Demonstration of Persistent Infections and Genome Stability by Whole-Genome Sequencing of Repeat-Positive, Same-Serovar Chlamydia trachomatis Collected From the Female Genital Tract

Summary

Genome sequence analysis of clinical samples demonstrated that identical or nearly identical Chlamydia trachomatis strains can be isolated from individual patients for up to 5 years. These data provide evidence for chlamydial persistence, even when patients are treated with antibiotics.

Abstract

Background.

The biology of recurrent or long-term infections of humans by Chlamydia trachomatis is poorly understood. Because repeated or persistent infections are correlated with serious complications in humans, understanding these processes may improve clinical management and public health disease control.

Methods.

We conducted whole-genome sequence analysis on C. trachomatis isolates collected from a previously described patient set in which individuals were shown to be infected with a single serovar over a lengthy period.

Results.

Data from 5 of 7 patients showed compelling evidence for the ability of these patients to harbor the same strain for 3–5 years. Mutations in these strains were cumulative, very uncommon, and not linked to any single protein or pathway. Serovar J strains isolated from 1 patient 3 years apart did not accumulate a single base change across the genome. In contrast, the sequence results of 2 patients, each infected only with serovar Ia strains, revealed multiple same-serovar infections over 1–5 years.

Conclusions.

These data demonstrate examples of long-term persistence in patients in the face of repeated antibiotic therapy and show that pathogen mutational strategies are not important in persistence of this pathogen in patients.

Chlamydia trachomatis infections are major health burdens throughout the world [1, 2]. In the female genital tract, repeated or persistent infections can lead to serious complications, including pelvic inflammatory disease, tubal infertility, chronic pelvic pain, and ectopic pregnancy. These unresolved infections significantly contribute to the public health burden, and it is therefore important to study their origins to improve management strategies and treatment regimens.

Various developing technologies have been applied to the challenge of distinguishing concordant from discordant repeat infections and have shown to be useful in studying the incidence of recurrent infections [3, 4]. These methods have demonstrated that reinfection is common, but the nature and significance of persistence in such infections remain unclear. In a previous study [5], 7 patients who were infected for long periods of time with single infrequent (J, Ia) or rare (D−, H, Ja, K) C. trachomatis serovars under conditions of regular antibiotic therapy were identified. These isolates were characterized using a serotyping system based on a major outer membrane protein (MOMP, OmpA [6]) that has been historically useful in characterization of epidemics and disease processes [7, 8].

In this study, we use a genomics-based approach to examine these patient samples to explore the nature of chlamydial persistence and to examine genomic changes that might be selected for in a strain in the face of possible antichlamydial immune responses. Additionally, we examine genome structure in uncultured versus cultured C. trachomatis from these patients. The results demonstrate that, in persistently infected patients, the genomes are remarkably stable over long periods of time. Additionally, we document individuals that are infected repeatedly with same-serovar but differing-genotype strains that evidently circulate in sexual networks within the studied community.

MATERIALS AND METHODS

Patient Population, Sample Collection, and Tissue Culture

The specimens and the resulting chlamydial cultures used in this study were initially collected from women attending the Seattle King County Health Department STD Clinics from 1986 through 1995. Patients were routinely screened for cervical chlamydial infection by culture and, in some cases, ligase chain reaction (LCR). Following standard clinic routine, patients with positive results were treated with doxycycline or azithromycin and were provided with counseling and sex partner follow-up. Selected chlamydial isolates were tested for minimum inhibitory concentrations against doxycycline and azithromycin [5]. The study was approved by the institutional review boards of the University of Washington and Oregon State University. Both primary clinical specimens (passage zero [p0]) and low-passage (n < 3), in vitro–cultured material (p1) were used for sequencing in this work.

