The OPCML Tumor Suppressor Functions as a Cell Surface Repressor–Adaptor, Negatively Regulating Receptor Tyrosine Kinases in Epithelial Ovarian Cancer

Ovarian cancer is the leading cause of death from gynecologic malignancy (1). The molecular basis of ovarian carcinogenesis is poorly understood but frequently involves lesions affecting p53 (2); BRCA1 and 2 (3); the phosphoinositide 3-kinase pathway, including dysregulation of AKT (4); growth factor signaling pathways, including the epidermal growth factor (EGF) and fibroblast growth factor (FGF) pathways (5–9); and neoangiogenesis (10).

We previously identified that opioid binding protein cell adhesion molecule (OPCML) was inactivated by LOH and epigenetic silencing in more than 80% of human epithelial ovarian cancers (11). We demonstrated that OPCML expression inhibited ovarian cancer cell growth, enhanced intercellular attachment, and abrogated both subcutaneous and intraperitoneal tumorigenicity in vivo (11). In recent publications, others have also confirmed OPCML to be frequently epigenetically inactivated in epithelial ovarian cancers (12–14), brain tumors (15), non–small cell lung carcinoma (16), bladder cancer (17), cholangiocarcinoma (18), primary nasopharyngeal, esophageal, gastric, hepatocellular, colorectal, breast, and cervical cancers, as well as lymphomas (19), indicating that OPCML has broad tumor suppressor activity in common cancers. In many tumor types, OPCML was ubiquitously nonexpressed. In several of these studies, investigators demonstrated a significant correlation between OPCML hypermethylation and loss of expression in cancer cell lines (11, 17, 19) and primary tumors (12, 14, 18), and OPCML methylation and loss of expression were associated with poor survival for the patient (17).

OPCML is a glycosyl phosphatidylinositol (GPI)-anchored cell adhesion-like molecule and a member of the IgLON family, denoting the immunoglobulin domain protein family that includes limbic system-associated membrane protein (20, 21), OPCML, neurotrimin (22), and more recently neuronal growth regulator 1 (23). The IgLONs are medium-sized proteins (∼55 kDa) comprising 3 conserved extracellular I-type immunoglobulin domains that share common molecular recognition properties enabling homo- and heterodimerization between family members (24). GPI-anchored proteins are trafficked to the plasma membrane and often are associated with detergent-insoluble fractions termed “lipid rafts” that mainly consist of sphingolipids and cholesterol (25). Lipid raft domains have also been shown to influence the distribution and signaling of many receptors from tyrosine kinases to integrins (26–28), although there is still some debate about the definition and existence of physiologically relevant lipid rafts (29).

Here, we describe the mechanism underlying the in vitro and in vivo tumor-suppression phenotype previously described for OPCML (11). Our results reveal that OPCML negatively regulates a specific spectrum of receptor tyrosine kinases (RTK) through physical interactions with their extracellular domains and promotion of their proteasomal degradation via trafficking redistribution, which in turn leads to altered RTK signaling. We also demonstrate that exogenous recombinant OPCML is a potent RTK suppressor in most ovarian cancer cell lines tested and provide a proof-of-concept of its therapeutic potential in vivo.

With several investigators demonstrating a correlation between OPCML methylation and loss of OPCML expression in cancer cell lines and primary tumors, we used the online database of the Cancer Genome Atlas (TCGA) HumanMethylation27 (Illumina) assay to confirm the frequency of OPCML methylation in human cancer. We found that 678 of 1537 (44%) of cancer patients with available methylation data (representing breast, ovarian, brain, leukemia, colon, renal, lung, and endometrial cancers) had OPCML-methylated tumors (with site-specific methylation rates ranging from 31% for breast cancer to 73% for colonic adenocarcinoma; Supplementary Fig. S1). The ovarian cancer TCGA dataset (30) also indicated loss of OPCML expression in 92% of serous high-grade ovarian cancers, suggesting that mechanisms other than somatic methylation (such as LOH) may also result in loss of OPCML expression. We then used the KMPlotter online ovarian and breast cancer meta-analysis and demonstrated that high OPCML expression (defined as above median expression for breast cancer and above lowest quartile expression for ovarian cancer) is a highly significant favorable prognostic factor for progression in 1,090 ovarian (HR 0.71, P = 4.3e⁻⁰⁵) and for relapse in 2,324 breast cancer (HR 0.57, P = 2.2e⁻¹⁶) patients (Supplementary Fig. S2). These findings underscore the clinical relevance of OPCML inactivation.

