FAK-ERK activation in cell/matrix adhesion induced by the loss of apolipoprotein E stimulates the malignant progression of ovarian cancer

The following antibodies were used: Col1a2 (14695), FN1 (15613) and GAPDH (10494) from ProteinTech (USA); MMP-9 (AF909, R&D); MMP-10 (NB100–92182, Novus); p-FAK (Y397, ab39967), LOX (ab174316) and Paxillin (ab32084) from Abcam; p-Erk1/2 (Thr202/Tyr204, 4377), p-Src (Y416, 2101), AlexaFluor goat anti-rabbit IgG (594 conjugate, 8889) and sheep anti-rabbit HRP-linked IgG (7074) from Cell Signaling. Reagents included β-aminopropionitrile (BAPN), PKH26 Red Fluorescent Cell Linker Mini Kit (Sigma), PD-325901 (MedChem Express), Hydroxyproline Detection Kit (Nanjing Jiancheng Bioengineering Institute), Rat tail collagen (Invitrogen), Cell counting Kit-8 (Dojindo) and Matrigel matrix (BD).

Wild Type (WT) (C57BL6) and ApoE−/− (B6.129-Apoe^tm1Smoc) female mice were purchased from Beijing HFK Bioscience. Mice were housed and maintained under specific pathogen-free conditions. For establishment of allografts, ID8-Luciferase cells (1*10⁷) were labeled using PKH26 Cell Linker, and resuspended in 100 μl PBS before intraperitoneal injection into 20-week-old WT and ApoE−/− mice. For BAPN treatment (30 mg/kg, i.p.), ApoE−/− mice were treated every day for one month. Two weeks after the last day of BAPN treatment, ID8-Luciferase cells (1*10⁷) were intraperitoneally injected. For PD-0325901 treatment, following tumor establishment, ApoE−/− mice were randomly assigned the second day and were treated with PD0325901 (25 mg/kg, p.o.) or PBS as a control every day for 2 weeks. The tumors were measured twice a week by quantification of luciferase expression and the animals were sacrificed at the indicated time.

ID8 cells were a kind gift from the The University of Kansas Medical Center, and SKOV3 and OV-90 cells were obtained from the American Type Culture Collection. DMEM, Fetal Bovine Serum (FBS), Insulin-Transferrin-Selenium (ITS), Mycoy’5A medium were purchased from Invitrogen. MCDB105 and 199 medium were purchased from Sigma. ID8 cells were maintained in DMEM supplemented with 4% FBS and 1*ITS. ID8-Luciferase cells were acquired through transfection of lentivirus overexpressing the luciferase reporter gene into ID8 cells. SKOV3 cells were cultured in McCoy’s 5A medium with 10% FBS. OV-90 cells were cultured in MCDB105/199 medium with 10% FBS.

Reverse transcription of 1 μg of total RNA isolated from the indicated tissues was performed using the PrimeScript™ RT reagent Kit (Takara). The cDNA was subjected to qPCR using an iTAQ™ Universal SYBR^® Green Supermix (BIO-RAD). The primer sets used were as follows: Col1a1 (Forward, GCTCCTCTTAGGGGCCACT; Reverse, CCACGTCTCACCATTGGGG), Col1a2 (Forward, GTAACTTCGTGCCTAGCAACA; Reverse, CCTTTGTCAGAATACTGAGCAGC), FN (Forward, ATGTGGACCCCTCCTGATAGT; Reverse, GCCCAGTGATTTCAGCAAAGG), LOX (Forward, TCTTCTGCTGCGTGACAACC; Reverse, GAGAAACCAGCTTGGAACCAG). The primer set for 18S rRNA (Forward, AGGGGAGAGCGGGTAAGAGA; Reverse, GGACAGGACTAGGCGGAACA) was used as a control. Real-time qPCR results were calculated using the delta-delta-Ct (ddCt) algorithm.

