β1 integrin–dependent Rac/group I PAK signaling mediates YAP activation of Yes-associated protein 1 (YAP1) via NF2/merlin

Cell adhesion to the extracellular matrix or to surrounding cells plays a key role in cell proliferation and differentiation and is critical for proper tissue homeostasis. An important pathway in adhesion-dependent cell proliferation is the Hippo signaling cascade, which is coregulated by the transcription factors Yes-associated protein 1 (YAP1) and transcriptional coactivator with PDZ-binding motif (TAZ). However, how cells integrate extracellular information at the molecular level to regulate YAP1's nuclear localization is still puzzling. Herein, we investigated the role of β1 integrins in regulating this process. We found that β1 integrin–dependent cell adhesion is critical for supporting cell proliferation in mesenchymal cells both in vivo and in vitro. β1 integrin–dependent cell adhesion relied on the relocation of YAP1 to the nucleus after the down-regulation of its phosphorylated state mediated by large tumor suppressor gene 1 and 2 (LATS1/2). We also found that this phenotype relies on β1 integrin–dependent local activation of the small GTPase RAC1 at the plasma membrane to control the activity of P21 (RAC1)-activated kinase (PAK) of group 1. We further report that the regulatory protein merlin (neurofibromin 2, NF2) interacts with both YAP1 and LATS1/2 via its C-terminal moiety and FERM domain, respectively. PAK1-mediated merlin phosphorylation on Ser-518 reduced merlin's interactions with both LATS1/2 and YAP1, resulting in YAP1 dephosphorylation and nuclear shuttling. Our results highlight RAC/PAK1 as major players in YAP1 regulation triggered by cell adhesion.

homeostasis (1). However, deregulation of this process often contributes to pathological disorders such as tumor formation, growth, and metastasis (2), exemplified by one of the hallmarks of cell transformation, the anchorage-independent growth (3,4).
Hippo signaling was identified as an important regulatory pathway that restricts cell proliferation, thereby controlling organ size and morphogenesis (5,6). This is achieved mainly through the control of two transcriptional coactivators: Yesassociated protein (YAP1) and transcriptional coactivator with PDZ-binding motif (TAZ/WWTR1). Upon restrictive proliferative conditions, these molecules are phosphorylated by the products of the large tumor suppressor gene 1 and 2 (LATS1/2). This phosphorylation creates a binding site for 14-3-3 proteins and promotes YAP/TAZ sequestration in the cytoplasm. As a consequence, phosphorylated forms of YAP/TAZ are sequestered in the cytoplasm (7,8).
Although originally described as the main switch to block cell proliferation when confluency is reached, it has become clear that the Hippo signaling pathway is integrating several inputs such as cell density, cell geometry, matrix stiffness, metabolic status, and serum composition (9). Indeed, activation of YAP/ TAZ-mediated transcriptions, which correlates with their nuclear localization, is tightly controlled by cell-matrix adhesion (10,11). This process explains why cells are dependent on matrix adhesion for a full mitogenic response to growth factor exposure (12). Loss of cell adhesion to the extracellular matrix is known to induce an increase in cyclic AMP (cAMP), which is correlated with the inhibition of mitogenic signaling (13). Because YAP/TAZ activities are also inhibited by cAMP, this rise in cAMP was proposed as the main factor to regulate YAP/ TAZ upon cell detachment, potentially mediated though downstream players such as RHOA or LATS1/2 (10,14). In contrast, cell adhesion to cell-matrix proteins was shown to trigger YAP nuclear localization through an integrin/SRC axis (11). Despite these findings, the molecular mechanism downstream of FAK 3 /SRC to control subcellular YAP/TAZ localization remains ill-defined and has never been directly investigated.
Initially, RHOA was identified as a critical regulator of YAP; however, more recently RAC1 and CDC42 were also found to be involved in its regulation (16 -18). So far, how small RHOA family GTPases regulate YAP nuclear translocation remains elusive. Knowing that integrins are key regulators of this GTPase class, we wondered whether ␤1 integrins might regulate YAP by controlling RHOA GTPases.
Herein, we address the mechanistic role of ␤1 integrins in the regulation of YAP localization and thereby cell proliferation. We found that ␤1 integrin-dependent cell adhesion was critical for supporting cell proliferation in mesenchymal cells both in vivo and in vitro by controlling YAP signaling rather than MAPK cascade. Mechanistically, we showed that ␤1 integrins are required for localizing the GTPase RAC1 at plasma membrane extensions. There, RAC1 activates its effector PAK1 and initiates in a merlin-dependent manner the nuclear translocation of YAP. Indeed, we found that merlin binds LATS via its FERM N-terminal domain but also interacts with YAP with its C-terminal moiety. The interactions between merlin and YAP or LATS are down-regulated upon phosphorylation by PAK1 at Ser-518. Altogether our data revealed a novel signaling pathway orchestrated by ␤1 integrins to locally activate a RAC/PAK1 cascade and negatively regulate the inhibitory protein merlin.

