Detection of Endometrial Cancer Using Tampon-Based Collection and Methylated DNA Markers

Abstract

Objective:

Alterations in DNA methylation are early events in endometrial cancer (EC) development and may have utility in EC detection via tampon-collected vaginal fluid.

Methods:

For discovery, DNA from frozen EC, benign endometrium (BE), and benign cervicovaginal (BCV) tissues underwent reduced representation bisulfite sequencing (RRBS) to identify differentially methylated regions (DMRs). Candidate DMRs were selected based on receiver operating characteristic (ROC) discrimination, methylation level fold-change between cancers and controls, and absence of background CpG methylation. Methylated DNA marker (MDM) validation was performed using qMSP on DNA from independent EC and BE FFPE tissue sets. Women ≥45 years of age with abnormal uterine bleeding (AUB) or postmenopausal bleeding (PMB) or any age with biopsy-proven EC self-collected vaginal fluid using a tampon prior to clinically indicated endometrial sampling or hysterectomy. Vaginal fluid DNA was assayed by qMSP for EC-associated MDMs. Random forest modeling analysis was performed to generate predictive probability of underlying disease; results were 500-fold in-silico cross-validated.

Results:

Thirty-three candidate MDMs met performance criteria in tissue. For the tampon pilot, 100 EC cases were frequency matched by menopausal status and tampon collection date to 92 BE controls. A 28-MDM panel highly discriminated between EC and BE (96% (95%CI 89–99%) specificity; 76% (66–84%) sensitivity (AUC 0.88). In PBS/EDTA tampon buffer, the panel yielded 96% (95% CI 87–99%) specificity and 82% (70–91%) sensitivity (AUC 0.91).

Conclusion:

Next generation methylome sequencing, stringent filtering criteria, and independent validation yielded excellent candidate MDMs for EC. EC-associated MDMs performed with promisingly high sensitivity and specificity in tampon-collected vaginal fluid; PBS-based tampon buffer with added EDTA improved sensitivity. Larger tampon-based EC MDM testing studies are warranted.

Introduction

Early detection and treatment of endometrial cancer (EC) portends an excellent prognosis with surgery alone often being curative in the setting of stage IA disease [1]. However, presentation of EC at advanced stages most often requires multimodal therapy and oncologic outcomes are less favorable [2]. The most common presenting symptom of EC is painless bleeding per vagina; however, the degree and duration of bleeding may not correlate with disease stage [3]. Additionally, while approximately 90% of women with EC present with abnormal uterine bleeding (AUB) or postmenopausal uterine bleeding (PMB), less than 10% of women worked up for perimenopausal AUB or PMB harbor EC [4]. Women with benign causes of AUB or PMB may undergo several invasive diagnostic procedures to rule out EC at substantial cost to the healthcare system [5] and yielding measurable discomfort to the patient [6, 7]. As such, there is a great clinical need for less invasive clinical diagnostic testing that can reliably distinguish between benign AUB/PMB and bleeding associated with an underlying EC or EC precursor.

The occasional incidental diagnosis of ECs by cervical cytology samples provides evidence that spontaneously shed EC cells reach the lower genital tract [8], although rarely with sufficient preservation to allow microscopic identification in asymptomatic women [9]. To overcome the limited sensitivity of microscopic detection of intact EC cells in lower genital tract specimens, we and others have conducted proof-of-concept studies that demonstrate vaginal fluid harbors unique DNA mutations, sequence alterations, or gene promotor methylation signaling the presence of EC or EC precursor lesions [10–13]. Thus, vaginal fluid molecular testing is proposed as an approach to facilitate minimally invasive triage of AUB and PMB. Further, patient self-collection with an intravaginal tampon or swab could potentially be applied in an at-home early detection approach for EC, similar to approaches evaluated for cervical cancer screening [14] and the current standard of care for colorectal cancer screening [15, 16].

Genes commonly mutated in EC were well-characterized along the spectrum of EC histologies through the efforts of The Cancer Genome Atlas [17]. However, without uniquely conserved mutations within EC, full gene sequencing would be required to leverage this molecular change thus potentially driving the cost of such modality. Alterations in DNA methylation are early events in cancer development and are currently leveraged as clinically utilized markers that indicate the presence of colorectal cancers and large colonic adenomas [15, 16]. Additionally, multiplexed assays developed to detect unique DNA methylation signals [15] carry the benefit of streamlining sample throughput. Within EC, most DNA methylation discovery efforts have employed array-based technologies with a preset selection of CpGs [17–20] which may limit the depth of discovery.

Accordingly, we aimed to broaden the repertoire of current candidate methylated DNA markers (MDMs) for EC using methylome sequencing discovery and independent sample validation experiments which included both the more common endometrioid histology and the less common, more aggressive EC histologies. We subsequently tested the performance of these novel EC MDMs in vaginal fluid obtained via patient self-collected tampons from women presenting with perimenopausal AUB, PMB, or a new diagnosis of biopsy-proven EC.

Methods

Study overview

This study included three phases (Figure 1). First, tissue-based discovery of methylated DNA markers (MDMs) was performed using reduced representation bisulfite sequencing (RRBS) on DNA extracted from frozen EC and benign tissues. Second, biological validation of EC-specific MDMs was performed using quantitative methylation specific PCR (qMSP) on DNA extracted from an independent group of formalin fixed paraffin embedded (FFPE) EC and benign endometrium (BE). The third phase involved clinical translation of MDM detection via qMSP in DNA extracted from vaginal fluid samples, obtained from women with EC, atypical endometrial hyperplasia (AEH), endometrial hyperplasia without atypia, or benign endometrium (BE) collected via patient self-collected intravaginal tampon. This study was approved by the Mayo Clinic Institutional Review Board.