Hydroxyproline was extracted and quantified in 20–30 mg of tissue (wet weight) or 300 μl of serum as previously described [18]. Paired primary and metastatic lesions from nine patients with grade III/IV high-grade serous ovarian cancer were included. The histology and grade were confirmed by two licensed pathologists. Fresh human and mouse tissues were fixed, paraffin-imbedded and prepared for sections. For picrosirius red analysis, sections were stained with 0.1% picrosirius red and counterstained with Weigert’s hematoxylin. Masson Trichrome was performed according to the manufacturer’s instruction (Sigma). Serial images were analyzed and quantified by the software Image-pro plus 6.0 (Media Cybernetics, Inc., USA). Three fields were randomly selected and the percentages of positive-stained area were calculated.

For in vivo adhesion assays, ID8 cells were labeled with PKH26 red fluorescent Cell Linker, and a single-cell suspension (5*10⁶ in 0.2 ml PBS) was intraperitoneally injected. After four hours, the omentum was excised after sacrifice. Tissues were rinsed three times with PBS to remove non-adhered cells, and images were captured using a fluorescence microscopy (Olympus DP73, Tokyo, Japan). Then, the omentum was digested in 5% NP-40 for 30 min at 37 °C. After scraping with a spatula, all removed cells were collected by centrifugation and the total fluorescence intensity was quantified. For the in vitro adhesion assay, SKOV3 and OV-90 cells were trypsinized and then suspended in cultured medium containing 50 mM Hepes and 1 mg/ml BSA at a concentration of 5*10⁵ cells/ml, from which 0.1 ml was added to each well of 6-well plate. The cells were allowed to adhere for 40 min. Subsequently, the unbound cells were removed by washing with PBS gently. Cells were fixed in 10% formalin and stained with 0.1% crystal violet. Dye was extracted with 10% acetic acid and the relative adhesion capacity was determined by reading at 595 nm [19].

For immunofluorescence staining, the tumor lesions were snap-frozen with OCT (Sakura Fineteck), and prepared as 7 μM cryo-sections. The sections were fixed with ice acetone and SKOV3/OV-90 cells were fixed with 4% paraformaldehyde. Following permeabilization with 0.1% Triton X-100 (Sigma), fixed sections and cells were blocked with 5% BSA in PBS and incubated with the indicated primary antibodies. Sections and cells were then incubated with AlexaFluor goat anti-rabbit IgG (594 conjugate) and nuclei were stained with DAPI (Invitrogen). Images were captured by fluorescence microscopy (Olympus, Japan). For IHC, paraffin sections were deparaffinized and rehydrated, followed by antigen retrieval and incubation with primary antibodies. HRP-based detection reagents were used, and the immunostaining intensity was scored as previously described [20]. Homogenized tissues and cells were lysed in RIPA buffer with protease inhibitors (Roche) and the protein preparations were subjected to western blot analysis as described previously.

SKOV3 and OV-90 cells were seeded on dishes with or without Collagen pre-coating. For the wound healing assay, the cellular layer was scratched using a plastic pipette tip. The migration of the cells at the edge of the scratch was analyzed after the indicated time. For the Matrigel invasion assay, following 36 h of culture and 12 h of serum-starved treatment, 1*10⁴ cells were plated into a Matrigel (dilution, 1:7) pre-coated trans-well insert (Corning) in serum-free DMEM, and the bottom chamber contained DMEM with 10% FBS. Cells were allowed to invade through the Matrigel-coated inserts for 36 h. The wound closure efficiency and the number of cells that had invaded were quantified using ImageJ software.

Cells (5000) were seeded in triplicates on 96-well plates with or without collagen pre-coating, and those cultured on pre-coated plates were treated with DMSO or PD-325901 (50 nM) for the indicated time. After treatment, cell numbers were quantified using the Cell Counting Kit-8 assay.