␤1 integrins regulate mesenchymal cell proliferation in a MAPK-independent manner
To explore the function of ␤1 integrins in bone tissue, we inactivated the ␤1 integrin gene in osteoblasts using Osterixdriven Cre recombinase expression. Mice with an osteoblastspecific ␤1-integrin deletion survived to adulthood but suffered from a growth deficit along with a significant decrease in the absolute number of osteoblasts (Fig. 1A). Because ␤1 integrins are known to regulate cell proliferation, we wondered whether the reduced osteoblast numbers observed could be due to a reduced proliferative capability of those cells. Although TUNEL staining did not reveal any significant difference in apoptotic cell number (Fig. 1, B and C), a significant reduction in BrdU incorporation was observed in mutant animals (Fig. 1, D  and E). Similarly, in vitro, the loss of ␤1 integrins in isolated osteoblasts resulted in a significant proliferation defect (Fig.  1F). To rule out any osteoblast-specific phenotypes, we isolated mouse embryonic fibroblasts (MEFs) and confirmed that the loss of ␤1 integrin expression was associated with a reduced proliferative capability of the cells (Fig. 1G). Although it was proposed that integrins are important regulators of ERK signaling (19), we could not rescue the proliferation defect of ␤1 integrin-deficient cells by activating the MAPK/ERK pathway (Fig. 1H). In addition, we did not detect any significant modification of ERK phosphorylation when ␤1 integrin-deficient cells were compared with wild-type cells (data not shown).

␤1 integrins are required for YAP nuclear localization and cell proliferation
YAP-dependent gene expression has emerged as an important pathway regulating cell proliferation (20). Moreover, it was recently reported that YAP nuclear localization is controlled in a cell adhesion manner through integrins and SRC/FAK (21); therefore, we first asked whether the loss of ␤1 integrin expression was indeed associated with a defect in YAP nuclear localization, which might account for the reduced proliferation observed in ␤1-deficient cells.
When compared with wild-type cells that displayed a prominent YAP/TAZ nuclear localization, the lack of ␤1 integrins was correlated with a strong relocation of these proteins within the cytoplasm (Fig. 2, A and B). This observation was confirmed with the other clones analyzed (Fig. 2C), as well as in cell lines stably expressing a flag-tagged YAP (Fig. 2D). To further evaluate the involvement of the ERK/MAPK pathway in controlling YAP nuclear localization downstream of ␤1 integrins, we isolated MEF cells from mice bearing a constitutively active allele of K-Ras (K-Ras G12D ) (22) with one or two deleted alleles of the ␤1 integrin gene. Even in the presence of the activated allele of K-Ras, YAP was mainly cytoplasmic as soon as the ␤1 integrin chain was genetically ablated (Fig. 2E). Notably, YAP/TAZ nuclear localization was completely restored upon re-expression of the human ␤1 integrin subunit in ␤1-deficient mouse cells, showing a direct relationship between ␤1 expression and the nuclear localization of YAP/TAZ (Fig. 2, A and B). In line with these observations, biochemical fractionations revealed a decrease in the nuclear pool of YAP in ␤1-deficient cells (Fig.  2F). Along with the reduced level of nuclear YAP, the expression of its target genes was also down-regulated in ␤1-deficient cells (Fig. 3A). YAP was shown to be sequestered in the cytoplasm by 14-3-3 proteins after the activation of its upstream protein kinase LATS (7). In good agreement with this, we noticed that the phosphorylation of YAP on Ser-127 (required for 14-3-3 binding) increased upon the deletion of ␤1 integrins as well as the ratio of activated (phosphorylated) versus total form of LATS (Fig. 3B). Therefore, taken together, these data confirmed that ␤1 integrins are important players in controlling YAP nuclear localization most likely in a LATS-dependent manner.
Having shown that ␤1 integrins regulate nuclear localization of YAP, we wondered whether the proliferation defect that we observed upon removal of ␤1 integrins was the consequence of YAP nuclear translocation. Our data highlighted that LATS-dependent phosphorylation of YAP was up-regulated in ␤1-deficient cells. Therefore, we stably expressed in those cells a nonphosphorylatable form of YAP. The expression of YAP 5SA in ␤1 integrin-null cells relocated YAP into the nucleus and up-regulated its target genes (Fig. 3, C and D). Importantly, the expression of YAP 5SA fully restored ␤1 Ϫ/Ϫ cell proliferation capabilities as quantified by BrdU incorporation (Fig. 3E). Therefore, our data highlighted the important role of YAP signaling in the control of cell proliferation downstream of ␤1 integrins.
Finally, to confirm this view, we performed unbiased transcriptomic analyses on wild-type and ␤1 integrin-deficient cells under optimal growth conditions. With the defined filtering and statistical criterion, 800 probes representing 555 well-annotated genes were identified as being differentially expressed between the two considered groups (supplemental Fig. S1). Known YAP/TEAD target genes were significantly down-regulated in ␤1 Ϫ/Ϫ cells. Among those were ankrd1, ctgf, and cyr61 Cell adhesion control of YAP activation (supplemental Fig. S2). These results confirmed our previous RT-qPCR analyses (Fig. 3, A and D). Consistent with the abovementioned proliferation defect, a number of cell-cycle regulators were also deregulated (supplemental Fig. S1). To directly estimate whether the YAP-TEAD complex might impact the expression of those important cell-cycle regulators, we analyzed p19 Arf , p21 CIP , and cyclin D2 expression by RT-qPCR in cell expressing YAP 5SA . In contrast to what was measured in ␤1-deficient cells, the expression of YAP 5SA in wild-type cells was able to up-regulate cyclin D2 and down-regulate p19 Arf but not p21 CIP when compared with control cells (Fig. 3E). This strongly suggests that some important cell-cycle regulators are transcriptionally modified by the YAP-TEAD complex.

