The Differential Role of L1 in Ovarian Carcinoma and Normal Ovarian Surface Epithelium

Research Article

Silvia Zecchini,1 Marco Bianchi,1 Nicoletta Colombo,2 Roberta Fasani,2 Giovanni Goisis,2 Chiara Casadio,2 Giuseppe Viale,2,3 Jinsong Liu,4 Meenhard Herlyn,5 Andrew K. Godwin,6 Paolo G. Nuciforo,1 and Ugo Cavallaro1

1FIRC Institute of Molecular Oncology, 2European Institute of Oncology, and 3University of Milano School of Medicine, Milano, Italy; 4Department of Pathology, The University of Texas M. D. Anderson Cancer, Houston, Texas; 5Wistar Institute and 6Fox Chase Cancer Center, Philadelphia, Pennsylvania

Abstract cytoreduction and chemotherapy usually induce tumor regression; however, most patients undergo a relapse of the disease with a Epithelial ovarian carcinoma (EOC) arises from the ovarian surface epithelium (OSE), a monolayer of poorly differentiated high mortality rate (1). Although a number of genetic lesions epithelial cells that lines the ovary. The molecular mecha- associated to higher risk for specific forms of EOC have been nisms underlying EOC invasion into the surrounding stroma identified, only very few biomarkers that contribute to the early and dissemination to the peritoneum and to retroperitoneal diagnosis and prognosis of the disease are currently available, lymph nodes are still unclear. Here, we analyzed the accounting for the difficult management of this tumor type. expression and the functional role of the cell adhesion Moreover, the molecular mechanisms that control the onset and molecule L1 during EOC development. In patient-derived progression of EOC have remained elusive, and their definition samples, L1 was expressed both in OSE and in a subset of EOC, would certainly open novel therapeutic avenues for the treatment of in the latter being mostly restricted to the invasive areas of the such a life-threatening disease. tumors. The expression of L1 correlated significantly with Cell adhesion molecules (CAM) are cell surface proteins mediating cell-cell and cell-matrix interactions. Alterations in poor outcome and with unfavorable clinicopathologic features CAM expression and/or function have been implicated in the of the disease. The peculiar expression pattern of L1 in normal development of various tumor types. In particular, the dysregu- OSE and invasive EOC raised the possibility that this adhesion lation of cell-cell adhesion caused by changes in the levels of molecule serves a different function in nontransformed versus cadherins and/or immunoglobulin-like CAMs (Ig-CAM) plays a neoplastic ovarian epithelial cells. Indeed, we showed that in causal role in the progression of several epithelial tumors (2). OSE cells L1 supports cell-cell adhesion and enhances The Ig-CAM L1 (also known as L1CAM or CD171) has been apoptosis, whereas it has no effect on cell proliferation and extensively characterized in the nervous system, where it invasion. In contrast, L1 inhibits cell-cell adhesion and mediates neuronal adhesion and migration, as well as axon apoptosis in ovarian carcinoma cells, where it promotes pathfinding and fasciculation (3). However, L1 is also expressed malignancy-related properties, such as cell proliferation, in nonneuronal tissues, including certain epithelia and some Erk1/2-dependent and phosphoinositide 3-kinase–dependent hematopoietic lineages, where its function remains elusive. invasion, and transendothelial migration. Interestingly, a Recent studies have reported an aberrant expression of L1 in crosstalk with the fibroblast growth factor receptor signaling various tumor types, including melanoma (4) and colon is implicated in the promalignant function of L1 in tumor carcinoma (5). In the latter, L1 has been characterized as a cells. Our findings point to L1 as an EOC biomarker transcriptional target of the Wnt/h-catenin pathway and as a correlating with poor prognosis, and highlight a switch in L1 factor involved in tumor cell invasion (5). Finally, an aberrant function associated to the neoplastic transformation of expression of L1 has been described in advanced EOC and has ovarian epithelial cells, thus implicating L1 as a potential been proposed to enhance the malignant phenotype of EOC cells therapeutic target. [Cancer Res 2008;68(4):1110–8] (6, 7). Here, we report the first demonstration that the expression of L1 is not restricted to EOC but is also abundant Introduction in the ovarian surface epithelium (OSE), the monolayer of poorly differentiated epithelial cells lining the ovaries that, upon Epithelial ovarian carcinoma (EOC) is the leading cause of neoplastic transformation, gives rise to most EOC forms (8). mortality among gynecological malignancies in the Western Furthermore, in agreement with previous reports, our analysis of world. Due to the lack of clinically relevant symptoms during clinical specimens revealed that the expression of L1 correlates the early phases of cancer development, the diagnosis is with poor prognosis in EOC patients. The expression pattern of frequently made at advanced stages of the disease. Surgical L1 and its association with specific clinicopathologic features of the disease suggested that this adhesion molecule serves a different function in normal versus neoplastic OSE. To verify this hypothesis, we have characterized the biological activities of L1 Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/). in OSE and in EOC cell lines. Our results indicated that L1 plays Requests for reprints: Ugo Cavallaro, FIRC Institute of Molecular Oncology, via markedly different, and often opposite, roles in nontumorigenic Adamello 16, I-20139 Milano, Italy. Phone: 39-02-574303-224; Fax: 39-02-574303-244; versus transformed ovarian epithelial cells. These findings E-mail: [email protected]. I2008 American Association for Cancer Research. underscore the importance of the cellular context in dictating doi:10.1158/0008-5472.CAN-07-2897 the functional role of L1 in OSE cells and provide the rationale