The supernatant of the ascites was collected, and snap-frozen in liquid nitrogen and stored at − 20 °C until analysis. Cytokines and chemokines were measured using the Mouse Cytokine Antibody Array (RayBiotech, QAM-CAA-4000), which includes 200 secreted proteins. The data were analyzed using the Quantibody® Q-Analyzer software, and the pathways were studied using the KEGG database.

Student’s t tests and Mann-Whitney tests were used to compare the statistical significance between two groups. One-way ANOVA (Dunnett post-test) was used for multiple comparison analysis. P < 0.05 was considered statistically significant. Values are expressed as the means and standard deviation unless otherwise stated. Graphs and analyses were performed using the GraphPad Prism software.

To explore the ECM alterations during tumorigenesis, we used the Oncomine database (https://​www.​oncomine.​org) to analyze published transcriptional data and found that the transcriptional levels of several ECM components, including collagen 1a1 (Col1a1), Col1a2, fibronectin 1 (FN1) and lysyl oxidase (LOX), were robustly increased in ovarian cancer compared to normal ovaries (Fig. 1, Additional file 1: Figure S1A). Also, the alterations of these genes showed clinical relevance. High mRNA expression of Col1a1, Col1a2, FN1 and LOX was linked to a poor prognosis in ovarian cancer, as shown by the reduced overall survival (OS) and progression free survival (PFS) (http://​kmplot.​com) (Fig. 1b, c). The improved OS and PFS in low-expressers implied a delay in recurrence. To investigate the ECM during ovarian cancer progression, we examined the histology of paired primary lesions and metastases from ovarian cancer patients. Indeed, compared to primary lesions, metastatic tumors showed a desmoplastic stromal response, which is characterized by dense ECM around tumor lesions (Fig. 1d). As collagen is the main structural protein in the extracellular space, we next investigated whether alterations of collagen composition occurred during tumor progression. By Masson’s Trichrome and Picro Sirius Red staining, we found a remarkable increase in fibrillar collagen in the metastatic tissues compared to primary tumors (Fig. 1e). Overall, these results indicate that ECM alterations, especially collagen deposition, frequently occurs during the malignant progression of ovarian cancer.

According to previous reports, the mRNA expression of ApoE showed a moderate increase during ovarian cancer tumorigenesis (Additional file 1: Figure S1B). To investigate whether ApoE loss mediates intraperitoneal ECM alterations, we measured the ECM components of the diaphragm and omentum, which are where ovarian cancer likely spreads (Fig. 2a). qPCR showed an ~ 2-fold increase in the expression of Col1a1, Col1a2, FN1 and LOX in ApoE−/− mice compared to WT controls (Fig. 2b). Consistently, western blot analysis indicated enhanced protein levels of Col1a2, LOX and FN1 in the diaphragm of ApoE−/− mice (Fig. 2c). The protein induction of LOX, an amine oxidase responsible for crosslinking and stabilization of adjacent collagens, was further confirmed by immunofluorescence analysis (Fig. 2d). We next explored the expression levels and structure of collagens using Masson’s Trichrome stain. The representative sections exhibited a remarkable increase in fibrous collagen (blue stain) in ApoE−/− mice compared to WT (Fig. 2e). Furthermore, hydroxyproline, a surrogate for collagen, was increased in both the plasma (157.4 ± 26.2 vs 59.5 ± 21.9 μg/ml; mean ± SD, n = 5) and diaphragm (1.29 ± 0.12 vs 0.94 ± 0.06 μg/mg; mean ± SD, n = 5) of ApoE−/− mice compared to WT mice (Fig. 2f). Taken together, ApoE−/− mice displays ECM accumulation, especially collagen deposition, in the peritoneal cavity.

