The stromal microenvironment is known to provide a niche that supports tumor growth ². To identify mediators produced by the stromal microenvironment that support ovarian tumor growth, we performed transcriptome profiling of microdissected stromal and epithelial components of ovarian tumor tissues. Among the differentially expressed genes, we identified 19 upregulated secretory protein coding genes in the stromal component of HGSC samples (Fig. 1A and Table 1). Microfibrillar-associated protein 5 (MFAP5), which we have shown to be amplified and overexpressed in a subset of ovarian cancer cells ⁶, was selected for further validation. Validation by quantitative real-time PCR revealed significantly higher levels of MFAP5 mRNA expression in the cancer stroma than the epithelial components and normal ovarian stroma (Fig. 1B). Immunohistochemistry also revealed a higher level of MFAP5 protein expression in the cancer-associated stroma, particularly in CAFs, than in normal ovarian fibroblasts (NOFs) and ovarian cancer cells (Fig. 1C). Western blot analysis confirmed a higher endogenous level of MFAP5 in ovarian CAFs when compared to normal fibroblasts and ovarian cancer cell lines (Fig. 1D).

To determine the clinical significance of stromal MFAP5 expression, we performed MFAP5 immunolocalization on 130 advanced HGSC samples. Kaplan-Meier analysis showed that high stromal MFAP5 expression is significantly associated with poor clinical outcomes (p < 0.001; log-rank test) (Fig. 1E). Cox regression analysis with adjustment for age and debulking status confirmed the prognostic significance of stromal MFAP5 expression (hazard ratio, 2.603; p=0.001).

Previous studies by our group revealed that MFAP5 could exert its effect on cancer cells in an autocrine manner via binding of its RGD domain to αVβ3 integrin. To determine CAF-derived MFAP5’s role in ovarian cancer progression, CAFs expressing high levels of MFAP5 were co-cultured with A224 and ALST ovarian cancer cells, which have undetectable levels of endogenous MFAP5 and high levels of αVβ3 integrin expression (Supplementary Fig. 1). Experimental results demonstrated that anti-MFAP5 and anti-αVβ3 integrin antibodies attenuated the stimulatory effect of MFAP5 on cancer cell motility (p < 0.001; two-tailed Student t-test) (Fig. 2A).

Real-time imaging of the HGSC cell line OVCA432 treated with purified recombinant MFAP5 protein (recMFAP5) revealed that exogenous recMFAP5 increased OVCA432 cell motility (Fig. 2B). To further validate the motility-stimulating effects of exogenous MFAP5, Boyden chamber based motility assays were performed on A224 and ALST cell lines. Increased migrated cells were observed in both recMFAP5-treated cell lines. Presence of anti-αVβ3 antibody attenuated the stimulatory effect of MFAP5 on cell motility (p < 0.01; two-tailed Student t-test) (Fig. 2C). Data from the co-culture studies and exogenous MFAP5 treatment confirmed that the stimulatory effect of MFAP5 on ovarian cancer cell motility was mediated by its binding to the αVβ3 integrin receptor.

In addition, A224 and ALST cells seeded onto Matrigel matrix with recMFAP5 resulted in three times the number of cells invaded through the matrix when compared to the control (Fig. 2D), suggesting that MFAP5 also stimulated the invasion potential of the ovarian cancer cells.

To determine the effects of exogenous MFAP5 on ovarian tumor growth in vivo, a MFAP5 overexpressing NOF line was generated. Overexpression of MFAP5 was confirmed by Western blot analysis (Supplementary Fig. 2A). Cell proliferation assay was performed to confirm similar proliferation rates in MFAP5 overexpressing and control cells (Supplementary Fig. 2B). For the mouse study, fibroblasts embedded Matrigel plugs were intraperitoneally implanted into nude mice followed by cancer cell injection (Fig. 2E). Matrigel plugs harvested were double-stained for cytokeratin 18 and MFAP5. Results showed that more cancer cells were found in Matrigel plugs embedded with MFAP5 overexpressing fibroblasts than those embedded with control fibroblasts (Figs. 2F and 2G), suggesting that stromal MFAP5 promoted ovarian cancer cell migration and invasion into the Matrigel plugs.