Figure 1.

Overall study flow diagram.

Endometrial cancer (EC), benign endometrium (BE), benign cervicovaginal tissue (BCV), reduced representation bisulfite sequencing (RRBS), quantitative methylation-specific PCR (qMSP), formalin fixed paraffin embedded (FFPE), atypical endometrial hyperplasia (AEH).

Discovery cohort

Primary fresh frozen EC tissues were identified from a prospectively maintained Mayo Clinic EC biorepository of >1,500 frozen samples collected from consenting patients at the time of hysterectomy for EC or AEH. ECs included in the discovery phase represented the five most common EC histologies (grade 1/2 endometrioid, grade 3 endometrioid, serous, and clear cell carcinomas, and uterine carcinosarcoma). Frozen tissue blocks were required to have at least 70% tumor purity for inclusion. Benign endometrium (BE) tissue was collected from consenting patients between February 2015 and March 2017 via Pipelle or EndoSampler as an additional sample for research immediately following a clinically indicated office endometrial biopsy in women ≥45 years of age presenting for a workup of AUB or PMB. EC histologies and BE menstrual phases or atrophic endometrium were confirmed by a gynecologic pathologist (JKS, SEK). Benign cervicovaginal (BCV) squamous tissue was collected prospectively specifically for this study between September 2016 and February 2017 from both premenopausal and postmenopausal women undergoing hysterectomy for benign indications. BE and BCV tissues were fresh frozen until DNA extraction. Buffy coats were collected from healthy control female donors without cancer who were current on cervical cancer screening and mammography. Women diagnosed with other cancers or who had received chemotherapy class drugs within the previous 5 years, had prior pelvic radiation, a synchronous cancer diagnosed at the time of EC, or a prior solid organ or bone marrow transplant were excluded. Clinical variables were abstracted from the electronic medical record for all included subjects.

Biological validation cohort

An independent set of women with a new diagnosis of EC who underwent hysterectomy for initial treatment was identified for the biological validation cohort. Formalin-fixed paraffin embedded (FFPE) EC tissues representing the same histologies as the discovery cohort were included. Additionally, FFPE BE tissues from women who underwent hysterectomy for benign indications, frequency-matched by age, and FFPE endometrial hyperplasia without atypia and AEH tissues from women who underwent hysterectomy were obtained. All FFPE tissues were obtained from clinically archived tissues within the Mayo Clinic Tissue Registry. All histologies were confirmed by a gynecologic pathologist (MES) who also selected the tissue block site for macro-dissection (see Biological validation—Laboratory methods below). Eligibility criteria were the same as in the discovery set.

Vaginal fluid clinical pilot cohort—Tampon pilot

Between February 2013 and August 2019, two groups of women were prospectively enrolled at Mayo Clinic in Rochester, Minnesota, to collect vaginal fluid via self-placed tampon (tampon pilot). One group included women ≥45 years of age presenting to the Mayo Clinic Division of Gynecology for workup of AUB or PMB or without bleeding but postmenopausal and referred for evaluation of a thickened endometrial stripe (ES) on pelvic ultrasound. Women were excluded if they did not undergo a clinical endometrial sampling or if they had undergone endometrial sampling within the prior 3 months. Final clinical pathology diagnosis was utilized. EC on endometrial sampling or on hysterectomy pathology were included as EC cases. All final clinical diagnoses of AEH or endometrial hyperplasia without atypia were included for exploratory analyses, and any with benign endometrial sampling were eligible as BE controls. The other group was comprised of women ≥18 years of age with biopsy-proven EC or AEH presenting to the Mayo Clinic Division of Gynecologic Oncology Surgery for clinically indicated hysterectomy. Eligibility criteria were the same as in the discovery and biological validation cohorts.

Discovery—Laboratory methods

Following verification of diagnosis and tissue block selection by one of the study gynecologic pathologists (JKS, SEK), frozen EC and BCV tissue embedded in optimal cutting temperature (OCT) compound underwent microtome cutting by the Mayo Clinic Pathology Research Core to provide ten 10-micron scrolls. BE whole frozen tissue samples, collected by 1 or 2 passes with an office biopsy Pipelle or EndoSampler, were used. Genomic DNA was purified from tissue and buffy coat specimens with the DNeasy Blood and Tissue protocol and QIAamp DNA blood protocol (Qiagen, Valencia, CA), respectively. DNA was re-purified with AMPure XP beads (Beckman-Coulter, Brea CA) and quantified by PicoGreen (Thermo-Fisher, Waltham MA). DNA quality was assessed using real time quantitative PCR. RRBS libraries were prepared as previously described [21]. Briefly, 300ng of DNA was 1) digested with MspI, 2) ligated to methylated sequencing adaptors, 3) treated with sodium bisulfite (Epitect Bisulfite protocol, Qiagen), 4) amplification enriched with adapter specific primers, and 5) size selected (160–280 bp) using AMPure beads to remove primer-dimers and larger CpG sparse regions. Libraries were sequenced using the Illumina HiSeq 2500 instrument (Illumina, San Diego CA) at the Mayo Clinic Medical Genomics Facility. Candidate genomic differentially methylated regions (DMRs) were selected as described below (Statistical analyses—Discovery).

