The composition of the vaginal microbiome, including both the presence of pathogens involved in sexually transmitted infections (STI) as well as commensal microbiota, has been shown to have important associations for a woman’s reproductive and general health. Currently, healthcare providers cannot offer comprehensive vaginal microbiome screening, but are limited to the detection of individual pathogens, such as high-risk human papillomavirus (hrHPV), the predominant cause of cervical cancer. There is no single test on the market that combines HPV, STI, and microbiome screening. Here, we describe a novel inclusive vaginal health assay that combines self-sampling with sequencing-based HPV detection and genotyping, vaginal microbiome analysis, and STI-associated pathogen detection. The assay includes genotyping and detection of 14 hrHPV types, 5 low-risk HPV types (lrHPV), as well as the relative abundance of 31 bacterial taxa of clinical importance, including Lactobacillus, Sneathia, Gardnerella, and 3 pathogens involved in STI, with high sensitivity, specificity, and reproducibility. For each of these taxa, reference ranges were determined in a group of 50 self-reported healthy women. The HPV sequencing portion of the test was evaluated against the digene High-Risk HPV HC2 DNA test. For hrHPV genotyping, agreement was 95.3% with a kappa of 0.804 (601 samples); after removal of samples in which the digene hrHPV probe showed cross-reactivity with lrHPV types, the sensitivity and specificity of the hrHPV genotyping assay were 94.5% and 96.6%, respectively, with a kappa of 0.841. For lrHPV genotyping, agreement was 93.9% with a kappa of 0.788 (148 samples), while sensitivity and specificity were 100% and 92.9%, respectively. This novel assay could be used to complement conventional cervical cancer screening, because its self-sampling format can expand access among women who would otherwise not participate, and because of its additional information about the composition of the vaginal microbiome and the presence of pathogens.
Citation: Bik EM, Bird SW, Bustamante JP, Leon LE, Nieto PA, Addae K, et al. (2019) A novel sequencing-based vaginal health assay combining self-sampling, HPV detection and genotyping, STI detection, and vaginal microbiome analysis. PLoS ONE 14(5): e0215945. https://doi.org/10.1371/journal.pone.0215945
Editor: Maria Lina Tornesello, Istituto Nazionale Tumori IRCCS Fondazione Pascale, ITALY
Received: May 3, 2018; Accepted: April 12, 2019; Published: May 1, 2019
Copyright: © 2019 Bik et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All bacterial and HPV sequence data from human samples as well as pools of human samples are available at the European Nucleotide Archive database under accession number PRJEB25853.
Funding: The funder, uBiome, Inc. provided support in the form of salaries for all authors, and decided which initial targets to include in the study design, but did not have any additional role in data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of each author are articulated in the ‘author contributions’ section.
Competing interests: All of the authors of the paper are current or past employees of uBiome, Inc. and have received stock options as well as other compensation. Some authors have patents pending in relation to this work: US Application No 15/198,818, Method and system for diagnostic testing, Application No 16/084,945, Method and system for microbiome-derived diagnostics and therapeutics for bacterial vaginosis, and Application No 16/115,542, Method and system for characterization for female reproductive system-related conditions associated with microorganisms. The data in this article were used in the development of a commercially available test product developed and marketed by uBiome. This does not alter our adherence to PLOS ONE policies on sharing data and materials.
A woman’s vaginal health is critical for her general well-being and reproductive success, and is in part determined by the vaginal microbiome composition, the presence of pathogens associated with sexually transmitted infections (STI), and the presence of human papillomavirus (HPV) types that can cause genital warts or cervical cancer. Current clinical vaginal health assays focus on the detection of STI or that of HPV, but there is no single test that combines these targets with vaginal microbiome analysis or with self-sampling.
The human vaginal microbiome has a unique composition compared to other microbial communities in the human body. In most healthy women, the vaginal microbiome is characterized by a low bacterial diversity and a dominance of lactobacilli, with low abundance of other bacterial genera [1,2]. Several vaginal microbial community types have been described, most of which are dominated by a single Lactobacillus species. Lactic acid produced by lactobacilli lowers the vaginal pH, which is believed to create an environment unfavorable for the growth of pathogenic bacteria . Low numbers of vaginal lactobacilli have been associated with many health conditions, such as bacterial vaginosis [1,4–6], aerobic vaginitis , cervicitis , and STI [9–12]. The composition of a woman’s vaginal microbiome thus plays an important role in women’s health and reproductive success. Yet, the analysis of this microbial community is not part of regular health care for women. In the US, as in many other countries, most healthcare providers instead focus on the detection of high-risk human papillomavirus (hrHPV), the predominant cause for cervical cancer.
Cervical cancer is one of the major causes of cancer-related deaths in women, with an annual worldwide mortality of 250,000 [13,14]. hrHPV DNA can be detected in almost all (>99%) cervical cancer specimens, and HPV is therefore considered the predominant causative agent for cervical cancer [15,16]. Although HPV infection is the most common STI worldwide, not all HPV infections will lead to cancer. Firstly, certain HPV types have higher oncogenic risks than others. Of the over 170 different HPV genotypes known to date, twelve types have been classified as Group 1 human carcinogens; these include types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, and 59 [17,18]. Together with other closely related HPV types, such as 66 and 68, which have been listed as probably or possibly carcinogenic, these are collectively called hrHPV types. hrHPV types 16 and 18 can be found in over 70% of cervical cancers  and the presence of these types is associated with the highest chance of developing cancer within 10 years . However, other hrHPV genotypes have also been shown to cause cervical cancer, and especially among women of non-European descent . In addition, many hrHPV infections are temporary and will be cleared within months of acquisition, without proceeding to pre-cancerous lesions . Other HPV types, collectively called low-risk HPV (lrHPV), are not implicated in cervical cancer, but instead cause genital warts .
National cervical cancer screening programs are offered to women worldwide with a starting age between 20 and 35 years old . Most of these programs involve an invitation for a Pap smear, in which a woman’s cervical cells are obtained by a physician for cytology , but additional molecular HPV testing is increasingly offered by health care providers as well . In the United States, most healthcare providers follow the American College of Obstetricians and Gynecologists (ACOG) guidelines  or the U.S. Preventive Services Task Force (USPSTF) guidelines for women  to come in for a Pap smear, often with HPV testing, every three to five years, depending on age and risk factors. The most recent USPSTF guidelines, released in August 2018, even include HPV testing alone as a recommended option for women over 30 .
Several commercial kits have FDA pre-market approval for the molecular detection of HPV . In a 2014 meta-analysis of 36 studies, the Qiagen digene Hybrid Capture 2 (HC2) assay was the most widely used . In the HC2 assay, cervical specimens are denatured with sodium hydroxide, denatured viral DNA is hybridized with specific RNA probes, and RNA:DNA hybrids are subsequently detected with antibodies . The HC2 test detects 13 hrHPV types, but does not report which specific type is present. Other HPV detection assays involve the amplification of viral DNA by Polymerase Chain Reaction (PCR). The most widely used primer pairs for HPV PCR detection are the GP5+/6+ primers  and the degenerate MY09/11 primers  or the PGMY09/11 primer pool , which are all based on conserved regions in the viral L1 open reading frame. The COBAS 4800 assay detects 14 hrHPV types using multiplex real-time PCR with specific probes; it reports the presence of HPV16, HPV18, or one of 12 remaining hrHPV types .
The vaginal microbiome is an emerging area of research in understanding the role of HPV infections and reducing the risk of cervical cancer . Several studies suggest a relationship between the composition of the vaginal microbiota and the acquisition and persistence of HPV infection. For example, vaginal microbial diversity is increased during an HPV infection, with decreased levels of Lactobacillus species and an increased presence of other microbial members such as Sneathia species or Gardnerella vaginalis [38–42]. In addition, certain microbiota compositions are associated with increased clearance of detectable HPV .
In this study, we tested the feasibility of a novel assay, that combines the detection and identification of HPV DNA, STI-associated pathogens, and microbiome analysis on samples obtained through self-sampling. We validated the performance of marker gene amplification and sequencing to detect the presence and relative abundance of 31 clinically important bacterial targets with high precision and accuracy. In addition to detecting Lactobacillus, Sneathia, and Gardnerella spp., this test detects STI-associated pathogens including Chlamydia trachomatis, Mycoplasma genitalium, and Neisseria gonorrhoeae, which combined infect over 150 million people each year and cause genital tract infections and cervicitis [10–12]. The performance of a novel amplification and sequencing-based strategy for HPV detection and type-specific identification was compared to that of the most widely used test for HPV detection in cervicovaginal specimens, the digene HC2 test. This assay was not developed to be employed in a clinical setting, but intended to complement, rather than replace, current healthcare guidelines for in-clinic cervical cancer screening.