Genome Sequencing

The first and last samples from each patient, as well as 3 intermediate patient samples (SQ02, SQ10, and SQ12) were selected for whole-genome sequencing (Figure 1). Both p0 and p1 samples were subjected to whole-genome sequencing, with the exception of SQ32 from patient 1 and SQ28 from patient 3, for which culture-independent sequencing was unsuccessful. Uncultured (p0) samples were processed using immuno-magnetic separation, as described in a previous study [9, 10]. Elementary bodies from cultured material were processed using the DNeasy Blood and Tissue Kit for DNA extraction (Qiagen). Genome sequencing was conducted on an Illumina HiSeq 2000 at the Oregon State University Center for Genome Research and Biocomputing.

Figure 1.

Timelines for individuals diagnosed multiple times with same-serovar Chlamydia trachomatis infections. A detailed presentation of these data can be found in [5]. Time 0 represents the first date the patient was culture-positive. The month and year of the initial positive diagnosis is indicated below each patient/serovar indication. Each specimen/strain number is indicated above its approximate time of isolation. Boldface strain names indicate specimens and resulting cultures that were whole-genome sequenced. Outlined strain names indicate cultured samples for which polymerase chain reaction/sequence analysis was conducted on specific regions of the chromosomes. Gray bars in each timeline indicate periods during which the patient was determined to be culture- or ligase chain reaction–negative.

Genome Assembly and Analysis

Illumina-generated reads were filtered and trimmed using the open source software Trimmomatic [11], discarding reads with ambiguous bases, and trimmed using a 4-base sliding window with 15 set as the Phred score threshold. Reference-guided and de novo assembly of genomes was conducted with Bowtie-2 [12] and Velvet [13] software packages, respectively. Samples from patient 7 were originally assembled using both the SotonK1 [14] and strain HUW/4 [15] sequences as reference genomes. Subsequent analysis of challenging areas were conducted using the Geneious software package, (v. 7.1.7; http://www.geneious.com [16]). Any ambiguous bases or polymorphisms were verified in the raw reads using a consensus threshold of 75% for that base. Accession numbers for all genome sequence data are provided in Supplementary Table 1. High-fidelity polymerase chain reaction (PCR) was conducted with target-specific primers (Supplementary Table 2) and Sanger sequencing at the Oregon State University Center for Genome Research and Biocomputing for analysis of ambiguous bases or to conduct single nucleotide polymorphism (SNP)–specific sequencing of intermediate strains.

Complete, unambiguous genome sequences from p1 cultures were generated for 14 of 16 strains (Figure 1, bold labels). Sequence analysis of cultured (p1) patient isolates SQ01 and SQ05 led to poor-quality data, and thus p0 genome data were used in the analyses. All genomes were compared with publicly available genome sequences, with a focus on strains that had been collected in the Seattle, Washington, area when possible [15, 17].

RESULTS

General Genomic Characterizations

A maximum likelihood–based approach was used to generate phylogenetic trees comparing all of the completed genomes from this study with those in the databases (Figure 2). The data show that each strain from within patients was highly linked, with the number of SNPs and insertions or deletions (indels) between the first and last isolate ranging from 0 (patient 2; Table 1) through approximately 45 (patients 4 and 5; Supplementary Tables 3 and 4). In most cases, genomes sequenced in other studies are not most closely linked with their serovar-matched genomes as observed in this patient set (Figure 2). For example, the serovar D minus (D−) strains collected from patient 1 share greater sequence identity with serovars E and F strains than to an independent Seattle area D− strain. These data support previous work by our group [17] and others [14], demonstrating that strains having identical ompA genotypes do not necessarily share a higher overall genomic-relatedness pattern.

Figure 2.

Whole genome–based phylogenetic trees showing the relationship of each strain in this study (boldface) to each other and to publicly available, serovar-matched strains. All strains were collected in the Seattle, Washington, area with the exception of Soton K1 and Soton Ia3, which were collected in the United Kingdom [14]. The assembled H/UW-4 was generated using sequence reads analyzed by Joseph and colleagues [15]. The expanded tree (right) shows the relationship of the serovar Ia strains cultured from patients 4 and 5. Scale bars represent substitutions per base pair for each tree. The arm leading to Ia/SotonIa3 is twice as long as indicated.