Assay Design and Testing—Laboratory methods

Quantitative methylation specific PCR assays were developed from the CpG methylation signatures of selected DMRs. Primers were designed using MethPrimer [22] to target the bisulfite-modified sequences for each gene identified, as well as a CpG-free reference region within the β-actin gene. Primers were quality control checked on 20 ng (~6250 genome equivalents) of positive and negative methylation controls. DNA was bisulfite converted using the EZ-96 DNA Methylation kit (Zymo Research, Irvine CA) and amplified using SYBR Green detection on Roche 480 LightCyclers (Roche, Basel Switzerland). Serially diluted universally methylated DNA samples were utilized as positive control standards, and negative controls included bisulfite converted and unconverted leukocyte-derived genomic DNA and converted whole genome amplified (unmethylated) DNA. MDM results were normalized to β-actin. Assay performance was verified using the discovery cohort samples. Markers that performed sub-optimally compared to the RRBS results and cut-offs (described below) were not considered further.

Biological validation—Laboratory methods

MDMs were then tested using qMSP on DNA extracted from independent FFPE EC and BE tissues. These MDMs were also tested in AEH and endometrial hyperplasia without atypia tissues. Following histologic verification and selection of macrodissection sites most representative of the diagnosis by a study gynecologic pathologist (MES), tissue blocks underwent macrodissection using a 1 mm or 2 mm core punch. DNA was purified using the Qiagen QIAmp FFPE DNA Tissue Kit (part# 56404) and bisulfite converted as described above. Samples were blinded, randomized, and assayed by qMSP, as above.

Tampon pilot—Biospecimen collection and laboratory methods

Consented women in both groups within the tampon pilot self-placed a regular sized, unscented Playtex^® tampon to collect vaginal fluid. Those enrolled in the group presenting for workup of AUB, PMB or thickened ES placed the tampon in the clinic prior to their gynecology consult and removed the tampon before clinically indicated pelvic examination and endometrial sampling. Those in the group that presented with a biopsy-proven EC or AEH placed the tampon in the preoperative area on the day of their hysterectomy and the tampon was removed in the operating room. Intravaginal tampon dwell time was recorded for both groups.

After removal, each tampon was placed in a 50mL conical tube containing sterile PBS buffer and transported to the laboratory. Each tampon was then suspended within a 50mL conical tube using a sterile insertable holder to keep it separate from the eluate, centrifuged, and separated into pellet and supernatant portions which were stored at –80 C until DNA extraction. Approximately half-way through prospective enrollment to the tampon pilot (starting on February 14, 2017), 50mM EDTA was added to the PBS buffer to enhance DNA recovery and reduce nuclease degradation. Tampon pellet DNA was extracted using the High Pure Viral Nucleic Acid Kit (Roche, Basel Switzerland) and quantified using a Qubit Fluorometer (Invitrogen, Walther MA). DNA was bisulfite converted and assayed by qMSP on selected MDMs as described above with β-actin as the reference gene [23].

Statistical analyses—Discovery

For discovery, we utilized a previously published approach [15, 21]. Briefly, Streamlined Analysis and Annotation Pipeline for RRBS (SAAP-RRBS), a Mayo Clinic in-house analysis software package, was used for quality scoring, sequence alignment, annotation to a University of California Santa Cruz reference genome, and differential analysis of DMRs [21, 24]. Candidate CpGs were excluded if the coverage of data within each sample group was <50%. CpG islands are typically biochemically defined by an observed to expected CpG ratio >0.6 [25]. However, for this model, DMRs were created based on the distance between CpG site locations for each chromosome with regions containing five or fewer CpGs excluded. DMRs were then selected for a background methylation ratio in the benign controls (BE, BCV, and buffy coat) of <2% and then ranked by AUCs for EC histologies referent to benign controls. Statistical significance was determined by over-dispersed logistic regression of the methylation percentage per candidate DMR, based on read counts. To account for varying read depths across individual subjects, an over-dispersed logistic regression model was used, where the dispersion parameter was estimated using a Pearson Chi-square statistic of the residuals from fitted model. Candidate genomic DMRs were ranked and selected for further testing according to their significance level, AUC, and fold-change difference between ECs and benign controls (BE, BCV, and buffy coat). Sample size estimates for the discovery are based on methods previously described [21].

A secondary DMR analysis was undertaken to identify endometrium specific MDMs methylated in both EC and BE and unmethylated in BCV and buffy coat. It was hypothesized that markers with these characteristics would allow for the normalization of EC-specific MDMs to total epithelium exfoliation and the future opportunity to explore quantifying spontaneous endometrial exfoliation in both EC and BE.

Statistical analyses—biological validation and tampon pilot

For independent tissue validation, sample sizes were chosen to increase precision (minimizing the widths of 95% confidence intervals (95% CIs)) of sensitivity and specificity. With an assumed specificity of 95%, a control set of 29 would provide a 95% CI no wider than ± 10%. To achieve a 95% CI that was no wider than ± 7% for a target sensitivity of 90%, a minimum of 84 samples was required. Distributions of individual markers were examined using boxplots and marker intensity maps. AUC values were generated for each marker to assess accuracy. Random forest (rForest) models were used to generate the predicted probability of a sample representing an EC case. Random forest uses 500 randomized unique training and test sets from a bootstrap selection (approximately 2/3 for the training set and 1/3 for the test set) to perform cross-validation of randomized marker sets to generate 500 models. Accuracy error based on out-of-bag samples is then averaged over the 500 models. Marker selection was performed using the VSURF package in R, which uses random forests in three steps to: 1) eliminate predictors with least importance, 2) select all predictors that relate to the response variable, and 3) reduce redundancy in final marker selection [26].