Materials and methods
Study participants and sample collection
This study was approved under a Human Subjects Protocol provided by an independent IRB (E&I Review Services, IRB Study #13044, 05/10/2013). E&I is fully accredited by the Association for the Accreditation of Human Research Protection Programs, Inc. (AAHRPP), with Registration # IRB 00007807. uBiome undergoes a yearly voluntary continuing review by E&I Review Services to determine that the project meets the same scrutiny of human subjects protections as projects conducted at research institutions. The IRB board membership of E&I Review Services is consistent with the Code of Federal Regulations (CFR) requirements of 21 CFR 56.107 and 45 CFR 46.107. Where required, specimens used in this study consisted of vaginal samples from women who had signed an informed consent to have their samples used for research. All participants were 18 years or older.
A vaginal self-collection kit was sent to each participant’s home address, consisting of a sterile swab, a tube with sterile water, a tube with zirconia beads in a proprietary lysis and stabilization buffer that preserves the DNA for transport at ambient temperatures, and sampling instructions (S1 Fig, included in the Supplementary Materials). Participants were instructed to wet the swab with the sterile water, insert the swab into the vagina as far as is comfortable, make circular movements around the swab’s axis for 1 minute (min), and then stir the swab for 1 min into the tube with lysis buffer and beads. After shaking the tube for 1 min to homogenize, the tube was then shipped by the participants to the laboratory by regular mail.
For the determination of the reference ranges of the 31 bacterial targets, a set of 50 vaginal specimens, each from a different woman (average age 48.4 ± 15.6 years; range 23 to 79 years), was selected. Inclusion criteria were the following: completion of the voluntary health survey that every woman was invited to participate in, and no self-report of any of the following conditions: bacterial vaginosis, cervical cancer, genital herpes or warts, urinary tract infection, or infection with HPV, C. trachomatis, T. pallidum, or yeast. In addition, all of these women reported no antibiotic usage in the six months before sampling.
A different set of specimens from 87 women was used to compare the performance of sampling with the digene collection device (Qiagen, Gaithersburg, MD, USA) and DNA extracted from samples collected with swabs. For this subset, women were asked to self-sample 2 vaginal specimens within 15 minutes. The first specimen was collected by using the digene collection device, which consists of a cervical brush and a digene transport tube with Specimen Transport Medium (STM). The second specimen was collected using a pre-wetted swab and resuspended in a collection tube with lysis buffer and beads, as described above, and used for DNA extraction.
For use in spiking and intra-run technical repeatability experiments described below, homogenized “vaginal pools” were created by combining 96 vaginal samples derived from 11 or 16 individuals who sampled themselves multiple times. These pools were created to make a relative large amount of a homogenized complex matrix similar to real biological samples. The pool created from samples taken by 16 subjects was used to test the synthetic DNA in spiking experiments, while the pool of samples taken by 11 subjects was used in the technical repeatability experiment (see below).
An additional set of 718 vaginal specimens were selected to compare the performance of the digene HC2 HPV test using the hrHPV probe (601 specimens) and lrHPV probe (148 specimens; overlap of 31 specimens), respectively, versus that of the amplification and sequence-based HPV type identification described in this study. Of these, 361 samples were from subjects who consented, while the remaining samples were residual clinical samples that were analyzed anonymously.
Positive STI control samples
Ten de-identified cervicovaginal swab specimens of known STI pathogen status were obtained through a commercial source (iSpecimen, Lexington, MA). Five of these samples were reported to be positive for C. trachomatis and negative for N. gonorrhoeae, while a second set of five samples were negative for C. trachomatis and positive for N. gonorrhoeae. Each sample was tested in five replicates for DNA extraction, 16S rRNA gene amplification, and target identification as described below. To confirm the presence of M. genitalium in 13 samples that tested positive and 9 samples that tested negative for this pathogen in the assay using the 16S rRNA gene amplification described in this study, a PCR was performed using primers on the MgPa adhesin gene .
In silico 16S rRNA gene target performance metrics
The assay includes 31 bacterial targets with clinical relevance for women’s reproductive tract health, which were identified through an exhaustive literature search (Fig 1). The most relevant associations between health conditions and the vaginal microbiota were narrowed down by choosing associations with high statistical significance that were found in humans subjects, not in laboratory animals or bioreactors, but performed on case/control, cohorts or randomized studied population. These include bacterial vaginosis [1,4–6], aerobic vaginitis , pelvic inflammatory disease , and sexually transmitted infections [10–12]. A complete list of these associations and references are provided in S1 Table. For each bacterial taxon intended to be included in this assay, using a process similar to that described in Almonacid et al. , we determined in silico performance metrics for identification of each taxa (sensitivity, specificity, positive and negative predictive value). Briefly, sequences assigned to each taxon in the SILVA database (Version 123)  were considered to be real positives for that taxon. Then, assuming amplification with up to two mismatches with the primers used, we identified for each taxa the sequences that would produce an amplicon, and evaluated whether that amplicon is unique to the taxon of interest (ti) or also shared by sequences from different taxa (dt). The number of true positives (TP), true negatives (TN), false positives (FP) and false negatives (FN) was computed for different tolerance ratios for the quotient dt/ti, and subsequently in silico performance metrics were assessed. Of the 72 bacterial targets initially selected, the 31 targets selected for the assay had all four in silico performance metrics above 90% (S2 Fig; S2 Table).
Dark blue dots indicate targets positively associated with the conditions, while pink diamonds indicate inverse associations. See S1 Table in the Supplementary Materials for more detailed information about e.g. the HPV genotypes included and a list of references.
In silico HPV target performance metrics
In addition to the 31 bacterial targets, hrHPV and lrHPV targets were selected for inclusion in the assay, based on their published association with cervical cancer lesions or genital warts (Fig 1, S1 Table). HPV reference genomes were downloaded in August 2017 from the PaVE database, which is a repository of curated and annotated HPV genomes [47,48]. Only revised and recognized sequences (180 HPV genomes) were used for an in silico PCR amplification using a set of 15 forward and 6 reverse primers (described below) targeting the L1 gene and allowing up to 4 mismatches between primers and target sequences. Under these conditions, the L1 genes from 118 HPV genomes could be amplified in silico. Of these, 19 HPV genomes, including 14 hrHPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, 68) and 5 lrHPV types (6, 11, 42, 43, 44) were selected based on their association with health conditions according to literature (S1 Table). In order to evaluate the performance metrics for identification of the HPV targets, sequences of the L1 segment of HPV genomes from the NCBI database were used. The search was filtered to sequences with length in the range 1,500–10,000 bp and with correct assignment of the type of the HPV (4177 sequences). These sequences were amplified in silico using the primers described below. Following these steps, we generated 161,398 amplicons. These sequences were mapped using VSEARCH  at 95% of identity against an HPV amplicon reference database consisting of the amplicons produced by the reference genomes in PaVE for the 19 HPV types included in our assay. The performance metrics were calculated as described above for the 16S rRNA gene targets. Briefly, the correct assignment of an in silico NCBI amplicon against the reference was counted as a true positive and an incorrect assignment was considered as a false negative. Also, we considered as a false negative any genome from NCBI that our primers could not amplify. According to this, the 19 HPV types obtained values for sensitivity, specificity, positive predictive value (PPV) and negative predictive values (NPV) above 90% (S3 Table).
In vitro validation of bacterial targets
To test our ability to identify members of each of the 31 bacterial targets, synthetic DNAs (sDNAs or gBlocks, Integrated DNA Technologies, Inc., Coralville, IA), were designed encompassing the V4 region of the 16S rRNA gene including primer regions, based on a SILVA representative sequence, plus 75 additional bases to both the 5’ and 3’ side, with one sDNA per target (S4 Table). The SILVA representative sequence per taxa was chosen by performing an all-against-all sequence comparison of all sequences in a taxa, and identifying as representative the sequence that shared the highest similarity with the largest number of sequences in the set.
To validate that each target could be detected in a vaginal swab specimen, 3 ng of each sDNA was spiked into 500 μl aliquots of a vaginal pool, created by combining 96 vaginal specimens from 16 women included in this study, and DNA was extracted from each spiked vaginal pool (see below). Each spike-in experiment was performed in triplicate. Subsequently, bacterial targets were detected by amplification using PCR targeting the 16S rRNA gene, sequencing, and a bioinformatics pipeline described below. Each target included in the final panel was detected above limit of detection (LOD) (see below) in each of the triplicate spiked-in amplification reactions performed on the extracted DNA from the vaginal pool (not shown).