Cancer Res 2008; 68: (4). February 15, 2008 1110 www.aacrjournals.org

Downloaded from cancerres.aacrjournals.org on September 24, 2021. © 2008 American Association for Cancer Research. The Dual Role of L1 in Ovarian Epithelial Cells for evaluating L1 both as a EOC biomarker and as a novel target maintained in less cohesive cells inside the tumor mass and at for the molecular therapy of ovarian carcinoma. the invasive front of the tumor (Fig. 1B and Supplementary Fig S2B and C). Because this pattern suggested that L1 could be implicated in loss of cell-cell adhesion and in cancer invasion Materials and Methods (and, hence, progression), we asked whether the expression of L1 Selection of the patients, tissue microarray analysis, and data in tumor specimens was associated with the clinicopathologic analysis. Clinicopathologic and follow-up data of ovarian cancer patients features of the disease. Indeed, the presence of L1 in primary EOC surgically treated at the European Institute of Oncology from 1995 to 2004 significantly correlated with the poorly differentiated, more were used to select the cases that were included in the study. Inclusion aggressive phenotype of the tumor cells (grade 3, P = 0.034; see criteria were (a) first surgery at the European Institute of Oncology, (b)no neoadjuvant treatment, and (c) common epithelial histologic subtypes Supplementary Table S2). In addition, tumors characterized by (serous, mucinous, and endometrioid cystoadenocarcinomas). advanced International Federation of Gynecology and Obstetrics For clinicopathologic details of the patients, construction of tissue stage (III–IV versus I–II) and by the presence of lymph node microarrays, and data analysis, see Supplementary Materials and Methods. metastasis exhibited higher L1 levels, although the correlation did Cell culture. Human ovarian carcinoma cell lines IGROV1 and OVCAR3 not reach statistical significance (P = 0.064 and 0.057, respective- were from American Type Culture Collection. ly). Kaplan-Meier survival analyses strengthened the prognostic HIO-80 cells (called HIOSE/A; ref. 9) and IOSE80 cells (called HIOSE/B; value of L1 by showing a significantly shorter 5-year overall (log ref. 10) were derived from primary OSE cells transfected with SV40 large rank, P = 0.0087) and disease-free (log rank, P = 0.029) survival in T antigen. patients with L1-positive tumors (Fig. 1D), thus confirming and For further details on culture conditions, adenoviral infection and small interfering RNA (siRNA) transfection, and additional cell lines used in this extending the results of a previous study performed on a lower study, see Supplementary Materials and Methods. number of cases (6). To determine if L1 expression was a For details on antibodies and chemicals, biochemical methods, cell prognostic factor independent from the other clinicopathologic biological assays, and immunostaining procedures, see Supplementary variables, we performed a multivariate proportional hazards Materials and Methods. modeling of overall and disease-free survival, including patient’s age, histologic grade, and tumor stage. Importantly, the presence of L1 was independently associated with shorter 5-year overall Results survival (Fig. 1D, left), whereas the correlation with shorter The Expression of L1 in Human OSE and EOC disease-free survival, despite showing a clear trend, did not reach The expression of L1 was investigated in a panel of surgically statistical significance (Fig. 1D, right). resected specimens, including 20 normal ovaries with a clearly In summary, our screening of human ovarian samples revealed detectable OSE, 4 benign lesions (cystadenoma), 211 primary EOC, that (a) L1 expression is significantly down-regulated in primary and 199 metastatic lesions (see Supplementary Materials and and metastatic EOC compared with normal OSE and (b) the Methods for details on selection and analysis of patients). The expression of L1 in ovarian cancer cells correlated with clinico- results are summarized in Fig. 1C and Supplementary Table S2. The pathologic features of aggressiveness and with poor outcome. immunohistochemical staining with the polyclonal antibody These apparently contradictory results, together with the frequent- pcytL1, directed against the cytoplasmic tail (11), showed high ly observed L1 immunoreactivity in less cohesive cells within the levels of L1 in 100% of the normal OSE samples, mainly localized at tumor mass and at the invasive front of EOC, suggest that a switch the intercellular boundaries and at the basolateral surface of the in L1 function might occur upon transformation of ovarian cells (Fig. 1A, left). The identity of L1-positive tissue as OSE was epithelial cells. confirmed by the positive staining of the same cells for cytokeratins To address the role of neoplastic transformation in L1 expres- (Supplementary Fig. S1C–E). By immunohistochemistry, benign sion, we took advantage of a genetically defined cellular model that lesions showed an L1 immunoreactivity similar to OSE (not recapitulates many features of EOC development. This consisted of shown). Because our observations were in contrast with a previous OSE cells that were immortalized by sequential transfection with study that reported the absence of L1 in normal OSE (6), we used SV40 T antigen (called IOSE29 cells) and human telomerase reverse additional antibodies (2C2, CE7, and L1-S; see Supplementary transcriptase (T29 cells), followed by full transformation with Table S1) that confirmed the L1 immunoreactivity in OSE tissue oncogenic H-RasV12 (T29H; ref. 10). By analogy to most cells in the (Supplementary Fig. S1A and data not shown). Furthermore, L1 EOC mass, neither IOSE-29 nor T29 cells expressed L1, whereas the was also detected in freshly isolated OSE cells where, by analogy to latter was clearly detectable in isogenic, fully transformed T29H normal ovary, it was enriched at cell-cell contacts of primary OSE cells (Supplementary Fig. S3). In agreement with the observations cells (Fig. 1A, right and Supplementary Fig. S1B). The immuno- in patient-derived specimens, these findings supported the notion blotting of OSE cell lysates showed that L1 migrated as a doublet of that the expression of L1 is specifically associated to a highly malig- 200 to 220 kDa (Supplementary Fig. S3), where the higher nant phenotype. molecular weight band is likely to represent the mature, surface- exposed protein, whereas the lower molecular weight form would The Functional Role of L1 in Nontransformed versus represent the intracellular pool of L1 (12). Neoplastic Ovarian Epithelial Cells Ninety primary EOC (43%) and 89 metastases (45%) exhibited a Cell proliferation. To verify the hypothesis of a dual role of clear L1 staining (Fig. 1C). However, in contrast to nonneoplastic L1 in normal versus transformed OSE, we used gain-of-function epithelium (where 100% of OSE cells exhibited L1 staining), the and loss-of-function strategies in cells derived either from vast majority of cancer cells within these tumor tissues showed normal OSE or from ovarian carcinoma. Primary OSE cells are no or very little expression of L1 in the tissue cores analyzed difficult to isolate and to maintain in culture, and they rapidly [primary EOC, mean 7.89% (0–80); metastatic cancer, mean 7.74 undergo cellular senescence (13). To overcome this problem, as a (0-100)]. The immunoreactivity of L1, instead, was specifically model for nontransformed ovarian epithelium, we used cells www.aacrjournals.org 1111 Cancer Res 2008; 68: (4). February 15, 2008