For the tampon pilot, sample size estimates were defined to detect an AUC of 0.70 (AUC= 0.50 represents chance). With 100 EC and 92 BE, there was greater than 90% power to detect this difference using a one-sided test at a 5% significance level. ECs were frequency matched to control BEs by subject menopausal status and tampon collection date.

All women in the clinic group that underwent evaluation for AUB, PMB or thickened ES and diagnosed with AEH or endometrial hyperplasia without atypia after tampon collection were included for exploratory analyses. Additionally, the following tampon pilot subanalyses were performed: 1) MDM sensitivity and specificity for EC when limited to vaginal fluid samples collected prior to endometrial sampling, a setting of presumed spontaneous endometrial DNA shedding, and 2) MDM performance specifically in PBS/EDTA buffered vaginal fluid samples.

Results

Endometrial cancer MDM discovery

RRBS was performed on 69 ECs (16 grade 1/2 endometrioid, 16 grade 3 endometrioid, 11 serous, and 11 clear cell carcinomas, and 15 uterine carcinosarcomas), 44 BE (14 proliferative, 18 disordered proliferative, and 12 atrophic), 18 BCV, and 18 buffy coat samples from healthy donor women. Clinicopathologic characteristics for discovery phase EC cases and BE controls are detailed in Supplemental Table 1.

Sequencing coverage depth across all the samples was approximately in the 40–50X range. On average, filtered Cs in the CpG context with at least 10X coverage (our minimum requirement for inclusion) averaged ~1.7 million/sample. Our DMR calling algorithm applied to multiple comparisons (all EC vs all controls, histologic EC subtype vs BE, histologic EC subtype vs BCV, etc.) yielded a total of 323 statistically significant DMRs. Imposing performance cut-offs (AUC >0.85; absolute average CpG methylation >20% in ECs; methylation fold-change ratio (cases/controls) >10; p-value < 0.001) reduced the number of DMRs to 54. Targeted qMSP assays were constructed and tested for the 54 selected DMRs. Twenty-one targeted qMSP assays were subsequently discarded due to QC failures or inferior performance relative to the respective sequencing data and/or the cut-offs indicated above.

Biological validation of EC MDMs

Independent tissue testing was performed on the remaining 33 MDMs: AIM1, c5orf52, CYTH2, DIDO1, EEF1A2, EMX2, EMX2OS, GATA2, GDF7, ITPKA, JSRP1, KANK1, LOC440925, LRRC8D, LRRC34, LRRC41, MAX.chr7.104624356, MAX.chr8.145103829, MAX.chr8.145104263, MAX.chr10.22624479, MAX.chr14.103021656, MDFI, MPZ, NBPF8, OBSCN, SEPT9, SFMBT2, SQSTM1, STX16, VILL, ZNF90, ZNF323, and ZNF506. Samples included 141 ECs (34 grade 1/2 endometrioid, 31 grade 3 endometrioid, 27 serous, 19 clear cell, and 30 uterine carcinosarcomas), 112 BEs (35 secretory, 30 proliferative, 19 disordered proliferative, and 28 atrophic), 35 AEHs, and 24 endometrial hyperplasias without atypia. See Supplemental Table 2 for clinicopathologic characteristics.

Several MDMs (EMX2OS, CYTH2, MPZ, NBPF8) demonstrated >0.85 AUCs in discriminating between EC (all histologic types combined) and BE (Supplemental Table 3), most were uniquely discriminatory between a specific EC histologic subtype and BE (Figure 2). For example, EEF1A2, which has an AUC of 0.59 when comparing all EC histologic subtypes to BE, discriminated clear cell EC from BE with an AUC of 0.91 (0.80–1). For analyses of clear cell EC versus BE, 14 of 33 MDMs had an AUC ≥0.85. In contrast, only 5 of 33 MDMs distinguished carcinosarcoma versus BE with an AUC ≥0.85. When assessing both EC histology-combined and histology-specific MDMs, only 10 MDMs fell below an AUC of 0.85 in all comparisons. As such, 23 MDMs had an AUC ≥0.85 on either histology-combined or histology-specific analyses.

Figure 2.

Heatmatrix for tissue-based biological validation. Increasing deciles of relative methylation intensity above a 95% cutoff for each methylated DNA marker (MDM) in benign endometrium (BE) samples are depicted on a yellow-red color spectrum with red depicting the highest intensity above the cutoff. Each row is a candidate MDM, each column is a patient tissue sample.

Endometrial hyperplasia without atypia (HPw/oA), atypical endometrial hyperplasia (AEH), uterine carcinosarcoma (CS), clear cell endometrial cancer (CC), grade 1/2 endometrioid endometrial cancer (EG1/2), grade 3 endometrioid endometrial cancer (EG3).

Testing candidate EC MDMs in vaginal fluid—Tampon pilot

For the tampon pilot, we included four MDM assays (CDH4, LYPLAL1, c17orf64, and MAX.chr12.52652301) developed from endometrium specific DMRs (i.e. methylated in both EC and BE, but unmethylated in BCV tissues). In addition to these four, 24 MDMs were brought forward from biological validation for a total of 28 MDMs tested in the tampon pilot: CDH4, c17orf64, CYTH2, DIDO1, EEF1A2, EMX2OS, GATA2, GDF7, JSRP1, LRRC8D, LRRC34, LRRC41, LYPLAL1, MAX.chr8.145103829, MAX.chr10.22624479, MAX.chr12.52652301, MAX.chr14.103021656, MDFI, MPZ, NBPF8, OBSCN, SEPT9, SFMBT2, SQSTM1, VILL, ZNF90, ZNF323, and ZNF506.

