Diagnostic-avoiding Chlamydia trachomatis variants detected in cervical and endometrial specimens from women during 16S microbiome profiling

T Cell Response against Chlamydia (TRAC) Cohort (246 participants)

This cohort was composed of young (median age, 21 years; range, 18-35 years) cis-gender women at high risk for the acquisition of CT [1]. Criteria indicating high-risk status included > 3 sexual partners in the previous 6 months, ≤ 14 years of age at sexual debut, history of pelvic inflammatory disease (PID), the presence of mucopurulent cervicitis on exam, or sexual contact with a partner known to be infected with CT or NG or non-gonococcal non-chlamydial urethritis. Women with a current diagnosis of PID were excluded. Additional exclusion criteria were pregnancy, uterine procedure or miscarriage in the preceding 60 days, menopause, hysterectomy, antibiotic therapy in the preceding 14 days, and allergy to study medications. TRAC participants were enrolled into a longitudinal study designed to investigate T cell responses important for protection from incident chlamydial infection over 12 months of follow-up between February 2011 and August 2014 and were recruited from the Allegheny County Health Department’s Sexually Transmitted Diseases Clinic, UPMC Magee-Womens Hospital (MWH) Ambulatory Care Clinic, and the Reproductive Infectious Disease Research Unit at MWH in Pittsburgh PA. Participants provided informed consent at the time of enrollment and completed questionnaires regarding obstetric/gynecologic history, behavioral practices, sex exposure, contraceptive methods, and symptoms. Clinical, histological, and microbiological testing was performed, and blood and endometrial samples were obtained after enrollment, after which all participants received single-dose agents for gonorrhea (ceftriaxone, 250 mg intramuscularly) and chlamydia (azithromycin, 1 g orally). Participants in this cohort were assessed for cervical and endometrial infection using AC2 (Hologic, Marlborough, MA) NAAT with overall infection rates of 67% and 8.5% for CT and NG, respectively. M. genitalium (MG) infection was determined using the Aptima MG NAAT (Hologic).

16S library construction and characterization

Cervicovaginal swab samples available from 246 women enrolled in TRAC were processed for 16S rRNA gene microbiome profiling [11]. In brief, cervicovaginal microbial communities present in participant specimens were profiled by sequencing the V4 region of 16S rRNA gene using Illumina sequencing. The 16S rRNA amplicon sequencing data was processed at the level of inferred amplicon sequence variants (ASVs) using USEARCH v11 [15]. The raw sequencing data supporting the conclusions of this article is publicly available at NCBI’s Short Read Archive (SRA) under BioProject accession number PRJNA1136868.

PCR template purification from residual diagnostic samples

DNA was extracted as previously published using a Quick-DNA Universal Kit (Zymo Research, Irvine, California, USA) from 100 μL to 300 μL of reserved cervical swab eluates. Chlamydial and/or mycoplasmal loads were measured by quantitative PCR [1] targeting the 16SrRNA gene [16] or the highly conserved MG190 locus [10], respectively. Primers directed against the parA pseudogene [17], 16S rRNA and 23S rRNA loci [18] were similarly used to validate and/or quantify NG abundance.

PCR amplification and sequencing of 23S rRNA alleles

Amplification conditions for detection of SNPs in C. trachomatis 23S were as described [19]. The melting temperature was raised to 98°C for optimal activation of the high fidelity, proof-reading enzyme Phusion (ThermoFisher). Samples were processed independently in a dedicated PCR hood to minimize risk of cross-contamination during analysis. Two rounds of amplification were required to recover sufficient product which was cleaned and concentrated using a DNA Clean and Concentrate kit (Zymo) before sequencing in forward and reverse directions using V3 region primers [19]. Two patient samples that had tested positively using the AC2 test were also analyzed as controls. Sequences generated via Sanger sequencing were aligned against the published sequence using Clone Manager Professional Version 11 (Sci Ed Software LLC, Westminster, CO). Residual cervical templates were used from profiled participants except for 3-209, for whom only endometrium-extracted DNA remained available for analysis.

Statistical analysis

The nonparametric Spearman rank correlation coefficient was calculated to assess the overall correlation between CT 16S rRNA gene DNA copy numbers measured by qPCR and the raw read counts of CT 16S rRNA gene from microbiome sequencing analysis using R function ‘cor’ [20]. The LOESS (Locally Weighted Scatter-plot Smoother) curve was estimated and plotted using a ‘geom_smooth’ function of the R ‘ggplot2’ package with default settings [21]. Abundance of NG-specific ASVs present in libraries from individuals testing positively with AC2 and those who tested negative were compared using a Mann-Whitney test (two tailed).

