Genomic Analysis of Uterine Lavage Fluid Detects Early Endometrial Cancers and Reveals a Prevalent Landscape of Driver Mutations in Women without Histopathologic Evidence of Cancer: A Prospective Cross-Sectional Study

Abstract

Background: Endometrial cancer is the most common gynecologic malignancy, and its incidence and associated mortality are increasing. Despite the immediate need to detect these cancers at an earlier stage, there is no effective screening methodology or protocol for endometrial cancer. The comprehensive, genomics-based analysis of endometrial cancer by The Cancer Genome Atlas (TCGA) revealed many of the molecular defects that define this cancer. Based on these cancer genome results, and in a prospective study, we hypothesized that the use of ultra-deep, targeted gene sequencing could detect somatic mutations in uterine lavage fluid obtained from women undergoing hysteroscopy as a means of molecular screening and diagnosis.

Methods and findings: Uterine lavage and paired blood samples were collected and analyzed from 107 consecutive patients who were undergoing hysteroscopy and curettage for diagnostic evaluation from this single-institution study. The lavage fluid was separated into cellular and acellular fractions by centrifugation. Cellular and cell-free DNA (cfDNA) were isolated from each lavage. Two targeted next-generation sequencing (NGS) gene panels, one composed of 56 genes and the other of 12 genes, were used for ultra-deep sequencing. To rule out potential NGS-based errors, orthogonal mutation validation was performed using digital PCR and Sanger sequencing. Seven patients were diagnosed with endometrial cancer based on classic histopathologic analysis. Six of these patients had stage IA cancer, and one of these cancers was only detectable as a microscopic focus within a polyp. All seven patients were found to have significant cancer-associated gene mutations in both cell pellet and cfDNA fractions. In the four patients in whom adequate tumor sample was available, all tumor mutations above a specific allele fraction were present in the uterine lavage DNA samples. Mutations originally only detected in lavage fluid fractions were later confirmed to be present in tumor but at allele fractions significantly less than 1%. Of the remaining 95 patients diagnosed with benign or non-cancer pathology, 44 had no significant cancer mutations detected. Intriguingly, 51 patients without histopathologic evidence of cancer had relatively high allele fraction (1.0%-30.4%), cancer-associated mutations. Participants with detected driver and potential driver mutations were significantly older (mean age mutated = 57.96, 95% confidence interval [CI]: 3.30-∞, mean age no mutations = 50.35; p-value = 0.002; Benjamini-Hochberg [BH] adjusted p-value = 0.015) and more likely to be post-menopausal (p-value = 0.004; BH-adjusted p-value = 0.015) than those without these mutations. No associations were detected between mutation status and race/ethnicity, body mass index, diabetes, parity, and smoking status. Long-term follow-up was not presently available in this prospective study for those women without histopathologic evidence of cancer.

Conclusions: Using ultra-deep NGS, we identified somatic mutations in DNA extracted both from cell pellets and a never previously reported cfDNA fraction from the uterine lavage. Using our targeted sequencing approach, endometrial driver mutations were identified in all seven women who received a cancer diagnosis based on classic histopathology of tissue curettage obtained at the time of hysteroscopy. In addition, relatively high allele fraction driver mutations were identified in the lavage fluid of approximately half of the women without a cancer diagnosis. Increasing age and post-menopausal status were associated with the presence of these cancer-associated mutations, suggesting the prevalent existence of a premalignant landscape in women without clinical evidence of cancer. Given that a uterine lavage can be easily and quickly performed even outside of the operating room and in a physician's office-based setting, our findings suggest the future possibility of this approach for screening women for the earliest stages of endometrial cancer. However, our findings suggest that further insight into development of cancer or its interruption are needed before translation to the clinic.

Conflict of interest statement

I have read the journal's policy and the authors of this manuscript have the following competing interests: TH, CAS, and JI are paid employees of Swift Biosciences.

Figures

Fig 1. Overview of the study pipeline beginning with collection of uterine lavage fluid at the initiation of hysteroscopy.

Figure designed by Jill Gregory.

Fig 2. Microscopic views of hematoxylin-eosin stained sections of all seven uterine cancer specimens diagnosed by classic histopathology.

(A) PT398: endometrioid, stage IA/grade 1 cancer. Arrow points to minute (<1 mm) focus of adenocarcinoma, adjacent to benign inactive endometrium (40x). (B) PT433: endometrioid, stage 1A/grade 1 cancer. Malignant glands organized in cribiform architecture. A small area of squamous differentiation is noted by the arrow (200x). (C) PT451: endometrioid, stage 1A/grade 1 cancer. Arrow points to a small focus of adenocarcinoma arising within the endometrial mucosa (40x). (D) PT484: endometrioid, stage 1A/grade 2 cancer arising within a polyp. Noted are cribiformed glands and complex architecture within the carcinoma (100x). (E1) PT468: fragment of high grade serous carcinoma with hyperchromatic nuclei (arrows) arranged in a complex glandular pattern (200x). (E2) PT468: another fragment of this tumor with grade 1–2 endometrioid carcinoma, with back-to-back glands (arrows) and lower-grade nuclei than the serous component presented in E1 (100x). (F) PT488: carcinosarcoma, stage 3A/grade 3 showing the biphasic features of high-grade carcinoma (top arrow pointing to the left of the image) and high-grade sarcoma (bottom arrow pointing to the right of the image) (40x). (G) PT492: endometrioid, stage IA/grade 1 cancer. High-power view of the cribiformed glands composed of malignant cells with atypical nuclei (arrow highlights one of these) depicting a grade 1 endometrioid adenocarcinoma ("L," lumen of the malignant glands) (400x).

Fig 3. Flowchart depicting the numbers of patients in each step of the study: enrollment, sample collection, ultra-deep sequencing of samples, and molecular and histopathological classification.

Fig 4. Mutation distributions detected by the 12-gene panel among the study cohort.

Patients are represented along the x-axis first by their histopathologic diagnosis (represented in top bar) then by total mutation number. Mutation types were color-coded hierarchically, displaying the most consequential mutation type (driver, potential driver, passenger) detected at each patient/gene intersection, as some genes carried multiple mutations. NGS-defined mutations validated by dPCR or Sanger sequencing are represented by a black dot. Note: a number of genes had multiple mutations validated.

Fig 5. Patients with somatic mutations identified by NGS using the targeted 12-gene panel.

Patients are ordered based on the sum of driver and potential driver mutations. Highlighted cases are those in which cancer was diagnosed by histopathology. Patient 433 is not represented because her samples were sequenced using the 56-gene panel.