OPCML is a nontransmembrane, external lipid leaflet-anchored protein with no direct route to affect proliferative intracellular signaling. We therefore hypothesized that it may mediate its tumor suppressor properties via transmembrane signaling proteins and analyzed the effect of RTK growth factor stimulation on OPCML gene expression. SKOV-3 cells treated with EGF or FGF 1/2 rapidly induced OPCML RNA and protein expression (Supplementary Fig. S3A and S3B). This observation was expanded to a range of cell lines, including the transformed normal ovarian surface epithelial cell line OSE-C2. Although maximal induction of OPCML in SKOV-3 and PEO-1 occurs at 30 minutes, the OSE-C2 and OVCAR5 lines achieve this at 90 minutes (Supplementary Fig. S3C). The previously described stable OPCML-expressing SKOV-3 cell lines (BKS-2.1 and SKOBS-3.5) and nonexpressing vector-only transfection control cells (SKOBS-V1.2) (11) were used to explore any interaction between OPCML and RTKs that were selected on the basis of being likely responders to EGF, FGF1, and FGF2. We also selected RTKs with potential importance in ovarian cancer [vascular endothelial growth factor receptors (VEGFR) and ephrin type A receptors (EPHA)]. BKS-2.1 and SKOBS-3.5 demonstrated a profound downregulation of a specific repertoire of RTKs in both the basal unstimulated or ligand-stimulated state. EPHA2, FGFR1, FGFR3, HER2, and HER4 were reproducibly downregulated by either stable OPCML transfection in SKOV-3 (Fig. 1A), or transient, polyclonal OPCML transfection in PEO1 ovarian cancer cells (Fig. 1B; ref. 31).

These same RTKs were found to be reciprocally upregulated when physiologic OPCML was knocked down by transient transfection with an siRNA pool in OSE-C2 (Fig. 1C; ref. 32). Endogenous protein expression of OPCML and a variety of RTKs is shown in Figure 1. The first (left) lanes for SKOV3, PEO1, and OSE-C2 (Fig. 1A–C) demonstrate that OPCML does not readily coexpress with the specific spectrum of RTKs identified previously while not affecting unassociated RTKs. Deconvolution of the siRNA duplexes in this pool demonstrated their specificity (Supplementary Fig. S4). Conversely, a group of other RTKs were unaltered by either OPCML overexpression or siRNA knockdown, including EPHA10, FGFR2, FGFR4, EGFR, HER3, VEGFR1, and VEGFR3 (Fig. 1 and summarized in Supplementary Table S1). We confirmed by quantification of immunofluorescence microscopy (IFM) that the expression of OPCML dramatically reduced the levels of EPHA2, FGFR1, and HER2 but not the levels of FGFR2 or EGFR in BKS-2.1 cells compared with SKOBS-V1.2 (Fig. 1D and E; Supplementary Fig. S5). Expression of the OPCML-regulated RTKs EphA2 and FGFR1 was analyzed in SKOV-3 (OPCML⁻) and BKS-2.1 (OPCML⁺) cells stimulated with EGF and FGF1. No significant change in overall RTK mRNA levels was observed between the OPCML^−/+ lines for these RTKs that show dramatic, OPCML-dependent changes in protein level, suggesting that OPCML regulates these RTKs posttranscriptionally (Supplementary Fig. S3D).

Cell proliferation assays in FBS-supplemented media that used stably transfected lower (SKOBS-3.5) and higher (BKS2.1) OPCML-expressing SKOV-3 clones showed significant growth inhibition compared with SKOBS-V1.2 (vector control) and displayed similar growth kinetics as the normal ovarian surface epithelial cell line OSE-C2 (Fig. 1F). To evaluate whether this growth inhibition was attributable to OPCML-mediated negative regulation of RTK phosphoactivation and relevant downstream signaling pathways, we further investigated the effects of acute ligand stimulation with EGF and FGF1. OPCML expression led to profound abrogation of phospho-HER2-Y1248, phospho-EGFR-Y1173, and phospho-FGFR1-Y766 (Fig. 2A and B). Although no EGFR loss was observed, loss of the activating dimerization partners of EGFR (i.e., HER2 and HER4), coupled with the continuing availability of the inactivating HER3 family member, manifests as downregulation of EGFR signaling. Analysis of downstream signaling revealed abrogation of phospho-ERK 1 and 2 (T202 and T204) and phospho-AKT-S473 (Fig. 2A), suggesting that both progrowth and prosurvival pathways are inhibited by OPCML re-expression.

To expand upon our siRNA observations (Fig. 1C), we developed stable short hairpin RNA (shRNA) OPCML knockdown lines in OSE-C2 (see Methods). OSE-C2 lines with 60% and 95% OPCML knockdown were generated (Fig. 2D). The 95% OPCML knockdown line demonstrated HER2 and EPHA2 upregulation (Fig. 2E). We analyzed the level of HER2 and EPHA2 proteins in the empty-vector control line PLKO-2 and 95% knockdown shRNA339–24 line after serum starvation followed by 1 hour of exposure to serum-free medium, 10% FBS-supplemented medium, or 50 ng/mL EGF stimulation on Western blots (Fig. 2E). The levels of HER2 and EPHA2 increased upon OPCML knockdown in a serum- and ligand-dependent fashion. Reciprocally to the overexpression of OPCML in cancer lines, the stable knockdown of OPCML in OSE-C2 resulted in increased proliferation. The empty-vector control line PLKO-2, the stably expressing sh339–24 and sh464–23 knockdown cell lines (exhibiting 95% and 60% knockdown, respectively), and an shRNA scrambled control line were directly compared in 10% FBS-supplemented medium, and a significant increase in proliferative rate was associated with the 95% knockdown of OPCML (P = 0.003, Fig. 2F). Thus, stably knocking down OPCML results in upregulation of RTKs from physiologically normal levels in ovarian surface epithelial cells to levels equivalent to those observed in SKOV-3 cancer cells (∼15-fold in the case of EPHA2) and is associated with accelerated growth in 10% serum.