Detection of chlamydia-derived amplicon sequence variant (ASV) in 16S libraries derived from vaginal swabs obtained from CT negative TRAC participants

Total DNA was extracted from “future use” vaginal swabs from a total of 246 TRAC participants for inclusion in a 16S rRNA gene-based sequencing analysis of the contribution of the cervicovaginal microbiome to ascending chlamydial infection. Three samples were excluded based on DNA low integrity values after extraction. We also excluded twenty-three samples due to the equivocal results of chlamydial infection at the cervix and/or endometrium from clinical diagnostic testing. The remaining 220 samples were processed using the ASV-level analysis, resulting in 1,990 ASVs detected in total from samples sequenced to the average depth of ∼151,000 reads. Taxonomy annotations of ASVs were based on sequence comparison against the 16S rRNA gene reference database RDP version 19 [22] using the SINTAX algorithm [23]. We assessed the concordance of STI detection by examining raw read counts of sexually transmitted pathogen-specific 16S rRNA genes vs infections previously assessed by NAAT diagnostic tests (CT, NG, MG) (Fig. 1a) and quantified via qPCR (CT [1], Fig. 1b). About 58% (83 of 143) samples from participants whose tests were positive for CT at enrollment had non-zero 16S rRNA gene read counts (Fig. 1a). Similarly, 85% (17 of 20) samples from participants whose tests were postive for NG, and 15.8% (6 of 38) samples from MG positive enrollees had non-zero 16S rRNA gene read counts. Overall, there was modest correlation (Spearman = 0.575) between the two CT quantification methods (Fig. 1b).

• Download figure
• Open in new tab

Figure 1. STI-derived 16S transcripts detected in vaginal swabs.

a) 16S rRNA gene read counts of STIs stratified by infection status at enrollment. b) Concordance between CT 16S rRNA gene read counts versus CT burden by qPCR, where samples are colored by infection status at enrollment. A LOESS curve is shown in red and Spearman correlation in the bottom-right. CT, Chlamydia trachomatis; GC, N. gonorrhoeae; MG, M. genitalium

Based on clinical diagnostic results obtained at enrollment, the TRAC participants were classified into three groups: 1) Uninfected; 2) Endo-participants with CT infection detected only in their cervix; 3) Endo+ participants who tested positively for CT in cervix and endometrium. There were no TRAC enrollees whose tests were positive only in the endometrial sample. Comparing ASV abundances between CT-positive (Endo- and Endo+) and CT-negative (control) women, we were surprised to detect 16S ASV within libraries derived from samples from women who had been assigned “uninfected” (N=7) based on the results of NAAT testing (Fig. 1a). These results correlated with estimated chlamydial abundance derived from a previously performed quantitative 16S PCR assay (Fig. 1b and Table 1), suggesting that attributing the original results to external or cross-contamination was incorrect. A review of these participants’ clinical data revealed that they were mostly asymptomatic (5/7), one had reported experiencing dysuria, consistent with >10 leucocytes detected per high power field in her urine sample with irregular vaginal bleeding and another reported dyspareunia and abnormal vaginal discharge. Clinical findings upon pelvic examination were unremarkable for all. Three of the seven participants within this group reported contact with a CT infected partner in the three months prior to their enrollment but had yet to be treated.

In the context of the TRAC cohort, reasons for false negative CT NAAT test results, other than testing failures detected at or during performance e.g., compromised reagents or automated run failure (batch specific), include inadequate sampling at the anatomical site or an unanticipated flaw in the diagnostic test. All NAAT tests were run within temporal parameters appropriate for clinical STI testing, so batch-specific effects did not serve as an explanation for these discrepancies because the participants were not recruited contemporaneously. Testing of an endometrial biopsy might result in a false negative because the entire endometrial lining is not sampled, the amounts of tissue obtained may vary [7], and the region of specimen dedicated to NAAT testing is small [10]. However, cervical testing, particularly when performed by highly trained practitioners, is robust [24, 25]. Several publications from Northern Europe have reported instances of false negative testing specifically associated with the AC2 test [12, 13, 26, 27] arising from probe-avoidant mutations at the chlamydial 23S rRNA locus [19]. Speculating that this might explain our discrepant results, we attempted to amplify the relevant V3 region from cervically or endometrially derived templates that had been extracted from residual diagnostic swab eluates. In all instances we were successful using the protocol and primers used by Hokynar et al. [19], further confirming the presence of chlamydial DNA in the clinical samples. We detected a G-A substitution at 1523 in the samples of all seven test-discrepant participants (Fig. 2) when these products were sequenced. NAAT positively-testing samples (N=2) from the cohort were wild type at this position (data not shown). These data indicated that at least one probe-avoidant variant of CT was circulating in the region between May 2012 and February 2014, the time window between enrollments of our earliest and latest discrepant TRAC participants.

