Inaugurating High-Throughput Profiling of Extracellular Vesicles for Earlier Ovarian Cancer Detection

Abstract

Detecting early cancer through liquid biopsy is challenging due to the lack of specific biomarkers for early lesions and potentially low levels of these markers. The current study systematically develops an extracellular-vesicle (EV)-based test for early detection, specifically focusing on high-grade serous ovarian carcinoma (HGSOC). The marker selection is based on emerging insights into HGSOC pathogenesis, notably that it arises from precursor lesions within the fallopian tube. This work thus establishes murine fallopian tube (mFT) cells with oncogenic mutations and performs proteomic analyses on mFT-derived EVs. The identified markers are then evaluated with an orthotopic HGSOC animal model. In serially-drawn blood of tumor-bearing mice, mFT-EV markers increase with tumor initiation, supporting their potential use in early cancer detection. A pilot clinical study (n = 51) further narrows EV markers to five candidates, EpCAM, CD24, VCAN, HE4, and TNC. The combined expression of these markers distinguishes HGSOC from non-cancer with 89% sensitivity and 93% specificity. The same markers are also effective in classifying three groups (non-cancer, early-stage HGSOC, and late-stage HGSOC). The developed approach, for the first time inaugurated in fallopian tube-derived EVs, could be a minimally invasive tool to monitor women at high risk of ovarian cancer for timely intervention.

Keywords: diagnostics; extracellular vesicles; ovarian cancer.

Scheme 1

Study design. High‐grade serous ovarian cancer (HGSOC) is considered to arise from precursor lesions within the fallopian tube (FT). Circulating EVs from FT precursor lesions thus can serve as early HGSOC biomarkers. In this study, we identified EV markers specific to FT carcinoma and detected them in the blood samples of HGSOC patients. Analyzing surface markers on FT‐derived EVs allowed for differentiating early‐ (stage I & II) and late‐stage (stage III & IV) HGSOC patients.

Figure 1

Sensitive, high‐throughput EV assay. A) SAViA (signal amplifying vesicles in array) scheme. EVs are captured on a multiwell plate via physical adsorption. Target EV protein is labeled with a primary antibody (1° Ab) which is further labeled with a secondary antibody (2° Ab) conjugated with horseradish peroxidase (HRP). With tyramide‐biotin and hydrogen peroxide (H₂O₂) added, the HRP catalyzes the dense deposition of biotin. Finally, the analytical signal is detected by adding fluorescent streptavidin (StAv‐BV510). B) Different EV‐assay formats were compared for the analytical signal. Physisorptive EV immobilization produced a higher signal (fivefold) than Ab‐based EV capture (left). Among the physisorptive EV assays, applying the tyramide amplification generated the highest signal. The overall signal increase was about 10³‐fold. Data are displayed as mean ± s.d. (n = 3). FL, fluorescent. C) The SAViA assay displayed superior sensitivity compared to conventional ELISA. Based on CD63 titration curves, the estimated detection limits were 2.4 × 10⁴ EV mL⁻¹ for SAViA and 3.0 × 10⁸ EV mL⁻¹ for ELISA. a.u., arbitrary units. Data are displayed as mean ± s.d. from technical triplicates.

Figure 2

Generation and characterization of mFT cell lines. (A) FT cells were isolated from genetically engineered mice harboring mutations in Brca1 or Brca2, as well as Tp53 and Pten. Isolated cells were rendered oncogenic through the doxycycline treatment. Tumor animal models were generated by implanting the transformed cells into mice. GEMM, genetically engineered mouse model. B) Oncogenic mFT cells expressed FT‐specific protein (PAX8) and HGSOC markers (CA125, γH2AX). Normal tissue samples (uterus, ovary, FT) lacked HGSOC markers, while PAX8 was positive only with FT tissue. C) Immunofluorescence microscopy confirmed that the oncogenic mFT cells (mFT3635 and mFT3666) expressed key HGSOC markers (CA125, p53, Ki67, WT1) at the cellular level. Scale bar, 50 µm. D) Under in vitro ultra‐low adherence culture conditions, oncogenic mFT cells (mFT3635 and mFT3666) formed tumor spheroids (left). When transferred to adhesion plates, tumor spheroids adhered to the surface and spread, demonstrating their capacity to engraft. Scale bar, 100 µm.