We selected EPHA2, FGFR1, and HER2 as examples of OPCML-regulated RTKs and EGFR as an example of an RTK not negatively regulated by OPCML. Immunoprecipitation of BKS-2.1 cell lysates with antibodies to EPHA2, FGFR1, and HER2 coprecipitated OPCML, suggesting an interaction with these 3 RTKs. A reciprocal immunoprecipitation, in which α-OPCML was used as the first antibody, confirmed that these RTKs all bound to OPCML. However, no interaction was seen between OPCML and EGFR, whose expression levels are not negatively regulated by OPCML (Fig. 3A). To confirm our initial immunoprecipitation studies, we constructed a recombinant GST-OPCML fusion protein consisting of Ig domains 1–3. Pull-down experiments with the use of SKOV-3 cell lysates revealed that FGFR1 and HER2 interacted with GST-OPCML D1–3 (Fig. 3B). We were able to confirm the specificity of these interactions by demonstrating again that EGFR was not pulled down in this assay, in agreement with our immunoprecipitation data in Figure 3A. We further determined that the interaction with GST-OPCML D1–3 was mediated by the extracellular domain of the RTKs for FGFR1 and HER2 (Fig. 3C).

Because this finding suggested that OPCML-mediated negative regulation of a specific repertoire of RTKs is defined by a physical binding event with the RTK extracellular domain, we sought to further demonstrate the functional consequences of OPCML binding to these RTKs by using HER2, an RTK highly expressed in SKOV-3 cells, as a paradigm. To understand the mechanism of OPCML-mediated downregulation of HER2, we transiently transfected BKS-2.1 or SKOBS-V1.2 cells with either a full-length, 185-kD rat HER2/Neu construct (Neu) or the previously reported 95-kD (Δ-5) Neu construct with a deleted extracellular domain sequence (33) as shown in Supplementary Figure S6 and Figure 3D to F. In SKOBS-V1.2 (OPCML⁻), transfection of full-length Neu and Δ-5 resulted in a clear increase in the amount of intact 185-kD protein and 95-kD truncated species compared with empty-vector control transfection.

In OPCML-expressing BKS-2.1 cells, however, both endogenous HER2 and the transfected full-length 185-kD Neu exhibited a depleted signal (visualized by immunoblotting for the His tag), whereas the 95-kD Δ-5 exhibited little change in band intensity between SKOBS-V1.2 and BKS-2.1 (Fig. 3D; left quadrants). Densitometric quantification revealed a 50% reduction in exogenous 185-kD Neu, with no reduction seen in the 95-kD Δ-5 species (Fig. 3E; upper chart).

To understand this interaction more clearly, we transiently cotransfected HER2/OPCML-null Cos cells with OPCML and the Neu constructs, which revealed a clear OPCML-induced downregulation of 185kD-Neu with unaltered expression of 95-kD Δ-5 (Fig. 3C; right quadrants). Quantitative densitometric analysis illustrated a significant (>75%) downregulation of 185-kD Neu and no alteration in 95-kD Δ-5 expression in OPCML cotransfected Cos cells (Fig. 3E; lower chart). Cell proliferation assays of the Cos cotransfections clearly demonstrated that 185-kD Neu transfection accelerated cell growth and that OPCML cotransfection significantly and completely abrogated this acceleration (Fig. 3F; upper graph). In the same assay, 95-kD Δ-5 transient transfection demonstrated similar growth acceleration that was unaffected by OPCML cotransfection (Fig. 3F; lower graph). These findings reveal that negative regulation of RTKs by OPCML is functionally mediated through extracellular domain protein binding.

To confirm the predicted “lipid-raft” localization of the GPI-anchored protein OPCML, we prepared detergent-soluble and -insoluble fractions from SKOBS-V1.2 and BKS-2.1 cells (see Methods), which revealed that OPCML was localized within the cholesterol-rich, detergent-insoluble fraction along with caveolin-1, a marker of a distinct form of lipid raft domain known as caveolae; Fig. 4A). We analyzed HER2 as an example of an OPCML-regulated RTK in this analysis. The OPCML-expressing cell line exhibited a reduced level of HER2 with sequestration of the remaining HER2 in the detergent-insoluble membrane fraction. However, in the OPCML nonexpressing line, HER2 is equally distributed between the 2 fractions. The total level of EGFR, a non–OPCML-regulated RTK, was not affected by OPCML expression, although there was a shift to the detergent-insoluble membrane fraction (Fig. 4A). These data indicate that OPCML expression leads not only to loss of HER2 expression but also HER2 redistribution on the plasma membrane.

IFM was used to examine the trafficking of OPCML in cells by the use of EEA-1 (a marker of the early endosome) and caveolin-1 (a marker of the raft-caveolar pathway) to investigate this apparent redistribution. OPCML was seen to colocalize with caveolin-1 rather than EEA1 (Fig. 4B). IFM colocalization of HER2 with EEA-1 or caveolin-1 (Fig. 4C; left) demonstrated a quantitative shift away from EEA-1/clathrin colocalization of HER2 to colocalization with the caveolin-1 detergent-insoluble membrane fraction in OPCML-expressing BKS2.1 cells. These observations suggest a shift from clathrin/EEA-1–associated HER2 trafficking in OPCML nonexpressing cells to a pathway mediating a different form of caveolin-1-associated trafficking (Fig. 4C; right).