Table 1.

Each Individual Mutation Captured From Chlamydia trachomatis Samples Collected From Persistently Infected Patients

Patient 1, serovar D−
Location	Gene	Change	SQ29	SQ30	SQ31	SQ32		Search target	Primers
851637	Hypothetical	S-> G	T	T	T	C		TGCTGAGAAAACAGGGACAC	1.1
Patient 2, serovar J
Location	Gene	Change	SQ01	SQ02	SQ05			Search target	Primers
No SNPs or indels between SQ01 and SQ05
Patient 3, serovar Ja
Location	Gene	Change	SQ25	SQ26	SQ27	SQ28		Search Target	Primers
32136	trmD	Silent	C	C	T/C	T		CAAGAAAGTGCAGAGTACGA	3.1
716420	ispA	Indel, FS	Insertion	…	…	…		TTCTTTTTAATAATCCCTAA	3.2
799178	Hypothetical	Silent	G	G	G	A		GTGAATGAGCAGCTCGTTGT	3.3
998750	Hypothetical	Silent	A	A	A	G		GCGACGAGAACAGACTGGTA	3.4
1020604	Hypothetical	S-> F	C	C	C	T		TCTTTGGAGAATTTTTTCTT	3.5
Patient 6, serovar K
Location	Gene	Change	SQ15	SQ16	SQ17	SQ18	SQ19	Search Target	Primers
200648	Inter	…	T(8)	T(7)	T(7)	T(7)	T(7)	ATATTATCAGACCTTTTTTT	6.1
592783	secY	Silent	T	T	A	A	A	AGAAACTGCCCAATCGTAGC	6.2
1019907	copB2	L-> I	A	A	A	T	T	TGGGAAGGATCGCTATCGTA	6.3
1033675	pmpE	A-> T	C	C	T	T	T	AAAAGAATTGCTCGTTCCAG	6.4
Patient 7, serovar H
Location	Gene	Change	SQ20	SQ21	SQ22	SQ23	SQ24	Search target	Primers
297826	Inter	…	C(9)				C(10)	GAAGAAGAAATTTTTATTCT	7.1
766013	Inter	Indel, FS	…	AAGAACG	AAGAACG	AAGAACG	AAGAACG	TGTTAAGCTGCGCTTAACCT	7.2

Patient number, serovar, and sequence accession number are indicated above each dataset. The columns in each individual table indicate the genome location, gene assignment (if available), the nature of the mutation, the strain being analyzed, a search target for finding the mutation, and the name of a relevant oligonucleotide primer pair used to amplify sequence surrounding the mutation (Supplementary Table 2). In the “Change” column, the single-letter amino-acid code is used to indicate protein sequence changes, with “silent” indicating no amino-acid change Gene abbreviations are indicated for convenience and may not be formally correct for the gene being identified. Strains that were fully genome sequenced are indicated in boldface, whereas strains analyzed by target-specific polymerase chain reaction are indicated in plain text. The search target indicated for each mutation can be used to quickly find a specific mutation within a sequence. For each individual mutation, the variable sequence is immediately 3’ of the search target. Abbreviations: FS, frame shift; IF, in-frame; indel, insertion and deletion; SNP, single nucleotide polymorphism.

A variety of individual mutations occurred in samples collected from these patients (Table 1; Supplementary Tables 3 and 4). However, there was no consistency in the pathway or individual protein affected by SNPs within or among patients. There were no changes in ompA in any of the examined lineages, including strains from patients persistently infected for >3 years. A total of 124 base changes or indels were identified, with 54 substitutions leading to amino-acid changes, 46 silent mutations within coding sequences, and 24 mutations in intergenic regions. A large majority of these mutations were found in strains infecting patients 4 and 5, with only 12 total changes identified in patients 1, 2, 3, 6, and 7 (Figures 3 and 4).