To evaluate the effect of stromal MFAP5 silencing on ovarian cancer metastasis, we silenced murine stromal Mfap5 in an orthotopic model of ovarian carcinoma ^{7, 8} by intravenous injection of Mfap5-specific siRNAs incorporated into chitosan nanoparticles (CH-NP) (Fig. 3A). We then evaluated the effect of these siRNA sequences on murine Mfap5 silencing was evaluated in murine fibroblasts using both quantitative RT-PCR and Western blot analyses (Supplementary Fig. 3A). One week after initial cancer cells injection, two Mfap5-targeting siRNAs and a non-targeting sequence were packaged in CH-NP and delivered via tail vein injection to luciferase labeled cancer cells bearing nude mice twice a week throughout the course of the experiment. Markedly lower luciferase activity was detected in Mfap5 knockdown groups compared to the control group at week 6 (Fig. 3B). At the end point of the study, necropsy revealed amelioration of cancer metastasis in the stromal Mfap5 silenced groups when compared to the control group (Fig. 3C). Mfap5 siRNAs/CH-NP resulted in a significant decrease in the number of metastatic tumor nodules found at mesentery, omentum, peritoneum, diaphragm, liver and spleen (p < 0.001; Mann-Whitney U test), tumor weight (p < 0.001; Mann-Whitney U test) and ascites volume (p < 0.001; Mann-Whitney U test) (Fig. 3D), suggesting that stromal Mfap5 silencing inhibited ovarian tumor growth and metastatic spread. Primary tumor burden was quantified by measuring the primary tumor area in the ovaries using images of H&E tissue sections prepared from tumors harvested from the three treatment groups. Similar primary tumor burdens were observed in the control and Mfap5-targeted animal groups suggesting that reduced metastasis observed in the MFAP5-targeting siRNA treatment group is the result from reduced cancer cell motility and invasive potential instead of reduced primary tumor burden (Fig. 3E). Immunostaining was performed on paraffin sections of tumor tissue harvested to validate in vivo silencing of murine stromal Mfap5 (Fig. 3F).

Transcriptome profiling on MFAP5 treated and untreated OVCA432 cells followed by pathway analysis identified a set of up-regulated motility-promoting genes associated with calcium signaling in recMFAP5-treated cells (Figs. 4A and 4B). Troponin C type 1 (TNNC1) was chosen for further studies. TNNC1 contains a calcium-binding subunit and facilitates the interaction between actin and myosin in skeletal and cardiac muscle cells and the formation of stress fibers, suggesting that it may play a key role in mediating MFAP5’s stimulatory effect on ovarian cancer cell motility. We performed quantitative RT-PCR analysis on five MFAP5-treated ovarian cancer cell lines and observed a significant increase in TNNC1 expression in all of them (Fig. 4C). Results of motility assays demonstrated that silencing of TNNC1 by a validated TNNC1-targeting siRNA (Supplementary Figs. 3B and 3C) abolished MFAP5’s stimulatory effect on A224 and ALST cell motility (Fig. 4D), suggesting that TNNC1 is the effector protein for MFAP5-enhanced cancer cell motility. In addition, F-actin staining showed that recMFAP5 treatment led to an F-actin cytoskeleton with increased density and organization in ovarian cancer cells (Fig. 4E), but that these effects were abrogated in TNNC1 siRNAs transfected cells (Supplementary Fig. 4) suggested that MFAP5 stimulated cell motility via induction of TNNC1 expression and F-actin cytoskeleton reorganization.

Transcriptome profiling and pathway analysis data suggest that cytosolic calcium mobilization is integral to MFAP5’s effect on ovarian cancer motility. The observations that MFAP5’s motility stimulatory effect was attenuated and F-actin reorganizing effect was abrogated in calcium chelator BAPTA/AM preloaded cancer cells (Figs. 5A and 5B), suggesting that calcium signaling is involved in modulating MFAP5’s functions. Using the calcium dye Fluo-4 AM and confocal fluorescence microscopy, we observed that exogenous MFAP5 mobilized intracellular Ca²⁺ in A224 and ALST cells (Fig. 5C). Further evaluation on the mechanisms involved in calcium mobilization in ALST cells showed that treatment with the L-type Ca²⁺ channel blocker nifedipine attenuated but did not abolish MFAP5-induced calcium mobilization (Supplementary Fig. 5A). The channel opener BAYK8644 triggered calcium mobilization in ALST cells (Supplementary Fig. 5B), suggesting that MFAP5 induced exogenous Ca²⁺ entry. Also, treatment with the IP₃ receptor inhibitor xestospongin C but not the ryanodine receptor blocker attenuated MFAP5-induced calcium mobilization, suggesting that IP₃-sensitive Ca²⁺ stores are mobilized by MFAP5 (Supplementary Figs. 5C and 5D). Because IP₃-sensitive Ca²⁺ store depletion can trigger store-operated Ca²⁺ entry and is a common entry pathway in nonexcitable cells, we confirmed that an exogenous Ca²⁺ entry mechanism resembling store-operated calcium entry was present in ALST cells (Supplementary Fig. 5E).