Figure 1. Expression of L1 in OSE and EOC and correlation with patient survival. Normal ovaries (A, left) and EOC tissues (B) were immunohistochemically stained with the polyclonal antibody pcytL1. A, right, primary OSE cells were subjected to immunofluorescence staining with the monoclonal antibody CE7. Arrows, accumulation of L1 at the basolateral surface of OSE cells in the normal ovary (A, left), at the sites of cell-cell contacts in cultured OSE (A, right), and at the invasive edges of EOC in tumor tissues (B). B, arrowheads, indicate noninvasive areas of the tumor with low levels of L1 (left). The insets in A (left) and B show a magnification of the boxed areas. S, stroma; T, tumor. Magnifications: A, 60Â; B, left, 10Â; B, right, 40Â; A, inset, 100Â; B, insets, 60Â. C, expression of L1 in patient-derived samples. D, overall (left) and disease-free survival (right) curves according to L1 expression in primary ovarian cancer. Hazard ratio (HR) and 95% confidence intervals (in brackets) obtained from Cox multivariate regression model before (crude) and after adjusting for age, pathologic stage, and tumor grade (adjusted) are shown. Dotted line, L1-negative tumors; solid line, L1-positive tumors.

isolated from normal OSE stably transfected with the SV40 early successfully used to reduce L1 expression in human colon cancer region, containing large T and small t antigens (SV40-TAg). This cells (5). Indeed, the levels of L1 in transfected IGROV1 were procedure has been reported to prolong the life span of OSE dramatically decreased compared with control cells transfected cells without inducing a tumorigenic phenotype (9, 10, 14, 15). In with siRNA oligonucleotides targeting the mRNA for green contrast with OSE cells in vivo, L1 was undetectable in four fluorescent protein (GFP; Supplementary Fig. S4B). Instead, different lines of SV40-TAg–transfected OSE cells (Supplementary OVCAR3 cells were transduced with adenovirally expressed Figs. S3 and S4A), possibly due to the inactivation of L1- human L1 or with GFP (Supplementary Fig. S4C). regulating transcription factors by SV40-TAg (see Discussion). We first determined the effect of L1 on serum-induced proli- For the characterization of L1 activity in nontransformed OSE feration of HIOSE and IGROV1 cells. After serum starvation, cells cells, we selected HIO-80 (called HIOSE/A; ref. 9) and IOSE80 were stimulated with fetal bovine serum (FBS) and cell growth (called HIOSE/B; ref. 10) cells to perform gain-of-function was determined every 24 h. The forced expression of L1 showed studies. Upon adenoviral transduction with the cDNA for human no significant effect on either HIOSE/A or HIOSE/B cells L1, HIOSE cells exhibited a remarkable expression of L1 which (Fig. 2A, top left), indicating that L1 is not involved in the was correctly localized at the cell surface (Supplementary proliferation of this cell type. In contrast, IGROV1 cells exhibited Fig. S4A). To study the role of L1 ovarian carcinoma cells, we a marked reduction in cell proliferation upon abrogation of selected two cell lines: IGROV1 cells, which express high levels of L1 expression compared with cells transfected with the siRNA L1 (Supplementary Fig. S4B), and OVCAR3 cells, where L1 is for GFP (Fig. 2A, top right). The decrease in the proliferation rate almost undetectable (Supplementary Fig. S4C). The former cell of L1-deficient IGROV1 cells was further validated by a reduction line was transfected with siRNA oligonucleotides that have been in serum-induced DNA replication, as shown by BrdUrd

Cancer Res 2008; 68: (4). February 15, 2008 1112 www.aacrjournals.org

Downloaded from cancerres.aacrjournals.org on September 24, 2021. © 2008 American Association for Cancer Research. The Dual Role of L1 in Ovarian Epithelial Cells incorporation assays (Supplementary Fig. S5). To confirm that was markedly reduced upon loss of L1 (Fig. 2B, right). Both control the inhibitory effect depended on L1, we incubated parental and L1 knockdown IGROV1 cells responded to serum stimulation IGROV1 cells with CE7, a monoclonal antibody that has been with an increase in Erk1/2 activation. In contrast, no constitutive reported to repress the tumorigenicity of ovarian cancer cells in activation of Erk1/2 was observed in HIOSE cells, and this was not immunodeficient mice (16). In agreement with the data on L1 changed by the forced expression of L1 (Fig. 2B, left). Erk1/2 knockdown cells, the CE7 monoclonal antibody, but not an activation in response to serum was retained in both control and isotype-matched irrelevant antibody, decreased the proliferation L1-transfected HIOSE cells. These findings imply that L1 regulates of IGROV1 cells (Fig. 2A, bottom left). As these results indicated the Erk1/2 signaling pathway in neoplastic, but not normal, ovarian that L1 is required for EOC cell proliferation, we asked whether epithelial cells. it is also sufficient. To address this question, we determined the effect of forcing L1 expression in OVCAR3 cells on their Crosstalk of L1 with Fibroblast Growth Factor growth capacity after serum stimulation. Indeed, the prolifera- Receptor Signaling tion rate of OVCAR3-L1 cells was markedly higher than control, Because L1 enhanced the growth of serum-stimulated ovarian GFP-expressing cells (Fig. 2A, bottom right). Overall, these carcinoma cells (Figs. 2A and 3A) but was unable to stimulate the findings implicated L1 in the proliferation of neoplastic, but proliferation of serum-starved cells (Fig. 3B and C), it is not normal, OSE cells. conceivable that L1 synergizes with a signaling machinery elicited To gain insight into the mechanism underlying L1-dependent by growth factors contained in the serum. In an attempt to identify ovarian cancer cell proliferation, we focused on the Erk1/2 such signaling partners, we focused on two growth factor receptors pathway, known to mediate cell proliferation in several exper- that have been previously implicated in L1 function in neurons, the imental models. As many transformed cell lines, serum-starved fibroblast growth factor receptor (FGFR) and the epidermal growth IGROV1 exhibited constitutive activation of Erk1/2 that, however, factor receptor (EGFR; refs. 17, 18). Aberrant expression and/or