Figure 3.

Patient histories for individuals with repeat or persistent infections at clinics in the Seattle, Washington, area. In all panels, time 0 is the first recorded date testing positive for Chlamydia trachomatis. In B and C, the blue boxes indicate the individual strain collected for sequence analysis. The culture and other diagnostic data for each cultured strain is in the vertical center of each box. A, Examples of repeatedly C. trachomatis–positive patients from the same patient population used in this study. Patients A, D, and E were infected with a variety of serovars, whereas patients B, C, and F were infected with different clusters of 3 same-serovar strains. B, Clinical data from patient 2 (see Figure 1), in which serotyping and genome sequencing identified a serovar J genome that did not vary at any base from the first to the last indicated culture. Strains collected for genomic or sequence-based analysis are indicated by SQ number, as discussed in the Materials and Methods. Culture +/− indicates whether the pathogen was successfully cultured out of clinical material at that visit. LCR indicates the results of ligase chain reaction–based diagnostics for that patient. Results of the blind passage of primary culture–negative specimens is indicated for several samples. C, Clinical data from patient 6 (see Figure 1), in which serotyping and genome sequencing identified a serovar K genome that varied at a total of 4 positions from the beginning to the end of the sampling. Individual sequences and strains are linked vertically within the blue boxes. Visits in which the individual was treated with doxycycline are indicated in the “Doxy” row. “Other” indicates that the individual was found to be positive for either Neisseria gonorrhoeae (GC) or Trichomonas vaginalis (TR). Documented partner culture status is indicated when determined, as well as visits in which the partner was treated with doxycycline. Each single nucleotide polymorphism position is identified relative to its position in the first sequenced genome for each identified patient (Supplementary Tables 3–5).

Figure 4.

Linear representation of the single nucleotide polymorphisms (red) and insertions and deletions (green) identified for the collected genomes sequenced from each individual patient shown in Figure 1. The numbers at the top represent genome positions in kilobases, with position 0 being as defined for the Chlamydia trachomatis D/UW3 genome [34].

Three independent, fully successful genome sequences were generated for patient 2: 2 from p0 samples (SQ01, SQ05) and 1 from a p1 sample (SQ02; not shown). Each of these genomes was identical, indicating there were no genome changes in the C. trachomatis collected from this patient over the course of 3 years.

There is a single example of recombination in a serovar Ia strain collected from patient 5 (sample SQ09), in which approximately 12290 base pairs (CT749−CT755) have 18 consecutive SNPs not found in any other Ia strain from this work (Figure 5). Sequence relatedness in this region demonstrates that, although sequences outside this region are more closely related with the collected strains in patients 4 and 5, this region is most similar to other non-Ia strains.

Figure 5.

Apparent lateral gene transfer leading to incorporation of novel base changes into the genome of strain SQ09 (patient 5, genovar Ia). A, Open reading frame map of a section of the SQ09 chromosome showing the region of lateral gene transfer (black orfs) inserted into the chromosome. The orf map is derived from the genome sequence data summarized in Supplementary Table 4. B and C, Maximum likelihood–based phylogenetic trees showing the relationship of regions in A acquired through horizontal transmission (B) and flanking sequence (C), represented by the orfs shown in gray in A. Scale bars in B and C indicate changes per base pair.

A total of 13 different indels were identified in the genomes from the collected strains (Supplementary Table 5). Analysis of these sequences demonstrated that all but 1 was the result of a duplication or deletion event within the preexisting sequence. These changes ranged from single base indels within homopolymeric repeats to the duplication of 73 base pairs within ispA in strain SQ28. Only 2 indels led to premature truncation of a protein sequence—the 73 base-pair insertion in SQ28 and the homopolymeric repeat within incA in samples taken from patients 4 and 5.