Since calcium mobilization is linked with increased cell traction force and with cell motility ^{9, 10}, we used traction force microscopy to evaluate MFAP5’s effects on changes in cancer cells’ traction force. RecMFAP5 exposure induced an amplitude increase and cell traction force redistribution in A224 and ALST cells (Figs. 5D and 5E), suggesting that that directional traction force induction is the physical mechanism of MFAP5-enhanced cell motility and invasion potential. Furthermore, TNNC1 silencing abrogated the amplitude increase of traction force induced by MFAP5 (Supplementary Fig. 3D), demonstrating that TNNC1 is the key molecule involved in traction force induction by MFAP5.

Western blot analysis of key signaling molecules implicated in calcium-dependent pathways and TNNC1 transcriptional regulation demonstrated increases in p-FAK (Y861), p-PLC-γ1 (Y783), p-PKCθ (T538), p-ERK1/2 (T202/Y204), p-CREB (S133), total Jun, and p-Jun (S63, S73 and S243) expression in MFAP5-treated cells (Figs. 6A and 6B). Based on our results, we hypothesized that binding of MFAP5 to αVβ3 integrin activates FAK. FAK activation activates PLC-γ1 and PKCθ, which regulate Ca²⁺ influx. Ca²⁺ mobilization leads to ERK1/2 phosphorylation and subsequent CREB activation. Translocation of activated CREB to the nucleus leads to c-Jun transcription via binding of activated CREB to cAMP-response element (CRE) in the c-Jun promoter, which then upregulates TNNC1 expression. Increased TNNC1 protein expression thereby enhances stress fiber formation and cell motility (Fig. 6C).

To test the hypothesis, Western blot analyses were performed. Pretreatment with an anti-αVβ3 integrin antibody abrogated MFAP5-induced FAK (Supplementary Figs. 6A and 7A) and PLC-γ1 phosphorylation (Supplementary Figs. 6B and 7B), suggesting that αVβ3 integrin is responsible for MFAP5-induced FAK and PLC-γ1 phosphorylation. Previous studies showed that PLC-γ1 phosphorylation is stimulated by both αVβ3 engagement and the formation of FAK/αVβ3 complex ^{11, 12}. We therefore treated cancer cells with an FAK inhibitor to determine whether MFAP5-induced PLC-γ1 phosphorylation is FAK-dependent. The results demonstrated that phosphorylation of p-PLC-γ1 was attenuated in FAK inhibitor pretreated cells (Supplementary Figs. 6C and 7C), suggesting that MFAP5-induced pPLC-γ1 phosphorylation is FAK-dependent.

Next, we sought to determine whether Ca²⁺ mobilization in MFAP5-treated ovarian cancer cells is dependent on p-FAK–mediated PLC-γ1 and PKCθ phosphorylation. Previous studies demonstrated that p-FAK induces PKCθ phosphorylation by activating PLC-γ1 ¹³. Activated PLC-γ1 increases hydrolysis of intracellular phosphatidylinositol-4, 5-bisphosphate to form two products: IP₃ and DAG. IP₃ induces endogenous Ca²⁺ release from the endoplasmic reticulum via IP₃ receptors, whereas DAG induces PKCθ phosphorylation and activation. Furthermore, p-PKCθ directly mobilizes Ca^{2+ 14, 15}. To determine whether PKCθ activation is mediated by p-FAK in MFAP5-treated ovarian cancer cells, Western blot analysis of p-PKCθ was performed. The level of p-PKCθ expression increased in MFAP5-treated cells only in the absence of the FAK inhibitor described above (Supplementary Figs. 6D and 7D). To determine whether the Ca²⁺ mobilization observed in MFAP5-treated cells was mediated by PKCθ activation, cells were treated with MFAP5 with or without a PKCθ inhibitor. The PKCθ inhibitor decreased but did not eliminate Ca²⁺ mobilization (Supplementary Fig. 5F), suggesting that Ca²⁺ mobilization is mediated only in part by PKCθ activation. To determine whether PLC activation is involved in Ca²⁺ mobilization, we studied the effect of PLC inhibitor U73122 on Ca²⁺ mobilization in MFAP5-treated ovarian cancer cells. PLC inhibition effectively suppressed Ca²⁺ mobilization (Supplementary Fig. 5G), suggesting that PLC activation is the major signaling event involved in MFAP5-mediated Ca²⁺ mobilization. However, Western blot analysis revealed that PLC inhibitor abolished phosphorylation of PKCθ and PKCθ inhibitor abolished phosphorylation of PLC-γ1 (Supplementary Figs. 6E, 6F, 7E and 7F), indicating that PLC and PKC are activated interdependently and affect Ca²⁺ mobilization in MFAP5-treated cancer cells.

