LOSS OF TUMOR SUPPRESSOR MERLIN IN ADVANCED BREAST CANCER IS DUE TO POST-TRANSLATIONAL REGULATION

Results: Merlin protein is degraded in advanced cancer

. The stability of Merlin protein is regulated, in part, by Akt-mediated phosphorylation at Threonine 230 and Serine 315 (15). Phosphorylation at these amino acids leads to Merlin degradation by ubiquitination. The reduced levels of Merlin in tumors of the nervous system are predominantly brought about by mutations or loss of heterozygosity (4,(16)(17)(18). However, Merlin's role in breast cancer has been largely ignored due to early, sporadic studies that did not detect mutations in tumor tissues (19,20). OPN is a secreted phosphoglycoprotein (21) that acts as an effector of tumor progression and metastasis at several levels (22,23). Elevated OPN is a marker for advanced breast cancer and multiple other cancer histotypes (24)(25)(26)(27)(28)(29)(30). OPNinitiated signaling activates NF-κB, PI-3-kinase and Akt pathways (31)(32)(33) and manifests as enhanced cell proliferation and survival, migration and adhesion (30). We report here that while the transcript levels of Merlin are unaltered in breast cancer tissues, there is loss of Merlin expression at the protein level in breast tumors, concomitant with an increase in OPN expression. Our studies revealed that OPN-initiated signaling induced Aktmediated phosphorylation and degradation of Merlin in breast cancer cells. Further, restoration of Merlin in breast cancer cells functionally impeded their malignant behavior. Logistic regression consistently identified decreased Merlin staining intensity in tumor tissues. It also showed that given the Merlin intensity, OPN ameliorates discrimination between normal and tumor tissue. Thus, our studies provide evidence that the availability of Merlin in breast tumors is regulated at the post-translational level. This is exciting from the perspective that Merlin was not found to be mutated or compromised at the transcript level in breast cancers. We have also defined a functional role for Merlin in limiting breast tumor growth and elucidated the utility of Merlin as an important biomarker in breast cancer.