Figure 3

mFT‐EV marker selection. A) EVs from oncogenic mFT cells were imaged via transmission electron microscopy (TEM). B) Size distribution of EVs in TEM images. No significant difference (P = 0.758, unpaired two‐sided t‐test) was observed. Each dot represents a single EV. Error bars, s.d. C) Bulk EV analysis confirmed that mFT EVs were enriched with canonical EV markers (i.e., CD63, CD9, CD81, TSG101) and devoid of a non‐EV marker (histone H2B), a.u., arbitrary units. Data are displayed as mean ± s.d. (n = 4). D) Marker selection algorithm. EVs from mFT3635 and mFT3666 cells were processed for proteomic analysis. Detected proteins were filtered for their location in the cell membrane (Uniprot) and presence in EVs (EVpedia), and the outcomes were further curated through a literature search (PubMed). E) Heatmap of proteins (n = 169) found in EVs from mFT3635 and mFT3666 cell lines. The data‐driven approach selected nine candidate markers (PODXL, JUP, TNC, VCAN, CD24, EpCAM, HE4, FOLR1, and CA125). F) Gene ontology (GO) analysis showed that the selected markers were strongly associated with cellular adhesion. GO analysis was performed with STRING v11.5.

Figure 4

Serial EV profiling with an HGSOC animal model. A) Study design. Oncogenic mFT cells were implanted into ovary fat mass in mice. Blood samples were collected from the animals before engraftment and up to 3 months thereafter. For each sample, EVs were collected and profiled by SAViA for nine HGSOC candidate markers. B) Survival analysis of mFT3635 (Brca1 ^−/−) and mFT3666 (Brca2 ^−/−) implanted animals. No significant difference (P = 0.989; log‐rank test) in survival was observed. The median survival was 107 (mFT3635, n = 6) and 114 days (mFT3666, n = 6). C) Immunohistochemical staining confirmed the expression of the FT epithelial (PAX8) and tumor markers (p53, WT1, STMN1) in a late‐stage (day 90) tumor. D) Longitudinal EV profiling in tumor‐bearing animals (n = 6). Nine HGSOC markers were measured in plasma EVs. The heatmap shows the z‐score of each marker. E) The expression of HGSOC markers increased after tumor initiation (day 9) and peaked 30 days after the mFT cell implant. Each data point is the average of fold changes from six animals. Data are displayed as mean ± s.d. (n = 6). F) Single EVs were imaged in plasma samples collected before the mFT cell engraftment (day 0) and during disease progression (day 30). EVs were stained for tetraspanins (CD63, CD9), PAX8 (FT epithelial marker), and CA125 marker (see Figure S8, Supporting Information for other markers). G) Tetraspanin‐positive EVs were present in both samples (day 0 and day 30) with no significant difference in numbers (P = 0.290; non‐paired, two‐sided t‐test). PAX8‐positive EVs, however, significantly increased in the tumor sample (P = 0.019; non‐paired, two‐sided t‐test). Data are displayed as mean ± s.d. (n = 15 field of views). H) EV imaging revealed that more EVs were both PAX8 and HGSOC‐marker positive in tumor samples. Data are displayed as mean ± s.d. (n = 3). The P‐values (non‐paired, two‐sided t‐test) are 0.002 (CA125), 0.0009 (VCAN), 0.014 (TNC), 0.105 (FOLR1), and 0.018 (HE4).

Figure 5

Profiling of plasma EVs from HGSOC patients. A) EVs from clinical plasma samples were profiled for HGSOC markers and CD63 (n = 14, non‐cancer individuals; n = 37, HGSOC patients). The expression of each marker was normalized (z‐score) and displayed in a heatmap. B) CD63 measurements confirmed the presence of EVs both in non‐cancer and HGSOC plasma samples. No significant difference was observed in CD63 expression between the two cohorts (P = 0.688; non‐paired, two‐sided t‐test). C) Five markers (EpCAM, CD24, HE4, VCAN, TNC) were chosen from a regression analysis, and their expressions were combined to define the EV_HGSOC score. In the receiver operating characteristic (ROC) analysis, the EV_HGSOC score achieved high accuracies in differentiating HGSOC patients from non‐cancer individuals. AUC, an area under the curve. D) EV_HGSOC scores were higher in early and late‐stage HGSOC patients than in non‐cancer individuals but were similar among HGSOC cohorts (Tukey's multiple comparisons test). Early, stages I & II; Late, stages III & IV. E) Linear discriminant analysis (LDA) model of top‐five markers (EpCAM, CD24, HE4, VCAN, TNC) differentiated three groups: non‐cancer individuals, early‐stage patients, and late‐stage patients. The overall classification accuracy was 74.5%. F) Tumor tissue of HGSOC patients was stained positive for HGSOC markers (EpCAM, CD24, HE4, VCAN, TNC). In the H&E micrograph, annotated are STIC lesions (1) and HGSOC (2 and 3). See Figure S12, Supporting Information, for other HGSOC markers.