Regulation of the neurofibromatosis type 2 tumor suppressor protein, merlin, by adhesion and growth arrest stimuli.

The neurofibromatosis type 2 tumor suppressor gene is inactivated in the development of familial and sporadic schwannomas and meningiomas. The encoded protein, Merlin, is closely related to the Ezrin, Radixin, and Moesin family of membrane/cytoskeletal linker proteins. Examination of Merlin in several cell lines revealed that the protein migrates as two distinct species near 70 kDa. Phosphatase treatment and orthophosphate labeling demonstrated that the species with decreased mobility is phosphorylated. Given Merlin's localization to cortical actin structures, we examined the effect of cell-cell contact or other forms of growth arrest on Merlin expression and post-translational modification. Under conditions of confluency or serum deprivation, the levels of phosphorylated and unphosphorylated Merlin species increased significantly. Cells arrested in G1 by other methods or other phases of the cell cycle did not show changes in Merlin levels. Furthermore, loss of adhesion resulted in a nearly complete dephosphorylation of Merlin, which was reversed upon re-plating of cells, suggesting Merlin phosphorylation may be responsive to cell spreading or changes in cell shape. Thus, the tumor suppressor function of Merlin may involve the regulation of cellular responses to cues such as cell-cell contact, growth factor microenvironment, or changes in cell shape.

Neurofibromatosis type II (NF2) 1 is an autosomal dominant cancer disorder characterized by the development of bilateral schwannomas of the eighth (auditory) cranial nerve. Other features of the disease are spinal nerve root schwannomas and cranial meningiomas (1). Genetic mapping and positional cloning led to the identification of the NF2 tumor suppressor gene, which was shown to be mutated in the germ line of NF2 patients and in sporadically occurring tumors of the type associated with the disease (2)(3)(4).
The NF2 gene encodes a 595-amino acid protein belonging to the band 4.1 protein superfamily (3,4). The NF2-encoded protein is most similar to three closely related genes in this family, Ezrin, Radixin, and Moesin (the "ERMs") and was hence dubbed Merlin for moesin, ezrin, radixin-like protein. Merlin is most similar to the ERMs in its N-terminal half (e.g. 61% identity to Ezrin), whereas the C-terminal half of the protein shares only 24% identity with the other members of this subfamily. NF2 is transcribed at moderate levels in most human and mouse tissues and is highly expressed in the mouse fetal brain (3)(4)(5)(6). 2 Several members of the band 4.1 superfamily, including the ERM proteins, are thought to serve as membrane-cytoskeletal linkers. Band 4.1 itself is a major component of the erythrocyte undercoat and binds to the integral membrane protein glycophorin C through its N-terminal domain and to actin via spectrin with the C-terminal domain (7). All three of the ERM proteins localize similarly to cortical actin structures near the plasma membrane such as microvilli, membrane ruffles, and lamellipodia (8 -12), and each can bind to the integral membrane protein CD44 via their N terminus (13,14) and to F-actin via their C terminus (15,16). The proposed actin-binding domain is highly conserved among the ERM proteins, but, importantly, it is not apparent in Merlin. Furthermore, it not known whether Merlin can bind any of the known ERM interactors nor have any novel Merlin binding partners been identified yet.
Depletion of ERM proteins by antisense oligonucleotide treatment, as well as experiments overexpressing Ezrin in insect cells, suggests a role for ERM proteins in regulating cell-substratum and cell-cell adhesion (17,18); a possible function for Merlin in regulating adhesion has also been reported (19). Additionally, studies of ERM proteins suggest they may serve to reorganize the actin cytoskeleton in response to growth factors (20). Because loss of NF2 function is associated with tumorigenesis and its localization to cortical actin (21)(22)(23), 3 we were interested in whether Merlin may be involved in responding to cell-cell contact, loss of substrate attachment, and other growth suppressive signals. Here, we examine the expression of Merlin under such conditions and find the protein is regulated in multiple ways during cellular response to microenvironmental changes. EXPERIMENTAL  bovine serum (0.2% calf serum for NIH3T3 cells) for 24 h except where indicated otherwise. For other growth arrest treatments, subconfluent cells in 10% serum were treated with 20 nM staurosporine, 750 M hydroxyurea (Sigma) for 24 h, 500 ng/ml nocodazole (Sigma) for 24 h, or 500 centigrays ␥-irradiation and then placed at 37°C for 18 h. For confluency experiments, exponentially growing cells were plated at various densities (as noted in figure legends) in fresh medium 24 h before Western lysates were prepared. All cells were cultured in 10-cm tissue culture dishes, and we determined that the saturation density of the NIH3T3 cells used on these dishes was 6 ϫ 10 6 cells per dish. All cells were incubated at 37°C under 5% CO 2 in a humidified chamber.