To determine whether MFAP5-induced Ca²⁺ mobilization activates FAK via a feedback loop, we evaluated the effect of the Ca²⁺ chelator BAPTA/AM on MFAP5-induced FAK phosphorylation. We found that MFAP5-induced FAK phosphorylation was abrogated in BAPTA/AM-loaded cells, suggesting that this phosphorylation is Ca²⁺-dependent (Supplementary Figs. 6G and 7G). Further experiments showed that phosphorylation of ERK1/2 and CREB induced by recMFAP5 was attenuated in cells preloaded with Ca²⁺ chelator BAPTA/AM (Supplementary Figs. 6H, 6I, 7H and 7I), and in cells pretreated with PKCθ inhibitor, and PLC inhibitor U73122 (Supplementary Figs. 6J, 6K, 7J and 7K). This suggested that MFAP5-induced ERK and CREB activation is calcium-dependent and mediated by αVβ3 integrin/FAK/PKCθ- and αVβ3 integrin/FAK/PLCδ activation.

CREB is known to be activated by ERK ¹⁶. To determine whether MFAP5 activates CREB via ERK, the effect of an ERK1/2 inhibitor on MFAP5-induced CREB activation was evaluated. Pretreatment of ovarian cancer cells with the ERK1/2 inhibitor abolished upregulation of p-CREB expression, suggesting that MFAP5 activates CREB via ERK (Supplementary Figs. 6L and 7L). CREB, together with co-activators, contributes to target gene transcription in response to Ca²⁺ and via binding to the CRE. Our data demonstrated upregulation of expression of c-Jun, which contains a CRE in its promoter, in MFAP5-treated cells. Also, promoter analysis revealed that the TNNC1 promoter consisted of potential AP-1 binding sites (Supplementary Fig. 8), suggesting that transcriptional upregulation of TNNC1 expression is mediated by CREB-mediated c-Jun expression. To verify this, we evaluated the effects of CBP/CREB interaction inhibitor and c-Jun inhibitor SP600125 on MFAP5-treated cells. Results demonstrated that CBP/CREB interaction inhibitor attenuated the upregulation of p-c-Jun (Supplementary Figs. 6M and 7M), while SP600125 abrogated the upregulation of TNNC1 (Supplementary Figs. 6N and 7N). MFAP5-induced TNNC1 mRNA expression was abrogated in BAPTA/AM, CBP/CREB interaction inhibitor, and c-Jun inhibitor pretreated cells (Fig. 7A). These data confirmed that upregulation of TNNC1 expression induced by MFAP5 is calcium-dependent and mediated via upregulation of c-Jun expression by CREB activation.

To validate the activation of ERK signaling and upregulation of TNNC1 by MFAP5 in vivo, immunostaining of p-ERK1/2 and TNNC1 was performed on paraffin-embedded sections of Matrigel plugs from the previously described animal study. We observed higher p-ERK1/2 and TNNC1 expression in cancer cells in the Matrigel plugs embedded with MFAP5-overexpressing fibroblasts than in cells in the control Matrigel plugs (Figs. 2H and 2I).

Following delineation of the underlying signaling mechanism of MFAP5 mediated cell motility, to further demonstrate that the tumor-supportive effects of MFAP5 are dependent on an intact FAK/ERK/CREB/TNNC1 signaling pathway in the cancer cells, we performed cell motility assays on MFAP5 treated ovarian cancer cells in the presence or absences of pathway inhibitors. While both MFAP5 treatment alone and MFAP5 treatment in the presence of DMSO solvent control enhanced motility of A224 and ALST ovarian cancer cells, pretreatment of cancer cells with a FAK inhibitor, ERK inhibitor, or CBP/CREB interaction inhibitor abrogated the motility promoting effects of MFAP5 (Fig. 6D). Taken together with the findings that TNNC1 silencing (Fig. 4D) and calcium depletion (Fig. 5A) markedly suppressed MFAP5 induced cell motility, these data suggested that MFAP5 induced cancer cell motility is mediated via an intact calcium-dependent FAK/ERK/CREB/TNNC1 signaling pathway in ovarian cancer cells.