The turnover of HER2 on the surface of cells was measured by pulse-chase biotinylation of reactive free amino or sulfhydryl groups on surface-accessible proteins of BKS-2.1 and SKOBS-V1.2 cell lines. Biotinylated fractions were then captured on a NeutrAvidin (Thermo Scientific) agarose resin, immunoblotted for HER2, and compared with the total input fraction. The biotinylated fractions demonstrated accelerated loss of surface HER2 protein, with an 85% decrease in the level of surface-labeled HER2 at 4 hours in OPCML⁻ expressing BKS-2.1 cells with only a 23% reduction in surface HER2 in OPCML⁻ negative SKOBS-V1.2 cells (Fig. 5A; right).

We transfected the OPCML-expressing and knockdown normal ovarian surface epithelial cells (OSE-C2) discussed previously with a hemagglutinin-tagged ubiquitin construct. The substantial HER2 ubiquitination observed in the control OSE-C2 PLKO cell line was profoundly suppressed in the OPCML stable knockdown line sh339–24 (Fig. 5B; top) and was associated with upregulation of total HER2 (Fig. 2E). We witnessed a reciprocal result upon transfection of the same hemagglutinin construct into SKOBS-V1.2 (OPCML⁻) and BKS-2.1 (OPCML⁺) cells with and without EGF stimulation. OPCML restoration in BKS-2.1 was associated with increased ubiquitination of HER2 (Fig. 5C; top).

We treated the normal OSE-C2 PLKO (tightly controlled/repressed HER2 expression) and OPCML nonexpressing OSE-C2 stable knockdown line sh339–24 with MG-132, a potent inhibitor of the proteasomal 26S proteinase. MG-132 treatment of PLKO resulted in strong upregulation of HER2 (Fig. 5B; bottom). In sh339–24, HER2 was strongly upregulated by OPCML knockdown, and treatment with MG-132 had no further impact on HER2 level (Fig. 5B; bottom).

We repeated this experiment in SKOV-3 cancer cells. OPCML-expressing BKS-2.1 cells were exposed to MG-132, and this prevented OPCML-mediated HER2 degradation with consequent HER2 upregulation. In contrast, chloroquine, a weak base that alkalinizes the lysosome, demonstrated no such inhibition of HER2 degradation (Fig. 5C; bottom). Furthermore, in SKOBS-V1.2 cells lacking OPCML expression (Fig. 5C; bottom), HER2 was strongly upregulated and MG132 had no additional impact on HER2 expression in these cells, in an analogous fashion to the sh339–24 cell line in Figure 5B. Taken together, these data suggest that RTK polyubiquitination followed by proteasomal degradation is preferentially used for OPCML-mediated negative regulation of HER2 both under physiologic conditions and in cancer.

OPCML therefore negatively regulates the activity of a specific repertoire of RTKs by binding the extracellular domains of those RTKs, inducing “lipid-raft” associated sequestration, mediating a switch away from clathrin internalization to caveolin-1–associated enhancement of RTK polyubquitination to ultimately result in proteasomal degradation of those RTKs. Consideration of our data in the context of recent publications (34) would suggest that clathrin-independent carriers/ GPI-enriched early endosomal compartment-mediated internalization is a possible pathway for the observed OPCML-mediated internalization and degradation of RTKs such as HER2 and this is linked to the observable strong tumor suppressor phenotype of OPCML.

The extracellular membrane location and mechanism of action of OPCML raised the possibility of direct extracellular tumor suppressor protein therapy, which would avoid the complexities of gene therapy for intracellular tumor suppressor replacement or intracellular delivery of protein therapies. We purified recombinant human OPCML domain 1–3 protein (rOPCML) by using a bacterial expression vector (pHis-Trx) subcloned with domains 1–3 of OPCML, excluding the signal peptide and GPI anchor sequences (Fig. 6A). The addition of rOPCML protein to cell culture supernatant demonstrated a specific, dose-dependent, highly significant inhibition of cell growth in OPCML nonexpressing SKOV-3 ovarian cancer cells without affecting normal OSE-C2 ovarian surface epithelial cells that express a physiologic level of OPCML (Fig. 6B).

We then confirmed that rOPCML profoundly and significantly inhibited cell growth in 6 of 7 non–OPCML-expressing epithelial ovarian cancer cell lines (Fig. 6C). IFM demonstrated that rOPCML treatment resulted in HER2 downregulation, mirroring the expression of OPCML in the same cell line (Fig. 6D). Immunoblotting demonstrated that the addition of rOPCML protein to culture media potently downregulated the same spectrum of RTKs as OPCML transfection and abrogated pERK and pAKT in both SKOV-3 and A2780 (Fig. 7A) cells. Exogenous application of rOPCML thus uses the same mechanism of action as transfection-induced re-expression of the normal GPI-anchored, glycosylated, OPCML protein. Noting that pAKT was abrogated and understanding that stable transfection of OPCML would by definition select cells that have survived and therefore be less apoptosis-prone (indeed, previous fluorescence-activated analysis demonstrated no evidence of apoptosis), we undertook annexin-V and caspase-glo assays and demonstrated modest but significant evidence of apoptosis in unselected SKOV-3 and A2780 (Supplementary Fig. S7A and S7B) cells. This finding suggests that both growth inhibition and apoptosis occur upon application of rOPCML.