Immunoblotting-Total cell lysates were obtained by two different methods with the same results observed with each methodology. Cultured cells were washed once with PBS and lysed in PBS containing 2% SDS, 0.5 mM phenylmethylsulfonyl fluoride, and 1 g/ml each aprotinin, leupeptinin, and pepstatin (51); or lysed in boiling 10 mM Tris, pH 7.5, 1% SDS, 50 mM NaF, 1 mM sodium orthovanadate. Cells were then scraped with a rubber policeman, boiled for 5 min, and then sheared with a 26-gauge needle three times, and an aliquot was taken for protein concentration determination using the Bio-Rad DC-Kit (Melville, NY) or using the BCA protein determination kit (Pierce) with identical results. Sample buffer (5 ϫ) was added to the remainder of the sample.
For immunoblotting, total cell extracts were separated on a 6% SDS-polyacrylamide gel with a 4% stack and electrophoretically transferred to polyvinylidene difluoride membranes (Schleicher & Schuell). For blotting with anti-Merlin sc331, membranes were blocked for 2 h at room temperature with 5% nonfat dry milk in TBST (10 mM Tris, pH 7.5, 150 mM NaCl, 0.05% Tween 20). sc331 was diluted to a final concentration of 1 g/ml in blocking buffer and incubated overnight at 4°C. Polyclonal antibody 457 was used as described previously (12). Each was washed extensively in TBST followed by probing with horseradish peroxidase-conjugated anti-mouse or anti-rabbit secondary antibodies (Amersham Corp.) and visualized by enhanced chemiluminescence (Amersham Corp.).
Immunoprecipitation-For methionine labeling, 1 ϫ 10 6 cells were plated in a 100-mm dish, and the next day the cells were serum-starved (as described above) for 42 h. Cells were then preincubated with 5 ml of methionine-free DMEM (Life Technologies, Inc.) for 40 min and then labeled for 6 h with 500 Ci of [ 35 S]methionine (NEN Life Science Products). For orthophosphate labeling, serum-starved cells were preincubated with phosphate-free DMEM (Life Technologies, Inc.) and then incubated for 5 h with 1 mCi of [ 32 P]orthophosphate (HCl-free, NEN Life Science Products). Cells were rinsed in PBS and then lysed in a modified RIPA buffer (10 mM Tris, pH 7.5, 0.5% sodium deoxycholate, 0.5% Nonidet P-40, 5 mM sodium orthovanadate, 50 mM NaF, 150 mM NaCl, 2 mM EDTA, 0.1% SDS, 1 mM phenylmethylsulfonyl fluoride, 5 g/ml each of aprotinin, pepstatin, and leupeptin). Lysed cells were rocked at 4°C for 30 min, then scraped and cleared by centrifugation at top speed in an Eppendorf centrifuge for 15 min. Supernatants were preincubated with irrelevant rabbit IgG in 50% protein A-Sepharose beads (Pierce) at 4°C for 2 h and then centrifuged for 1 min in an Eppendorf microcentrifuge. The resulting supernatant was used for immunoprecipitation with 0.5 g of sc331 or sc332 antibody (per sample) overnight at 4°C. 50 l of protein A-Sepharose was added and rocked 2 h further at 4°C. Immunocomplexes were recovered by centifugation and washed three times in RIPA buffer; sample buffer was added, and samples were boiled 5 min. Equal number of trichloroacetic acid-precipitable counts were immunoprecipitated for each sample within a given experiment. Samples were electrophoresed on SDSpolyacrylamide gel electrophoresis as described above.