The relationship between tumor TNNC1 and stromal MFAP5 expression in 107 HGSC tumor samples was evaluated by immunohistochemistry (Fig. 7B) and a significant positive correlation between tumor TNNC1 and stromal MFAP5 expression (r=0.515, p < 0.001) was observed (Fig. 7C). Furthermore, Kaplan-Meier analysis revealed that low tumor TNNC1 expression was significantly associated with an improved survival (p=0.018; log-rank test) (Fig. 7D). Cox regression analysis adjusted for age and debulking status (hazard ratio: 1.705; p=0.029) confirmed the prognostic significance of tumor TNNC1 expression.

Ovarian tissue samples were obtained from the ovarian cancer repository of the Department of Gynecologic Oncology and Reproductive Medicine at The University of Texas MD Anderson Cancer Center under protocols approved by the MD Anderson Institutional Review Board. All tumor samples were obtained from the primary ovarian sites of pre-treatment cases. Microdissection was performed as described by Yeung et. al. to isolate the stromal and epithelial components of normal and malignant ovarian tissue for RNA extraction ³². In brief, tissue sections were fixed in 70% ethanol and stained them with 1% methyl green to visualize the histologic features. During dissection, areas with immune cell and blood vessel infiltration were excluded to minimize contamination. Purified RNA samples were amplified, labeled, and hybridized onto GeneChip Human Genome U133 Plus 2.0 microarrays (Affymetrix Inc., Santa Clara, CA) according to the manufacturer’s protocol. After hybridization, arrays were washed and stained using a Fluidics Station 450 and then scanned using a GeneChip Scanner 3000 7G (Affymetrix Inc., Santa Clara, CA).

Human ovarian adenocarcinoma cell lines A224, ALST (gifts from Dr. Michael Birrer’s laboratory at the Massachusetts General Hospital), OVCA432 (gift from Dr. Robert Bast’s laboratory at the MD Anderson Cancer Center), and SKOV3ipluc (obtained from American Type Culture Collection; ATCC) were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum, 2 mM glutamine, and penicillin/streptomycin. The immortalized human NOF line NOF151-hTERT and primary CAF lines were cultured in 1:1 MCDB105/199 medium supplemented with 10% fetal bovine serum, 1 ng ml⁻¹ epidermal growth factor, and penicillin/streptomycin.

The relative expression of each target gene was calculated using the 2^−ΔΔC_T method to average the C_T value of the housekeeping gene for a single reference gene value. Pre-designed human MFAP5 (Hs00185803_m1), TNNC1 (Hs00896999_g1), cyclophilin A (Hs99999904_m1), and murine Mfap5 (Mm00489404_m1) TaqMan gene expression assays (Life Technologies Corp., Grand Island, NY) were used in the PCR.

Protein lysates from CAFs, normal fibroblasts, and ovarian cancer cells were separated on 10% sodium dodecyl sulfate NuPAGE gels under denaturing conditions and transferred onto nitrocellulose membranes using an iBlot Western blotting system (Life Technologies Corp., Grand Island, NY) before being incubated with primary antibodies. Anti-human MFAP5 (1:1000; #HPA010553) was purchased from Sigma-Aldrich (St. Louis, MO). An anti-p-FAK antibody (Y861; 1:1000 ; #44-626G) was purchased from Life Technologies Corp. (Grand Island, NY). All other antibodies, including anti-FAK (1:1000; #3285), anti-PLC-γ1 (1:1000; #2822), anti-p-PLC-γ1 (T783; 1:1000; #2821), anti-PKCθ (1:1000; #2059), anti-p-PKCθ (T538; 1:1000; #9377), anti-ERK1/2 (1:1000; #9102), anti-p-ERK1/2 (T202/204; 1:1000; #9101), anti-CREB (1:1000; #9197), anti-p-CREB (S133; 1:1000; #9198), anti-c-Jun (1:1000; #9165), and anti-p-cJun (S63, S73, and S243; 1:1000; #2361, #9164, and #2994), were purchased from Cell Signaling Technology (Beverly, MA). After being washed with Tris-buffered saline with Tween, the membranes were incubated with a goat anti-rabbit IR dye-conjugated secondary antibody (1:5000; LI-COR Biosciences, Lincoln, NE). Protein bands were detected using an Odyssey infrared imaging system (LI-COR Biosciences, Lincoln, NE). The ImageJ software program (National Institutes of Health, Bethesda, MD) was used to quantify band intensities of Western images. Protein expression levels in terms of band intensities were normalized against protein loading using the housekeeping gene as reference. Relative protein expression levels were presented with respect to their corresponding controls, which had the baseline of 1.