EXPERIMENTAL PROCEDURES
Cell Culture -MCF10AT, MCF7, MDA-MB-231 and MDA-MB-435 cells were cultured as previously described (34). SUM159 cells were grown in DMEM/F-12 supplemented with FBS, insulin, and hydrocortisone in a humidified 5% CO 2 environment. The lineage infidelity of MDA-MB-435 cells has been discussed in several papers (35)(36)(37). We used this cell line as a model due to the fact that it naturally expresses copious OPN. Stable Merlin-expressing transfectants of MDA-MB-231 and SUM159 cells were generated by transfecting a Merlinexpressing construct.
Empty-vector was transfected as control; stable transfectants were selected on G418 (Invitrogen, Carlsbad, CA).
Transfection and Drug Treatment -Cells were transfected with empty vector, Merlin (WT; wild-type) or T230A S315A Merlin mutant and treated with clasto-Lactacystin β-Lactone (Sigma, St. Louis, MO) for 2 hours. Recombinant OPN (100 ng/mL) (R&D Systems, Minneapolis, MN) was added and cells were lysed after 6 hours. Where indicated, cells were first treated with Akt inhibitor IV (Calbiochem) in serum free media for 30 min followed by 100 ng/mL human rOPN for 16 hours.
Immunoprecipitation -Cells were transfected with pcDNA3.1 HA-ubiquitin alone or in combination with pIRES2-myc-Merlin and incubated for 24 hrs. Cells were treated with 10μM Lactacystin, 100ng/ml OPN and 10uM AKT inhibitor IV for 12 hrs and lysed in NP-40 buffer. The lysate was immunoprecipitaed with anti-Merlin antibody and the immunoprecipitate was assessed by immunoblotting.
Real-time quantitative PCR of tissue array -TissueScan plates (Origene, Rockville, MD) were assessed for the expression of OPN and Merlin transcripts using the manufacturer's protocol. The reaction was carried out in a Bio-by guest on March 24, 2020 http://www.jbc.org/ Downloaded from Rad iCycler iQ5 using the following program: activation step of 50°C for 2 minutes, then 42 cycles of 95°C for 5 minutes, 95°C for 15 seconds, and 60°C for 1 minute. Data was expressed as fold change (2 -ΔΔCT ). Statistical analysis was conducted using JMP version 7.0.1 (SAS, Inc., Cary, NC). A 5% level of significance was used to determine significance of results. The data were summarized using mean, standard deviation, and standard error of mean. The Pearson's correlation coefficient was used to determine correlation between numerical variables such as age. Wilcoxon test was used to compare C T levels of Merlin and OPN by group (normal or tumor), grade, and stage. p-value of < 0.05 was considered significant between groups.
Soft agar colonization assay -Cells were seeded in soft agar in triplicate in a 6-well plate, allowed to grow for 2-3 weeks, stained with crystal violet solution. Colonies with > 50 cells were microscopically counted.
Animal studies -Cells (1 million) suspended in HBSS (Invitrogen) were injected into the exposed third mammary fat pad of female athymic nude mice. Orthogonal tumor measurements were recorded twice-weekly. Mean tumor diameter was calculated as the square-root of the product of orthogonal measurements. These studies were conducted under IACUC-approved protocol. Statistical Analyses -Associations between intensities of Merlin and OPN expressions and patient's clinicopathologic data were assessed using the Wilcoxon rank test for categorical data and the Pearson's correlation coefficient for numerical data. The percentages of normal and tumor tissues expressing Merlin or OPN were compared using a Chi-square test. The significance of percentages of samples expressing Merlin or OPN as compared to the chance occurrence was determined using the exact binomial test. The univariate and multiple logistic regression models were fit to a binary variable normal versus tumor with Merlin and OPN as possible predictors. The possibility of developing a model using the relationship between OPN and Merlin was tested with a logistic regression model on a selected cohort of the data, scoring only the positive staining events from normal tissues for Merlin and the positive staining events from tumor tissue for OPN. The selection criteria were based on the fact that Merlin is a tumor suppressor, with a strong expression in normal tissue, whereas OPN -a tumor promoting protein, is known to be overexpressed in tumor tissue. The Chi-square test was used to assess the usefulness of model for prediction of likelihood of tumor. The effect likelihood ratio test was used to assess the usefulness of predictor variables in the model. The area under the ROC curve was used to determine the predictive ability of models and in model selection. All statistical analyses were performed using software JMP v 7 (SAS Inc.). All results with p-value < 0.05 were considered statistically significant. Statistical analyses of in vitro data: Statistical differences between groups were assessed using the Mann-Whitney test, t-test or ANOVA, using GraphPad Prism 4 software. Statistical significance was determined if the analysis reached 95% confidence. The precise p-values are listed in the corresponding figure legends.  (Fig. 1A, images c-f (tumor) relative to a-b (normal)). The expression of Merlin was significantly lower regardless of the nodal involvement. Notably, of the 75 carcinoma tissues, 56 (75%) tissues had lost expression of Merlin (p=0.0000097) (Fig. 1, B & C). In contrast, the expression of OPN was increased in breast cancer cases compared to normal breast tissues (Fig. 1A, images i-l (tumor) relative to gh (normal)) (p = 0.0097) (Fig. 1, D). Relative to normal tissues, a greater proportion of the tumor tissues showed OPN expression ( Fig. 1, E). Overall, of all the 3 grades combined, 43 tissues out of the 56 tissues showed no staining for Merlin simultaneous with increased staining for OPN. Thus, 77% (43 out of 56 Merlin-negative tissues) of the tissues that had lost Merlin expression showed increased OPN expression (p=0.000031). Specifically, the primary tumors from 23 out of the 24 cases with distant metastasis showed no staining for Merlin (p=0.000001). Of these, 20 cases (80%) showed increased OPN staining (p=0.00077) (Fig. 1, F). Thus, our studies showed that Merlin protein expression is lost in invasive breast cancer and the loss of Merlin is accompanied by an increased expression of OPN.