In addition, the LOD of each target in a complex background of other targets was determined according to published guidelines . First, to check for potential contamination, we calculated a limit of blank (LOB), which was calculated using a set of 77 blank wells of a 96-well PCR plate where wells of the first row and first column of the plate each contained 200 pg/μl of synthetic 16S rRNA gene DNA from different targets. The LOB was set as the average number of reads in these blank wells (18.57 reads) plus 1.65 standard deviations (29.70 reads), thus at 48.27 reads. To calculate the LOD of the bacterial target, pools of bacterial sDNAs were mixed in different ratios. To create these mixes, each bacterial sDNA was randomly assigned to one of two pools, A and B, that each contained sDNAs in equimolar amount. Each pool was serially diluted in PCR grade water. Pool A dilutions were mixed 1:1 with undiluted Pool B and vice versa. All pool A/B combinations were used in triplicate for DNA extraction, amplification, and sequencing as described below. For each target, the LOD was defined as the lowest concentration of sDNA where at least two of the three replicates contained at least 2 reads for that target in a sample with 10,000 reads or more. Using this LOD, we calculated a lower threshold for detection for each taxa at its LOD as the LOB (48.27) plus the standard deviation of the taxa at LOD * 1.65. This threshold is used to correctly assign a taxa as identified in a sample at or above its LOD.
For targets that had both a species and a genus level sDNA present in the mixed pools A and B, a bioinformatic correction was applied. The total reads for a genus-level target for which a species within that genus was also present in the mixed pools, was defined as the total measured reads for the genus and subtracting all those reads corresponding to species-level targets belonging to that genus in the same pool mix, i.e., only reads that match to a genus and not to a species level were finally assigned to the genus.
In vitro validation of HPV targets
To test the ability of our assay to detect and genotype HPV targets, fragments of the L1 gene of approximately 600 bp long were ordered for each of the representative sequences of 19 HPV types in the PaVE database as sDNAs (gBlocks, Integrated DNA Technologies, Inc.). To represent hrHPV type 68, two sDNAs were ordered, 68a and 68b. The sequences of the 20 gBlocks representing 19 HPV types (14 hrHPV and 5 lrHPV) are listed in S5 Table. To validate that each target could be detected in a vaginal swab specimen, 3 ng of each HPV sDNA was spiked into 500 μl aliquots of a vaginal pool created by combining 96 vaginal specimens from 16 women included in this study, and DNA was extracted from each spiked vaginal pool (see below). Subsequently, the spiked HPV targets were detected by amplification using the PCR targeting the L1 gene and bioinformatics pipeline described below. Each spike-in experiment was performed in triplicate. Each HPV target was detected above the LOD (see below) in each of the triplicate spiked-in amplification reactions performed on the extracted DNA from the vaginal pool (not shown). Each target had a ratio > 0.1 for the number of HPV-assigned reads divided by the total number of normalized reads assigned to an internal spike-in control (see below).
To determine the LOD of HPV targets, 10-fold serial dilutions of the sDNAs representing HPV targets were made in nuclease-free water, ranging from 105 to 102 molecules per μl. Dilutions of one target were inversely combined with dilutions of another target, forming different pairs of HPV sDNAs. Each dilution pair was used directly as template for PCR in triplicate as described below.
DNA extraction and amplification targeting 16S rRNA and HPV L1 genes
DNA was extracted from vaginal specimens, pools thereof, or sDNA dilutions in tubes containing lysis/stabilization buffer as described previously . For 16S rRNA gene amplification, extracted DNA was used as the input of a one-step PCR protocol to amplify the V4 variable region of the 16S rRNA gene. This PCR contained universal primers 515F and 806R [45,51], both with sample-specific indices and Illumina tags. PCR was performed as described before . Following amplification, DNA was pooled by taking the same volume from each reaction.
For HPV amplification, extracted DNA was used as the input of a PCR protocol to amplify the HPV L1 gene. To each sample, sDNA with a randomized HPV type 16 sequence was added as an internal positive control. The first PCR mix contained a pool of previously described HPV specific primers [35,52], and two new primers, HPV_RSMY09-LvJJ_Forward: 5’ CGTCCTAAAGGGAATTGATC, and HPV_PGMY11-CvJJ_Reverse: 5’ CACAAGGCCATAATAATGG. All these primers contained sequencing adaptor regions. The PCR products from the first amplification round were used as input for a second PCR containing sample-specific forward and reverse indices and Illumina tags. PCR products from this second step were pooled for sequencing.
The 16S rRNA gene and HPV PCR consolidated library pools were separately quantified by qPCR using the KAPA Library Quant Kit (Bio-Rad iCycler qPCR Mix) following the manufacturer’s instructions using a BioRad MyiQ iCycler. Sequencing was performed in a paired-end modality on the Illumina NextSeq 500 platform rendering 2 x 150 bp paired-end sequences.
Sequence analysis and taxonomic annotation for bacterial targets
After sequencing, demultiplexing of reads according to sample-specific barcodes was performed using Illumina’s BCL2FASTQ algorithm. Reads were filtered using an average Q-score > 30. Forward and reverse 16S rRNA gene reads were appended together after removal of primers and any leading bases, and clustered using version 2.1.5 of the Swarm algorithm  using a distance of one nucleotide and the “fastidious” and “usearch-abundance” flags. The most abundant sequence per cluster was considered the real biological sequence and was assigned the count of all reads in the cluster. The representative reads from all clusters were subjected to chimera removal using the VSEARCH algorithm . Reads passing all above filters (filtered reads) were aligned using 100% identity over 100% of the length against the true positive 16S rRNA gene sequences identified in silico from SILVA for each of the 31 taxonomic groups targeted by the assay as described above (S4 Table). The relative abundance of each taxon was determined by dividing the count linked to that taxa by the total number of filtered reads.
Sequence analysis and taxonomic annotation for HPV targets
Raw sequencing reads were demultiplexed using BCL2FASTQ. Primers were removed using cutadapt . Trimmomatic  was used to remove reads with a length less than 125 bp, and a mean quality score below 30. After that, forward and reverse paired reads were joined using custom in-house scripts and converted to a fasta file. Identical sequences were merged and written to a file in fasta format and sorted by decreasing abundance using—derep_fulllength option in VSEARCH . Target sequences in the fasta files were compared to the fasta-formatted query database sequences (19 HPV target sequences) using the global pairwise alignment option with VSEARCH, using 95 percent sequence identity, to obtain the counts for each HPV type within a different sample.
The HPV portion of the assay was considered positive if the number of sequence reads assigned to the specific HPV types was above the threshold at the limit of detection, and greater than a previously defined cutoff. To set this cutoff, two normalization steps were employed. First, according to in silico PCR amplification, a different number of combinations of primers amplify different HPV targets (e.g. HPV16 is amplified using 66 different combinations, while HPV43 is amplified with just 10 combinations), reflecting the sequence variability within the primer binding site among HPVs. This also means that the spiked-in internal control and the target HPV have different amplification efficiencies. To avoid this bias, the internal control (which has the primer sites for HPV16) is normalized for the amplification factor (number of primer combinations that generate an amplicon) of each HPV type. The number of HPV-assigned reads was divided by the total number of normalized reads assigned to the spike, and a sample was considered HPV-positive if that ratio was above 0.1, which corresponds to approximately 500 target molecules.
Intra- and inter-run precision
Intra-run technical repeatability was assessed by including nine replicates of the same vaginal pool (consisting of 96 vaginal samples derived from 11 individuals) into the same DNA extraction, 16S rRNA gene amplification, and sequencing run. This experiment was then repeated in two additional sequencing runs to yield three sets of nine replicate samples analyzed within the same run. In addition, inter-run technical reproducibility was performed by processing three replicates of a set of 18 vaginal samples on three different days by three different operators. Samples included in the analysis were those that had at least 10,000 reads and where at least two of the three replicates were present (11 sets).
Comparison of the results, both intra- and inter-run, were done using the raw counts of the 31 bacterial species- and genus-level targets. Data was processed using the R-package Phyloseq , and visualized using Principal Coordinates Analysis (PCoA), based on a distance matrix calculated using the Bray-Curtis method.
digene HC2 hrHPV test on digene tubes or on extracted DNA
The digene HC2 HPV detection assay (Qiagen) was used as a reference to validate the hr- and lrHPV portions of the assay. The High-Risk HPV Probe in the digene HC2 HPV test detects hrHPV types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68, while the Low-Risk HPV Probe in the digene test detects lrHPV types 6, 11, 42, 43, and 44. The vaginal health test described here detects all these types plus hrHPV type 66. The digene HC2 assay is intended to be used directly on cervical samples collected in the digene STM transport tube. In order to validate the use of the digene kit on extracted vaginal DNA, we first compared the performance of the digene HC2 assay on a set of 87 self-sampled, paired vaginal samples, i.e. a digene brush resuspended in an STM transport tube, as well as DNA extracted from a vaginal swab resuspended in lysis and stabilization buffer according to the manufacturer’s instructions. Negative and positive controls and calibrators included in the kit were processed within each 96-well assay, and used for assay validation and cutoff, as per instructions. A specimen was considered positive if its chemiluminescence measurement (Relative Light Units, RLU) was higher than or equal to that of the assay’s Positive Calibrator cutoff (RLU ratio of 1 or more), as specified in the digene HC2 assay instructions.