The inhibitors used in pathway analyses included a PLC inhibitor (U73122; #sc-3574), PKCθ pseudosubstrate inhibitor (#sc-3097), and ERK inhibitor II (FR180204; #sc-203945), which were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). An FAK inhibitor (SU6656; #S9692) and a c-Jun inhibitor (#SP600125) were purchased from Sigma-Aldrich Co. (St.. Louis, MO), and a CBP/CREB interaction inhibitor (#217505) was purchased from EMD Millipore Corp. (Billerica, MA). Original images for essential blots are supplied as supplementary information.

Immunolocalization of MFAP5 and TNNC1 was performed using 130 and 107 FFPE ovarian tumor sections, respectively, obtained from HGSC patients. Slides containing the sections were stained with commercially available anti-MFAP5 (1:500; #HPA010553) and anti-TNNC1 (1:100; #WH0007134M1) antibodies (Sigma-Aldrich Co., St. Louis, MO). Target protein expression was visualized using a Betazoid 3,3'-diaminobenzidine or Warp Red chromogen kit (Biocare Medical, Concord, CA).

To quantify target protein expression in the FFPE sections, slides were scored according to the staining intensity and percentage of MFAP5- and TNNC1- positive cells. To quantify stromal MFAP5 expression, the expression scores were calculated by multiplying the staining intensity (range, 0–3) by the percentage of stroma with MFAP5-positive staining. TNNC1 tumor expression scores were calculated as the products of staining intensity and percentage of TNNC1-positive staining.

CAF793092 cells were seeded onto a 24-well plate at a density of 5×10⁴ cells/ well. The culture medium was then removed from the plate, and cells were washed with phosphate-buffered saline (PBS) the next day. Serum-free M105/199 media supplemented with 1 ng mL⁻¹ epidermal growth factor and 10 µg mL⁻¹ control IgG or anti-MFAP5, anti-αVβ3, or anti-α5 antibodies were added to the wells. Seventy-two hours later, 5 × 10⁴ A224 or ALST cells were seeded onto each 8-µm insert (BD Biosciences, San Jose, MA) in the serum-free medium and co-cultured with CAFs. After 15 hours of incubation, the cells were stained with calcein AM (Life Technologies Corp., Grand Island, NY), and nonmigrating cells in the upper surface of the insert were removed. The migrated cells in each chamber were counted by obtaining images of the stained cells in nine random fields of view per membrane using a fluorescent microscope and quantified using the Image-Pro Plus 7.0 software program (Media Cybernetics Inc., Rockville, MD).

OVCA432 cells were seeded onto chamber slides in 10% serum RPMI medium. After they attached to the slides, they were washed with PBS, and serum-free RPMI medium with or without 500 ng mL⁻¹ recMFAP5 was added to the wells. The cells were then placed in an imaging chamber maintained at 37°C and supplemented with 5% CO₂. Cell images were taken every 15 minutes for 48 hours using confocal microscopy and the resulting series of images was imported into the MetaMorph software program to track the movement of individual cells.

To confirm exogenous MFAP5’s effects on ovarian cancer cell motility, a motility assay was performed using a Boyden chamber with two additional ovarian cancer cell lines: A224 and ALST. Serum-free RPMI medium alone or with 500 ng mL⁻¹ recMFAP5 was incubated with 10 µg mL⁻¹ control IgG or anti-MFAP5, anti-αVβ3, or anti-α5 antibodies for 1 hour in the wells of 24-well plates. Monoclonal anti-αVβ3 antibody (#MAB1976Z) was purchased from EMD Millipore Corp. (Billerica, MA), and an anti-α5 antibody (#555615) was purchased from BD Bioscience (San Jose, CA). A224 or ALST cells (5×10⁴) were then seeded onto each 8 µm insert (BD Biosciences) in serum-free medium and incubated for 15 hours. Cells were stained with calcein AM (Life Technologies Corp., Grand Island, NY), and nonmigrating cells were removed from the upper surface of the inserts. The migrated cells in each chamber were counted by obtaining images of stained cells in nine random fields of view per membrane using a fluorescent microscope and quantified using the Image-Pro Plus 7.0 software program (Media Cybernetics Inc., Rockville, MD).