There were 100 women with EC, 92 with BE, 11 with AEH, and 25 with endometrial hyperplasia without atypia included in the tampon pilot. EC cases included 31 women who presented for AUB or PMB clinical evaluation whose EC was diagnosed after tampon collection and 69 women with biopsy-proven EC known prior to tampon collection. All women with BE, all but 3 with AEH, and all but 1 with endometrial hyperplasia without atypia were from the group of women with perimenopausal AUB or PMB and those subjects’ tampons were collected prior to endometrial sampling. EC cases included 49 grade 1/2 endometrioid, 9 grade 3 endometrioid, 24 serous, 4 clear cell carcinomas, 9 uterine carcinosarcomas, and 5 mixed EC histologies. Clinicopathologic characteristics of the EC case, BE control, and AEH and endometrial hyperplasia without atypia groups are detailed in Table 1.

Table 1.

Clinicopathologic characteristics of tampon pilot cohort EC cases, benign endometrium (BE) controls, endometrial hyperplasia without atypia, and atypical endometrial hyperplasia (AEH).

	Endometrial cancer (EC) N=100	Benign endometrium (BE) N=92	Endometrial hyperplasia w/o atypia N=25	Atypical endometrial hyperplasia (AEH) N=11
Age, years; median [IQR]	64 [58–69]	63 [55–68]	54 [52–60]	62 [57–65]
BMI, kg/m2; median [IQR]	33.8 (29.1, 38.9)	29.9 [24.8–37.7]	31.6 [25.8–38.8]	41.8 [37.2–46.3]
Pregnancies; median [IQR]	3 [2, 3]	3 [2, 4]	2 [2, 3]	2 [1.5, 3]
Live births; median [IQR]	2 [1, 3]	2 [1, 3]	2 [1.8, 2]	2 [2, 2.5]
Race; N (%)
White	91 (91%)	88 (95.7%)	23 (92%)	11 (100%)
Non-White	3 (3%)	3 (3.3%)	1 (4%)	0 (0%)
Unknown	6 (6%)	1 (1.1%)	1 (4%)	0 (0%)
Tobacco Use; N (%)
Current	3 (3%)	1 (1.1%)	1 (4%)	0 (0%)
Previous	18 (18%)	27 (29.3)	6 (24%)	1 (9.1%)
Never	78 (78%)	63 (68.5%)	18 (72%)	10 (90.9%)
Unknown	1 (1%)	1 (1.1%)	0 (0%)	0 (0%)
Menopausal status; N (%)
Postmenopausal	82 (82%)	70 (76.1%)	14 (56%)	9 (81.8%)
Perimenopausal	3 (3%)	3 (3.3%)	5 (20%)	1 (9.1%)
Premenopausal	13 (13%)	16 (17.4%)	3 (12%)	1 (9.1%)
Unknown	2 (2%)	3 (3.3%)	3 (12%)	0 (0%)
Diabetes mellitus; N (%)	13 (13%)	5 (5.4%)	1 (4%)	3 (27.3%)
Hypertension; N (%)	51 (51%)	35 (38%)	8 (32%)	8 (72.7%)
Hyperlipidemia; N (%)	35 (35%)	38 (41.3%)	3 (12%)	5 (45.5%)
EC Histology; N (%)
Grade 1/2 endometrioid	49 (49%)	-	-	-
Grade 3 endometrioid	9 (9%)	-	-	-
Serous	24 (24%)	-	-	-
Clear cell	4 (4%)	-	-	-
Uterine carcinosarcoma	9 (9%)	-	-	-
Mixed	5 (5%)
EC Stage; N (%)		-	-	-
I	73 (73%)	-	-	-
II	1 (1%)	-	-	-
III	15 (15%)	-	-	-
IV	8 (8%)	-	-	-
Unknown	3 (3%)	-	-	-
Tampon intravaginal dwell time, minutes; median [IQR]	92.5 [54.5–115.8] *	44 [35.3–61.5]	45 [30–58]	40 [30–62]

^*Median dwell time in the pre-biopsy cohort: 49 [27.50–58.50]

When comparing the combined EC cases from both the AUB/PMB and biopsy-proven EC groups to BE controls, the 28 MDMs individually had marked methylation fold changes compared to controls (Figure 3A). Table 2 lists the AUCs in discriminating between EC and BE for each of the 28 MDMs tested in the tampon pilot. The 28-MDM panel discriminated between EC and BE at a set 96% (95% CI 89–99%) specificity with 76% (66–84%) sensitivity (AUC 0.88 [0.82–0.93]). When reducing the number of MDMs in a post-hoc analysis to a 3-MDM panel, the combination of SFMBT2, NBPF8, and MAX.chr10.22624479 yielded the same AUC as the 28-MDM panel (Figure 3B). When considering age ≥ 64 years v. < 64 (median age in the tampon pilot) and BMI ≥ 30 v. < 30 kg/m2, neither covariate was statistically significantly different when comparing stratified AUCs. When limiting the analysis to tampon samples collected prior to endometrial sampling, the 28-MDM panel distinguished between EC (n=31) and BE at a set 96% (95% CI 89–99%) specificity with similar sensitivity of 74% (95% CI 55–88%) (AUC 0.87 [0.77–0.98]).

Figure 3.

A) Heatmatrix of 28 EC MDMs tested via qMSP on DNA from vaginal fluid collected via the tampon pilot. The tampon pilot included women with histologic confirmation of benign endometrium (BE) (N=92), endometrial hyperplasia without atypia (HPw/oA) (N= 25), atypical endometrial hyperplasia (AEH) (N=11), and endometrial cancer (EC) (N=100). CDH4, LYPLAL1, c17orf64, and MAX.chr12.52652301 are endometrium specific MDMs as they are similarly methylated in both EC and BE, but unmethylated in benign cervicovaginal tissues. B) Area under the receiver operating characteristics curve (AUC) for a reduced 3-MDM panel (SFMBT2, NBPF8, MAX.chr10.22624479).