RAC1, but not CDC42 nor RHOA, controls YAP nuclear localization downstream of ␤1 integrins
Although accumulating evidence pinpointed a role for FAK and SRC in integrin-dependent control of the Hippo pathway (21), little is known about the downstream effectors. However, members of the Rho GTPase family were shown to be involved in the control of YAP activity (10,23,24). It is well known that integrins also play a critical role in A, histomorphometric analysis of osteoblast number on wild-type (␤1 f/f ) and Osx-Cre;␤1 f/f (␤1 Ost-KO ) 30-day-old mouse tibias. Graphs show the mean Ϯ S.D. from five independent experiments. B, quantification of apoptotic (TUNELpositive) and proliferating (BrdU-positive) cells in periosteum and trabecular bone in wild-type and mutant 30-day-old mouse tibias. n ϭ 50. Statistical significance of differences was assessed by a two-tailed unpaired Student's t test for three independent experiments. C, representative TUNEL staining. D, BrdU staining on trabecular bone sections from wild-type and mutant mouse tibias (hc, hypertrophic cartilage; tb, trabecular bone; bm, bone marrow). Scale bar, 40 m. E, images of BrdU staining of trabecular bone sections. F, BrdU-based quantification of the proliferation rate of ␤1 f/f , ␤1 Ϫ/Ϫ , primary mouse embryonic fibroblasts. (statistical significance of differences assessed by a two-tailed unpaired Student's t test, three independent experiments). G, in vitro proliferation rate of wild-type (␤1 f/f ) and ␤1 integrin-deficient (␤1 Ϫ/Ϫ ) osteoblasts. n ϭ 50. Statistical significance of differences was assessed by a two-tailed unpaired Student's t test for three independent experiments. H, BrdU-based quantification of the proliferation rate of ␤1 f/f and ␤1 Ϫ/Ϫ osteoblasts or the ␤1 Ϫ/Ϫ osteoblasts expressing human ␤1 integrin (rescue), constitutively active MEK (MEKQ56P), or nuclear-active ERK fusion mutant (MEK/ERKLA). Statistical significance of differences was assessed by a two-tailed unpaired Student's t test for three independent experiments.    . Results are from four independent experiments. Statistical significance of differences was assessed by a two-tailed unpaired Student's t test. B, analysis of YAP and LATS phosphorylation (YAP pSer-127 and LATS1/2 pSer-909 ), and total YAP and LATS in ␤1 f/f , ␤1 Ϫ/Ϫ , and ␤1 resc (rescued) osteoblasts. Actin is shown as loading control. C, immunostaining of YAP on ␤1 f/f and ␤1 Ϫ/Ϫ osteoblasts expressing or not the YAP5SA mutant. Scale bar, 10 m. D, RT-qPCR analysis of Cyr61 and CTGF mRNA in ␤1 f/f and ␤1 Ϫ/Ϫ osteoblasts and ␤1 f/f and ␤1 Ϫ/Ϫ osteoblasts expressing FLAG-YAP5SA. Statistical significance of differences was assessed by a two-tailed unpaired Student's t test for three independent experiments. E, BrdU-based quantification of the proliferation rate of ␤1 f/f and ␤1 Ϫ/Ϫ osteoblasts expressing or not the YAP 5SA mutant. n ϭ 30. Statistical significance of differences was assessed by a two-tailed unpaired Student's t test for three independent experiments. F, RT-qPCR analysis of cyclinD1 (CCND1), cyclinD2 (CCND2), p19 Arf , and p21 CIP (CDKN1A) mRNAs in ␤1 Ϫ/Ϫ and ␤1 f/f osteoblasts expressing FLAG-YAP 5SA normalized to ␤1 f/f osteoblasts (set to 1, green line). Statistical significance of differences was assessed by a two-tailed unpaired Student's t test for three independent experiments.