Merlin and OPN are inversely expressed in
The transcript levels of Merlin are unaltered in breast cancer tissues while those of OPN are increased -We queried the expression of Merlin in breast tumor tissues at two levels: amount of the transcript and the extent of protein expression. We assessed the transcript levels in tissues from 41 breast cancer patients and 7 normal control tissues. The transcript levels of Merlin did not show any appreciable changes (p > 0.05) between normal and breast tumorderived tissues; there was also no change in the Merlin transcript levels across the different grades of tumors or the disease stage (Fig. 2, A-C). In contrast, the transcript levels of OPN were significantly (p < 0.01) greater in the tumor tissues relative to normal tissues. The OPN transcript levels also increased significantly in tissues derived from grades II and III tumors and with progression of the disease stage (Fig. 2, D-F).

Merlin suppresses malignant behavior of breast cancer cells -Merlin's role as a tumor suppressor
is characterized in tumors of the nervous system. In order to determine the role of Merlin in impacting malignant behavior of breast cancer cells, we restored the expression of Merlin in two human breast cancer cell lines, SUM159 (Fig. 3, A) and MDA-MB-231 (Fig. 3, B). We assessed the malignant attributes of the resultant Merlinexpressing transfectants. Expression of Merlin caused a significant reduction in the ability of breast cancer cells to form foci ( Merlin for proteasome-mediated degradation. OPN interacts with a variety of cell surface receptors including CD44 and multiple integrins to activate signaling via the Akt pathway (31,38,39). To assess the role of Akt in OPNinitiated degradation of endogenous Merlin, we treated MCF7 cells (which express Merlin but do not express detectable levels of OPN) with recombinant OPN. Treatment with OPN caused phosphorylation of Akt concomitant with a decrease in the levels of endogenous Merlin suggesting that degradation of Merlin can be initiated by signaling downstream of OPN via Akt (Fig. 4, B, Lane 3). MCF7 cells were also treated with Akt inhibitor IV in addition to OPN (Lanes 3, 4, 5). While the levels of Akt phosphorylation predictably decreased after treatment, the levels of endogenous Merlin were restored by the inhibition of Akt phosphorylation even in the presence of OPN (Lanes 4 & 5) suggesting that inhibition of Akt activation downstream of OPN blocks the effects on degradation of Merlin. As seen in the accompanying Table, co-treatment with the Akt inhibitor blocks the effects of OPN allowing for a total recovery of endogenous Merlin (Lanes 4 & 5). Phosphorylation of Merlin via Akt targets it for degradation by the proteasome (15,40,41). Thus, in order to determine if OPN can induce ubiquitination of endogenous Merlin leading to its proteasomal degradation, MCF10AT cells were transfected with a HA-ubiquitin expressing construct. In the presence of OPN, endogenous Merlin undergoes some ubiquitination that is evident as a smear (Fig. 4, C, Lane 2). This smear persisted in the presence of Lactacystin

OPN initiated signaling causes phosphorylation of Merlin at Serine 315 -
The loss of Merlin in presence of OPN is caused by the phosphorylation of Merlin at the Ser315 position (Fig. 5, A). Specifically, phosphorylation of Merlin at this residue has been reported to target it for proteasome-mediated degradation (15,40). This form of Merlin was detectable upon inhibition of proteasomal degradation with Lactacystin in presence of OPN. We further determined that while OPN is able to induce degradation of Merlin, the Merlin mutant T230A S315A (that cannot be phosphorylated by Akt) is resistant to the effects of OPN (Fig. 5, B). Thus, cumulatively, our results suggest that OPN activates Akt-mediated signaling that causes phosphorylation of Merlin at Ser315. This event targets Merlin for ubiquitin-mediated degradation in breast cancer cells.