• Download figure
• Open in new tab

Figure 2. Portion of sequence alignment of V3 amplified PCR products from discrepant TRAC participant specimens revealing a g to a substitution at 1523 in the 23S rDNA locus.

PCR products from two representative CT+ samples, as determined by AC2 diagnostic test were wild type at this position (data not shown).

16S rRNA gene microbiome analysis also revealed a total of five participants with discrepant NG AC2 testing enrolled from April 2011 to July 2013 (Fig. 1A, Table 2), including one participant, 3-209, who also tested discrepantly for CT (Table 1). One participant reported mild urinary symptoms with abnormal vaginal discharge and irregular bleeding and one other participant reported bleeding with intercourse. Clinical findings on pelvic examination were noteworthy for bleeding with cervical contact for four of five participants but were otherwise unremarkable. Only one of these participants reported contact with an infected partner within the preceding three months but had not received treatment.

The genus Neisseria includes a group of commensal species, N. subflava, N. mucosa, N. elongata, in addition to the important human pathogens, N. gonorrhoeae and N. meningitidis with the potential for genetic exchange between these naturally competent microorganisms [28] so we initially speculated that the ASV detected lacked discriminatory capacity. However, alignment of the 16S V4 sequence from N. gonorrhoeae with the 35 Neisseria species included in the RDP v18 database, revealed that ASV ZOtu95, detected in the discrepant samples, was highly specific, aligning with 100% identity between bases 535-787 of the 1544 nucleotide V4 region. This suggested that the 16S ASV detected in libraries generated from the vaginal swabs reflected true gonococcal infection. Next, we compared the number of ZOtu95 reads from discordant individuals with those detected in libraries generated from TRAC participants with concordant NG detection (Table 2) as a surrogate for pathogen load. The number of reads counted in NG+ samples ranged across four orders of magnitude but participant swab samples yielding libraries with ≤6 counts still tested positively via AC2 and no difference was detected between the groups using the non-parametric Mann-Whitney test (P=0.4116). Low levels of infection may have contributed to differential sampling of bacterial populations present at the cervix (diagnostic) and in the vagina (16S analysis). Quantifying NG by qPCR in residual template was difficult because of the minimal amount of cervical DNA remaining and the potential for high genetic variability between strains. Using primers that targeted multicopy (16S, 23S rRNA) and single copy (ParA pseudogene) loci, with genomic DNA extracted from N. gonorrhoeae strain FA19 as a positive control [29], we performed qPCR on diagnostic swab eluates or the 16S template that remained from discrepant NG+ participants (Table 2). Evaluating the performance of the primer pairs using material obtained from participants testing positively for NG infection, we observed that only 8 of the 17 samples consistently yielded PCR product for all 3 loci, while 6, were positive for 1-2. We were unsuccessful in amplifying material from the remaining 3 samples. Overall, successful amplification of the targets did not reflect abundance of 16S read counts in participants’ 16S rRNA libraries. For samples derived from participants who had tested negatively clinically but with ZOtu95 reads detected via 16S rRNA microbiome profiling, all tested positively for at least one multicopy locus.

With no source for cross-contamination available in the lab performing the 16S rRNA analysis, we also looked retrospectively at participant enrollment dates to exclude the remote possibility of cross- or environmental contamination [30] occurring during sample collection or NAAT testing. In only one instance did we observe enrollment of a discrepant NG- (3-121) and a confirmed clinically NG+ participant within the same 48-hour window (3-120). However, the presence of additional, neisserial species ASVs in the library generated from 3-120 only (Table 2), is inconsistent with contamination of the discrepant NG-sample by material derived from this diagnostic swab.