Cell adhesion control of YAP activation
the activation or the coupling of Rho GTPases with their effectors. Therefore, we wondered whether YAP nuclear localization driven by ␤1 integrins might also be regulated by Rho GTPases and, if so, which one.
To address this question, we used ␤1-deficient osteoblasts that displayed a dramatic defect in YAP nuclear localization to generate stable cell lines expressing constitutively activated forms of RHOA, RAC1, and CDC42. Then, we analyzed which  , and YAP subcellular localization was analyzed. Statistical significance of differences was assessed by a two-tailed unpaired Student's t test, and the box plot is representative of two independent experiments. C, GST-Rhotekin pulldown assay was performed on ␤1 f/f and ␤1 Ϫ/Ϫ cells. RHOA was analyzed by immunoblotting from the pulldown fraction (RHOA-GTP) and the input (total RHOA). D, ␤1 f/f and ␤1 Ϫ/Ϫ cells were transiently transfected with pYFP-RHOA construct, and 48-h cells were fixed and analyzed by confocal microscopy. E, statistical analysis of YAP nuclear to cytoplasm ratio, n Ͼ50 cells, represented in a logarithmic scale. Control (Ctl) ␤1 f/f and ␤1 Ϫ/Ϫ cells or ␤1 Ϫ/Ϫ osteoblasts stably expressing RAC G12V or CDC42 G12V were spread on fibronectin (10 g/ml), and YAP subcellular localization was analyzed. Statistical significance of differences was assessed by a two-tailed unpaired Student's t test, and the box plot is representative of three independent experiments. F, statistical analysis of cell spreading. The projected areas of ␤1 f/f and ␤1 Ϫ/Ϫ stably expressing Rac G12V cells spread on fibronectin (10 g/ml) for 2 h were estimated after labeling of the cell with Vybrant TM Dil and thresholding the image to fit the mask size to the cell geometry. Measurements were performed with Metamorph software. n Ͼ50 cells; statistical significance of differences was assessed by a two-tailed unpaired Student's t test, and the box plot is representative of two independent experiments.