Figure 2. The effect of L1 on cell proliferation and on Erk1/2 activation. A, cell proliferation in HIOSE (top left), IGROV1 (top right and bottom left), or OVCAR3 cells (bottom right) was measured as described in Supplementary Materials and Methods. Bottom left, cell proliferation of IGROV1 cells in the presence of 20 Ag/mL CE7 antibody or of an isotype-matched, anti-hemagglutinin (HA) antibody. Points, mean from three independent experiments, each performed in triplicate; bars, SE. *, P < 0.005, relative to control cells. B, HIOSE/A (left) or IGROV1 cells (right) were serum-starved and then stimulated with 10% FBS for 10 min before immunoblotting for phosphorylated Erk1/2 (top). The blots were stripped and reprobed for total Erk1/2 (bottom).

www.aacrjournals.org 1113 Cancer Res 2008; 68: (4). February 15, 2008

activities of both receptors have been associated with ovarian cancer malignancy (19–21). Serum-induced proliferation of OVCAR-3 cells expressing either GFP or L1 was determined after a pretreatment of the cells with PD173074 or AG1478, which specifically inhibit FGFR and EGFR activity, respectively. Interest- ingly, PD173074 specifically abrogated the positive effect of L1 on serum-induced proliferation, without affecting the growth of OVCAR3-GFP cells (Fig. 3A), thus pointing to a functional cooperation between L1 and FGFR. In contrast, the growth of both GFP-expressing and L1-expressing OVCAR-3 cells was reduced by AG1478 to a very low level (Supplementary Fig. S6A), implicating EGFR in the proliferative capacity of OVCAR-3 cells but ruling out any specific effect on L1-dependent proliferation. To confirm the interplay between L1 and FGFR, serum-starved OVCAR3-GFP or OVCAR3-L1 cells were stimulated with FGF-2 and subjected to cell proliferation assay. Notably, only L1- expressing cells exhibited a remarkable proliferative response to FGF-2 (Fig. 3B), thus confirming that L1 cooperates with FGFR signaling. In agreement with the data on AG1478, EGF stimulated the proliferation of both OVCAR3-GFP and OVCAR3-L1 cells to the same extent (Supplementary Fig. S6B), implying the lack of a crosstalk between L1 and EGFR. The role of L1 in FGFR activity was also confirmed in IGROV1 cells by a loss-of-function approach. Indeed, the abrogation of L1 expression in these cells led to a marked decrease in FGF-induced proliferation, compared with cells transfected with a control siRNA (Fig. 3C). Apoptosis Resistance Previous work implicated L1 in the protection of ovarian carcinoma cells from apoptosis (22). To verify whether this antiapoptotic response occurs also in nontransformed OSE cells, we compared the effect of the drug staurosporine, which induces caspase-mediated apoptosis, in HIOSE and IGROV1 cells. The activation of caspase-3 after staurosporine treatment, a widely used readout for apoptosis, was monitored by immunoblotting with an antibody specific for the activated (i.e., cleaved) form of caspase-3. Staurosporine induced caspase-3 cleavage in L1- negative HIOSE cells with low efficiency, with maximal effect at 2 h of treatment in HIOSE/A and at 4 h in HIOSE/B cells. However, upon expression of L1, the proapoptotic effect of staurosporine increased remarkably in both cell lines (Supple- mentary Fig. S7). In contrast, L1 counteracted staurosporine- induced apoptosis in IGROV1 cells. Indeed, whereas control cells showed no caspase-3 cleavage during a 4-h treatment, cleaved caspase-3 in response to staurosporine became evident upon siRNA-mediated down-regulation of L1 (Supplementary Fig. S7). Thus, L1 acted as a survival factor in EOC cells, confirming and extending previous observations (22), while it enhanced drug-induced apoptosis in nontransformed OSE cells. Cell-Cell and Cell-Matrix Adhesion L1 has been originally characterized as an adhesion molecule that Figure 3. Crosstalk of L1 with FGFR. A, OVCAR3-GFP or OVCAR3-L1 promotes the physical interaction between adjacent cells. In cells were pretreated with 100 nmol/L PD173074 (PD) or with vehicle addition, cell-cell adhesion has long been known to affect tumor (DMSO) for 2 h before serum stimulation for the indicated time lengths and cell proliferation assay. *, P < 0.05 relative to PD173074-treated OVCAR3-L1 cells. malignancy and, in particular, may counteract the peritoneal B, serum-starved OVCAR3-GFP or OVCAR3-L1 cells were stimulated with dissemination of ovarian carcinoma (23). Based on these consid- 20 ng/mL FGF-2 for the indicated time lengths before measuring cell proliferation. erations and on our observation that, in EOC tissues, high levels of L1 *, P < 0.05; **, P < 0.005, relative to untreated OVCAR3-L1 and to FGF-treated OVCAR3-GFP cells. C, serum-starved IGROV1-siGFP or IGROV1-siL1 cells correlated with a less cohesive phenotype (e.g., see Supplementary were stimulated with 20 ng/mL FGF-2 for the indicated time lengths before Fig. S2B), we checked whether L1 regulates cell-cell adhesion in measuring cell proliferation. The decline in cell number observed at 72 h in untreated cells is likely due to apoptosis after serum deprivation. *, P <0.05; nontransformed versus neoplastic ovarian epithelial cells. As shown **, P < 0.005, relative to FGF-treated IGROV1-siGFP cells. in Fig. 4A, the forced expression of L1 in either HIOSE/A or HIOSE/B