Flow Cytometry-Flow cytometry analysis was performed using a FACScan (Becton Dickinson). Cells were trypsinized, washed in media, rinsed in PBS, and fixed with cold 95% ethanol. Cells were pelleted and resuspended in 20 g/ml propidium iodide and 200 g/ml RNase A in PBS and then incubated at 37°C for 30 min and placed at 4°C overnight. A total of 10,000 cells were analyzed for each arrest sample per experiment.
Phosphoamino Acid Analysis-The phosphoamino acid composition was analyzed essentially as described (52). Briefly, [ 32 P]orthophosphate-labeled immunoprecipitates were separated on a 6% SDS-polyacrylamide electrophoresis gel and transferred to polyvinylidene difluoride as described. Bands containing 32 P-labeled Merlin were located by autoradiography, excised, and subjected to direct hydrolysis in 100 l of 6 N HCl at 100°C for 1 h. Supernatants were lyophilized and resuspended in distilled H 2 0 and spotted onto plastic-backed cellulose thin layer chromatography plates (EM Sciences, Gibbstown, NJ). A mixture of 3 mg/ml phosphoserine, phosphothreonine, and phosphotyrosine (Sigma) containing a trace of phenol red was spotted on top of each sample. Two-dimensional high voltage TLC electrophoresis was performed at 1000 V for 20 min each on a Multiphor II (Pharmacia) electrophoretic apparatus. Buffers and visualization of standards were as described (50).
Adhesion Assays-For adhesion assays, confluent NIH3T3 cells were serum-starved in DMEM (0% calf serum) for 18 h, washed with PBS, and trypsinized. Cells were then washed twice in DMEM containing 0.5% soybean trypsin inhibitor and 0.1% bovine serum albumin (essentially fatty acid-free, Sigma) and suspended in DMEM for 30 min at 37°C. Samples were plated on fibronectin (Biocoat dishes, Becton Dickinson) or tissue culture plastic plates for 15-60 min, except where Similarly for experiments in presence of serum, calf serum was added to 10% in DMEM at time of plating. All plates were washed with PBS prior to making protein extracts to ensure that only adherent cells were examined.

RESULTS
Endogenous Merlin Is Differentially Phosphorylated-To begin to investigate the expression and post-translational modification of Merlin, we characterized the ability of multiple antisera to recognize endogenous Merlin by immunoprecipitation and immunoblotting of total cell lysates. As shown in Fig.  1A, Western blot analysis revealed that in mouse embryonic stem (ES) cells the endogenous protein migrated as a distinct doublet around 70 kDa. The identity of these species as Merlin was confirmed by using multiple antisera, cells overexpressing hemagglutinin-tagged Merlin, Nf2-deficient cells derived from tumors from Nf2-heterozygous mutant mice, and ES cells bearing a homozygous disruption of the Nf2 gene ( Fig. 1A; Ref. 24).
As Merlin had been previously reported to be a phosphoprotein (25), we were interested in whether the nature of the two migrating species was due to phosphorylation. We first characterized the ability of three polyclonal antisera (directed at three distinct peptide epitopes; see "Experimental Procedures") to immunoprecipitate Merlin. Because preliminary experiments indicated that these antisera were more efficient in immunoprecipitating Merlin from human as opposed to mouse cells, human osteosarcoma U2OS cells were employed. As shown in Fig. 1B, Merlin was immunoprecipitated as a doublet around 70 kDa in [ 35 S]methionine metabolically labeled U2OS cell lysates by all three antibodies. We also detected two species in immunoprecipitates from mouse, although the two species of human Merlin migrate more closely together (also observed in immunoblot analysis, not shown). Importantly, no immunoreactivity was observed in Merlin-deficient tumor cells (Fig. 1B), confirming the specificity of these antibodies.