Sensitivity, specificity, and accuracy of the hr- and lrHPV portions of the vaginal health assay were evaluated using the digene HC2 HPV DNA assay as the gold standard, and extracted DNA from 601 (hrHPV) and 148 (lrHPV) vaginal swabs, respectively, as the input for both tests. The vaginal health assay was considered to be positive for a HPV type if the number of reads assigned to that HPV type divided by the normalized number of reads assigned to a spiked-in control (see above) was greater than 0.1. Agreement between the two methods was evaluated using Cohen’s kappa , where the level of agreement is defined by the range: 0–0.20, poor; 0.21–0.40, fair; 0.41–0.60, moderate; 0.61–0.80, good; 0.81–1.00, very good.
Limit of detection of bacterial and HPV targets
The vaginal health assay described here is based on a list of 31 bacterial 16S rRNA gene targets and 19 HPV types that were identified through an exhaustive literature search to play important roles in health and disease of women’s reproductive tracts (Fig 1, S1 Table). For each bacterial target, the LOD was determined by combining different dilutions of pools of sDNAs, followed by DNA extraction, amplification of the V4 region of the 16S rRNA gene using broad-range primers, and sequencing (Fig 2A). The LOB was set as the average number of reads in 77 blank wells (18.57 reads) plus 1.65 standard deviations (29.70 reads). Using this value, we calculated the threshold of identification for each taxon as the LOB + 1.65 standard deviations (48.27) plus the standard deviation of the taxon at LOD * 1.65. For the 31 taxa targeted by the assay, the threshold related to LODs was in the range 49.0 to 59.1reads (S6 Table).
Dilutions of two pools of sDNAs were mixed in different amounts, and microbial targets were amplified and sequenced. For each dilution and target, the relative abundance in samples with 10,000 reads or more are shown. A. LOD of bacterial targets. LOD read thresholds are provided in S6 Table. B. LOD of HPV targets. Two different sDNAs were used to represent hrHPV type 68. For each dilution and HPV type, the relative abundance in samples with 10,000 reads or more are shown. The LOD read thresholds for each HPV target are provided in S7 Table.
To determine the LOD for the HPV targets, different dilutions of pools of sDNAs were mixed as done for the bacterial targets. The molecules were then amplified, sequenced, and analyzed by the HPV bioinformatics pipeline. For all HPV targets analyzed, the threshold related to LODs was in the range 40.8 to 224.8 reads (Fig 2B, S7 Table).
Intra- and inter-run variability
Intra-run technical variability was evaluated in a combined set of 18 replicates of the same vaginal pool, each of which yielded 10,000 reads or more. Ordination plots of both genus and species level bacterial taxa (Fig 3) showed a tight clustering of intra-run technical replicates, indicating that within a single sequencing run, results generated by the laboratory process and the bioinformatics analysis were consistent.
PCoA ordination showing clustering of inter-run (11 vaginal samples analyzed in triplicate on three independent sequencing runs) and intra-run (18 aliquots of the same vaginal pool) data, at the genus and species taxonomic level. Shapes indicate the sequencing run, while colors indicate sample replicates. For each category (Species and Genus) samples were processed together, sharing the same scale in the visualization.
For the inter-run analysis, a total set of 11 groups of replicates (at least two samples) passed the filtering criteria (over 10,000 reads). The PCoA visualization at genus and species level showed a dispersion of the different samples, but with a clustering according to the respective replicates (Fig 3). This suggests that there is limited within-sample variation when the same samples are processed on different days by different operators.
Relative abundance of bacterial targets in healthy vaginal samples
To determine reference ranges for the 31 bacterial targets in the assay, we selected a set of 50 vaginal samples from our database. These represent self-reported healthy individuals from the uBiome microbiome research study. In addition to health status, additional selection criteria included no usage of antibiotics six months prior, and no current urinary tract or vaginal infections, including the presence of STDs. The 50 samples were processed with the target database, and the relative abundance ranges for each bacterial target in the cohort are shown (Fig 4).
A set of 50 vaginal samples, each from a different woman, was selected based on the self-reported answers given to survey questions indicating general and vaginal health. Each dot represents the relative abundance of a different bacterial target on genus level (top) or species level (bottom) within a different vaginal sample. Boxes indicate the 25th-76th percentile, with the median indicated inside each colored box. Red lines indicate the 99% percentile of each distribution, and are also the cutoff for the reference ranges. Not all of the taxa used in the assay were plotted, as some had no abundance values for this healthy cohort (Papillibacter, C. trachomatis, M. mulieris, M. genitalium, N. gonorrhoeae, P. amnii), based on the exclusion criteria.
As expected, given the nature of the samples, Lactobacillus was the most abundant genus, with the widest abundance distribution. At the species level, a similar distribution of the relative abundances was found, including a wide range and a high relative abundance for Lactobacillus iners.
Among the 31 bacterial targets in the assay are three pathogens implicated in STI: C. trachomatis, N. gonorrhoeae, and M. genitalium. The performance of the assay to detect two of these pathogens was confirmed on a set of ten clinical samples available through a commercial source, five of which were positive for C. trachomatis, and five of which were positive for N. gonorrhoeae. A vaginal pool consisting of samples derived from 11 healthy individuals was included as a control sample, and was found to be negative (Fig 5).
Ten de-identified clinical verification specimens (iSpecimen) containing either C. trachomatis (n = 5) or N. gonorrhoeae (n = 5), as well as a vaginal pool (VP) constructed by combining 96 vaginal samples from 11 individuals, were tested for the presence of either pathogen using 16S rRNA gene amplification and sequencing. Five replicates of each specimen were tested. The heatmap shows the relative abundance of the two pathogens in each replicate experiment, on a scale from light yellow (absent) to dark purple (100% relative abundance).
The three STI-associated targets (C. trachomatis, M. genitalium, and N. gonorrhoeae) were not present in any of the 50 samples from the healthy subject set (see also below), nor in a set of 87 vaginal samples used to validate the performance of the digene test on extracted DNA (see below). Twelve of thirteen positive M. genitalium samples found in a larger set of samples used to compare the HPV genotyping part of the assay (see below) were confirmed to be positive in an M. genitalium specific adhesin PCR described by others , while eight out of nine negative samples were confirmed to be negative, corresponding to a sensitivity of 92.3% and a specificity of 88.9% (not shown).
Performance of digene HC2 HPV test on extracted DNA
In order to validate the use of the digene kit on extracted vaginal DNA, we compared the performance of the digene HC2 HPV assay on a set of 87 self-obtained, paired cervicovaginal samples, i.e. a digene brush resuspended in digene STM, as well as DNA extracted from a paired vaginal swab resuspended in lysis transport medium. Of the 87 samples, 84 showed concordant results (69 were negative and 15 were positive for HPV in both tests) (Table 1). Three sample pairs that were positive with the digene STM sample were HPV negative when the corresponding test was performed on extracted DNA. These three samples had an average digene RLU ratio of 1.94, suggesting that these contained low levels of HPV (Fig 6). Agreement between digene STM and digene DNA was of 96.6%, with a sensitivity of 83.3%, a specificity of 100%, and a Cohen’s Kappa of 0.89 ± 0.12 (Z = 8.39, p-value = 0.0001).
Samples were tested directly from STM tubes or from a paired sample after DNA extraction. The purple lines show the cutoff of the digene assay (RLU ratio = 1). Three STM specimens were positive for hrHPV with an average RLU ratio of <2 (low positive), but below RLU ratio = 1 for their corresponding extracted DNA specimen. The results for all other 84 specimens were concordant. TN, true negative; TP, true positive; FN, false negative; FP, false positive.
One set of samples was collected using a digene brush resuspended in digene Specimen Transport medium (“digene STM”), and the second set was extracted DNA from swabs suspended in tubes with lysis/stabilization buffer (“digene DNA”). Samples were considered to be HPV positive if the RLU ratio was 1 or more, as instructed in the digene protocol.