Cancer Res 2008; 68: (4). February 15, 2008 1114 www.aacrjournals.org

Cell Invasion Previous studies have shown that L1 is implicated in the migration of ovarian carcinoma cells (7). However, it has remained elusive whether L1 also modulates the invasion of tumor cells through a three-dimensional extracellular matrix, a key step in cancer progression. To address this issue, we subjected L1-positive and L1-negative HIOSE and IGROV1 cells to invasion assays through Matrigel, a reconstituted basement membrane. Both HIOSE/A and HIOSE/B cells exhibited a weak invasive activity that was not significantly affected by the ectopic expression of L1 (Fig. 5A, left and middle columns). In contrast, L1 was required for Matrigel invasion by IGROV1 cells, as shown by the f75% reduction of invasive activity upon ablation of L1 expression (Fig. 5A, right columns). Given the biological implications of these findings for the progression of ovarian carcinoma, we used the L1- negative OVCAR-3 cellular model to verify whether L1 was also sufficient for ovarian cancer cell invasion. As shown in Fig. 5B, the forced expression of L1 resulted in a 4-fold increase in OVCAR-3 cell invasion, confirming that this adhesion molecule contributes to the malignant properties of ovarian cancer cells. Based on our observation of a L1/FGFR crosstalk in ovarian cancer cell proliferation, we asked whether a similar mechanism was involved in L1-induced OVCAR-3 cell invasion. Indeed, a pretreatment with the FGFR inhibitor PD173074 efficiently inhibited Matrigel invasion of L1-expressing OVCAR-3 cells to the level of control, untreated cells, whereas the invasion of OVCAR3- GFP cells was not affected (Fig. 5B). This specifically implicates FGFR signaling in L1-induced invasion. In contrast, the EGFR inhibitor AG1478 reduced the invasive capacity of both OVCAR3- GFP and OVCAR3-L1 cells by f3.5-fold (Supplementary Fig. S6C), which suggests that EGFR is involved in ovarian cancer cell invasion in a L1-independent manner. In an attempt to identify the signaling mediator(s) of L1- dependent tumor cell invasion, we assessed the Matrigel invasion Figure 4. The effect of L1 on cell-cell adhesion. HIOSE (A) or IGROV1 cells of L1-transfected OVCAR-3 cells in the presence of specific (B) were subjected to cell-cell adhesion assays as described in Supplementary inhibitors of either the Erk1/2 or the phosphoinositide 3-kinase Materials and Methods. Columns, mean from a representative experiment performed in triplicate; bars, SD. *, P < 0.05, relative to control cells. (PI3K) pathways, two major biochemical cascades previously implicated in cancer cell malignancy. Both U0126 and LY294002, which inhibit Erk1/2 and PI3K activity, respectively, efficiently cells enhanced their intercellular adhesion. We tested the effect of repressed the invasion of L1-expressing OVCAR-3 cells (Fig. 5B). L1 on both calcium-independent and dependent cell-cell adhesion. These findings indicate that L1 enhances the invasive potential of Whereas the former is mediated by Ig-CAMs, calcium-dependent ovarian carcinoma cells via the Erk1/2 and PI3K signaling intercellular adhesion is initiated by members of the cadherin family pathways. (2). Thus, the fact that L1 stimulated also calcium-dependent adhesion in HIOSE cells (Fig. 4A) suggested a functional interaction Transendothelial Migration with cadherins in this cell type. In contrast, L1 exhibited a negative Ovarian carcinoma disseminates predominantly through detach- effect on cell-cell adhesion in IGROV1 cells, in that L1 knockdown ment of tumor masses from the primary site, which then spread cells formed f3.5-fold and 10-fold more clusters than control cells into the peritoneal cavity and/or colonize peritoneal organs (25). in calcium-independent and calcium-dependent conditions, respec- However, a significant proportion of ovarian carcinomas forms tively (Fig. 4B). Therefore, L1 stimulates cell-cell adhesion in HIOSE metastases in the retroperitoneal lymph nodes, most likely cells, whereas it exerts an inhibitory effect on intercellular adhesion disseminating through the lymphatic circulation (26, 27). To verify in IGROV1 cells. whether L1 is implicated in this route of EOC metastatic spread, we L1 has been reported to interact with integrins, the key regulators assayed for the ability of L1 to modulate the transmigration of wild- of cell adhesion to the extracellular matrix (24). Hence, we asked type versus L1-deficient IGROV1 cells through a monolayer of whether L1 affected cell-matrix adhesion in nontumorigenic and lymphatic endothelial cells. We used HDLEC1 cells, a lymphatic neoplastic ovarian epithelial cells. Adhesion assays on different endothelial cell line, whose life span was prolonged by telomerase extracellular matrix components revealed that L1 stimulates cell- expression, and HMEC-1 cells, a widely used endothelial cell line matrix adhesion, although the efficiency of this proadhesive activity that was recently reported to express several lymphatic endothelial and the specific integrins involved (as reflected by the adhesion to markers (28). Reducing L1 expression in IGROV1 cells caused a specific ECM components) show some differences between HIOSE 2-fold decrease in their ability to cross the lymphatic endothelial and IGROV1 cells (Supplementary Fig. S8). barrier constituted by either HMEC-1 or HDLEC1 cells (Fig. 5C). An www.aacrjournals.org 1115 Cancer Res 2008; 68: (4). February 15, 2008

