeutic concentrations of AZM and select for T. pallidum resistance in at-risk persons who are coinfected or recently infected with syphilis during the period of waning drug levels (10, 35, 36). Patients presenting with GUD to PHCs may be coinfected with discharge-causing STI pathogens and treated with macrolide therapy as per national STI syndromic management guidelines. In a study of trends in the relative prevalence of GUD pathogens over a 9-year surveillance period in Johannesburg, South Africa, a concomitant genital discharge syndrome was present in over 30% of patients presenting with GUD (3).

Macrolides are also used in the treatment of other bacterial, non-STI-related infections (39, 40). In our setting, HIV-infected individuals in particular may have increased exposure to macrolides used in the treatment of nontyphoidal Salmonella infections and severe community-acquired pneumonia (40–42). Hence, we investigated for an association between macrolide-resistant T. pallidum and HIV infection in our study. Patients are twice as likely to develop macrolide-resistant T. pallidum infections if they have been treated with this class of antibiotics within the previous 12 months (21). Once macrolide resistance has been established within a community, it spreads rapidly (10, 11, 21, 43), and patients who were treated with AZM for primary syphilis should be closely monitored for clinical treatment failure (21).

In a previous South African T. pallidum surveillance study from 2005 to 2010, Müller et al. showed that 14d was the predominant subtype using the limited CDC tpr and arp gene subtyping method (1). Certain T. pallidum strain types, such as 14a/a and 14d/f, are associated with neuroinvasive disease (26). The predominant strain type identified in South Africa is 14 d/f, and this was the second most common type found in Zimbabwe. Strain type 14d/f is also the predominant strain type circulating in the United States (26), France (44), and China (45). Additionally, we detected three new tpr gene RFLP patterns, which did not fit any of the previously published T. pallidum RFLP banding patterns, and seven new tp0548 gene sequences, resulting in the identification of 14 novel individual strain types. Two of these 14 newly identified strain types harbored the A2058G mutation. Based on the tp0548 gene sequences, strains could be grouped into the SS14-like or Nichols-like clade. The majority of T. pallidum strains in our study, and more than 80% of those harboring the A2058G mutation, clustered into the SS14-like clade. SS14 was also the predominant clade in a multilocus sequence typing (MLST) study of T. pallidum specimens from outpatients with early syphilis or syphilis of unknown stage in Europe (46). In a study conducted among heterosexual and MSM populations in Japan, the predominant clade, SS14, was also associated with macrolide resistance (47).

One of the newly identified sequence types from the South African cohort grouped with the Samoa D strain of T. pallidum subsp. pertenue, the causative agent of yaws, a tropical disease not typically associated with genital ulcers. As T. pallidum subsp. pertenue is closely related to Treponema pallidum subsp. pallidum, with less than <0.2% difference in their genomes (48), this was not a surprising find. Human cases of yaws are periodically reported from some Central and West African countries (49). Yaws is hypothesized to be enzootic in nonhuman primates (baboons and vervet monkeys) in regions of East Africa in which it had previously been endemic in humans, and this may be a risk for future zoonotic spillover (50–53). One of the Botswana cohort specimens was closely related to the Bosnia A strain of T. pallidum subsp. endemicum, the causative agent of bejel, or endemic syphilis. Cases of bejel are usually reported in arid, desert regions of Africa (49). It is not considered a venereal disease; however, T. pallidum subsp. endemicum was detected using MLST in patients presenting with anogenital and skin lesions and serologically diagnosed as having syphilis in Cuba (34). In Japan, where syphilis is reemerging, MSM who had been diagnosed clinically, serologically, and through nucleic acid amplification tests as having T. pallidum subsp. pallidum infection were subsequently confirmed to have bejel following molecular genotyping of ulcer material. Interestingly, azithromycin resistance-associated mutations were detected in all T. pallidum subsp. endemicum strains. The patients presented with features suggestive of primary or secondary syphilis, suggesting that clinical manifestations caused by endemic treponematoses may overlap those of classical venereal syphilis (54).