Table 2:

AUCs for 28 MDMs included in the panel tested in the tampon pilot. Analysis performed on all samples (100 ECs, 92 BE), including tampons collected into PBS alone + tampons collected into PBS/EDTA. The subanalysis on tampons collected into PBS/EDTA included 57 ECs and 52 BE. AUCs are listed in descending rank based on analysis of PBS alone + PBS/EDTA.

MDM	PBS alone + PBS/EDTA AUC (95% CI)	PBS/EDTA AUC (95% CI)
MAX.chr12	0.87 (0.82–0.92)	0.92 (0.86–0.97)
CDH4	0.86 (0.81–0.92)	0.89 (0.82–0.95)
c17orf64	0.86 (0.81–0.92)	0.85 (0.78–0.93)
EMX2OS	0.86 (0.8–0.91)	0.91 (0.85–0.97)
NBPF8	0.86 (0.8–0.91)	0.89 (0.83–0.96)
SFMBT2	0.85 (0.8–0.91)	0.84 (0.77–0.92)
JSRP1	0.83 (0.77–0.89)	0.87 (0.8–0.94)
DIDO1	0.82 (0.76–0.88)	0.92 (0.86–0.97)
MAX.chr10	0.81 (0.75–0.87)	0.78 (0.69–0.87)
MPZ	0.79 (0.73–0.86)	0.79 (0.7–0.88)
ZNF506	0.79 (0.72–0.86)	0.76 (0.66–0.85)
GATA2	0.79 (0.72–0.85)	0.8 (0.72–0.88)
VILL	0.78 (0.72–0.85)	0.82 (0.74–0.91)
MAX.chr14	0.78 (0.71–0.85)	0.82 (0.73–0.9)
CYTH2	0.76 (0.7–0.83)	0.85 (0.78–0.92)
LRRC8D	0.76 (0.69–0.83)	0.85 (0.77–0.93)
LYPLAL1	0.75 (0.68–0.82)	0.8 (0.72–0.89)
MAX.chr8	0.74 (0.67–0.81)	0.8 (0.71–0.89)
SQSTM1	0.71 (0.64–0.79)	0.81 (0.73–0.89)
ZNF323	0.71 (0.64–0.79)	0.78 (0.69–0.86)
OBSCN	0.69 (0.61–0.77)	0.79 (0.7–0.88)
ZNF90	0.65 (0.57–0.73)	0.75 (0.66–0.84)
LRRC34	0.64 (0.59–0.68)	0.68 (0.62–0.75)
GDF7	0.63 (0.55–0.71)	0.71 (0.61–0.82)
MDFI	0.63 (0.55–0.71)	0.62 (0.51–0.73)
EEF1A2	0.62 (0.54–0.7)	0.72 (0.63–0.82)
LRRC41	0.61 (0.53–0.69)	0.75 (0.66–0.84)
SEPT9	0.52 (0.44–0.6)	0.48 (0.37–0.6)

Exploration of the performance of the 28 MDMs in tampon specimens from women subsequently diagnosed with AEH or endometrial hyperplasia without atypia revealed lower methylation intensities compared to EC and higher intensities compared to BE (Figure 3A).

Impact of tampon buffer on MDM performance—Tampon pilot subanalysis

As previously noted, approximately half-way through the prospective vaginal fluid collection study period, 50mM EDTA was added to the PBS tampon buffer with the goal of improved DNA stabilization. No eligible tampon specimens were excluded due to insufficient DNA quantity. DNA quantification utilizing β-actin within PBS v. PBS/EDTA buffer was comparable both before bisulfite conversion and after conversion (Supplemental Figure 1). Among the total EC cases and BE controls in the tampon pilot, tampons were collected into PBS/EDTA buffer for 57 ECs and 52 BEs. The AUCs for each of the 28 individual MDMs in discriminating between EC and BE based on tampons collected into PBS/EDTA buffer are listed in Table 2. The combined 28-MDM panel demonstrated improved sensitivity when tested on tampon specimens collected into PBS/EDTA buffer (set 96% (95% CI 87–99%) specificity; 82% (70–91%) sensitivity (AUC 0.91 [0.85–0.97]) compared to the full tampon pilot including both PBS alone and PBS/EDTA buffered vaginal fluid (Table 2). Additionally, in the PBS/EDTA buffer subanalysis with a set 95% specificity, the 28-MDM panel correctly identified: 17 (85%) of the 20 endometrioid ECs, 18 (78%) of the 23 serous ECs, all (100%) of the 9 uterine carcinosarcomas, 2 (67%) of the 3 clear cell ECs, and 1 (50%) of the 2 mixed EC histologies.

Characterization of select identified EC MDMs

To explore a potential functional role of the identified EC MDMs, we sampled a random selection of the 323 initial DMRs, used genomic coordinates to map to highly annotated genes (RefSeq), and then queried Uniprot for molecular and biological roles. DMRs most often mapped to either 5-prime regulatory sequences or intronic gene body locations. Gene-protein function for all DMRs included pathways known to contribute to driving tumorigenesis, such as transcriptional regulation, cell cycle, growth, signaling, and apoptosis. For the 9 EC MDMs with AUC ≥0.85 in PBS/EDTA buffered tampon samples, we confirmed pathway associations relevant to cancer and identified previously published evidence of cancer-related actions for each these genes (Table 3) [27–34].