Cell adhesion control of YAP activation
Rho GTPase was able to induce the relocation of YAP within the nucleus. Surprisingly, activated RHOA was unable to restore the nuclear localization of YAP (Fig. 4, A and B). In line with this observation, we neither observed any significant difference in RHOA-GTP levels in wild-type versus ␤1-deficient cells nor any change in its cellular distribution (Fig. 4, C and D).
In sharp contrast, RAC G12V expression (and to a lower extent CDC42 G12V ) was associated with a significant increase in YAP nuclear localization in ␤1 Ϫ/Ϫ cells (Fig. 4, A and E). Importantly, RAC G12V expression restored YAP nuclear localization in ␤1-deficient cells but not their spreading defect (Fig. 4F). This strongly suggested that RAC1 and CDC42 might act downstream of ␤1 integrins in the signaling pathway that regulates YAP nuclear localization. Because the expression of constitutively activated CDC42 and RAC1 could activate common effectors such as PAK, we wondered whether both small GTPase proteins were physiologically involved in this regulation. Hence, to discriminate between RAC1 and CDC42, we specifically inhibited their activities and analyzed YAP subcellular localization. The expression of a dominant-negative form of CDC42 (CDC42 Asn-17 ) did not result in any significant YAP redistribution in wild-type cells ( Fig. 5A and supplemental Fig. S3). In sharp contrast, the inhibition of RAC1 activity either pharmacologically with ETH1864 or by the expression of a dominant-negative form (RAC Asn-17 ) led to a significant YAP redistribution to the cytoplasm ( Fig. 5A and supplemental Fig. S3).
As our previous data pointed out for a major role of RAC1, we next analyzed whether the expression of constitutively active RAC1 could reverse the proliferation defect observed in ␤1-deficient cells. Indeed, RAC G12V expression in ␤1-deficient cells restored proliferation up to the level of control cells (Fig. 5B). This suggested that both RAC1 and YAP were in a common pathway to regulate ␤1-dependent cell proliferation. Indeed, the expression of RAC G12V in ␤1-deficient cells induced a decrease in YAP phosphorylation to a comparable level to the one observed in ␤1 f/f cells (Fig. 5C). From these results, we concluded that RAC1 is involved in YAP signaling upon ␤1-dependent cell adhesion. Surprisingly, quantification of RAC GTP levels in ␤1 f/f and ␤1 Ϫ/Ϫ in whole-cell lysates did not reveal any significant difference (Fig. 5D). This discrepancy suggested that an altered RAC1 localization, rather than a global defect in its activity per se, resulted in YAP mislocalization observed in ␤1-deficient cells. Supporting this hypothesis, it was reported that cell adhesion regulates RAC1 plasma membrane localization (25), although the role of ␤1 integrin in this process was not addressed. To address this question directly, we performed immunostaining to analyze endogenous RAC1 localization in ␤1 f/f and ␤1 Ϫ/Ϫ cells. As expected, in wild-type cells RAC1 frequently accumulated at protrusive cell edges; however, this localization was strongly reduced upon ␤1 removal (Fig. 5, E and F). To orchestrate actin dynamics, it was previously shown that RAC1 recruits cortactin at cell lamellipodia (26). Further supporting the defect in RAC1 membrane localization in ␤1-deficient cells, we observed that cortactin localization at cell edges was also significantly reduced in mutant cells (Fig. 5, G  and H).

PAK1, a merlin inhibitor, acts downstream of RAC1
On our search of how RAC1 could affect YAP nuclear localization, we focused our attention on its effectors, the PAK family. Indeed, PAK1/2 are activated at the plasma membrane by RAC1 and/or CDC42, where they regulate membrane dynamics (27,28). First, we asked whether impaired RAC1 targeting to the plasma membrane was translated into a defect in PAK1 activity. As expected, this activity, monitored by its phosphorylation, was reduced in ␤1-deficient cells but restored upon the re-expression of ␤1 integrins (Fig. 6A). Consistent with the defect in RAC1 recruitment to the plasma membrane, PAK membrane localization was reduced in ␤1-deficient cells but restored upon the expression of the constitutively active RAC G12V (Fig. 6B). Altogether, these results supported the view that ␤1 integrins by regulating RAC1 localization at the plasma membrane would promote the recruitment and activation of its downstream effectors such as PAK and cortactin.
Having shown that RAC1 was required for controlling YAP nuclear translocation downstream of ␤1 integrins, and that PAK1 activity was reduced in ␤1-deficient cells, we wondered whether YAP nuclear translocation was also dependent on PAK activity. We transiently expressed in both wild-type and ␤1-deficient cells the constitutively active mutant of PAK1 (PAK1 T423E ) and analyzed YAP subcellular localization. As observed previously with the expression of activated RAC1, the expression of constitutively active PAK1 also rescued the defective YAP nuclear localization in ␤1-deficient cells (Fig. 6C), whereas the pharmacological inhibition of group I PAK using IPA3 (a specific inhibitor of this class) in wild-type cells, significantly reduced it (Fig. 6D). Similarly, the expression of a dominant-negative form of PAK1 (PAK1 K299R ) resulted in the delocalization of YAP out of the nucleus (Fig. 6E). Altogether, our data strongly suggest that ␤1 integrins control YAP nuclear localization in a RAC/ PAK1-dependent manner.