Assessment of Persistence Versus Reinfection in Same Serovar–Infected Patients

An expected typical distribution of serovars in patients infected multiple times over a lengthy course (approximately 3 years) will include a variety of serovars, with the most common ompA types being present more often than the less common or rare types (Figure 3A, patients A, B, D, and E). This study examines a set of patients repeatedly diagnosed with strains expressing single, rare serovars during the entire course of the study period. Six of the 7 women had C-class serovar (H, Ia, J, or K) infections. Analysis of the large dataset that included the strains examined in this study demonstrated that C-class serovars were strongly correlated with apparent persistence [5], and that is supported by this work. Similar patterns with C-class serovars can also be seen in patients C and F in Figure 3A, where different C-class strains were repeatedly found in individuals.

Our criteria for assessing whether an individual was persistently infected included multiple same-serovar isolations of C. trachomatis with very limited numbers of mutations and evidence that mutations accumulated with time (Figure 4). The clearest example of this were strains cultured from patient 2, in which no differences in genome sequence were observed in 3 independent serovar J strains collected over a 30-month period (SQ01, SQ02, SQ05) (Figure 1). Patient 1 had a single base-pair difference in 2 genovar D− sequenced strains and 2 strains examined by PCR collected over a 56-month period. All 4 strains were genovar D−, a relatively rare OmpA type (2.1% of the studied population) that carries a serologically distinct variant of serovar D OmpA [8]. A single base difference at position 851462 was identified in the sequenced strains (SQ29 and SQ32), and PCR analysis of this position showed the change was only found in the latest isolated strain (Figure 1; Table 1). A control serovar D− genome (UW-3991/D−) (Figure 2) was not closely related to the sequence of the D− strains in this patient.

Strains isolated from patients 3 and 6 had 5 and 4 mutations, respectively, during the period of study with each strain set showing an accumulation of mutations with time (Figure 3;Table 1). Polymerase chain reaction–based Sanger sequence analysis of samples from patient 3 showed that 1 mutation within trmD (Table 1) was identified as a mixed population in material from patient SQ27 (Supplementary Figure 1) as the population transitioned from the PCR-based clonal population in SQ26 to the strain fully genome sequenced in SQ28. This was the only example of a mixed genomic population in our collected genome sequences.

Patients 2, 6, and 7 had diagnostic evidence of pathogen clearance during these analyses, only to be found to be C. trachomatis–positive at a later date. In each case, the subsequent strain was genetically identical or very nearly identical to the strain identified before being clinically diagnosed as negative (Figure 1). For example, patient 6 was culture-negative multiple times, including 4 negative tests of cure cultures, but virtually identical strains were found both before and after the negative cultures, and the rare mutations in these strains accumulated from beginning to end (Figure 3C). In patient 2, 5 specimens identified as negative by LCR and culture were shown to contain low levels of viable organism after 3 blind passages of the original culture material. Chlamydia trachomatis isolated in these blind-passaged samples all serotyped as serovar J, although material was no longer available for further characterization. These data suggest that persistently infected patients can maintain very low levels of pathogen, which can expand to detectable levels in tested tissues.

Evidence for Repeated Infections Within the Patient Population

Patients 4 and 5 have 2 and 6, respectively, culture-confirmed infections with serovar Ia C. trachomatis. Sequence analysis of these patients, however, yielded a very different infectious process from that observed with patients 1, 2, 3, 6, and 7. Strains from these patients carried a larger number of what appeared to be random mutations across their genomes (Figure 4, Supplementary Tables 3 and 4). In contrast with the persistently infected patients, no single SNP accumulated over time within either of these individuals.