Table 3.

Functional description of the MDMs with AUC ≥0.85 in tampon pilot PBS/EDTA buffered samples.

MDM	Gene Name	Gene/Protein Function	Site of Methylation	Cancer-related associations in the literature
c17orf64	Chromosome 17 Open Reading Frame 64	NA	Promoter	Hypermethylated in ovarian cancers [30]
CDH4	Cadherin-4	Calcium dependent cell adhesion protein	Promoter	Hypermethylated in colorectal and gastric cancers [28]
CYTH2	Cytohesin 2	Guanine nucleotide exchange factor (GTPase)	3’ UTR	Activates EGF and IGF-1 pathways. Overexpressed in CRC [27]
DIDO1	Death-inducer obliterator 1	Transcription factor	5’ UTR	Overexpressed in bladder cancer [33]
EMX2OS	EMX2 Opposite Strand/Antisense RNA	Long non-coding RNA enhancer	NA	Regulates proliferation, migration, and invasion in prostate cancer [34]
JSRP1	Junctional sarcoplasmic reticulum protein 1	Calcium channel regulation	Gene body/intron	Highly overexpressed in endometrial cancers [29]
LRRC8D	Leucine rich repeat containing 8 VRAC subunit D	Anion channel subunit	Gene body/intron	Implicated in the progression and migration of multiple cancers, including cervical, colon, and ovarian cancers [32]
MAX.chr12	MAX.chr12.52652301	NA	NA	NA
NBPF8	Neuroblastoma breakpoint family member 8	Potential pseudogene	NA	Overexpressed in NSCLC [31]

Discussion

Through rigorous discovery and validation in tissue, we identified unique EC MDMs, which are detectable in vaginal fluid collected with tampons and demonstrate promise for triaging patients with perimenopausal AUB or PMB using self-collected samples. In fact, translation to tampon-collected vaginal fluid samples indicated the 28-MDM EC panel tested in the tampon pilot had high sensitivity and specificity in discriminating between underlying EC and BE. This high sensitivity and specificity also appeared to be maintained when we evaluated a smaller, 3-marker panel. Reassuringly, the sensitivity to detect EC also remained high in subanalyses including only vaginal fluid samples collected from women presenting with perimenopausal AUB or PMB before underlying endometrial pathology was determined via endometrial sampling, supporting the view that EC-associated MDMs are spontaneously shed into the vagina.

Liquid biopsy approaches for cancer screening, diagnostic, and prognostic purposes are attractive as minimally invasive modalities. They allow for frequent, repeated assessments and often leverage biospecimens that are readily available, such as blood, stool, urine, or other body fluids [15, 16, 35–38]. In specifically targeting EC detection, vaginal or cervicovaginal biospecimens collected either via Pap methodology by a care provider or self-collected by the patient using a tampon have both demonstrated promising utility [10–13, 37, 39]. Detection of methylated EC markers in urine [36, 40] is also a potentially useful approach for self-collection of a liquid biopsy. In this present study, all tampon-collected vaginal fluid specimens were self-collected by the consented person. While collection was completed in a supervised clinical setting for this study, the patient successfully collecting their own sample suggests biospecimen collection could be done in locations remote from a medical provider’s office. This enhances the potential for greater access to diagnostic care in the setting of perimenopausal AUB and PMB, including the option for at-home collection of vaginal fluid samples. Self-collecting biospecimens in gynecologic cancer screening is attractive in that, as evident in patient vaginal fluid self-sampling for high-risk human papilloma virus (HR-HPV) [41–43], it may provide a patient-preferred approach compared to clinician-collected biospecimens and result in an increased uptake of cancer screening, early detection, and interception. At-home stool-based collection for colon cancer screening provides a successful model [15, 16] and is endorsed by the United States Preventative Services Task Force (USPSTF) [44]. This success in colorectal cancer screening suggests a similar approach focused on EC could ultimately foster the development of a screening test for EC.

Our findings in this present study, however, predominantly support the potential diagnostic value of a self-collected vaginal fluid-based test among symptomatic women presenting with PMB or perimenopausal AUB. Given that only approximately 1 in 10 women presenting with PMB will have an underlying EC [4], DNA methylation findings suggesting a benign cause for PMB could complement ultrasound-only evaluations of the endometrium or substantially decrease the relatively invasive diagnostic testing approaches currently utilized [5, 46]. With a high sensitivity for an underlying EC, the detection of EC-associated MDMs in vaginal fluid would streamline the triage of patients directly to diagnostic endometrial sampling, while deferring endometrial biopsy in those testing negative.

While exploratory in nature, the relatively greater methylation intensity in DNA from vaginal fluid in the setting of endometrial hyperplasia without atypia as well as AEH compared to BE suggests MDMs also signal biological tissue changes along the spectrum of carcinogenesis. Indeed, when we explored the biological relevance of EC-associated MDMs that performed best in vaginal fluid, several MDMs are associated with cancer-related pathways such as cellular proliferation, migration, and invasion [28, 32, 34]. These findings also suggest an EC-associated MDM panel could have the potential to identify EC precursors, allowing for clinical interception before the development of EC.

While an a priori change to the tampon buffer was not planned within the prospective tampon pilot, the addition of EDTA to the PBS buffer did allow us to test the 28-MDM panel in two different buffers. Indeed, we demonstrated improved AUCs in samples collected into PBS/EDTA buffer compared to the collective PBS alone plus PBS/EDTA collected samples, suggesting that the addition of EDTA provided improved DNA stabilization. We selected a PBS-based buffer given its relative safety if ingested; however, there may be other buffers that are safe for at-home biospecimen collection that provide even greater DNA recovery. Subsequent studies are needed to test other buffers, assess DNA yield with at-home tampon self-collection, and determine the stability of tampon-collected DNA when transported at ambient temperatures to a centralized testing center.