To address directly the nature of the difference between the two Merlin isoforms, lysates from U2OS cells were immunoprecipitated with the sc331 antibody, and aliquots were treated with calf intestinal phosphatase (CIP). As shown in Fig. 1C, CIP treatment eliminated the slower migrating species (lane 3), whereas CIP buffer alone had no effect (lane 2). More significantly, immunoprecipitation of Merlin from parallel dishes of [ 32 P]orthophosphate-labeled cells yielded a single species, which migrated with the same mobility as the slower form in the 35 S-labeled immunoprecipitates (Fig. 1C, lane 4). A faster migrating species was never observed in any 32 P-labeled immunoprecipitates, even upon exposure times 200 times that shown in lane 4 of Fig. 1C. These data demonstrate that the two Merlin species are differentially phosphorylated, with the slower migrating form representing a phosphorylated species and the faster migrating form most likely a fully unphosphorylated form.
Up-regulation of Merlin by Serum Starvation and Confluency-Given Merlin's localization to the cell periphery and its status as tumor suppressor, we were interested in whether the regulation of the protein might be affected by cell-cell contact, which can result in growth inhibition of untransformed cells, as first described in NIH3T3 cells (26). NIH3T3 cells were plated at increasing degrees of confluency ( Fig. 2A), and Merlin levels were examined by immunoblotting of total cell lysates normalized for total protein levels. As shown in Fig. 2B, lanes 1-5, with increasing confluency, the levels of both the phosphorylated and unphosphorylated Merlin species increased, with the faster migrating, unphosphorylated form of the protein being most prominent when the majority of the cells in the population had established cell-cell contact ( Fig. 2A). To examine the specificity of the response, we then determined whether other cell cycle arrest-inducing agents might up-regulate Merlin levels. We first utilized serum starvation to induce quiescence in subconfluent NIH3T3 cells (same density as in Fig. 2A, 2nd panel) and, once again, observed the up-regulation of both phosphorylated and unphosphorylated Merlin species to approximately equimolar amounts (lane 7). Strikingly, restimulation of the serum-starved cells resulted in a rapid disappearance of the unphosphorylated form of the protein as follows: within 5-15 min following serum readdition, only a fraction of the faster migrating form remained (Fig. 2B, lanes 8 and 9). The level of the phosphorylated, slower migratory species also appeared to decrease modestly following restimulation with 10% serum (Fig. 2B, lanes 7-9). These results were confirmed using a second polyclonal antibody (1398) (data not shown).
Given that both serum starvation and confluency cause an arrest in the G 0 /G 1 phase of the cell cycle, we were interested in whether Merlin levels or phosphorylation status might be affected by growth arrest stimuli generally, or only G 1 -specific arrests, or if this response was specific to serum deprivation and confluency. To address this question, duplicate sets of sparsely plated NIH3T3 cells were treated with a variety of growth-suppressive agents or were left untreated for 24 h. One set was then subjected to cell cycle characterization utilizing propidium iodide labeling and fluorescent activated cell sorter analysis (Fig. 3A). Total cell extracts were made from the other set of plates, equilibrated for total protein, and immunoblotted as before (Fig. 3B). While serum deprivation for 24 h led to a significant increase in protein levels, no up-regulation of phosphorylated or unphosphorylated Merlin was observed in cells treated with another G 1 arresting agent, staurosporine (Fig.  3B, 3rd lane) (27). Furthermore, no up-regulation of Merlin was observed in cells treated with ␥-irradiation, nocodazole, or hydroxyurea (Fig. 3B, 4th and 5th lanes; data not shown). As shown in Fig. 3A, hydroxyurea and nocodazole led to early S phase and G 2 /M arrest, respectively. Similar results were obtained with U2OS cells (data not shown). Thus, the changes in Merlin expression associated with confluency and serum deprivation appear to be specific and not a secondary consequence of a G 1 phase or more general cell cycle arrest.