SRC acts downstream of ␤1 integrin but upstream of RAC/PAK1
As mentioned previously, SRC was described to mediate YAP nuclear translocation downstream of integrins. SRC was also shown to be required for RAC1 activity (29,30). Therefore, we wondered whether the loss of ␤1 integrin expression also impaired SRC activity. Indeed, we observed that in ␤1 Ϫ/Ϫ cells SRC was not properly activated (Fig. 7A). Consistent with this observation, the expression of a constitutively activated SRC rescued ␤1-deficient cell proliferation as well as YAP nuclear accumulation (Fig. 7, B-D) and PAK1 activation (Fig. 7E). Yet, the inhibition of RAC1 was still able to block YAP nuclear translocation (Fig. 7D). Altogether, these data clearly indicated that SRC is downstream of ␤1 integrins and upstream of RAC1 in this signaling pathway.

␤1 integrins control YAP nuclear translocation through merlin
Next, we hypothesized that merlin, a known regulator of LATS1/2 (31), might be regulated by PAK1, and this might be an important step in how ␤1 integrin mediated YAP nuclear translocation. Indeed, together with PKA, PAK was described to induce merlin inhibition by phosphorylation of the Ser-518 Cortactin Fluorescence instensity (Arb.units)

Cell adhesion control of YAP activation
residue. Monitoring merlin phosphorylation by Western blotting revealed a decrease in Ser-518 phosphorylation in ␤1-deficient cells when compared with parental ␤1 f/f cells (Fig. 8A). Such modification should favor the recruitment and activation of LATS. Indeed, merlin phosphorylation at Ser-518 was proposed to stimulate the intramolecular FERM to C-terminal interaction (32). In its non-phosphorylated form, merlin adopts an open conformation in which the FERM domain interacts

Ln ([YAP]n/[YAP]c)
p<0.0001 PAK1 K299R Ctl E Figure 6. Rac1 effector PAK1 acts downstream of ␤1 integrins to control adhesion-dependent YAP localization. A, Western blot analysis of PAK1 and PAK1 pT423 in ␤1 f/f , ␤1 Ϫ/Ϫ , and ␤1 resc (rescue) osteoblasts. Actin was used as loading control. B, Western blot analysis of endogenous RAC1, GFP-RAC1 (GFP), and PAK1 in ␤1 f/f ␤1 Ϫ/Ϫ and ␤1 Ϫ/Ϫ stably expressing GFP-RAC G12V after cell fractionation to isolate total cell membranes (M) and cytoplasm (C). Tubulin and RalA were used as cytoplasmic and membrane markers, respectively. C, right panel, immunolocalization of YAP in ␤1 Ϫ/Ϫ osteoblasts transiently transfected with the constitutively active eGFP-PAK1 (PAK1 T423E ). Scale bar, 10 m. Left panel, statistical analysis of YAP nuclear to cytoplasm ratio in ␤1 Ϫ/Ϫ and ␤1 Ϫ/Ϫ transfected with activated PAK1. Cells were spread overnight on fibronectin (10 g/ml). n Ͼ20 cells, represented in a logarithmic scale; statistical significance of differences was assessed by a two-tailed unpaired Student's t test, and the box plot is representative of two independent experiments. D, statistical analysis of YAP nuclear to cytoplasm ratio, represented in a logarithmic scale. ␤1 f/f were spread on fibronectin (10 g/ml) in the absence or presence of IPA3 (PAK Inh) and then stained for YAP. n Ͼ50 cells; statistical significance of differences assessed by a two-tailed unpaired Student's t test, and the box plot is representative of two independent experiments. E, statistical analysis of YAP nuclear to cytoplasm ratio represented in a logarithmic scale. ␤1 f/f were mock-transfected (Ctl) or transfected with the dominant-negative form of PAK1 (PAK1 K299R ), and after 48 h cells were seeded on fibronectin (10 g/ml) for 1 h and processed for PAK1 and YAP immunostaining. PAK1-positive cells were selected to quantify YAP nuclear to cytoplasm ratio. n Ͼ30 cells; statistical significance of differences assessed by a two-tailed unpaired Student's t test, and the box plot is representative of two independent experiments.