To evaluate the role of calcium and TNNC1 in cell motility induced by MFAP5, ovarian cancer cells were pretreated with the cell-permeant calcium chelator BAPTA/AM (B-6769; Life Technologies) at a concentration of 10 µM for 1 hour or transfected with efficiency-evaluated TNNC1-specific siRNA (Hs00896999_g1; Life Technologies Corp., Grand Island NY) (Supplementary Figs. 3B and 3C) before performing the motility assay with recMFAP5.

The invasion potential of ovarian cancer cells treated with recMFAP5 was evaluated using a Matrigel invasion assay with a BD BioCoat Matrigel invasion chamber (BD Biosciences, San Jose, CA) according to the manufacturer’s protocol. In brief, 3×10⁴ A224 or ALST cells were seeded onto an 8 µm BD BioCoat Matrigel invasion chamber in serum-free medium and inserted into a BD Falcon TC companion plate with or without 200 ng mL⁻¹ recMFAP5. The cells were incubated in a humidified tissue culture incubator at 37°C in a 5% CO₂ atmosphere for 15 hours. Cells were then stained with calcein AM, and non-invading cells were removed from the upper surface of the invasion chambers. Counting of the invaded cells was facilitated by photographing the membrane under a fluorescent microscope, and the numbers of invaded cells were analyzed using the Image-Pro Plus 7.0 software program (Media Cybernetics Inc., Rockville, MD).

The telomerase-immortalized human NOF line NOF151-hTERT was transduced with lentiviruses packaged with an open reading frame lentiviral expression clone (#EX-D0360-Lv105; GeneCopoeia, Rockville, MD) for 48 hours. Transduced cells were selected using complete medium supplemented with puromycin at a working concentration of 1 µg mL⁻¹ for 5 days. MFAP5 expression in the NOFs was measured using Western blot analysis, in which the lysate and conditioned medium from the resultant NOF line NOF151-LvMFAP5 were compared with those from the mock-transduced control line. A proliferation assay was performed using an xCELLigence system (Roche Applied Bioscience, Indianapolis, IN) to ensure that the NOF151-LvMFAP5 and NOF151-hTERT cells had similar growth rates.

5×10⁴ normal parental NOF151 cells or MFAP5-overexpressing NOF151 cells (NOF151-LvMFAP5) resuspended in one part of PBS was mixed with two parts of Matrigel, and the mixture was injected into the intraperitoneal cavity of anesthetized female nude mice at the age of 6 weeks. 500µl mixture was injected into each animal and allowed to warm to body temperature to form a Matrigel plug attached to the peritoneal wall. The next day, 1 million A224 or OVCA432 cells were injected intraperitoneally into the mice. Matrigel plugs were harvested from A224- and OVCA432-injected mice at 48 and 72 hours post cancer cell injection respectively. The resected plugs were fixed in formalin and processed for histological evaluation. The described animal procedures have been reviewed and approved by the institutional animal care and use committee of the MD Anderson Cancer Center.

Luciferase labeled OVCA432ip ovarian cancer cells were suspended in 50µl of HBSS and injected directly into the left ovary of anesthetized female nude mice at the age of 6 weeks through a 1.5cm intraperitoneal incision. One week after cell injection, chitosan nanoparticles packaged with Mfap5-targeting siRNA or nontargeting siRNA were injected intravenously into nude mice twice a week for a total of seven weeks. Six weeks after initial injection, tumor volume of each treatment group was quantified by the IVIS® −200 bioluminescence and fluorescence imaging system (Caliper Life Sciences, Inc., Hopkinton, MA). At week 8, the animals were sacrificed and necropsy was performed. The number, weight, and location of individual tumor nodules were identified and recorded. Tumor tissues were fixed in formalin and processed for histological evaluation. The described animal procedures have been reviewed and approved by the institutional animal care and use committee of the MD Anderson Cancer Center.