In view of these in vitro findings and the detailed in vivo experiments we previously published (11) that clearly demonstrated the in vivo tumor suppressor phenotype of stably transfected OPCML, we tested whether the rOPCML protein was effective as an in vivo anticancer therapy. Mice were injected intraperitoneally with either SKOV-3 or A2780 cancer cells, and after 1 week they received twice-weekly intraperitoneal injections of either 1 mL (10 μM) bovine serum albumin (BSA) or 1 mL (10 μM) rOPCML. The experiment was terminated after 3 weeks because of obvious extensive intraperitoneal tumor growth and deteriorating condition of BSA-treated control animals, whereas rOPCML-treated mice remained healthy throughout the duration of the experiment (Supplementary Fig. S8A). rOPCML significantly and profoundly suppressed intraperitoneal tumor growth (Fig. 7B and E). In addition, the formation of ascites was profoundly and significantly inhibited by rOPCML in vivo in both immunoprecipitation models (Fig. 7C), and in A2780 tumor-bearing mice, rOPCML significantly inhibited the number of intraperitoneal deposits compared with the BSA control (Fig. 7D).

Western blotting of the recovered SKOV-3 intraperitoneal xenograft from control and rOPCML-treated animals (there was insufficient A2780 xenograft recovered for Western blotting because of the suppression of tumorigenicity) confirmed the same spectrum of rOPCML-mediated RTK downregulation as previously shown in vitro (Fig. 7F), including the lack of EGFR downregulation. Immunohistochemical staining of tumor sections from animals treated with rOPCML by the use of an OPCML antibody showed peripheral cell surface staining of OPCML in contrast to the weak or absent cytoplasmic OPCML staining seen in tumor sections from BSA-treated control animals (Supplementary Fig. S8B).

OPCML is frequently inactivated by somatic methylation and LOH in more than 80% of epithelial ovarian cancers (11) and in many other cancers (ref. 19; also see Supplementary Figure S1 and TCGA http://tcga-portal.nci.nih.gov/tcga-portal/AnomalySearch.jsp) with evidence of prognostic importance and near-ubiquitous loss of expression in cell lines and clinical biopsies (ref. 17; Supplementary Fig. S2 and Kaplan-Meier Plotter: http://kmplot.com/breast/index.php?p=1). We demonstrate here the tumor suppressor mechanism of action of OPCML.

OPCML negatively regulates a specific RTK repertoire consisting of EPHA2, FGFR1, FGFR3, HER2, and HER4 receptors and does not regulate EGFR, HER3, the remaining FGF receptors, VEGFR1/3, and many of the other EPHA receptors (see Supplementary Table S1). Immunoprecipitation and cell-free pulldown experiments with RTKs demonstrated that OPCML physically interacts with EPHA2, FGFR1, and HER2 via their extracellular domains but not with EGFR (the levels of which are unchanged by OPCML). The structural basis for this specificity is currently under investigation. We further explored the mechanism of OPCML action by using HER2 as a paradigm in the cancer SKOV-3 and the normal OSE-C2 cell model systems.

To demonstrate that OPCML mediates its function by interaction with the target RTK extracellular domain as a prerequisite for RTK downregulation, we transiently transfected full-length and truncated (extracellular domain-deleted) rat HER2/Neu constructs in the presence or absence of OPCML. We demonstrated clear downregulation of the intact 185-kD Neu receptor by greater than 75% in response to OPCML in contrast to the 95-kD extracellular domain-less truncated neu that remained unaffected by OPCML expression. In addition, we demonstrated that the extracellular domain-containing RTK's negative regulation by OPCML was functional and responsible for the observed tumor suppressor phenotype. OPCML-specific sequestration of HER2 to the detergent-resistant membrane fraction (detergent-insoluble fraction, or cholesterol-rich “lipid-raft” domain) was observed in OPCML-expressing SKOV-3 cells (BKS-2.1) as well as enriched colocalization of HER2 and OPCML. Furthermore, IFM also demonstrated the redistribution of HER2 to lipid-resistant membrane domains consistent with OPCML being localized and therefore internalized via a “lipid-raft” non–clathrin-dependent endocytic pathway such as the clathrin-independent carriers/GPI-enriched early endosomal compartment pathway (34).

These data are supported by the findings from pulse-chase experiments that demonstrate accelerated loss/nonrecirculation of biotinylated surface HER2 protein in OPCML-expressing cells. We also demonstrated that OPCML binding mediated polyubiquitination of this specific spectrum of RTKs in both OPCML-transfected cancer cells and normal ovarian surface epithelial cells expressing physiologic levels of OPCML. MG-132 studies in OPCML-expressing cancer and normal cells confirmed that these polyubiquitinated RTKs were then targeted for proteasomal degradation, thereby explaining how OPCML negatively regulates its RTK binding partners.

Our hypothesis is therefore that by binding to specific RTKs via their extracellular domains, OPCML sequesters these specific RTKs in detergent-resistant membrane domains and diverts those RTKs (in this case HER2) away from the canonical clathrin endocytic route, where rapid RTK recycling is prevalent, to a raft-dependent bulk endocytic pathway. This leads to OPCML-binding-specific polyubiquitination that results in degradation via a proteasomal route, presumably via a specific E3 ubiquitin ligase. OPCML can therefore abrogate the ligand-activated and steady-state phosphorylation of ERK 1 and 2 and AKT by regulating the availability of a spectrum of specific RTKs.