Although the data from the isolates of patients 4 and 5 suggest repeat infections, examination of intermediate strains suggests that some genotypes did persist briefly in patient 5. Strains SQ11 and SQ13 (Figure 1) were collected from patient 5 but were not genome sequenced. The collection dates for each of these strains is flanked by the collection of strains that were fully genome sequenced (SQ10, SQ12, SQ14). The genome sequences of strains SQ10 and SQ12 differ by 45 SNPs or indels, whereas strains SQ12 and SQ14 differ at approximately 40 positions. A PCR-based approach was used to examine 8 variable SNPs across the genome that discriminate among the different strains in patient 5 (Supplementary Table 4). The data suggest that strain SQ11 is most similar to strain SQ10, whereas SQ13 is most similar to SQ14. Therefore, there is evidence of short-term (4–5 month) strain persistence in patient 5, against a background of repeated reinfection with new strains.

Although the strains in patients 4 and 5 do not appear to represent long-term persistence within these patients, there are 2 examples of very similar, near clonal serovar Ia lineages being perpetuated in the community during the sampling period. Strain SQ06 in patient 4 is very closely related to strain SQ12 in patient 5, and strain SQ07 in patient 4 is highly related to strain SQ14 in patient 5 (Figure 2). Strain SQ06 differs from strain SQ12 at only 3 genomic locations: (1) a repeat within-histone analog hct2 [18]; (2) a repeat in tarP, which encodes translocated actin recruiting phosphoprotein (TARP) [19]; and (3) a short region of insertion between the rRNA operons (approximately position 847040) (Supplementary Tables 3 and 4). Similarly, strains SQ07 and SQ14 also show differences in hct2 and tarP and have a difference at a single additional position in the genome (position 196184). Examination of the time points at which these strains were collected from the patients indicates that it is not likely that direct transmission was involved in these similarities. For example, strain SQ07 was collected from patient 4 at a time point when patient 5 was infected with SQ12, a largely dissimilar strain. Instead, the data suggest that these strains were circulating and stable in the local population during the study period.

Plasmid sequences were assembled for 6 p0 specimens and each of the sequenced p1 cultured strains. With only 1 exception, there were no differences between plasmid sequences at the beginning and end of the analysis (Supplementary Figure 2). The single exception was an ambiguous in-frame insertion within the ORF encoding Pgp6 in patient 6 (not shown). There were also no examples of differences in plasmid sequence between p0 and p1 in any examined patient sample. Variation in plasmid sequence among strains from the different patients was unremarkable and completely consistent with work published by others [14].

DISCUSSION

Chlamydial persistence in vitro has been well documented. Very different stimuli can lead to chlamydial in vitro persistence in a variety of model systems [20–23], which suggests that persistence is a stress response in the chlamydia–host interactions in vivo. Recently, microscopic evidence of aberrant forms has been documented in the female genital tract in infected patients [24]. However, the occurrence and importance of long-term persistence in vivo remains enigmatic, and the distinction between long-term persistence and repeated infections in sexually active individuals is not clear. In this article, we used whole-genome sequencing to examine strains collected from individuals infected for a long period of time with single uncommon or rare serovars to document and characterize persistent strains in human sexually transmitted infections. The results provide evidence of both persistence and same-serovar repeat infections in these individuals. Evidence for persistence included (1) zero (patient 2) or a very small number of mutations (patients 1, 3, 6, 7) within isolates from a single patient and (2) accumulation of mutations from the beginning to the end of the sampling period. For example, patient 6 was culture-positive several times over the course of 5 years (Figure 3C), and any mutations that were found in any single intermediate strain were also found in the last strain collected from the patient. In contrast, patients 4 and 5 showed clear evidence of repeat infections with genovar Ia strains that had a larger number of mutations that, in no single case, accumulated from intermediate strains to the final isolate. Additional evidence for reinfection in these patients included the presence of a likely horizontally acquired region of the chromosome in strain SQ09 only (Figure 5) and the demonstration that there was a much higher level of identity in strains between patients than within patients (Figure 2). Polymerase chain reaction–based analysis of intermediate strains in patient 5 provided evidence that individual strains did persist for a limited number of months but eventually a new and independent infection event, also with a serovar Ia strain, replaced the previous strain found in this patient. The lack of or very modest levels of mutations in the persistently infected individuals suggest that C. trachomatis can persist in individuals in spite of repeated doxycycline therapy and in the face of an immune response, both of which were apparently ineffective at removing the organism. There was no evidence for antigenic shifts within the persistent strains because there was not a single mutation in any persistent strain that was associated with identified chlamydial antigens. This is surprising because it is known that modest changes in OmpA can lead to significant immune avoidance in animal models, and variants of OmpA/MOMP, generated by either recombination or mutation, are thought to be important mediators that allow chlamydiae to avoid immunological protection [25, 26]. Additionally, patients 2, 6, and 7 were negative by both cell culture and LCR at some point during the study period but were then determined to be positive again with either identical (patient 2) or very nearly identical (patients 6, 7) isolates. Interestingly, during 2 visits, patient 6 and the documented partner were both culture-negative, providing evidence that both had resolution of detectable viable Chlamydia (Figure 3C) during these visits. In contrast, in 1 case (patient 2), viable C. trachomatis was recovered after blind passage of samples that were initially culture- and LCR-negative. Therefore, these individuals were likely positive during the entire period of this study, and the negative diagnostic tests resulted from either chlamydial abundance being lower than the limits of detection and/or sequestration of chlamydiae in different, unsampled body sites. A recently proposed model, which is supported by data collected in both animal and human systems, suggests that the gastrointestinal tract may be a site of refuge for chlamydiae that can then disseminate and recolonize other classical sites of infection (reviewed in [27]). It is therefore possible that infection of the rectum and/or colon might have participated in the maintenance of C. trachomatis within these persistently infected patients.