Performance of the hrHPV sequencing test on clinical samples
Using 601 vaginal specimens, the performance of the assay to detect hrHPV was compared to that of the digene HC2 hrHPV assay. Of the 601 samples, 504 were negative in both tests, while 69 were positive in both tests (Table 2). Unexpectedly, three samples that were positive by sequencing for hrHPV 66, which is not covered by the digene hrHPV probe, were still positive in the digene assay. Ten digene-positive samples did not yield any validated hrHPV reads after amplification and sequencing by our assay. Of these ten false negatives, six samples were found to contain single or mixed non-high risk HPV strains by genotyping, including HPV 30, 61, 40, 42, 53, and 67. In addition, eighteen samples were negative in the digene HC2 hrHPV assay but yielded sufficient hrHPV reads (of types 16, 31, 35, 51, 52, 56, 59, and 68b) to be identified as positives by our genotyping assay. Thus, in comparison to the digene HC2 hrHPV test, the hrHPV sequencing assay had a sensitivity of 87.3% and a specificity of 96.6%, with an overall agreement of 95.3%, and a Cohen’s kappa of 0.804 ± 0.070 (Z = 19.8; p-value <0.0001). After removal of the six samples where cross-reactivity of the digene hrHPV probe with lrHPV sequences was suspected, the sensitivity and specificity of the hrHPV genotyping assay were 94.5% and 96.6%, respectively, with a kappa of 0.841 ± 0.065.
The genotyping assay was considered positive if the normalized number of reads assigned to any of the validated hrHPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68) divided by the number of reads assigned to a spike in control was greater than 0.1. The digene test was considered positive if the measured RLU was equal to or greater than the assay’s cutoff (RLU ratio of 1 or higher), as per the manufacturer’s instructions.
Good correlation was found between the number of normalized hrHPV sequencing reads and the digene HC2 hrHPV RLU ratios, confirming that the PCR and sequencing hrHPV assay described here can not only detect hrHPV types but also assess their relative abundance (Fig 7A).
DNA from 718 vaginal samples was extracted and tested by PCR amplification and sequencing using HPV primers, and additionally used directly in the digene assay using the HC2 hrHPV (panel A) or lrHPV (panel B) probe mix. For each sample, the x-axis shows the normalized ratio of reads assigned to validated HPV types over reads assigned to a spiked-in internal control, while the Y-axis shows the digene HPV probe RLU values normalized over the assay’s cut-off RLU. The two purple lines show the cutoff for each of the assays. A. Comparison of hrHPV test results in a subset of 601 samples. Six samples that were positive in the digene hrHPV assay and negative in the hrHPV genotyping assay, but in which lrHPV sequences were detected by genotyping, are shown as light purple triangles. B. Comparison of lrHPV test results in a subset of 148 samples.
Performance of the lrHPV sequencing test on clinical samples
Using 148 vaginal specimens, the performance of the assay to detect lrHPV by genotyping was compared to that of the digene HC2 lrHPV assay. Of the 148 samples, 118 were negative in both tests, while 21 were positive in both tests (Table 3). No false negatives were found, but nine samples that were negative in the digene lrHPV assay were found to have lrHPV sequences by genotyping, of types 42, 43, and 44. Using the digene HC2 lrHPV test as the gold standard, the lrHPV genotyping assay was found to have a sensitivity of 100% and a specificity of 92.9%, with an overall agreement of 93.9%, Cohen’s kappa = 0.788 ± 0.131 (Z = 9.81, p-value < 0.0001).
The genotyping assay was considered positive if the normalized number of reads assigned to any of the validated lrHPV types (6, 11, 42, 43, 44) divided by the number of reads assigned to a spike in control was greater than 0.1. The digene test was considered positive if the measured RLU was equal to or greater than the assay’s cutoff (RLU ratio of 1 or higher), as per the manufacturer’s instructions.
As with the hrHPV assay, the number of normalized lrHPV sequencing reads was positively correlated to the digene HC2 lrHPV RLU ratios, confirming that the PCR and sequencing lrHPV assay described here can not only detect lrHPV types but also assess their relative abundance (Fig 7B).
Positive clinical samples for hrHPV and lrHPV types
In the combined 718 samples used in the comparison of the vaginal health assay to the digene HC2 HPV assay, 142 samples were found to be positive for 1 or more of the 19 HPV types. Of these, 107 samples contained only a single HPV type, while 35 samples contained 2 or more HPV types. Among the 100 samples that contained at least one of the 14 validated hrHPV types, HPV types 52 (13%), 68 (13%), 59 (12%), and 16 (11%) were the most common. Within the 60 samples positive for at least one of the 5 validated lrHPV types, HPV type 42 (43%) and 6 (30%) were the most common, while type 11 was not found (Fig 8).
A set of 718 vaginal samples was tested in the HPV genotyping assay described in this study. Of these, 142 samples were positive for at least one of the 19 HPV types validated in this study, with 100 samples positive for hrHPV, and 60 for lrHPV. In 35 of the 142 positive samples, two or more HPV types were found. Positive samples are plotted as percentage found in the total set of 718 samples.
Here, we describe a novel vaginal health assay combining vaginal microbiome analysis, STI-associated pathogen detection, and HPV detection and identification in a self-sampling format. Although each of these components have been described before, to our knowledge, this assay is the first to combine all of these parts, thus offering women a unique opportunity to gain a broad perspective into their vaginal and reproductive health.
The detection of hrHPV in combination with vaginal self-sampling has been proposed as an effective method for cervical cancer risk screening . Although the sensitivity of probe hybridization-based hrHPV detection, such as the digene HC2 assay, in self-obtained vaginal swabs has been found to be slightly lower than that in clinician-obtained cervical specimens, hrHPV detection based on PCR was shown to be equally sensitive in self-sampled specimens . While the vaginal health assay described in this study is not intended to replace regular cervical cancer screening programs, offering women the opportunity to self-collect vaginal specimens poses fewer barriers for women to be screened, and thus could lead to increased participation rates [59,60]. Therefore, allowing self-collection of vaginal samples for hrHPV screening in parallel to regular screening programs, as already implemented in numerous countries, and recommending women to seek further physician examination in case of a positive result may have a positive impact on rates of detection of cervical cancer, and potentially save lives .
The vaginal health assay described here not only detects whether HPV is present in a sample, but also identifies the presence of specific type(s) by using sequencing analysis. Several other HPV genotyping assays based on PCR and sequencing have been reported [30, 61–65]. These studies detected HPV types not found by traditional methods , as well as infections with multiple types [63–65] with high sensitivity. As recommended by the VALGENT study framework , we compared the performance of the HPV component of the test to that of the widely used digene HC2 HPV assay. Because the novel test reported here is performed on extracted DNA, we first validated the use of the digene assay on extracted DNA. The digene performance on the extracted DNA was slightly less sensitive than that directly performed on the digene STM tubes. The digene HC2 assay has been found to give discordant results in about 8% of paired tests [67,68], where, for example, a positive sample will test negative at retesting, most often in samples with a low RLU ratio in the positive test. A cutoff ratio of 2 or 3 instead of 1 has been proposed to serve as a better indicator for reproducible positive results [67–70]. In our study, all specimens with RLU ratios of 2 or higher in the direct digene HC2 test on STM tubes were also correctly identified as positive when the test was performed on extracted DNA, suggesting that the digene assay can be applied to extracted DNA as well.
Using extracted DNA from 718 vaginal specimens as the template, the performance of the HPV genotyping parts of the vaginal health assay was compared to that of the digene HC2 HPV assay. The hrHPV genotyping assay showed good correlation with the digene assay, with a sensitivity and specificity of 87.3% and 96.6%, respectively, and a Cohen’s kappa of 0.804. Three samples containing hrHPV type 66 were found to be positive in both tests, even though they are supposedly not covered by the digene assay. Among a set of ten samples that were reported positive by the hrHPV digene assay but that did not contain hrHPV sequences as determined by our assay, six samples were found to contain single or mixed lrHPV types, including HPV 30, 61, 40, 42, 53, and 67. Cross-reactivity of the digene HC2 hrHPV probe mix with lrHPV sequences has been demonstrated by several others [69,71–74]. Thus, even though these samples had to be classified as false-negative because the digene HC2 HPV assay was taken as the gold standard, it is likely they were actually false-positives in the digene test. Removing these six samples from the comparison, the sensitivity and specificity of the hrHPV genotyping assay were 94.5% and 96.6%, respectively, with a kappa of 0.841 ± 0.065. The lrHPV genotyping assay had a sensitivity of 100% and a specificity of 92.9%, with a Cohen’s kappa of 0.788. In addition, the two tests were in good general agreement about the relative amount of HPV molecules detected. The specific genotyping of HPV by sequencing therefore can detect the presence of lrHPV that are known to give false-positive results in conventional clinical hrHPV tests. In addition, HPV genotyping provides clinicians and patients with more detailed information about which strain(s) are present in the vagina, allowing for more precise tracking of infection and clearance than most conventional assays .