Cell adhesion control of YAP activation
with LATS, and the C-terminal interacts with AMOT (another important effector of YAP signaling).
To confirm this, we performed GFP-trap experiments with several merlin domains with or without mutations. As reported by others, LATS was shown to interact with the FERM domain of merlin but not with its C-terminal moiety. Moreover, this interaction was significantly reduced with the phospho-mimetic S158D mutant (Fig. 8B). The deletion of a stretch of 7 amino acids within the FERM domain (named "blue-box") was reported to act as a dominant-negative form when expressed both in Drosophila and mammals. Of importance, LATS interaction with merlin was strongly reduced in the blue-box mutant as compared with the full-length or FERM domain of merlin (Fig. 8B). We took advantage of this mutant (named NF2 BB in this study) and generated stable cell lines expressing either a wild-type or a blue-box-mutated form of merlin. Expression of the NF2 BB mutant in ␤1-deficient cells restored YAP nuclear localization (Fig. 8, C and D) suggesting (i) that merlin was ␤1 f/f and ␤1 Ϫ/Ϫ osteoblasts were serum-starved overnight, and then serum (10%) was added to the cells. Phosphorylation of SRC and YAP as well as the total amount was analyzed by Western blotting. Actin was used as loading control. B, EdU-based quantification of the proliferation rate of ␤1 f/f and ␤1 Ϫ/Ϫ osteoblasts expressing or not (Ctl) the constitutively active form of SRC (SRC YF ). n ϭ 50; statistical significance of differences was assessed by a two-tailed unpaired Student's t test for two independent experiments. C, immunostaining of YAP (red) in ␤1 f/f and ␤1 Ϫ/Ϫ osteoblasts stably expressing or not the constitutively active form of SRC (SRC YF ). Most right panel, ␤1 Ϫ/Ϫ osteoblasts expressing SRC YF were treated with ETH1864 for 3 h prior to YAP staining. Scale bar, 10 m. D, statistical analysis of YAP nuclear to cytoplasm ratio, represented in a logarithmic scale. ␤1 f/f , ␤1 Ϫ/Ϫ , ␤1 Ϫ/Ϫ expressing a constitutively active form of SRC (SRC YF ) without (Ctl) or with ETH1864 (Rac1 inh) treatment. n Ͼ50 cells; statistical significance of differences was assessed by a two-tailed unpaired Student's t test, and the box plot is representative of two independent experiments. E, Western blot analysis of PAK1 activation. ␤1 f/f and ␤1 Ϫ/Ϫ osteoblasts expressing or not the constitutively active form of SRC (SRC YF ) were analyzed for phosphorylated PAK1 (p423, activated form) and total PAK1. Actin was used as loading control.

Cell adhesion control of YAP activation
required for YAP nuclear translocation downstream of ␤1 integrins and (ii) that the recruitment of LATS by merlin was important in this regulation. Consistent with these results, expressing the NF2 BB mutant in ␤1-deficient cells also reduced YAP phosphorylation (Fig. 8E) and restored cell proliferation (Fig. 8F) and YAP target genes (Fig. 8G). Altogether, our data strongly support the view that during cell spreading, ␤1 integrins mediated the activation of the RAC/PAK1 axis to phosphorylate and inactivate merlin. Consequently, inactivated merlin results in LATS release and thereby YAP dephosphorylation. Finally, if this assumption was correct, the inhibition of PAK1 activity should block YAP nuclear translocation in a merlin-and LATS-dependent manner. We analyzed YAP subcellular localization in IPA3-treated ␤1 f/f cells expressing or not the NF2 BB mutant. As expected, the inhibition of PAK1 significantly reduced YAP nuclear localization in control cells, although it had no significant effect in NF2 BB -expressing cells, indicating that PAK1 was involved in YAP nuclear translocation upstream of merlin (Fig. 8H). An identical result was also observed when a specific Rac1 inhibitor was used (Fig. 8H).