We speculate that during carcinogenesis, the frequent loss of OPCML expression in epithelial ovarian cancers (and many other types of cancer) by somatic methylation and LOH (11) may therefore deregulate RTKs, conferring a signaling-mediated selective growth advantage on those cells (as seen in OSE-C2 OPCML knockdown cells discussed previously). The lack of identification of a single frequently activated RTK candidate in ovarian cancer to date leads us to speculate that this may be a plausible mechanism. We further speculate that OPCML action may have even more relevance in other cancer types where a more oncogenic “addicted” signaling state for these RTKs exists (such as HER2 amplification in breast cancer), and this is currently under investigation.

These findings have immediate relevance for human cancer. OPCML is an unusual example of an extracellular tumor suppressor protein and therefore the possibility of direct tumor suppressor protein therapy at the cell surface could be considered rather than gene therapy with all its attendant complexities. The construction of a protein consisting of domains 1–3 without the N-terminus signal peptide or the C-terminus GPI anchor signal, and without eukaryotic glycosylation, allowed us to test the hypothesis that a recombinant OPCML-like protein (rOPCML) might have potential as a cancer therapeutic. In vitro, this recombinant protein therapeutic demonstrated dose-dependent growth inhibition in 6 of 7 epithelial ovarian cancers cell lines tested with no effect on a normal ovarian surface epithelial cell line (that expresses physiologic OPCML). The mechanism of this pharmacologic inhibition by rOPCML was shown to be identical to that of the transfected OPCML described in this study, with abrogation of an identical repertoire of RTKs, similar downstream pERK and pAKT effects, evidence of both growth inhibition and apoptosis, and no discernable effect on EGFR. Furthermore, in 2 separate intraperitoneal models of ovarian cancer in vivo, twice-weekly rOPCML intraperitoneal therapy inhibited peritoneal tumor growth and the formation of ascites and inhibited peritoneal dissemination in the A2780 model. Future studies will define the potency, optimal dose, and schedule of rOPCML in these models.

In summary, we have defined a novel mechanism of action for the frequently inactivated and prognostically important tumor suppressor OPCML as a systems-level repressor of a defined spectrum of receptor tyrosine kinases. We have also developed a recombinant OPCML-derived protein that has potential as a treatment approach for ovarian and other cancers. More generally, the discovery of the mechanism of action of OPCML may have serendipitously uncovered a spectrum of RTKs that could be coinhibited with RTK inhibitor combinations to avoid the problem of signaling redundancy and produce a more profound anticancer effect in ovarian and other cancers.

Polyclonal goat and monoclonal mouse anti-OPCML antibodies were purchased from R&D Systems. Anti-HER2 antibodies were purchased from Calbiochem [anti-ErbB2 (Ab-4) and (3B5) mouse monoclonal antibodies]. Anti-EGFR antibody (goat polyclonal antibody; catalog number AF-231) was from R&D Systems. Antihemagglutinin antibody was from Santa Cruz Biotechnology. Anti–EEA-1 and anti–caveolin-1 were purchased from Abcam. Phospho-EGFR, HER2, FGFR1 phospho-FGFR1 (Y766), phospho-ERK total ERK, phospho-AKT, total AKT, EPHA2, FGFR3, HER4, HER3, FGFR2, EphA10, VEGFR1, and VEGFR 3 b-tubulin were all purchased from AbCam. Horseradish peroxidase-conjugated secondary antibodies were from Dako. Alexa Fluor 488 goat antirabbit IgG and Alexa Fluor 555 goat antimouse were from Molecular Probes.

PEO1, PEA1, and PEA2 were recently authenticated and tested by Dr. Katherine Hale Stemke in the laboratory of Dr. Gordon Mills (Cancer Center Support grant for Characterized Cell Line core at University of Texas MD Anderson Cancer Center, National Cancer Institute grant number CA16672). These 3 cell lines were originally developed by Dr. Simon Langdon in 1992 and obtained by us in that laboratory. For PEO1, PEA1, and PEA2, we used Applied Biosystem's Identifyler kit, which tests 16 STR loci, and then we compared the output to public databases. Only 8 loci are published to avoid identification of individuals (Supplementary Table S2). The testing was undertaken in March 2011.

OVISE and OSE-C2 were received from their originating laboratories but not tested. OVISE was received from Dr. Junzo Kigawa (Tottori University, Japan) in November 2004. OSE-C2 was received from Dr. Richard Edmondson (Newcastle University, UK) in 2008. We have not undertaken authentication of the remaining cell lines, all of which were received from the American Type Culture Collection.

The SKOV-3-derived OPCML-expressing lines (SKOBS-3.5, BKS2.1, and empty-vector SKOBS-V1.2) have been described previously (11). Stimulation time courses were undertaken with 50 ng/mL human recombinant EGF (Promega) or 10 ng/mL acidic and basic FGFs (FGF1/2; R&D Systems) following serum-starvation overnight. The immortalized normal ovarian surface epithelium (OSE) line OSE-C2 (32) was held in culture at the permissive temperature of 33°C in 95% air and5% CO₂. We peformed transient transfections with the use of the Effectene lipofectamine reagent (QIAGEN), following the manufacturer's guidelines.