Among the samples tested positive with hrHPV genotyping, HPV types 16, 52, 59, and 68 were most common. HPV types 6 and 42 were the most prevalent lrHPV types in the samples tested in this study. Although our sample set was small (718 samples, of which 142 tested positive) and only screened for the presence of 19 HPV types, the high proportion of positive samples for types 6, 16, 52, 59 has been found by others [75,76], while the results for hrHPV type 68 appears to be unique in our dataset.
In addition to the HPV portion of the novel vaginal health assay described here, the assay also reports the relative abundance of commensal and pathogenic bacteria in vaginal samples, and compares these to reference ranges. Self-collection has been shown to be well-suited for vaginal microbiome analysis as reported by Forney et al. who showed that microbial diversity is similar between self-collected and physician collected vaginal samples .
Several bacteria have been associated with vaginal health conditions, such as bacterial vaginosis [1,4–6], aerobic vaginitis , pelvic inflammatory disease , and sexually transmitted infections [9–12]. The vaginal health assay described here detects the relative abundance of bacteria positively associated with bacterial vaginosis, such as Sneathia or Gardnerella species, as well as those negatively associated with that condition such as Lactobacillus species. In addition, it detects the presence of three common STI-associated pathogens, i.e., C. trachomatis, N. gonorrhoeae, and M. genitalium. Of these, M. genitalium has been recently recognized as an important pathogen implicated in pelvic inflammatory disease and infertility [12,78]. Although some early diagnostic tests have been described [79,80], very few clinicians test for its presence. Furthermore, the vaginal microbiota composition has been reported to be associated with the progression of HPV infection, from early states to cervical cancer [37,40]. Vaginal microbiome analysis therefore not only can be used to detect STI-associated pathogens and bacteria involved in bacterial vaginosis, but also to assess a woman’s microbiome similarity to the microbiome of a group of individuals with progressed HPV infection. This brings about the opportunity to leverage microbiome information to understand HPV infection progression and women's susceptibility to cancer development. Future versions of this assay could also include additional microorganisms associated with pathogenic outcomes such as Trichomonas vaginalis, Mycoplasma parvum, and Ureaplasma urealithicum.
In conclusion, we here present a vaginal health assay that for the first time combines the detection of the most important bacterial and viral indicators of vaginal health and disease. We envision that this test has the potential to provide clinicians and patients with a more comprehensive understanding of the vaginal microbiome, and to encourage women to take an active role in related conversations with their doctors. We also hope that by improving accessibility through self-sampling, we may encourage more women to engage with current screening and treatment guidelines for vaginal pathogens and cervical cancer.
S1 Fig. Vaginal sampling instructions.
Participants in this study were sent a vaginal sampling kit containing a swab, sterile water to pre-wet the swab, a tube containing zirconia beads and a lysis and stabilization buffer, and sampling instructions such as the one shown above. After sampling according to the instructions, participants could ship their sample back by regular mail.
S2 Fig. In silico performance metrics of the bacterial targets.
Initially, 72 bacterial targets were identified based on their association with vaginal and reproductive health, comprised of 57 species and 15 genera. The following performance metrics were evaluated based on the number of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) detected in a manually curated amplicon database (described in S1 Doc in Almonacid et al., 2017). The target performance are plotted as follows: specificity = TN / (TN + FP); sensitivity = TP / (TP + FN); positive predictive value (PPV) = TP / (TP + FP); and negative predictive value (NPV) = TN / (TN + FN). Based on a cutoff of 90% (red vertical line), 31/72 preliminary targets passed for each of the parameters, resulting in the accurate in silico detection of 16 bacterial species (light purple), and 15 bacterial genera (dark purple).
S1 Table. Assay targets and associated health conditions.
List of all 31 bacterial targets and 19 HPV targets, their associations with different health conditions, and references.
S2 Table. In silico performance metrics for the 72 bacterial targets (genus and species level) that were initially selected.
PPV, positive predictive value; NPV, negative predictive value. All values in the sensitivity, specificity, PPV, and NPV columns are given as percentages. Of the 72 selected targets, 31 passed selection criteria of all values above 90%. Values below 90% are shown in red.
S3 Table. In silico performance metrics for the 19 HPV targets.
HPV, human papillomavirus; TP, true positive; FN, false negative; FP, false positive; TN, true negative; Sens, sensitivity (in %); Spec, specificity (in %); PPV, positive predictive value (in %); NPV, negative predictive value (in %).
S4 Table. List of the 31 synthetic DNAs created to represent the bacterial targets included in the assay.
Sequences are based on the 16S rRNA gene.
S5 Table. List of the 20 synthetic DNA sequences representing 5 lrHPV and 14 hrHPV types included in the assay.
Because of sequence variability, hrHPV type 68 was represented by 2 different sDNAs.
S6 Table. Limit of detection (LOD) assay for the bacterial targets.
The table shows the lowest dilution at which at least two of the three replicates had 2 or more reads per taxon, and the calculated threshold for identification per taxa at the LOD in number of reads.
S7 Table. Limit of detection (LOD) assay for the HPV targets.
The table shows the lowest dilution at which at least two of the three replicates had 2 or more reads per HPV type, and the calculated threshold for identification per HPV type at the LOD in number of reads.
- 1. Ravel J, Gajer P, Abdo Z, Schneider GM, Koenig SSK, McCulle SL, et al. Vaginal microbiome of reproductive-age women. Proc Natl Acad Sci USA. 2011 Mar 15;108 Suppl 1:4680–7.
- 2. Younes JA, Lievens E, Hummelen R, van der Westen R, Reid G, Petrova MI. Women and their microbes: the unexpected friendship. Trends Microbiol. 2018 Jan;26(1):16–32. pmid:28844447
- 3. O’Hanlon DE, Moench TR, Cone RA. Vaginal pH and microbicidal lactic acid when lactobacilli dominate the microbiota. PLoS ONE. 2013 Nov 6;8(11):e80074. pmid:24223212
- 4. Ling Z, Kong J, Liu F, Zhu H, Chen X, Wang Y, et al. Molecular analysis of the diversity of vaginal microbiota associated with bacterial vaginosis. BMC Genomics. 2010 Sep 7;11:488. pmid:20819230
- 5. Srinivasan S, Hoffman NG, Morgan MT, Matsen FA, Fiedler TL, Hall RW, et al. Bacterial communities in women with bacterial vaginosis: high resolution phylogenetic analyses reveal relationships of microbiota to clinical criteria. PLoS ONE. 2012 Jun 18;7(6):e37818. pmid:22719852
- 6. Ravel J, Brotman RM, Gajer P, Ma B, Nandy M, Fadrosh DW, et al. Daily temporal dynamics of vaginal microbiota before, during and after episodes of bacterial vaginosis. Microbiome. 2013 Dec 2;1(1):29. pmid:24451163
- 7. Donders GGG, Bellen G, Grinceviciene S, Ruban K, Vieira-Baptista P. Aerobic vaginitis: no longer a stranger. Res Microbiol. 2017 May 11;168(9–10):845–858 pmid:28502874
- 8. Gorgos LM, Sycuro LK, Srinivasan S, Fiedler TL, Morgan MT, Balkus JE, et al. Relationship of specific bacteria in the cervical and vaginal microbiotas with cervicitis. Sex Transm Dis. 2015 Sep;42(9):475–81. pmid:26267872
- 9. Petrova MI, Lievens E, Malik S, Imholz N, Lebeer S. Lactobacillus species as biomarkers and agents that can promote various aspects of vaginal health. Front Physiol. 2015 Mar 25;6:81. pmid:25859220
- 10. Hill SA, Masters TL, Wachter J. Gonorrhea—an evolving disease of the new millennium. Microb Cell. 2016 Sep 5;3(9):371–89. pmid:28357376
- 11. Ziklo N, Huston WM, Hocking JS, Timms P. Chlamydia trachomatis Genital Tract Infections: When Host Immune Response and the Microbiome Collide. Trends Microbiol. 2016 Sep;24(9):750–65. pmid:27320172
- 12. Jensen JS. Mycoplasma genitalium: yet another challenging STI. Lancet Infect Dis. 2017 Aug;17(8):795–6. pmid:28701270
- 13. Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. CA Cancer J Clin. 2011 Apr;61(2):69–90. pmid:21296855
- 14. Bray F, Ren J-S, Masuyer E, Ferlay J. Global estimates of cancer prevalence for 27 sites in the adult population in 2008. Int J Cancer. 2013 Mar 1;132(5):1133–45. pmid:22752881
- 15. Walboomers JMM, Jacobs MV, Manos MM, Bosch XF, Kummer AJ, Shah KV, et al. Human papillomavirus is a necessary cause of invasive cervical cancer worldwide. The Journal of Pathology. 1999 Sep 1;189(1):12–9. pmid:10451482
- 16. Bosch FX, Muñoz N. The viral etiology of cervical cancer. Virus Res. 2002 Nov;89(2):183–90. pmid:12445658
- 17. Bouvard V, Baan R, Straif K, Grosse Y, Secretan B, El Ghissassi F, et al. A review of human carcinogens—Part B: biological agents. Lancet Oncol. 2009 Apr;10(4):321–2. pmid:19350698
- 18. IARC Working Group on the Evaluation of Carcinogenic Risks to Humans. Biological agents. Volume 100 B. A review of human carcinogens. IARC Monogr Eval Carcinog Risks Hum. 2012;100(Pt B):1–441. pmid:23189750
- 19. de Sanjose S, Quint WG, Alemany L, Geraets DT, Klaustermeier JE, Lloveras B, et al. Human papillomavirus genotype attribution in invasive cervical cancer: a retrospective cross-sectional worldwide study. Lancet Oncol. 2010 Nov;11(11):1048–56. pmid:20952254
- 20. Khan MJ, Castle PE, Lorincz AT, Wacholder S, Sherman M, Scott DR, et al. The elevated 10-year risk of cervical precancer and cancer in women with human papillomavirus (HPV) type 16 or 18 and the possible utility of type-specific HPV testing in clinical practice. J Natl Cancer Inst. 2005 Jul 20;97(14):1072–9. pmid:16030305
- 21. Vidal AC, Smith JS, Valea F, Bentley R, Gradison M, Yarnall KSH, et al. HPV genotypes and cervical intraepithelial neoplasia in a multiethnic cohort in the southeastern USA. Cancer Causes Control. 2014 Aug;25(8):1055–62. pmid:24928693
- 22. Rosa MI, Fachel JMG, Rosa DD, Medeiros LR, Igansi CN, Bozzetti MC. Persistence and clearance of human papillomavirus infection: a prospective cohort study. Am J Obstet Gynecol. 2008 Dec;199(6):617.e1–7.