To explore whether the altered ECM in the peritoneal microenvironment of ApoE−/− mice could modulate the progression of ovarian cancer, we used an ID8 intraperitoneal allograft model. ID8 is a spontaneously tumorigenic mouse ovarian surface epithelial cell line that is not rejected by the host immune system. To monitor tumor progression, ID8 cells were stably transfected with a luciferase reporter, and the cellular membrane was labeled with PKH26 Cell Tracking Dye (Fig. 3a). Two weeks after tumor engraftment, in vivo bioluminescent imaging showed increased luciferase expression in ApoE−/− mice compared to WT (mean radiance: 5.67*10⁶ vs 0.92*10⁶ p/s/cm²/sr, P < 0.005) (Fig. 3b). Then a cohort of mice was sacrificed and dissected under a fluorescent microscope. Likewise, the peritoneal cavity of ApoE−/− mice displayed ~ 2-fold and 2.64-fold increases in tumor number and weight, respectively (Fig. 3c). In the advanced stage of malignancy, morbid ApoE−/− mice presented obvious abdominal distension, caused by excessive hemorrhagic ascites, with a 42 mm abdominal circumference and an 11.5 ml ascites volume on average, compared to 30 mm (P < 0.005) and 7.5 ml (P < 0.05) in WT mice (Fig. 3d). By anatomy and quantification, morbid ApoE−/− mice displayed a heavier tumor burden compared to WT control (Fig. 3e). Supporting those observations, ApoE−/− mice had a shortened survival compared to WT mice (Fig. 3f). Together these results revealed that ApoE loss accelerated the malignant progression of ID8 intraperitoneal allografts.

To assess the mechanism by which ECM alterations promote ovarian cancer progression, we further evaluated the tumor specimens from ApoE−/− and WT mice. Hematoxylin and eosin (H&E) staining showed that the sizes of tumor lesion at two weeks were significantly larger in ApoE−/− mice (Fig. 4a). Also, lesions in the ApoE−/− mice exhibited increased collagenous tissues, especially around the tumor lesions, indicated by Masson’s Trichrome stain (Fig. 4b). Therefore, we reasoned that the altered ECM might stimulate tumor progression by providing a favorable niche for cancer cells. Because ascites facilitate peritoneal spreading of ovarian cancer [2], we then analyzed the cytokines/chemokines in the ascites using mouse-specific cytokine profiling arrays. Though there might be cellular contamination of the secretome due to the increase of E-cadherin, a group of molecules that potentiate ECM reorganization and MMP activation stood out (Fig. 4c, d and Additional file 2: Table S1). Since MMPs are important regulators of cell invasion, we measured MMP-10 and MMP-9 expressions using IHC. Our results confirmed the markedly elevated expression of MMP-10 and MMP-9 in the tumor lesions of ApoE−/− mice compared to that in WT mice (Fig. 4e). Although ApoE could suppress angiogenesis in melanoma [13], we did not observe significant differences of angiogenesis in tumor lesions from ApoE−/− and WT mice (Additional file 3: Figure S2). In vitro, collagen pre-coating of substrates could stimulate the proliferation of SKOV3 and OV-90 cells, two human ovarian cancer cell lines (Additional file 4: Figure S3A). The protein expressions of MMP-10 and MMP-9 were increased in cells cultured on collagen pre-coated substrates (Additional file 4: Figure S3B), and these cells exhibited enhanced invasive potentials (Additional file 4: Figure S3C-F). Overall, ECM accumulation, especially collagen accumulation, enhances the invasive potentials of ovarian cancer cells.