Additionally, there are several molecular marker families that are or have been targeted in the quest to develop a liquid biopsy-based EC diagnostic or screening test. The combination of targets has the potential to complement and enhance sensitivity and specificity. In addition to MDMs, EC-associated mutations and aneuploidy have shown great promise as markers for the presence of EC [12, 13]. In fact, assaying a combination of mutations and aneuploidy in the PapSEEK test, for example, resulted in a sensitivity of nearly 80% to detect early stage EC with a specificity of 99% when the assay utilized fluid from endocervical Pap tests. The sensitivity was successfully driven higher to 93% when the assay was perfomed on Tao brush collected intrauterine samples [13]. One challenge of this approach as well as others that have leveraged Pap tests and/or Tao brush collected specimens [11, 13, 20, 37, 45] is the requirement for provider-collected biospecimens that are not readily translatable to an at-home self-collected biospecimen. Despite that challenge, recognizing that such molecular biomarkers can identify the presence of EC, translation to less invasive sampling is a patient-centered future focus that could provide greater access to EC diagnostic strategies.

The strengths of this study include the utilization of RRBS, perhaps the most comprehensive MDM identification methodology, in the MDM discovery process. Additionally, all tissues from which DNA was extracted were selected and diagnostically reviewed by a gynecologic pathologist, including within the discovery set which consisted of only pure EC histologies. The RRBS process included a robust approach to quality control, RRBS was performed in a single batch, and technical validation of novel MDMs was performed using qMSP as a second methodology. Within the tampon pilot, subjects were matched by menopausal status and the date the tampon was collected with the goals of minimizing methylation variability that could potentially be attributed to underlying BE biological activity and DNA stability between cohorts’ biospecimens.

The translation to testing MDMs in vaginal fluid in this study was a clinical pilot and larger clinical studies are needed to 1) evaluate clinical performance, 2) optimize the number of MDMs needed, 3) set algorithms for objective tampon-based test cutoffs, 4) determine the test’s negative predictive value and false positive rates, and 5) further assess the detection in vaginal fluid of MDMs identified within the less common EC histologies. We also recognize there were age differences in both the discovery and biological validation cohorts which may have resulted in the inclusion of MDM candidates that reflect increased age as opposed to cancer. However, this issue was mitigated in part by appropriately age-matching between EC and BE in the tampon clinical pilot. Additionally, while the 28-MDM panel identified nearly 80% of serous ECs, improved sensitivity is needed in this high risk histology and the addition of complementary markers that are well-recognized as drivers of carcinogenesis in type II EC, such as p53 mutations [47], may improve sensitivity. Additionally, our tampon pilot had limited numbers of endometrial hyperplasia without atypia and AEH. With such well-defined EC precursor lesions [48], endometrial hyperplasia states represent an opportunity in EC risk reduction and these histologies should be included in subsequent phases of biomarker development. The value of incorporating clinical EC risk factors into the biomarker model was also not addressed in the clinical pilot. This study was also limited by the exclusion of patients who received chemotherapy within the five years prior to enrollment, had prior pelvic therapeutic radiation, a synchronous cancer diagnosed at the same time as their EC, or had received a transplant given the potential for these factors to interfere with methylation levels. Further research is needed to assess the applicability of the identified EC MDMs in these populations as well as further test the specificity for EC in the setting of other neoplastic and benign gynecologic conditions. All women included in this study were from a single institution and were predominantly White. Subsequent studies in a greater diversity of women are imperative in developing a generalizable test, assessing for unique racial differences in vaginal fluid MDMs, and further understanding preferences in biospecimen sampling modality.

Given the favorable signals observed in the clinical tampon pilot, a phase II biomarker clinical study has been developed (NCT05051722) and is open and enrolling with the goals of testing the EC MDMs in an expanded diversity of women as well as streamlining EC MDMs to a reduced number of markers. Leveraging tampon-collected vaginal fluid and molecular markers in the detection of EC is also the focus of the ongoing DETECT clinical trial (NCT03538665). In this present study, we found that there was a similar AUC in vaginal fluid collected exclusively from women prior to endometrial sampling compared to the combined analysis of tampons collected prior to endometrial sampling plus those collected after a biopsy-proven EC or AEH diagnosis. This is reassuring in that there appears to be minimal artifact introduced by studying vaginal fluid from women who have had recent endometrial instrumentation. This suggests that in subsequent phases of EC biomarker development, it is reasonable to include biopsy-proven EC or AEH as it allows for focused clinical trial resources and optimized sample sizes for cases and controls.

In conclusion, through a systematic discovery approach and independent validation, we identified novel EC-associated MDMs that highly discriminated between the presence of EC or BE in vaginal fluid self-collected by patients using a tampon. Larger clinical studies are warranted to independently validate the translational findings in vaginal fluid, expand to include a greater diversity of women, test vaginal fluid for complementary biomarkers of other gynecologic cancers, and optimize the MDM panel to still encompass the heterogeneity of EC histologies while minimizing MDM redundancy.

Highlights.

• Whole methylome sequencing identified novel endometrial cancer (EC) methylated DNA markers (MDMs).

• A 28-MDM EC panel discriminated between EC and benign endometrium in vaginal fluid collected by tampons.

• An abbreviated 3-MDM EC panel performed with similar sensitivity and specificity as the 28-MDM panel.