- 23. Egawa N, Doorbar J. The low-risk papillomaviruses. Virus Res. 2017 Mar 2;231:119–27. pmid:28040475
- 24. Mendes D, Bains I, Vanni T, Jit M. Systematic review of model-based cervical screening evaluations. BMC Cancer. 2015 May 1;15:334. pmid:25924871
- 25. Tambouret RH. The evolution of the Papanicolaou smear. Clin Obstet Gynecol. 2013 Mar;56(1):3–9. pmid:23314726
- 26. Tota JE, Bentley J, Blake J, Coutlée F, Duggan MA, Ferenczy A, et al. Introduction of molecular HPV testing as the primary technology in cervical cancer screening: Acting on evidence to change the current paradigm. Prev Med. 2017 May;98:5–14. pmid:28279264
- 27. Committee on Practice Bulletins—Gynecology. Practice bulletin no. 168: cervical cancer screening and prevention. Obstet Gynecol. 2016;128(4):e111–30. pmid:27661651
- 28. US Preventive Services Task Force. Final Update Summary: Cervical Cancer: Screening [Internet]. Final Update Summary: Cervical Cancer: Screening. U.S. Preventive Services Task Force. September 2016. 2016 [cited 2017 Nov 2]. Available from: https://www.uspreventiveservicestaskforce.org/Page/Document/UpdateSummaryFinal/cervical-cancer-screening
- 29. Curry SJ, Krist AH, Owens DK, Barry MJ, Caughey AB, et al. Screening for Cervical Cancer: US Preventive Services Task Force Recommendation Statement. JAMA 2018 Aug;320(7):674–686 pmid:30140884
- 30. Gradíssimo A, Burk RD. Molecular tests potentially improving HPV screening and genotyping for cervical cancer prevention. Expert Rev Mol Diagn. 2017 Apr;17(4):379–91. pmid:28277144
- 31. Arbyn M, Verdoodt F, Snijders PJF, Verhoef VMJ, Suonio E, Dillner L, et al. Accuracy of human papillomavirus testing on self-collected versus clinician-collected samples: a meta-analysis. Lancet Oncol. 2014 Feb;15(2):172–83. pmid:24433684
- 32. Lörincz AT. Hybrid captureTM method for detection of human papillomavirus DNA in clinical specimens: A tool for clinical management of equivocal pap smears and for population screening. J Obstet Gynaecol Res. 1996 Dec;22(6):629–36. pmid:9037955
- 33. de Roda Husman AM, Walboomers JM, van den Brule AJ, Meijer CJ, Snijders PJ. The use of general primers GP5 and GP6 elongated at their 3’ ends with adjacent highly conserved sequences improves human papillomavirus detection by PCR. J Gen Virol. 1995 Apr;76 (Pt 4):1057–62.
- 34. Manos MM, Ting Y, Lewis AJ, Broker TR, Wolinsky SM. The use of polymerase chain reaction amplification for the detection of genital human papillomaviruses. Cancer Cells 1989;7:209–214.
- 35. Gravitt PE, Peyton CL, Alessi TQ, Wheeler CM, Coutlée F, Hildesheim A, et al. Improved amplification of genital human papillomaviruses. J Clin Microbiol. 2000 Jan;38(1):357–61. pmid:10618116
- 36. Heideman DAM, Hesselink AT, Berkhof J, van Kemenade F, Melchers WJG, Daalmeijer NF, et al. Clinical validation of the cobas 4800 HPV test for cervical screening purposes. J Clin Microbiol. 2011 Nov;49(11):3983–5. pmid:21880968
- 37. Mitra A, MacIntyre DA, Marchesi JR, Lee YS, Bennett PR, Kyrgiou M. The vaginal microbiota, human papillomavirus infection and cervical intraepithelial neoplasia: what do we know and where are we going next? Microbiome. 2016 Nov 1;4(1):58. pmid:27802830
- 38. Gao W, Weng J, Gao Y, Chen X. Comparison of the vaginal microbiota diversity of women with and without human papillomavirus infection: a cross-sectional study. BMC Infect Dis. 2013 Jun 10;13:271. pmid:23758857
- 39. Lee JE, Lee S, Lee H, Song Y-M, Lee K, Han MJ, et al. Association of the vaginal microbiota with human papillomavirus infection in a Korean twin cohort. PLoS ONE. 2013 May 22;8(5):e63514. pmid:23717441
- 40. Brotman RM, Shardell MD, Gajer P, Tracy JK, Zenilman JM, Ravel J, et al. Interplay between the temporal dynamics of the vaginal microbiota and human papillomavirus detection. J Infect Dis. 2014 Dec 1;210(11):1723–33. pmid:24943724
- 41. Reimers LL, Mehta SD, Massad LS, Burk RD, Xie X, Ravel J, et al. The Cervicovaginal Microbiota and Its Associations With Human Papillomavirus Detection in HIV-Infected and HIV-Uninfected Women. J Infect Dis. 2016 Nov 1;214(9):1361–9. pmid:27521363
- 42. Shannon B, Yi TJ, Perusini S, Gajer P, Ma B, Humphrys MS, et al. Association of HPV infection and clearance with cervicovaginal immunology and the vaginal microbiota. Mucosal Immunol. 2017 Jan 25;10(5):1310–9. pmid:28120845
- 43. Dutro SM, Hebb JK, Garin CA, Hughes JP, Kenny GE, Totten PA. Development and performance of a microwell-plate-based polymerase chain reaction assay for Mycoplasma genitalium. Sexually Transmitted Diseases 2003 Oct;30(10): 756–63. pmid:14520174
- 44. Brunham RC, Gottlieb SL, Paavonen J. Pelvic inflammatory disease. N Engl J Med. 2015 May 21;372(21):2039–48. pmid:25992748
- 45. Almonacid DE, Kraal L, Ossandon FJ, Budovskaya YV, Cardenas JP, Bik EM, et al. 16S rRNA gene sequencing and healthy reference ranges for 28 clinically relevant microbial taxa from the human gut microbiome. PLoS ONE. 2017 May 3;12(5):e0176555. pmid:28467461
- 46. Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, et al. The SILVA ribosomal RNA gene database project: improved data processing and web-based tools. Nucleic Acids Res. 2013 Jan;41(Database issue):D590–6. pmid:23193283
- 47. Van Doorslaer K, Tan Q, Xirasagar S, Bandaru S, Gopalan V, Mohamoud Y, et al. The Papillomavirus Episteme: a central resource for papillomavirus sequence data and analysis. Nucleic Acids Res. 2013 Jan;41(Database issue):D571–8. pmid:23093593
- 48. Van Doorslaer K, Li Z, Xirasagar S, Maes P, Kaminsky D, Liou D, et al. The Papillomavirus Episteme: a major update to the papillomavirus sequence database. Nucleic Acids Res. 2017 Jan 4;45(D1):D499–506. pmid:28053164
- 49. Rognes T, Flouri T, Nichols B, Quince C, Mahé F. VSEARCH: a versatile open source tool for metagenomics. PeerJ. 2016 Oct 18;4:e2584. pmid:27781170
- 50. Clinical and Laboratory Standards Institute. Protocols for Determination of Limits of Detection and Limits of Quantitation, Approved Guideline. Wayne, PA USA: CLSI 2004;CLSI document EP17.