One possibility for each of the strains that appear to demonstrate persistent infections in patients is that they are members of a genotype that is very tightly conserved and exhibits identity or minor differences in the population in general. To address this, we examined published genomes for serovar J [17], K [14], and H [15] and completed a draft sequence of an additional D− strain [28] and included these sequences in tree-building analysis. Our phylogenetic data demonstrate that strains from several patients were more closely related to strains expressing a different OmpA type than they were to genomes expressing the same serovar, even when the strains were collected in the same geographic region (Figure 2). In each case, there were hundreds of differences between genome sequences when comparing sequence collected from persistently infected patients, in contrast with comparisons with genomes from patients outside of this study population (D−, H, and J). Therefore, variation within even rare serovars is relatively high across the species, and it is not likely that individuals are repeatedly infected with very highly conserved strains.

The sequences of the apparent persistently infected patients in this study showed remarkable genome stability. Other investigators have demonstrated that certain chlamydial genes are variably altered or inactivated during in vitro culture [29–31]. Our sequence analyses of both in vitro−cultured and uncultured strains did not identify any consistent differences in any genes (not shown). However, the number of in vitro passages prior to sequencing of cultured strains was fewer than those examined by others.

Our results add to the understanding of the survival of C. trachomatis in patients and will enhance future studies of chlamydial transmission, recurrence, and persistence. Although seemingly unlikely, the possibility that the patients in this study were continually reinfected with the same strain by the same infected partner for many years cannot be ruled out. Despite this caveat, this study provides the strongest evidence for chlamydial persistence reported to date. It is not yet apparent whether this persistence results from the ability of chlamydia to promote treatment failure by surviving in the presence of antibiotics [32, 33], the inability of the host to resolve the infection through the immune response, variation in genital tract microbiota, or a combination of these factors. Future areas of study should include prospective studies focusing on patients with apparent treatment failures and recurrent infections along with documented partner treatment and test of cure. In addition, other anatomic sites, including the rectum and pharynx could be included as sites possibly harboring infections.

Supplementary Data

Supplementary materials are available at The Journal of 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. The authors thank Dr P. Scott Hefty, Dr Kevin Hybiske, Dr Rajesh Chahota, Dr Julia Dombrowski, and Ms Lotisha Garvin for critical reading of the article.

Financial support. This research was supported by Public Health Service (USA) awards R21AI088540, R01A1099278, and R01AI126785

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.