To investigate whether this was a general property of Merlin regulation or a phenomenon specific to NIH3T3 cells, we examined many cell types from different species, including murine embryonic and Rat1 fibroblasts, normal rat kidney (NRK), and human osteosarcoma (MG63 and U20S) cells ( Fig. 4A and data not shown). In all cell types examined, both confluency and serum starvation were associated with the presence of the two migrating forms of Merlin, of which the faster species was diminished upon addition of serum or in low density cells. To examine whether this form of regulation was specific to Merlin, we stripped and reprobed the anti-Merlin immunoblot from Fig. 4B (of NIH3T3 total cell lysates) with an anti-Radixin polyclonal antibody (12). Radixin was also up-regulated by confluency but did not migrate as two distinct species (Fig. 4B). We determined that there was no cross-reactivity of the Radixin antibody for Merlin. Serum starvation/restimulation did not affect Radixin levels or mobility (data not shown). These same lysates were run in parallel and probed with an anti-c-src antibody, which confirmed that equivalent amounts of protein were loaded in each lane (Fig. 4B).
To determine whether the increase in Merlin levels at higher cell confluency might just be due to exhaustion of critical serum factors in the media, we examined whether treating cells plated at high density with fresh serum might reduce Merlin levels. As shown in Fig. 5, cells plated at near-saturation density showed a preferential loss of the unphosphorylated form of Merlin when treated with serum (compare lanes 3 and 4 of Fig.  5). Importantly, however, the levels of Merlin after serum treatment were still significantly higher than the level in more sparsely plated cells (compare lane 4 to lanes 1 and 2 of Fig. 5). These results suggest that confluency and serum deprivation synergize in their up-regulation of Merlin. Consistent with this idea, serum-starvation of the densely plated cells resulted in even further up-regulation of Merlin levels (compare lanes 3  and 5, Fig. 5). On the contrary, however, cells plated at saturation density (6 ϫ 10 6 for our NIH3T3 cells on a 10-cm dish) did not show a loss of Merlin levels when treated with serum, arguing that at the highest cell densities, these cells were no longer responsive to the serum factors that induce down-regulation of Merlin (Fig. 5, lanes 6 and 7).
Merlin Is Phosphorylated on Serine and Threonine-To identify which residues are phosphorylated in these cells and to determine if the loss of the unphosphorylated form of Merlin following serum treatment results in the appearance of a novel phosphorylated form, we performed phosphoamino acid analysis on serum-starved and serum-treated U2OS cells. As shown in Fig. 6, Merlin is phosphorylated on both serine and threonine residues but apparently not on tyrosine residues; moreover, the ratio of serine to threonine phosphorylation was not affected by serum treatment.
Merlin Phosphorylation Is Also Regulated by Loss of Adhesion and Cell Spreading-Antisense oligonucleotide experiments have suggested a role for Merlin and the other ERM proteins in regulating cell adhesion (17,19). Given our results with serum deprivation and confluency, we were interested in whether loss of adhesion might also modulate Merlin protein levels or phosphorylation. To test this idea, total cell lysates were made from serum-starved, confluent NIH3T3 cells before or after trypsinization and placement in suspension in the absence of serum for 30 min (28). As shown in Fig. 7A, placement of the cells in suspension resulted in a rapid loss of the phosphorylated form of Merlin and a concomitant increase in the level of unphosphorylated Merlin. Parallel cultures kept in suspension for 30 min were then replated on tissue culture plastic plates in the absence of serum to address whether adhesion might be capable of re-inducing Merlin phosphorylation. Indeed, upon replating, the slower migrating species of Merlin returned, with the ratio of phosphorylated to unphosphorylated reaching nearly one to one at 1 h post-plating (Fig.  7A). It should be noted for all of these experiments that the plates were washed prior to preparing total cell lysates, so only adherent cells were collected.
To examine further a possible connection between Merlin regulation and cell shape, cells in suspension were plated on fibronectin in the absence or presence of CCD, which disrupts the actin cytoskeleton. CCD treatment allows for cell attachment but prevents cell spreading (29). Plating of cells on fibronectin resulted in clear phosphorylation of Merlin when compared with suspended cells, but this was prevented when cells were plated in the presence of 1 M CCD (Fig. 7B). Microscopic analysis revealed that the CCD-treated cells were well attached but not spread, as described previously (29). Given the ability of cell spreading to induce Merlin phosphorylation in the absence of serum, we were interested in whether serum factors could have this effect alone on cells in suspension and whether they might synergize with cell spreading to induce Merlin phosphorylation. As shown in Fig. 7B, treatment of suspended cells with serum did not result in a significant phosphorylation of Merlin. However, when combined with plating (on fibronectin or tissue culture plastic plates), serum addition led to an enhancement of the phosphorylation of Merlin (Fig. 7B). These results suggest that serum factors and cell spreading synergize in regulating the ratio of phosphorylated to unphosphorylated Merlin but that a cell spreading or shapespecific signal is needed in order for Merlin to be responsive to growth factor-mediated phosphorylation.