Degradation-resistant
Merlin functionally restricts malignant behavior -We assessed the ability of the Merlin mutant T230A S315A for its ability to impact the properties of breast cancer cells in the perspective of OPN signaling. Both, the wild-type Merlin and the T230A S315A Merlin mutant significantly (p < 0.05) reduced the numbers of foci formed by the SUM159 cells (Fig. 5, C). In order to test the effectiveness of T230A S315A Merlin mutant under conditions of elevated OPN expression, we tested the ability of Merlin to impact the foci formation capability of SUM159-OPN (stably expressing OPN) cells. While wild-type Merlin cannot impact the foci formation capability of the SUM159-OPN cells, the T230A S315A Merlin mutant brings about a significant (p < 0.05) reduction in the numbers of foci formed (Fig. 5, D). Similar results were obtained in the assessment of anchorageindependent growth in a soft-agar colonization assay (Fig. 5, E), suggesting that the degradationresistant T230A S315A Merlin mutant retains its ability to effectively blunt malignant attributes in presence of OPN.
OPN enhances tissue identification and discriminatory power of Merlin -In order to assess the discriminatory power of Merlin and OPN, we applied a logistic regression model to a binary variable of normal & tumor tissue to our data. The Chi-square test for appropriateness of model (p = 0.0448; ROC (Receiver Operating Characteristic) curve area = 0.7220) indicates that Merlin has a discriminatory power for distinguishing between normal and tumor tissues (Fig. 6, A). The logistic regression also showed that OPN by itself is not a good discriminator between normal and tumor tissues (p = 0.2878; ROC area = 0.6040) (Fig. 6, B). Further, multiple logistic regression showed that OPN does not increase the discriminatory power of Merlin (p = 0.162; ROC area = 0.723) (Fig. 6, C). Towards the possibility of developing a model that uses the unique inverse relationship between OPN and Merlin, we applied a logistic regression model to a selected cohort of the data, scoring only the positive staining events from normal tissues for Merlin and the positive staining events from tumor tissue for OPN. As seen in Figure 6, D, it is apparent that the logistic model for Merlin alone, using this data set is very good at discriminating between normal and breast tumor tissues (p < 0.0001; R 2 = 0.43; ROC area = 0.93). Furthermore, given the Merlin intensity, OPN expression ameliorates tissue identification with increased discriminative power of the model (n=46; p < 0.0001; R 2 = 0.81; ROC area = 0.9917) (Fig. 6, E). We then applied model developed from this training set to our selected data and we found that out of the 46 samples queried, only 2 samples were misclassified (Fig.  6, F) resulting in 96% probability of correct classification.

DISCUSSION
While Merlin has been extensively explored in tumors arising from the nervous system, its role in breast cancer is understudied. Early studies reported that mutations in Merlin were not detected in breast cancer (19). In a separate study, Yaegashi et al reported infrequent involvement of mutations in the NF2 gene (encoding for Merlin) in an independent cohort of 60 breast cancer patients (20). Dai et al reported that the estrogen-response gene and tumor suppressor, NHREF, likely acts in conjunction with Merlin to transduce a growth suppressive signal (42). Thus, while there are sporadic references regarding Merlin in breast cancer, the functional and biological roles of Merlin in breast cancer have largely been ignored due to the absence of detectable mutations and the lack of reports of change at the transcript level. In this study, we have seen that the level of Merlin transcript does not appreciably change in breast tumor tissues. Thus, it was intriguing to note a significant decrease in the immunohistochemical staining for Merlin, suggestive of the fact that Merlin protein expression is lost in breast cancer. In contrast, the oncoprotein, OPN showed an increase in expression at the transcript levels as well as at the protein level. OPN binding to cell surface receptors, such as the integrins, cause several signal transduction pathways to turn on culminating in enhanced proliferation, migration and survival (22). Our studies demonstrate that OPN induces Akt-mediated phosphorylation of Merlin that targets Merlin for ubiquitin-mediated degradation in breast cancer cells resulting in decreased overall cellular pools of endogenous Merlin. Ubiquitin-mediated degradation of tumor suppressors such as p53, PML, PTEN and VHL has also been documented to be responsible for the decreased availability of the respective proteins in tumor cells (43,44). We showed that degradation of endogenous Merlin is one of the    breast tissue (a, b); but is lost in invasive breast cancer (c-f). Conversely, OPN is expressed at very low levels in normal breast (g, h); but is upregulated in invasive breast cancer (i-l). Immunohistochemical staining was performed for Merlin and OPN on serial sections from 75 cases of invasive breast cancer and 9 normal breast tissues. We recorded loss of Merlin in 75% (56 cases) of invasive breast cancer cases. Of these 56 tissues, 43 Table and is represented as the ratio of Integrated Density Values (IDV) and also as a relative percent IDV ratio (relative to Lane 1, untreated cells). Densitometry was done using the AlphaEase program.