RNAs were isolated from OVCA432 cells and treated with PBS or 200 ng mL⁻¹ recMFAP5. 100 ng of total RNA from each group were used to generate biotin-labeled aRNA and then subjected to whole-genome transcriptome profiling using GeneChip Human Genome U133 Plus 2.0 microarrays (Affymetrix Inc., Santa Clara, CA). Differentially expressed genes, which were expressed at levels more than 1.5-fold higher than those of the corresponding controls p < 0.05), were selected for further analyses using the Ingenuity Pathway Analysis software program.

A224 and ALST cells were seeded onto Lab-Tek chamber slides (Thermo Fisher Scientific, Pittsburgh, PA), and serum-free medium was added to the wells the next day to fast the cells overnight. Cells were then treated with fresh serum-free medium supplemented with PBS or 200 ng mL⁻¹ recMFAP5 for 24 hours. After incubation, cells were fixed with 3.7% formaldehyde and stained with Alexa Fluor 594 phalloidin (Life Technologies Corp., Grand Island, NY) according to the manufacturer’s instructions. Fluorescent microscopy was used to evaluate the F-actin rearrangement in cells stimulated by MFAP5.

To evaluate the role of calcium and TNNC1 in F-actin reorganization induced by MFAP5, ovarian cancer cells were pretreated with the cell-permeant calcium chelator BAPTA/AM (B-6769; Life Technologies Corp., Grand Island, NY) for 1 hour and transfected with TNNC1-specific siRNA, respectively, before MFAP5 treatment and staining with Alexa Fluor 594 phalloidin.

Knockdown of TNNC1 protein expression in A224 and ALST cells by siRNA transfection was confirmed using Western blot analysis. Prior to this analysis, cell pellets were collected from non-target scrambled and TNNC1-targeting siRNA-transfected A224 and ALST cells, and cytoplasmic protein fractions of the lysates were isolated using the NE-PER nuclear and cytoplasmic extraction reagents (Thermo Fisher Scientific Inc., Waltham, MA) according to the manufacturer’s protocol. A polyclonal rabbit anti-human TNNC1 antibody (1:500; 13504-1-AP; Proteintech, Chicago, IL) was used to determine TNNC1 expression levels in the cytoplasmic protein factions via Western blot analysis.

The cell traction force exerted on a flexible substrate impregnated with fluorescent microspheres by A224 or ALST cells growing on the substrate was determined according to measured substrate displacements using the LIBTRC analysis library developed by Micah Dembo of Boston University ^33–35. In brief, polyacrylamide gel (5% acrylamide and 0.1% Bis-acrylamide, 75 µm thick; Young’s modulus = 5.8 kN m⁻²) was prepared on glass coverslips with fluorescent microspheres (0.2 µm diameter, yellow-green FluoSpheres; Life Technologies) impregnated under the top surface. The surface was then coated with 0.2 mg mL⁻¹ collagen type I (BD Bioscience, San Jose, CA) after photoactivation of the surface molecular linker sulfosuccinimidyl-6-(4'-azido-2'-nitrophenylamino) hexanoate (sulfo-SANPAH; Pierce Chemical, Dallas, TX). Cells were seeded and cultured overnight. To evaluate the effect of exogenous MFAP5 on the traction force exerted by cancer cells on the substratum, cancer cells were treated with exogenous MFAP5 in Hanks balanced salt solution (HBSS) and a series of images of a single individual cancer cell was captured during the course of each experiment. A background image of the fluorescent beads was also taken after the cell was removed from the substrate. The LIBTRC library was then used to track the displacement of microspheres and compute the cell traction force. Total traction force (cell surface area×traction force per unit area) was normalized by the total traction force of the same cell before exposure to MFAP5 or HBSS. One cell was studied for each coverslip. All images were collected using a TCS SP5 confocal microscope (Leica Microsystems, Buffalo Grove, IL).

A224 or ALST cells that grew on collagen-coated Petri dishes with glass bottoms were loaded with Fluo-4 AM in HBSS for 30 minutes followed by 20 minutes to de-esterify the dye. Fluo-4 AM was excited at 488 nm, and its emission was collected using a bandpass filter at 522/35 nm. Fluorescent images of the cells were collected using a Leica TCS SP5 confocal microscope at 0.25 Hz.

The SPSS 19 (IBM Corp., Armonk, NY) and Prism 5.0 (GraphPad Software, La Jolla, CA) software programs were used to perform statistical tests. All in vitro experiments were repeated independently in triplicate. A two-tailed Student t-test was used to determine differences in sample means for data with normally distributed means. The Mann-Whitney U test was used for analysis of nonparametric data. A p-value less than 0.05 was considered statistically significant.