LOX is a key regulator of collagen homeostasis and fibronectin expression [21]. β-Aminopropionitrile (BAPN), a specific inhibitor of LOX activity, can ameliorate fibrosis by suppressing collagen synthesis [22] and reducing FN expression [23]. Therefore, we used BAPN to decrease the synthesis and secretion of ECM for reversing ECM reprogramming in ApoE−/− mice. BAPN treatment was initiated in 20-week-old ApoE−/− mice and sustained for one month prior to tumor establishment (Fig. 5a). To evaluate the efficiency of BAPN treatment, a group of ApoE−/− mice was sacrificed and compared to PBS-treated mice (CTRL). Hydroxyproline was measured and indicated that BAPN treatment curtailed the collagen content in the serum (106 ± 9.4 vs 168 ± 17.9 μg/ml; mean ± SD, n = 5) and diaphragm (0.91 ± 0.16 vs 1.28 ± 0.06 μg/mg; mean ± SD, n = 5) (Fig. 5b). Also, Masson’s Trichrome stain exhibited decreased collagen deposition (Fig. 5c). ID8 allografts were established two weeks after the last day of BAPN pre-treatment. Since the first step of the metastatic cascade is the adhesion of cancer cells, we sacrificed a cohort of mice four hours after ID8 intraperitoneal injection to assess adhesions in vivo. Whereas more cancer cells were attached to the tissues surfaces in ApoE−/− mice compared to WT controls, pre-treatment with BAPN robustly attenuated the adhesions compared to CTRL mice (Fig. 5d). By monitoring tumor progression using in vivo bioluminescence, we found a significant decrease in the tumor burden in the BAPN-treated group compared to the CTRL group, both at two weeks and two months post allografts establishment (2 weeks: 2.11 vs 3.13*10⁶ p/s/cm²/sr, P < 0.05; 2 months: 0.93 vs 2.32*10⁸ p/s/cm²/sr, P < 0.005) (Fig. 5e). Furthermore, IHC staining showed significantly diminished expression of MMP-9 and MMP-10 with BAPN treatment (Fig. 5f). Overall, these results suggest that ECM accumulation stimulates ovarian cancer progression by promoting adhesion.

Increased collagen-matrix density enhances adhesion and activates the FAK-Rho-ERK pathway, which enhances the invasive phenotype of mammary epithelial cells [24]. Focal adhesion kinase (FAK), a mediator of cell adhesion, motility and angiogenesis, can transmit extracellular signals into cells and thus facilitate various neoplastic processes [25]. In our model, immunofluorescence analysis at the early stage of the tumor establishment showed significantly increased FAK activation in the tumor cells from ApoE−/− mice compared to WT controls (Fig. 6a). Also, adhesions of SKOV3 and OV-90 cells were dramatically enhanced with collagen pre-coated substrates, and immunofluorescence analysis confirmed the collagen signaling-induced Paxillin-positive adhesions (Additional file 5: Figure S4A, B). Notably, phospho-FAK was significantly increased, and its signaling partner, Src, was also activated (Additional file 5: Figure S4B, C). To explore the FAK-mediated signaling underpinning the malignant phenotype of ovarian cancer cells, we measured the activity of ERK in tumor lesions with IHC staining. Compared to WT mice, ApoE−/− mice showed active ERK signaling in tumor lesions (Fig. 6b). Consistently, SKOV3 and OV-90 cells seeded on collagen pre-coated substrates also exhibited ERK activation (Additional file 5: Figure S4B, C). To explore the FAK-ERK activation in ovarian cancer progression, we treated ApoE−/− mice after ID8 engraftment with the MEK inhibitor (MEKi) PD-325901 to selectively inhibit ERK phosphorylation. The treatment continued for one month. By IHC staining, we found that PD-325901 robustly diminished MMP-9 protein expression in the tumor lesions (Fig. 6c). Also, PD-325901 significantly curtailed the collagen signaling-induced cell proliferation, and decreased the synthesis of MMP-9 and MMP-10 (Additional file 5: Figure S4D, E). These results were consistent with previous reports showing that MMPs are regulated by the MAPK/ERK pathway [26, 27]. Consistently, BAPN pre-treatment in ApoE−/− mice caused decreased FAK activation accompanied by diminished cancer cells adhesions (Fig. 6a). Also, BAPN-treated mice developed smaller metastases with weak p-ERK expression (Fig. 6b). More importantly, our results demonstrated that PD-325901 treatment significantly delayed the tumor progression in ApoE−/− mice compared to PBS-treated mice (Fig. 6d). Overall, our results strongly suggest that the remodeled microenvironment in ApoE−/− mice enhances malignancy of ovarian cancer by increasing adhesion-mediated FAK-ERK-MMP signaling.