Keratin-14 (KRT14) Positive Leader Cells Mediate Mesothelial Clearance and Invasion by Ovarian Cancer Cells

Abstract

Epithelial ovarian cancer metastasis is driven by spheroids, which are heterogeneous cancer cell aggregates released from the primary tumour mass that passively disseminate throughout the peritoneal cavity to promote tumour spread, disease recurrence, and acquired chemoresistance. Despite their clinical importance, the molecular events that control spheroid attachment and invasion into underlying healthy tissues remain poorly understood. We examined a novel in vitro invasion model using imaging mass spectrometry to establish a "snapshot" of the spheroid/mesothelial interface. Amongst numerous adhesion-related proteins, we identified a sub-population of highly motile, invasive cells that expressed the basal epithelial marker KRT14 as an absolute determinant of invasive potential. The loss of KRT14 completely abrogated the invasive capacity, but had no impact on cell viability or proliferation, suggesting an invasion-specific role. Our data demonstrate KRT14 cells as an ovarian cancer "leader cell" phenotype underlying tumor invasion, and suggest their importance as a clinically relevant target in directed anti-tumour therapies.

Keywords: cancer metastasis; invasion; keratin 14; leader cells; ovarian cancer.

Figure 1

(A) Peritoneal microenvironment model. A mesothelial monolayer is grown on the basement membrane in the upper chamber of an xCELLigence CIM plate; the lower chamber contains media and microelectrodes. Disruption of the mesothelial cell layer (clearance) and the invasion of ovarian cancer cells alters the current flow and creates impedance, resulting in an increased cell index as a measure of invasion. (B) Real-time cell analysis (RTCA) invasion assay. Three patient-derived high-grade serous (HGS) ovarian cancer/mesothelial co-cultures (A—blue, B—green, C—red), and one benign fibroma/mesothelial (pink) co-culture were monitored in real time over an 18-h period, mean 15-min impedance readings with lower standard deviation are shown; n = 2 wells/sample of a representative experiment are shown. (C) Mesothelial clearance. Parallel assays demonstrating mesothelial clearance by the three patient-derived ovarian cancer spheroids (A–C) but not benign fibroma spheroids over 48 h with a representative image are shown. (D) RTCA adhesion and proliferation. The adhesion and proliferation of patient-derived ovarian cancer cells (A–C) and the benign control on uncoated and fibronectin-coated wells was measured by RTCA assay. Samples were monitored over a 13-h period with mean 5-min impedance and lower standard deviation shown; n = 2 wells/sample of one representative experiment.

Figure 2

(A) Parallel endpoint Boyden chamber assays. Boyden chamber assays using labelled mesothelial cells overlaid with individual patient-derived ovarian cancer spheroids; we observe no invasion of the mesothelial cells at MALDI imaging collection points; n = 3 wells/sample of one representative experiment. (B) Haemotoxylin and Eosin (H&E) staining of the invasive interface. H&E staining identifying the invading interface of ovarian cancer spheroid mesothelial co-cultures and the interface used for MALDI imaging mass spectrometry (IMS). (C) MALDI IMS of the invading interface. MALDI IMS identifies: CDCA8, HNRN, keratin-14 (KRT14) and FNDC3B expressed at the invading interface of ovarian/mesothelial co-cultures. (D) Representative qRT-PCR of MALDI identified candidates using fresh-frozen confirmed primary high-grade serous ovarian tumours or normal whole ovary (n = 3/group) where individual data points represent individual patient samples. (E) Representative IHC of individual MALDI identified candidates in HGSC primary ovarian samples.

Figure 3

(A) Western blot analyses demonstrating the endogenous expression of MALDI identified candidates FNDC3B, HNRN, CDCA8, or KRT14 in ovarian cancer cell lines or the mesothelium LP-9. (B) RTCA proliferation assays. Representative results from OVCAR4 real-time proliferation assays conducted in complete culture medium. Data are mean impedance readings from triplicate wells taken every 15 min over a 30 h assay period with lower standard deviation shown, where the non-targeting control is NT and individual CRISPR gene knock-outs are indicated. (C) RTCA invasion assays. Representative results from OVCAR4 real-time invasion assays. Data are mean impedance readings from duplicate wells taken every 15 min over a 26 h assay period with lower standard deviation shown. Individual CRISPR gene knock-outs are indicated on the graph. (D) Wound-healing assays. Wounded OVCAR4 cell lines (UT = untreated, NT= non-targeting and KRT14^KO = KRT14 CRISPR knockout) were cultured in complete medium. Cells were imaged at regular intervals ranging from 0 to 48 h with the original wound area indicated by dotted lines to aid in the assessment of wound closure. Images presented are representative of the observations of three separate experiments at assay commencement and 48 h, the assay was completed on a minimum of three separate occasions.