DISCUSSION
Neurofibromatosis type 2 is a severe inherited cancer disorder that predisposes individuals to bilateral vestibular schwannomas, as well as meningiomas and spinal schwannomas. In 1993, two groups used positional cloning to isolate the NF2 gene, which has since been found to be mutated in sporadically occurring schwannomas, meningiomas, and malignant mesotheliomas (2). Further supporting a broad role for NF2 in tumor suppression, mice heterozygous for a targeted mutation at the Nf2 locus are predisposed to a wide range of tumor types. 4 The NF2-encoded protein, Merlin, is structurally related to a group of proposed membrane/cytoskeletal linker proteins, particularly the ERM proteins. However, Merlin has not been shown to be functionally related to these proteins. Moreover, it remains unclear how such a protein may serve to regulate proliferation or how its loss might contribute to tumorigenesis. Key in the early understanding of other important tumor suppressors were studies that examined the expression and posttranslational modification of these proteins (e.g. cell cycle-dependent phosphorylation of the retinoblastoma protein (30 -32) and up-regulation of p53 by DNA damaging agents (33)), which ultimately proved instrumental in pointing the way toward understanding the function of those genes. We and others (21,23) 3 have demonstrated that Merlin localizes similarly to the ERM proteins at the membrane/actin interface, with Merlin being particularly enriched in membrane ruffles. Given the localization of Merlin to regions of dynamic actin at the cell periphery, and its role as a tumor suppressor, we were interested in whether growth inhibition associated with cell-cell contact might cause a change in the pattern of Merlin expression. We observed that the level of endogenous Merlin protein is increased by confluency as well as by serum deprivation and that the two stimuli can synergize in their up-regulation of Merlin levels. Furthermore, we have discovered that these stimuli are tightly associated with the appearance of a novel, unphosphorylated species of Merlin. This unphosphorylated form disappears rapidly (Ͻ5 min) following serum addition to serum-starved cells, suggesting a tight post-translational regulation of this species leading to its rapid degradation by growth factor-activated pathways. The up-regulation of Merlin was not observed in cells arrested in G 1 by staurosporine or in cells arrested in S or G 2 /M induced by hydroxyurea or nocodazole, respectively, suggesting that the effects of cell-cell contact and serum deprivation stimuli are specific.
Merlin appears to be constitutively phosphorylated on both serine and threonine residues, and no obvious changes in patterns of phosphorylation were discerned following serum treatment or in confluent versus subconfluent conditions (data not shown). The protein was previously reported to be phosphorylated specifically on serine in human U251-MG glioblastoma cells (25). Given the abundant threonine phosphorylation observed in our experiments, it is not clear whether there may be a cell type-specific difference or that the threonine phosphorylation was below the level of detection in the previous experiments. The fact that we detect threonine phosphorylation is of particular interest as a germ line missense mutation of Thr-352 of Merlin has been detected in a severely affected NF2 patient (34), and this residue bears a possible consensus recognition sequence for protein kinase C (3,35). All three ERM proteins have been shown to be serine/threonine, as well as tyrosine, phosphorylated in response to specific growth factors in various cellular contexts (20,36,37), although Merlin lacks almost all FIG. 6. Merlin is constitutively phosphorylated on serine and threonine residues. Phosphoamino acid analysis from serum-starved and serum-stimulated U2OS cells, performed as described under "Experimental Procedures," is shown. Positions of phosphoserine (P-SER), phosphothreonine (P-THR), and phosphotyrosine (P-TYR) standards are indicated.

FIG. 7. Merlin phosphorylation is regulated by cell spreading, which synergizes with growth factor-induced phosphorylation.