Downloaded from cancerres.aacrjournals.org on September 24, 2021. © 2008 American Association for Cancer Research. Cancer Research even stronger effect was observed on the transmigration through Discussion blood vessel endothelial cells (Fig. 5C). Therefore, L1 is required for This study reports for the first time that the adhesion molecule the transendothelial migration of ovarian carcinoma cells, L1 exerts different functions in nontransformed versus neoplastic emerging as a potential player in the formation of EOC lymph OSE cells, promoting a nonmalignant phenotype in the former and node metastasis. an invasive one in the latter. Previous studies have implicated L1 in EOC progression, based on the correlation between L1 expression and poor prognosis in patients (6), an observation that we have confirmed and extended in the present report, and on the L1- dependent growth of EOC xenografts in the peritoneum of immunodeficient mice (16). However, the role of L1 in OSE cells has not been addressed, likely due to previous negative results on L1 expression in this tissue. Indeed, the groups of Altevogt and Fogel have reported the absence of L1 in normal OSE (6, 29), a result that is in sharp contrast with our data. Those studies were performed using two monoclonal antibodies (L1-11A and L1-14-10, both raised against the ectodomain of L1) different from the antibodies used in our screening, which probably accounts for the discrepancy between the results of Altevogt and Fogel and ours. This highlights the importance of the antibody selection when analyzing the expression and localization of L1, also based on the notion that L1 can undergo extensive posttranslational modifications (including glycosylation, proteolytic cleavage, and ectodomain shedding; ref. 30), which probably change its immunoreactivity. Overall, the finding that OSE cells do express L1 is noteworthy because, combined with the specific L1 enrichment in highly invasive EOC cells, it was at the origin of our hypothesis that L1 plays a dual role in nontransformed versus neoplastic OSE cells. In nontumorigenic OSE cells, L1 supported cell-cell adhesion and enhanced drug-induced apoptosis, whereas it showed no effect on cellular processes associated with tumor malignancy. In contrast, ovarian cancer cells exhibited L1-dependent cell proliferation, invasion, resistance to apoptosis, and transendothelial migration, all representing important steps in cancer progression, whereas cell-cell adhesion was repressed by L1. The inhibitory effect of L1 on intercellular adhesion is not restricted to ovarian cancer cells, as it was recently reported in the breast cancer cell line MCF7, where L1-mediated disruption of adherens junctions resulted in enhanced cell motility (31). On one hand, our findings confirm and extend previous observations on the role of L1 in enhancing the malignant phenotype of ovarian carcinoma. For example, L1 has been reported to support the i.p. growth of ovarian cancer cells in immunodeficient mice (16) and to induce their migration and resistance to apoptosis (7, 22). On the other hand, our results highlight a novel function of L1 in nontumorigenic OSE cells, namely the induction of both calcium- dependent and independent cell-cell adhesion. Based on the notion that intercellular adhesion, and in particular the calcium- dependent one, efficiently represses tumor invasion (2), it is Figure 5. The effect of L1 on cellular invasion and on transendothelial migration. tempting to speculate that L1 contributes to inhibit or restrict the A, HIOSE (left and middle columns) or IGROV1 cells (right columns) were subjected to Matrigel invasion assays as described in Supplementary Materials malignant transformation of OSE cells. In agreement with this and Methods. Values represent the percentage of invasion relative to control hypothesis, L1 is down-regulated in the majority of the tumor IGROV-siGFP cells. B, OVCAR3 cells expressing GFP or L1 were left untreated cells during EOC development (this study and ref. 6). Yet, high or treated with 100 nmol/L PD173074, 20 Amol/L U0126, or 10 Amol/L LY294002 for 30 min before the Matrigel invasion assay. Values represent the percentage of levels of L1 are present in less cohesive cells within the tumor invasion relative to control OVCAR3-GFP cells. Columns, mean from three mass, as well as at the tumor-stroma interface, namely where independent experiments, each performed in triplicate; bars, SE. *P < 0.005. C, IGROV1 cells transfected with siRNA against L1 or GFP were subjected to cancer cells are actively invading the surrounding tissue. Overall, transmigration assays through monolayers of either, HDLEC1, HMEC-1, or the picture emerging from the expression pattern of L1 in vivo human umbilical vascular endothelial cell (HUVEC) cells, as described in and from our functional studies in cultured cells is consistent Supplementary Materials and Methods. Columns, mean from a representative experiment performed in triplicate; bars, SD. *, P < 0.05, relative to control, with a model, whereby L1 enhances intercellular adhesion in OSE siGFP-transfected cells. while it acts as a tumor promoter in advanced EOC.