Figure 4

(A) Representative qRT-PCR of KRT14 expression in LP9 cells, whole normal ovary, benign fibroma, primary ovarian tumour, or ascites-derived cells; and migratory leader cell populations (n = 3 separate patient isolations where individual data points represent individual patient samples). (B) KRT14 expression in cultured cells by immunofluorescence. KRT14 expression was detected as a granular cytoplasmic stain expressed by a small proportion of ovarian cancer cells in the OVCAR4 line cultured as a two-dimensional (2D) monolayer. Scale bar = 100 µm. (C) KRT14 expression was detected at the periphery of OVCAR4 ovarian cancer cells cultured as three-dimensional (3D) spheroids with no signal detected in the core of the spheroid. Scale bar = 100 µm. (D) Localisation of KRT14 at the periphery of ovarian cancer spheroids by comparison to the intracellular staining observed for N-cadherin. Scale bar = 100 µm. (E) Vector control (NT), KRT14^KO, and KRT14^OE cells were suspended in SFM/0.25% methylcellulose in U-bottom 96-well plates and imaged at regular intervals (0–48 h) to observe spheroid aggregation. Images are representative of multiple wells observed at the 13-hour and 48-hour time points. Scale bar = 100 µm. (F) Spheroid attachment to the target mesothelium. Vector control (NT) and KRT14^KO spheroids were co-cultured with the target mesothelium, and attachment was measured 6 h post-addition. The results presented are from one representative assay where data points are the attachment counts of three individual wells per cell line.

Figure 5

(A) Mesothelial displacement assays. Ovarian cancer spheroids were generated from wild-type (WT), KRT14^KO and KRT14^OE lines and overlaid onto a confluent layer of mesothelial cells and imaged by light microscope. Representative images following overnight and 48-h co-culture are presented where dotted lines indicate spheroid outgrowth and mesothelial displacement. (B,C) Embedding and outgrowth assays. OVCAR4 ovarian cancer spheroids were formed and overlaid into (B) Collagen I (24 h) or (C) Matrigel (48 h) samples, and subsequently fixed and stained for KRT14 expression. (D) Wound-healing assays OVCAR4 cells were grown as a monolayer, cells were wounded, fixed, and stained for KRT14 expression at time-points preceding total wound closure. Images from a representative well are shown with scale bars = 100 µm.

Figure 6

(A) Table of the immunohistochemical assessment of KRT14 expression in ovarian cancer tissue microarrays. Staining intensity was scored according to 0 (no stain), 1 (low), 2 (medium), or 3 (high). (B) Representative immunohistochemical staining of KRT14 in the various ovarian cancer subtypes (as labelled in figure) from TMA assessments. Scale bar = 200 µm. (C–F) Association of KRT14 expression with progression-free survival (PFS) according to (C) overall PFS (HR 1.17; 95% CI 1.03–1.33 p < 0.015; B); (D) early stage (I/II) diagnosis (HR 1.96; 95% CI 1.08–3.56 p < 0.025); (E) following platinum and taxol-based chemotherapy (HR 1.27; 95% CI 1.07–1.51 p < 0.006); and (F) following optimal debulk (HR 1.24; 95% CI 1.03–1.5 p < 0.026).

Figure 6

(A) Table of the immunohistochemical assessment of KRT14 expression in ovarian cancer tissue microarrays. Staining intensity was scored according to 0 (no stain), 1 (low), 2 (medium), or 3 (high). (B) Representative immunohistochemical staining of KRT14 in the various ovarian cancer subtypes (as labelled in figure) from TMA assessments. Scale bar = 200 µm. (C–F) Association of KRT14 expression with progression-free survival (PFS) according to (C) overall PFS (HR 1.17; 95% CI 1.03–1.33 p < 0.015; B); (D) early stage (I/II) diagnosis (HR 1.96; 95% CI 1.08–3.56 p < 0.025); (E) following platinum and taxol-based chemotherapy (HR 1.27; 95% CI 1.07–1.51 p < 0.006); and (F) following optimal debulk (HR 1.24; 95% CI 1.03–1.5 p < 0.026).