- 51. Caporaso JG, Lauber CL, Walters WA, Berg-Lyons D, Lozupone CA, Turnbaugh PJ, et al. Global patterns of 16S rRNA diversity at a depth of millions of sequences per sample. Proc Natl Acad Sci USA. 2011 Mar 15;108 Suppl 1:4516–22.
- 52. Estrade C, Sahli R. Comparison of Seegene Anyplex II HPV28 with the PGMY-CHUV assay for human papillomavirus genotyping. J Clin Microbiol. 2014 Feb;52(2):607–12. pmid:24478495
- 53. Mahé F, Rognes T, Quince C, de Vargas C, Dunthorn M. Swarm: robust and fast clustering method for amplicon-based studies. PeerJ. 2014 Sep 25;2:e593. pmid:25276506
- 54. Martin M. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet j. 2011 May 2;17(1):10.
- 55. Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics. 2014 Aug 1;30(15):2114–20. pmid:24695404
- 56. McMurdie PJ, Holmes S. phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data. PLoS ONE. 2013 Apr 22;8(4):e61217. pmid:23630581
- 57. Cohen J. A Coefficient of Agreement for Nominal Scales. Educ Psychol Meas. 1960 Apr 1;20(1):37–46.
- 58. Gupta S, Palmer C, Bik EM, Cardenas JP, Nuñez H, Kraal L, et al. Self-sampling for HPV testing: Increased cervical cancer screening participation and incorporation in international screening programs. Frontiers in Public Health. 2018 April 9;6:77. pmid:29686981
- 59. Racey CS, Withrow DR, Gesink D. Self-collected HPV testing improves participation in cervical cancer screening: a systematic review and meta-analysis. Can J Public Health. 2013 Feb 11;104(2):e159–66. pmid:23618210
- 60. Verdoodt F, Jentschke M, Hillemanns P, Racey CS, Snijders PJF, Arbyn M. Reaching women who do not participate in the regular cervical cancer screening programme by offering self-sampling kits: a systematic review and meta-analysis of randomised trials. Eur J Cancer. 2015 Nov;51(16):2375–85. pmid:26296294
- 61. Carvalho N de O, del Castillo DM, Perone C, Januário JN, Melo VH de, Brasileiro Filho G. Comparison of HPV genotyping by type-specific PCR and sequencing. Mem Inst Oswaldo Cruz. 2010 Feb;105(1):73–8. pmid:20209333
- 62. Conway C, Chalkley R, High A, Maclennan K, Berri S, Chengot P, et al. Next-generation sequencing for simultaneous determination of human papillomavirus load, subtype, and associated genomic copy number changes in tumors. J Mol Diagn. 2012 Apr 1;14(2):104–11. pmid:22240447
- 63. Militello V, Lavezzo E, Costanzi G, Franchin E, Di Camillo B, Toppo S, et al. Accurate human papillomavirus genotyping by 454 pyrosequencing. Clin Microbiol Infect. 2013 Oct;19(10):E428–34. pmid:23573945
- 64. Nowak RG, Ambulos NP, Schumaker LM, Mathias TJ, White RA, Troyer J, et al. Genotyping of high-risk anal human papillomavirus (HPV): ion torrent-next generation sequencing vs. linear array. Virol J. 2017 Jun 13;14(1):112. pmid:28610586
- 65. Ambulos NP, Schumaker LM, Mathias TJ, White R, Troyer J, Wells D, et al. Next-Generation Sequencing-Based HPV Genotyping Assay Validated in Formalin-Fixed, Paraffin-Embedded Oropharyngeal and Cervical Cancer Specimens. J Biomol Tech. 2016 Jul;27(2):46–52. pmid:27006646
- 66. Arbyn M, Depuydt C, Benoy I, Bogers J, Cuschieri K, Schmitt M, et al. VALGENT: A protocol for clinical validation of human papillomavirus assays. J Clin Virol. 2016 Mar;76 Suppl 1:S14–21.
- 67. Castle PE, Lorincz AT, Mielzynska-Lohnas I, Scott DR, Glass AG, Sherman ME, et al. Results of human papillomavirus DNA testing with the hybrid capture 2 assay are reproducible. J Clin Microbiol. 2002 Mar;40(3):1088–90. pmid:11880448
- 68. Carozzi FM, Mistro AD, Confortini M, Sani C, Puliti D, Trevisan R, et al. Reproducibility of HPV DNA testing by hybrid capture 2 in a screening setting. Am J Clin Pathol. 2005 Nov;124(5):716–21. pmid:16203283
- 69. de Cremoux P, Coste J, Sastre-Garau X, Thioux M, Bouillac C, Labbé S, et al. Efficiency of the hybrid capture 2 HPV DNA test in cervical cancer screening. Am J Clin Pathol. 2003 Oct;120(4):492–9. pmid:14560561
- 70. Moss SM, Bailey A, Cubie H, Denton K, Sargent A, Muir P, et al. Comparison of the performance of HPV tests in women with abnormal cytology: results of a study within the NHS cervical screening programme. Cytopathology. 2015 Dec;26(6):373–80. pmid:25274541
- 71. Vernon SD, Unger ER, Williams D. Comparison of human papillomavirus detection and typing by cycle sequencing, line blotting, and hybrid capture. J Clin Microbiol. 2000 Feb;38(2):651–5. pmid:10655362
- 72. Ginocchio CC, Barth D, Zhang F. Comparison of the Third Wave Invader human papillomavirus (HPV) assay and the Digene HPV hybrid capture 2 assay for detection of high-risk HPV DNA. J Clin Microbiol. 2008 May;46(5):1641–6. pmid:18367578
- 73. Gillio-Tos A, De Marco L, Carozzi FM, Del Mistro A, Girlando S, Burroni E, et al. Clinical impact of the analytical specificity of the hybrid capture 2 test: data from the New Technologies for Cervical Cancer (NTCC) study. J Clin Microbiol. 2013 Sep;51(9):2901–7. pmid:23804385
- 74. Boehmer G, Wang L, Iftner A, Holz B, Haedicke J, von Wasielewski R, et al. A population-based observational study comparing Cervista and Hybrid Capture 2 methods: improved relative specificity of the Cervista assay by increasing its cut-off. BMC Infect Dis. 2014 Dec 9;14:674. pmid:25487281
- 75. Dunne EF, Unger ER, Sternberg M, McQuillan G, Swan DC, Patel SS, et al. Prevalence of HPV infection among females in the United States. JAMA. 2007 Feb 28;297(8):813–9. pmid:17327523
- 76. Dickson EL, Vogel RI, Bliss RL, Downs LS. Multiple-type human papillomavirus (HPV) infections: a cross-sectional analysis of the prevalence of specific types in 309,000 women referred for HPV testing at the time of cervical cytology. Int J Gynecol Cancer. 2013 Sep;23(7):1295–302. pmid:23970156
- 77. Forney LJ, Gajer P, Williams CJ, Schneider GM, Koenig SSK, McCulle SL, et al. Comparison of self-collected and physician-collected vaginal swabs for microbiome analysis. J Clin Microbiol. 2010 May 1;48(5):1741–8. pmid:20200290
- 78. Wiesenfeld HC, Manhart LE. Mycoplasma genitalium in Women: Current Knowledge and Research Priorities for This Recently Emerged Pathogen. J Infect Dis. 2017 Jul 15;216(suppl_2):S389–95. pmid:28838078
- 79. Gaydos CA. Mycoplasma genitalium: Accurate Diagnosis Is Necessary for Adequate Treatment. J Infect Dis. 2017 Jul 15;216(suppl_2):S406–11. pmid:28838072
- 80. Munson E. Molecular Diagnostics Update for the Emerging (If Not Already Widespread) Sexually Transmitted Infection Agent Mycoplasma genitalium: Just About Ready for Prime Time. J Clin Microbiol. 2017 Oct;55(10):2894–902. pmid:28724558