Cancer Res 2008; 68: (4). February 15, 2008 1116 www.aacrjournals.org

The cellular and molecular factors that determine the changes treatment of EOC. In agreement with our observations, the in L1 expression during EOC progression remain elusive. In adhesion molecule N-cadherin has been shown to potentiate FGFR advanced colon carcinoma, the Wnt/h-catenin/TCF pathway has signaling in breast cancer cells by favoring a sustained activation of been proposed to induce the expression of L1 at the invasive the receptor by FGF-2 (35). It is noteworthy that the crosstalk of an front of the tumor (5). Unlike colon carcinoma, whose normal adhesion molecule with FGFR can also lead to the repression of counterpart is negative for L1 (5),7 the latter is found both in FGF-induced signaling, as we have recently reported for another Ig- normal OSE and in a subset of EOC cells, implying that different CAM, neural cell adhesion molecule, that abolishes the cellular mechanisms could regulate the expression of L1 in different response to FGF-2 in fibroblasts (36). Hence, the outcome of the tumors. Because most of the tumor cells in EOC masses show no adhesion molecule/FGFR interaction is likely to depend on the L1 expression, ovarian tumorigenesis is accompanied by a general specific adhesion molecule involved and on the cellular context. On loss of L1. Such a loss could result from the inactivation of the other hand, also the repertoire of growth factor receptors tumor-suppressing genes, such as p53, an event occurring in 50% involved in crosstalk with L1 seems to be cell type–specific, to 70% of advanced ovarian cancers (32). Indeed, an inactivating because, unlikely previous observations in neurons (18), we mutation in the p53 gene is accompanied by the down-regulation obtained no evidence of L1 interacting with the EGFR signaling of L1 in small cell carcinoma of the prostate.8 Along the same machinery in ovarian cancer cells. line, we report the loss of L1 in OSE cells expressing the SV40 T Other signaling mediators that have been implicated in L1- antigen, a well-characterized antagonist of p53 function (33). The induced axonal growth (37) and, based on our studies, are also presence of L1 at the invasive front of more advanced EOC could involved in L1-dependent ovarian cancer progression include then depend on its reexpression as part of the transition toward mitogen-activated protein kinase and PI3K pathways. But why an invasive phenotype. Based on our results in Ras-transformed are these signaling cascades activated by L1 specifically in ovarian OSE cells (see Supplementary Fig. S3) and on the frequency of cancer and not in OSE cells? A possible scenario is that the oncogenic Ras mutations in advanced EOC (34), it is conceivable repertoire of proteins interacting with L1 changes upon neoplastic that this oncogene is causally involved in the expression of L1 in transformation of OSE cells. Indeed, many studies on neuronal invasive EOC cells. Along this line, we have observed induction systems have documented the heterophilic interactions of L1 with of L1 in OVCAR3 cells upon forced expression of activated Ras.9 a broad spectrum of molecules. The interacting partners of L1 At the cellular level, one possibility is that microenvironment- include several components of the extracellular matrix, cell surface derived factors induce L1 expression in the cancer cells located at molecules, such as EGFR, neuropilin, and various integrins (besides the tumor-stroma interface. Alternatively, the L1-positive cells at FGFR, as mentioned above), intracellular signaling effectors, such the edge of EOC might derive from the selection of a subset of as Src, Numb, and RanBPM, and cytoskeletal components, such as transformed OSE cells that have maintained the expression of L1, ankyrins and ezrin (reviewed in refs. 3, 38). Many of these accompanied by a functional switch of the protein to a molecules are involved in cellular processes that, once deregulated, proinvasive activity, thus enhancing tumor malignancy. contribute to cancer progression, thus providing a potential basis The dual role of L1 in OSE cells indicates that the cellular for L1-dependent function in EOC. For example, FGFRs are context and, in particular, the acquisition of a transformed abundantly expressed in EOC (19, 20, 39), and our data support phenotype, has a major effect on the function of L1 and can the hypothesis that the simultaneous expression of L1 leads to actually switch it from a bona fide cell-cell adhesion molecule to a excessive FGFR signaling that, in turn, enhance tumor invasion. tumor-promoting factor. The molecular events that determine this In summary, we have shown that L1 plays a dual role in OSE functional switch remain unknown. An attractive candidate as a cells, consistent with a tumor-suppressive function in nontrans- mediator of L1’s effect on EOC development is the FGFR, given that formed cells and with a proinvasive function in cancer cells. The (a) an aberrant FGFR signaling has been associated to EOC characterization of the cellular and molecular determinants malignancy (19, 20) and (b) the crosstalk between L1 and FGFR has responsible for this shift in L1 activity will contribute to the long been described in neuronal cells, where it stimulates axonal identification of novel regulatory mechanisms implicated in EOC growth (17), although the physical association between the two progression, hopefully providing new therapeutic targets for this molecules has not been reported. Our results showed for the first neoplastic disease. time a functional interaction of L1 with FGFR in ovarian cancer cells, where it plays an important role in L1-dependent cell Acknowledgments proliferation and invasion. Thus, interfering with the L1/FGFR Received 7/30/2007; revised 11/8/2007; accepted 12/19/2007. crosstalk might prove a suitable therapeutic approach for the Grant support: Associazione Italiana per la Ricerca sul Cancro grant 1378 (U. Cavallaro), Telethon Foundation grant GGP04078 (U. Cavallaro), and Cariplo Foundation grant 2004.1587/11.5437 (U. Cavallaro). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance 7 Our unpublished observation. with 18 U.S.C. Section 1734 solely to indicate this fact. 8 D.E. Hansel, personal communication. We thank V. Lemmon, M. Schachner, K. Blaser, T.J. Lawley, A. Insinga, R. Nisato, and 9 S. Zecchini and U. Cavallaro, unpublished data. M. Pepper for providing antibodies and cell lines.

References by cadherins and Ig-CAMs in cancer. Nat Rev Cancer 4. Thies A, Schachner M, Moll I, et al. Overexpression of 2004;4:118–32. the cell adhesion molecule L1 is associated with 1. Sun CC, Ramirez PT, Bodurka DC. Quality of life for 3. Kenwrick S, Watkins A, Angelis ED. Neural cell metastasis in cutaneous malignant melanoma. Eur J patients with epithelial ovarian cancer. Nat Clin Pract recognition molecule L1: relating biological complexity Cancer 2002;38:1708–16. Oncol 2007;4:18–29. to human disease mutations. Hum Mol Genet 2000;9: 5. Gavert N, Conacci-Sorrell M, Gast D, et al. L1, a novel 2. Cavallaro U, Christofori G. Cell adhesion and signalling 879–86. target of {h}-catenin signaling, transforms cells and is www.aacrjournals.org 1117 Cancer Res 2008; 68: (4). February 15, 2008