A, sc331 immunoblot of confluent NIH3T3 cells, serum-starved for 18 h, placed in suspension for 30 min, and then replated on tissue culture plastic plates. Total cell lysates were made prior to suspension (ss ϩ confl), after 30 min of suspension (sus), or from cells 15, 30, or 60 min after replating. Replated cells were washed with PBS to remove nonadherent cells, and then total cell lysates were prepared. Note: 2nd lane is overloaded. B, sc331 immunoblot of confluent, starved NIH3T3 cells kept in suspension or plated on fibronectin (FN) or tissue culture plastic plates (plast) in the absence or presence of 10% calf serum (ϩ ser) or 1 M cytochalasin D (ϩ CCD). of the previously described phosphorylation sites found in the other family members. Furthermore, there was no increase in the level of phosphorylation on Merlin following serum treatment (data not shown); it remains to be determined whether individual growth factors might induce phosphorylation of Merlin.
In addition to the regulation of Merlin levels and phosphorylation by serum and confluency, we found that loss of adhesion leads to a rapid, nearly complete dephosphorylation of Merlin. This effect was reversed by replating the cells which led to rephosphorylation of Merlin to control levels. The phosphorylation of Merlin upon replating showed a close correlation with degree of cell spreading, and, in fact, blocking cell spreading by cytochalasin D treatment prevented the phosphorylation effect. Interestingly, experiments using antisense oligonucleotides have demonstrated that loss of ERM proteins leads to a loss of cell-cell and cell-substratum attachment (17), and a similar approach has suggested a role for Merlin in cell-substratum attachment (19). Furthermore, overexpression of Ezrin has been shown to increase the adhesion of insect cells to their substratum and to override contact inhibition of confluent NIH3T3 cells (18,38). Whether Merlin directly modulates the activity of the ERM proteins in these processes or responds to the same cues that stimulate ERM-mediated cytoskeletal reorganization, but performs a distinct function in regulating proliferation, remains to be determined.
The responsiveness of Merlin to confluency, serum deprivation, and loss of substratum attachment may be due to a common upstream signal regulating the organization of the actin cytoskeleton. It is interesting that other microfilament-associated proteins such as Vinculin, ZO-1, Gas2, and the MARCKS protein are all up-regulated upon confluency and/or serum starvation-induced arrest (39 -42), and in fact, Vinculin is regulated by cell shape as well (39). One possible component of the upstream signal could be members of the Rho subfamily of small GTPases. These proteins are known to regulate the actin cytoskeleton as well as mitogenic signaling and cell cycle progression in G 0 /G 1 in fibroblasts in response to extracellular signals (43,44). Intriguingly, the activity of RhoA in Swiss 3T3 cells is reduced by serum deprivation, confluency, and placement of cells into suspension (45)(46)(47). The ERM proteins (48) as well as Merlin (49) have been reported to colocalize with RhoA in various cellular contexts, and Rho modulates the affinity of ERM proteins for their membrane targets (14). Moreover, we have recently demonstrated that RhoA activity is necessary and sufficient for the phosphorylation and relocalization of the ERM proteins (50). Perhaps the phosphorylation and/or levels of Merlin are responsive to changes in RhoA activity as well, or alternatively, Merlin may serve to downmodulate the activity of RhoA or a related GTPase in response to these environmental signals.
Although the function of Merlin is still largely unknown, the data presented here represent a first step toward understanding what stimuli may regulate its activity. Taken together, our results suggest Merlin may serve its tumor suppressor role in the response of cells to loss of cell-substratum attachment, increased cell-cell contact, or poor growth factor microenvironment, perhaps preventing inappropriate cell-cycle entry under these circumstances. Furthermore, the unphosphorylated form of the protein may represent the active tumor suppressor conformation. Given that the protein is up-regulated by some growth arrest stimuli, it is possible that Merlin participates in these arrest responses and that loss of Merlin function might lead to loss of growth arrest capacity under certain conditions and, ultimately, to tumor formation. An interesting test of these ideas would be to examine the ability of cells deficient for Merlin to arrest under conditions of confluency, serum deprivation, or growth in suspension.