The OPCML cDNA expression plasmids in pcDNA3.1zeo previously described (11) were used for transient transfections. The cDNA encoding all 3 OPCML Ig domains was generated by PCR and introduced into the bacterial GST-fusion expression vector pGEX-6P-1 (GE Healthcare) and pHisTRx (kindly provided by Dr Edward McKenzie, University of Manchester). The hemagglutinin-tagged Ubiquitin vector (pRK5-HA-Ubiquitin-WT) was obtained from Dr. Luke Gaughan (Newcastle University), and the EGFR and HER2 cDNA in pcDNA-3.1zeo was provided by Prof. Bill Gullick (University of Kent). Rat Neu full-length cDNA (185-kD pNeu) and deletion mutant form with no extracelluar domain (95-kD Δ5) were cloned in pSecTagB2 as previously described (33). A full-length FGFR1 cDNA clone and extracellular domain constructs were provided by Prof. Graeme Guy and Prof. Kyung Hyun Kim, respectively.

Cells grown on glass slides were fixed in 4% paraformaldehyde and permeabilized for 20 minutes with PBS containing 0.2% saponin before blocking in PBS containing 10% goat serum, 2% albumin and 2% fetal calf serum for 1 hour. Slides were incubated with appropriate combinations of mAb OPCML, mAb HER2, and pAb EGFR primary antibodies for 1 hour at room temperature, followed by incubation for 1 hour with animal antimouse Alexa-555 (OPCML) and animal antirabbit Alexa 488 (HER2) before they were mounted and imaged on a Zeiss LSM 510 confocal microscope.

Endogenous OPCML was knocked down in OSE-C2 cells by transient transfection of a specific pool of 3 siRNAs (StealthKD-Invitrogen) with Lipofectamine RNAiMAX reagent following protocol guidelines.

Stable expression of shRNAs against physiologic OPCML was achieved in OSE-C2 cells via the use of the MISSION cloned shRNAs in pLKO.1-puro, supplied as glycerol stocks from Sigma-Aldrich. Five such cloned shRNAs were transfected into OSE-C2 and selected on 3 μg/mL puromycin for 3 to 4 weeks.

Cell proliferation assays were performed in quadruplicate with the MTT assay. Cells were plated out in 96-well plates at a density of 2,000 cells/well and cultured in low-serum medium (0.25% fetal calf serum) or low-serum medium supplemented with 50 ng/mL EGF. Cells were incubated MTT for 2 hours at 37°C and the purple fomazan product was solubilized in 100 μL of dimethyl sulfoxide resuspended and read on a plate reader at 540 nm.

Cell layers were washed in PBS and incubated for 30 minutes in lysis buffer (1% Triton X-100; 10 mM Tris, pH 8.0; 150 mM NaCl; 2.5 mM MgCl₂; 5 mM EGTA; 1mM Na₃VO₄; 50 mM sodium fluoride; and protein inhibitor cocktail; Roche). Cell lysates were then cleared by centrifugation at 13,000 rpm (16,000 3 g) for 20 minutes at 4°C, and aliquots containing equal amounts of protein were incubated with the appropriate antibody before addition of secondary antibody conjugated to Sepharose resin. Beads were then washed 3 times with lysis buffer and eluted by heating for 5 minutes in 50 μL of SDS sample buffer. Pull-down assays were performed with recombinant GST-OPCML fusion proteins bound to magnetic glutathione beads (Promega). Cell lysates that were prepared as for immunoprecipitation and proteins that were produced with the use of the in vitro TNT Rabbit Reticulocyte Lysate Expression System (Promega) or expressed in bacteria were analyzed for interactions.

Recombinant proteins were produced in the BL21 bacterial cell line (Promega). Protein expression was induced by the addition of 1 mM final concentration IPTG at 37°C for 4 hours. Cells were then harvested by centrifugation at 5000 rpm for 20 minutes at 4°C. Cell pellets were resuspended in PBS pH 7.4 and prepared for lysis with the use of lysozyme from the egg white of a chicken (1 mg/mL; Sigma-Aldrich), 10 units/mL Dnase I (Sigma-Aldrich), and 1% Triton X-100. After 1 hour of incubation at room temperature, the lysate was subject to freeze-thaw to enable complete lysis.

Inclusion bodies were solubilized in denaturation buffer (8 M Urea; 20 mM Tris-HCl, pH 8.0; 150 mM NaCl; and 10 mM DTT) to a final concentration of 5 mg/mL. Refolding of proteins was undertaken by extensive dialysis against cold PBS in 10-kDa MWCO dialysis tubing. Protein concentrations were monitored throughout the experiment with protein assay reagent (Bio-Rad Laboratories) with the use of bovine serum albumen as a standard.