expressed at the invasive front of colon cancers. J Cell mice by anti-L1-cell adhesion molecule monoclonal lymphatic endothelial cells with an extended life span. Biol 2005;168:633–42. antibody treatment. Cancer Res 2006;66:936–43. Am J Pathol 2004;165:11–24. 6. Fogel M, Gutwein P, Mechtersheimer S, et al. L1 17. Williams EJ, Furness J, Walsh FS, Doherty P. 29. Huszar M, Moldenhauer G, Gschwend V, Ben-Arie A, expression as a predictor of progression and survival in Activation of the FGF receptor underlies neurite Altevogt P, Fogel M. Expression profile analysis in patients with uterine and ovarian carcinomas. Lancet outgrowth stimulated by L1, N-CAM, N-cadherin. multiple human tumors identifies L1 (CD171) as a 2003;362:869–75. Neuron 1994;13:583–94. molecular marker for differential diagnosis and targeted 7. Gast D, Riedle S, Riedle S, et al. L1 augments cell 18. Islam R, Kristiansen LV, Romani S, Garcia-Alonso L, therapy. Hum Pathol 2006;37:1000–8. migration and tumor growth but not h3integrin Hortsch M. Activation of EGF receptor kinase by L1- 30. Haspel J, Grumet M. The L1CAM extracellular region: expression in ovarian carcinomas. Int J Cancer 2005; mediated homophilic cell interactions. Mol Biol Cell a multi-domain protein with modular and cooperative 115:658–65. 2004;15:2003–12. binding modes. Front Biosci 2003;8:s1210–25. 8. Shih IM, Kurman RJ. Ovarian tumorigenesis: a 19. De Cecco L, Marchionni L, Gariboldi M, et al. Gene 31. Shtutman M, Levina E, Ohouo P, Baig M, Roninson proposed model based on morphological and molecular expression profiling of advanced ovarian cancer: IB. Cell adhesion molecule L1 disrupts E-cadherin- genetic analysis. Am J Pathol 2004;164:1511–8. characterization of a molecular signature involving containing adherens junctions and increases scattering 9. Yang WL, Godwin AK, Xu XX. Tumor necrosis factor- fibroblast growth factor 2. Oncogene 2004;23:8171–83. and motility of MCF7 breast carcinoma cells. Cancer Res a-induced matrix proteolytic enzyme production and 20. Valve E, Martikainen P, Seppanen J, et al. Expression 2006;66:11370–80. basement membrane remodeling by human ovarian of fibroblast growth factor (FGF)-8 isoforms and FGF 32. Okamoto A, Sameshima Y, Yokoyama S, et al. surface epithelial cells: molecular basis linking ovulation receptors in human ovarian tumors. Int J Cancer 2000; Frequent allelic losses and mutations of the p53 and cancer risk. Cancer Res 2004;64:1534–40. 88:718–25. gene in human ovarian cancer. Cancer Res 1991;51: 10. Liu J, Yang G, Thompson-Lanza JA, et al. A genetically 21. Hennessy BT, Mills GB. Ovarian cancer: Homeobox 5171–6. defined model for human ovarian cancer. Cancer Res genes, autocrine/paracrine growth, and kinase signaling. 33. Ahuja D, Saenz-Robles MT, Pipas JM. SV40 large T 2004;64:1655–63. Int J Biochem Cell Biol 2006;38:1450–6. antigen targets multiple cellular pathways to elicit 11. Mechtersheimer S, Gutwein P, Agmon-Levin N, et al. 22. Stoeck A, Gast D, Sanderson MP, Issa Y, Gutwein P, cellular transformation. Oncogene 2005;24:7729–45. Ectodomain shedding of L1 adhesion molecule pro- Altevogt P. L1-CAM in a membrane-bound or soluble 34. Mammas IN, Zafiropoulos A, Spandidos DA. Involve- motes cell migration by autocrine binding to integrins. form augments protection from apoptosis in ovarian ment of the ras genes in female genital tract cancer. Int J J Cell Biol 2001;155:661–73. carcinoma cells. Gynecol Oncol 2007;104:461–9. Oncol 2005;26:1241–55. 12. Zisch AH, Stallcup WB, Chong LD, et al. Tyrosine 23. Sundfeldt K. Cell-cell adhesion in the normal ovary 35. Suyama K, Shapiro I, Guttman M, Hazan RB. A phosphorylation of L1 family adhesion molecules: and ovarian tumors of epithelial origin; an exception to signaling pathway leading to metastasis is controlled by implication of the Eph kinase Cek5. J Neurosci Res the rule. Mol Cell Endocrinol 2003;202:89–96. N-cadherin and the FGF receptor. Cancer Cell 2002;2: 1997;47:655–65. 24. Silletti S, Mei F, Sheppard D, Montgomery AM. 301–14. 13. Auersperg N, Wong AS, Choi KC, Kang SK, Leung PC. Plasmin-sensitive dibasic sequences in the third fibro- 36. Francavilla C, Loeffler S, Piccini D, Kren A, Christo- Ovarian surface epithelium: biology, endocrinology, and nectin-like domain of L1-cell adhesion molecule (CAM) fori G, Cavallaro U. Neural cell adhesion molecule pathology. Endocr Rev 2001;22:255–88. facilitate homomultimerization and concomitant integ- regulates the cellular response to fibroblast growth 14. Leung EH, Leung PC, Auersperg N. Differentiation rin recruitment. J Cell Biol 2000;149:1485–502. factor. J Cell Sci 2007;120:4388–94. and growth potential of human ovarian surface 25. Tan DSP, Agarwal R, Kaye SB. Mechanisms of 37. Schmid RS, Pruitt WM, Maness PF. A MAP kinase- epithelial cells expressing temperature-sensitive SV40 transcoelomic metastasis in ovarian cancer. Lancet signaling pathway mediates neurite outgrowth on L1 T antigen. In vitro Cell Dev Biol Anim 2001;37:515–21. Oncol 2006;7:925–34. and requires Src-dependent endocytosis. J Neurosci 15. Capo-Chichi CD, Smith ER, Yang DH, et al. 26. SundarSS,ZhangH,BrownP,etal.Roleof 2000;20:4177–88. Dynamic alterations of the extracellular environment lymphangiogenesis in epithelial ovarian cancer. Br J 38. Maness PF, Schachner M. Neural recognition mole- of ovarian surface epithelial cells in premalignant Cancer 2006;94:1650–7. cules of the immunoglobulin superfamily: signaling transformation, tumorigenicity, and metastasis. Cancer 27. Ueda M, Hung Y-C, Terai Y, et al. Vascular endothelial transducers of axon guidance and neuronal migration. 2002;95:1802–15. growth factor-C expression and invasive phenotype in Nat Neurosci 2007;10:19–26. 16. Arlt MJ, Novak-Hofer I, Gast D, et al. Efficient ovarian carcinomas. Clin Cancer Res 2005;11:3225–32. 39. Zhang Y, Guo KJ, Shang H, Wang YJ, Sun LG. inhibition of intra-peritoneal tumor growth and dis- 28. Nisato RE, Harrison JA, Buser R, et al. Generation Expression of aFGF, bFGF, FGFR1 in ovarian epithelial semination of human ovarian carcinoma cells in nude and characterization of telomerase-transfected human neoplasm. Chin Med J (Engl) 2004;117:601–3.

Cancer Res 2008; 68: (4). February 15, 2008 1118 www.aacrjournals.org

Downloaded from cancerres.aacrjournals.org on September 24, 2021. © 2008 American Association for Cancer Research. The Differential Role of L1 in Ovarian Carcinoma and Normal Ovarian Surface Epithelium

Silvia Zecchini, Marco Bianchi, Nicoletta Colombo, et al.

Cancer Res 2008;68:1110-1118.

Updated version Access the most recent version of this article at: http://cancerres.aacrjournals.org/content/68/4/1110

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2008/02/22/68.4.1110.DC1

Cited articles This article cites 39 articles, 12 of which you can access for free at: http://cancerres.aacrjournals.org/content/68/4/1110.full#ref-list-1

Citing articles This article has been cited by 11 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/68/4/1110.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/68/4/1110. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.