The association between strain types 14d/f and 14d/g and macrolide resistance observed in this study has also been described in the United States, where the A2058G mutation was also observed to a lesser degree in strain types 13d/d, 14d/i, 15d/e, and 15e/e (55). In Sydney, Australia, the predominant strain type, 14d/g, was found to be significantly associated with A2058G macrolide resistance compared to non-14d/g strain types (56). Similarly, in the Netherlands, the A2058G mutation was present in 88% of T. pallidum-positive specimens and the dominant circulating strain type was 14d/g (57). A2058G macrolide resistance has also been associated with tp0548 gene sequence type g in London, United Kingdom (12), and sequence types d and g in Brno and Prague, Czech Republic (58). A relatively small proportion of macrolide-resistant T. pallidum strains from the United States (59) and Czech Republic (13, 58) harbored the A2059G point mutation; this mutation has not yet been identified in southern African T. pallidum strains.

Our results show a high level of strain type diversity in South Africa and Zimbabwe. There may be several reasons for this. The enhanced CDC typing methodology used in this study, which included tp0548 gene sequencing, increased discriminatory power compared to the original CDC method used by Müller et al. in 2012 for typing southern African strains (1). This resulted in the identification of additional strain types. Furthermore, the study period for the South African cohort spanned 9 years and samples were collected from all provinces of the country. Additionally, because it is the most industrialized economy in the region, migrants and workers from other sub-Saharan African countries travel to South Africa in search of economic opportunities. These migrants return periodically to their countries of origin and may have concurrent sexual partnerships in different geographical regions. South Africa is also a popular tourist destination. All these factors may lead to importation of T. pallidum strains circulating in other countries into the region. In Zimbabwe, the surveillance specimens were collected from the two most populous and industrialized cities of the country. The sample size from Botswana was too small to make any conclusion about strain diversity. Within South Africa, strain type distribution found in Gauteng province was more diverse than in other provinces. This may be attributed to the regular annual STI surveys conducted in Gauteng with a greater number of GUD specimens collected over several years. Gauteng is also the most densely populated province and the financial hub of South Africa, where there is a mixing of population groups from other provinces and countries in the region.

This study’s main limitation was that we had relatively few T. pallidum-positive specimens for analysis during the surveillance period, particularly from neighboring Zimbabwe and Botswana. Additionally, a continuous annual surveillance data set was not available for all the South African provinces between 2010 and 2018. Consequently, we could not conduct comprehensive T. pallidum epidemiology and ascertain whether there is a temporal trend for macrolide resistance throughout the country. However, considering that the national STI syndromic management guidelines are used at all PHCs across the country, it is possible that an increase in macrolide-resistant prevalence, similar to that observed in Gauteng, would occur in other provinces. We had no available data on travel history, sexual behaviors, or clinical outcome. We could therefore not link T. pallidum strain type to specific risk groups or to travel from a specific geographical region. A limitation of the enhanced CDC strain typing method may be attributed to genetic instability of arp and tpr loci, leading to T. pallidum intrastrain variability when multiple samples are tested in parallel from a patient (27). Whole-genome sequencing, or the recently published MLST system (46, 60), will enhance T. pallidum subspecies classification, distinguishing genetically different clades and subtypes within these clades. However, it should be noted that the longer amplicons required for higher discriminatory power in MLST may lead to lower amplification efficiency (46). Therefore, when choosing a molecular typing method, it is important to establish a balance between increased discriminatory power and typing efficiency.

Our findings reveal that the prevalence of macrolide resistance in T. pallidum has increased significantly. This is possibly linked to the inclusion of 1 g of AZM in national STI treatment guidelines and therefore increased use of AZM in at-risk persons. Macrolide resistance in T. pallidum is likely to increase in the future, and AZM will not be considered a therapeutic option for syphilis in South Africa. The collection of additional demographic, sexual risk behavior, and travel information from surveillance participants in future surveys would be useful in identifying strain type clusters for targeted intervention and control measures.