Cells were harvested by treatment with EDTA and centrifugated at 3000 rpm for 5 minutes, and then washed with ice-cold PBS and followed by a second centrifugation step. Cells were then osmotically lysed by the use of a low-salt buffer (20 mM Tris-HCl, pH 7.0, containing complete protease inhibitor cocktail; Roche); then, we passed the cell suspension through a 22-gauge needle (an aliquot was retained for analysis; whole-cell lysate). After that, we performed a second centrifugation at 45,000 rpm, 4°C for 1 hour. The membrane pellet was resuspended (total membrane fraction; aliquot retained) in ice-cold PBS plus inhibitors and centrifuged again 45,000 rpm, 4°C for 30 minutes. The membrane pellet was resuspended in ice-cold PBS, 1% Triton-X100, plus inhibitors, and allowed to incubate for 30 minutes at 4°C. The detergent-treated membrane fraction was then centrifuged at 45,000 rpm at 4°C for 1 hour, the supernatant was removed and retained for analysis (detergent-soluble membranes), and the remaining pellet was solubilized in 1% SDS (detergent-resistant membrane fraction).

SKOBS-V1.2 and BKS-2.1 were transiently transfected with a hemagglutinin-tagged ubiquitin vector. At 24 hours after transfection, cells were serum starved for an additional 24 hours before being treated with the MG-132 for 1 hour at 37°C. Cells were then treated with EGF (50 ng/mL) for 1 hour at 37°C (+EGF) or left untreated (–serum). A rabbit anti-HER2 monoclonal antibody was used for immunoprecipitation. Ubiquitinated proteins were detected by immunoblotting with an antihemagglutinin antibody. Samples were run in parallel to probe for HER2 with a mouse anti-HER2 monoclonal antibody.

Cell-surface proteins were labeled via the Pierce cell-surface protein isolation kit where cell-impermeable, cleavable biotinylation reagent (Sulfo-NHS-SS-Biotin) labels exposed primary amines of proteins on the surface of cells. After a 30-minute “pulse” incubation with biotinylation reagent at 4°C, the cells are returned to growth media and incubated for required periods at 37°C in 5% CO². Cells are then harvested, lysed, and the labeled surface proteins are affinity-purified using Thermo Scientific NeutrAvidin Agarose resin, followed by elution in 2× SDS lysis buffer. Before spun-column purification of the biotinylated protein, 25% of the cell lysate was kept as a total cellular input sample and added to SDS-gel loading buffer for subsequent Western analysis.

A total of 10⁵ cells were seeded into 6-well plates, with 1 mL of RPMI (10% fetal calf serum, 2 mM L-glutamine, and 50 units/mL penicillin/streptomycin) for 24 hours. Then, 24 hours later they were either treated with BSA (3 mg/mL) or OPCML (3 mg/mL) for 6 hours. Cells were analyzed by flow cytometry after harvesting by gentle cell scraping and dual staining with annexin V and PI (FITC Annexin V Apoptosis Detection Kit II; BD Pharmingen) as per manufacturer's protocol. Flow cytometric analysis was performed using a FACSCalibur and CellQuest software (Becton Dickinson) and FlowJo (Tree Star) data analysis.

Casapse-3/7 apoptosis assays were performed in quadruplet by using a commercially available kit following manufacturer's instructions (Caspase-Glo 3/7 assay, Promega) according to the aforementioned scheme.

Two groups of 10 animals were intraperitoneally administered either 1 × 10⁷ cells of the A2780 or SKOV-3 cell line suspending in 0.5 mL of sterile PBS. Tumors were allowed to develop for 10 days, and 2 animals from each cohort were euthanized to analyze tumor take. The remaining animals were segregated into 2 groups: control animals (4 animals per cell line) receiving twice-weekly 1-mL injections of 10 μM BSA in PBS and trial animals (4 animals per cell line) receiving twice-weekly 1-mL injections of 10 μM rOPCML in PBS. The experiment was terminated after 3 weeks of treatment because of disease progression in control animals.

Sections (2.5 microns thick) were cut from formalin-fixed, paraffin-embedded tissue. The tissue sections were dewaxed and rehydrated by immersing in 3 changes of xylene, 3 changes of 74 OP, and 1 70% alcohol for 2 minutes each change, then in tap water. Endogenous peroxidase blocking was done by immersing the slides in hydrogen peroxide block (H₂O₂)–0.6% hydrogen peroxide in tap water for 15 minutes. The sections were then stained, developed, and counterstained on the 16000 Biogenex machine set according to manufacturer protocol and using super sensitive polymer HRP IHC detection kit (Biogenex).

Data are expressed as mean ± SEM. Differences were analyzed by the Fishers exact or Student t test. P < 0.05 was considered significant.

No potential conflicts of interest were disclosed.

We wish to acknowledge the support of the NIHR Biomedical Research Centre, the Cancer Research UK Clinical Centre, and the Experimental Cancer Medicine Centre at Imperial College London. We thank Prof. K. H. Kim, from the University of South Korea, for His-tagged FGFR1 ECD; Dr. Luke Gaughan, of Newcastle University, for HA-tagged ubiquitin expression plasmid; and Prof. Bill Gullick, of the University of Kent, for EGFR and HER2 cDNA in pcDNA-3.1zeo. The immortalized normal OSE line, OSE-C2, was a kind gift from Dr. Richard Edmondson of Newcastle University. We would also like to thank Mr. Feras Al-Jayoosi for technical assistance with some of these experiments. We are extremely grateful to Dr. Fiona Simpson and Prof. Henning Walczak for critical review of this work.

Supported by the Ovarian Cancer Action Research Centre Core Grant (principal investigator H. Gabra) and Cancer Research UK Discovery Award C8220/A14254 (principal investigator H. Gabra).