Merlin acts as a scaffold to bring LATS and YAP in close vicinity
Together with others, our data favor the view that the assembly of a merlin-centered inhibitory complex, including LATS and YAP within the cells, inhibits YAP activity. The presence of such a complex at plasma membrane extensions was supported by immunofluorescence staining of the cells. Indeed, we observed a clear colocalization between YAP and LATS on the one hand, and YAP and merlin/NF2 on the other hand (Fig. 9A). Next, the staining of YAP and pYAP indicated that this membrane pool of YAP was phosphorylated. Indeed the staining of phosphorylated YAP was much stronger in ␤1 integrin-deficient cells, in line with the view that plasma membrane extensions were important sites of YAP phosphorylation (Fig. 9B).
It is noteworthy that in wild-type cells, RAC1 colocalized together with YAP at the cell edges, whereas this colocalization was strongly reduced upon removal of ␤1 integrins (Fig. 9C). We wondered whether proteins recruited at cell edges in a RAC1-dependent manner could also colocalize with YAP. Cells stably expressing RFP-cortactin showed an extensive colocalization during cell spreading, showing that YAP is enriched in a protrusive membrane region and suggested a proximity between a RAC1-based signaling with YAP (Fig. 9D).
To further characterize such a complex, we mapped the interactions between merlin and the different partners involved in YAP regulation. As mentioned above, LATS interaction with merlin was previously reported on the N-terminal FERM domain, whereas the AMOT was mapped to the C-terminal part of merlin (33). Because YAP could interact with both LATS and AMOT, we wondered which part of merlin was required for its putative interaction with YAP. We used HEK293 cells to express either the full-length GFP-merlin, the GFP-merlin FERM domain, or GFP-merlin C-terminal moiety. YAP and LATS association with merlin was analyzed after GFP pulldown. In contrast to LATS that interacts with merlin via its FERM domain, YAP was coimmunoprecipitated with the C-terminal moiety of merlin or full-length merlin but not by the N-terminal FERM domain (Fig. 10A). Altogether, these results strongly suggested that merlin serves as a scaffolding protein to bring into close contact the protein kinase LATS with its substrate YAP. Indeed, the phosphorylation of merlin at Ser-518 favors the close conformation of merlin and reduced its interaction with YAP compared with wild-type merlin (Fig. 10A). Because the loss of ␤1 integrin was associated with a reduced merlin phosphorylation at Ser-518, this could be translated into a differential interaction between YAP and merlin. Indeed, immunoprecipitating YAP in ␤1-deficient cells recovered a significantly greater amount of merlin from the membrane pool when compared with wild-type cells (Fig. 10B). Again, inhibition of PAK1 led to a similar observation (Fig. 10C), strongly suggesting that PAK1 activation downstream of ␤1 integrins was involved in the control of an inhibitory complex encompassing LATS, merlin, and YAP according to the model presented in Fig. 10D.

␤1-integrins control cell proliferation in a RAC1/PAK/YAP-dependent manner
Although pioneering works have shown that integrins orchestrate the recruitment of growth factors and clustering of their receptors at the plasma membrane (likely via cytoplasmic effectors such as FAK and SRC), a clear picture of how integrins are involved in the control of cell proliferation is still missing (34). Recently, integrins and cell-matrix adhesion were proposed to participate in the regulation of the Hippo signaling pathway via SRC and FAK (21). However, from these data it was not established what are the downstream effectors of integrin/ FAK/SRC and how this could be molecularly translated into YAP activation. Herein, we provide a molecular basis of the integrin control YAP nuclear translocation, and we decipher the final stage of this regulation.
The loss of ␤1 integrins was associated with a defect in osteoblast proliferation both in vivo and in vitro. Our work also confirmed ␤1 integrins as the main cell-surface receptors by which these cells are capable of linking YAP nuclear translocation in response to cell adhesion. Our observations are in line with previous reports showing that ␤1 integrins control cell proliferation in other tissues (4). Although our main data were obtained with osteoblasts, we observed a similar behavior in MEFs suggesting that the signaling pathway described herein applies to other adherent cell types. Although we focused our work on non-transformed cells, we recently reported that both ␤1 integrins and YAP are overexpressed in primary bone tumors in which they have been identified as poor prognostic markers (35). Knowing the role of ␤1 integrins during tumor progression and their capability to sense the extracellular environment, we can envision that ␤1-dependent YAP nuclear translocation may play an important role in the tumorigenesis of solid tumors.
Although the loss of ␤1 integrins in osteoblasts was clearly associated with a strong defect in YAP signaling that was responsible for the reduced proliferation observed in mutant cells, an important question that remains to be solved is why ␤1 integrins are so critically involved in controlling YAP nuclear localization. Indeed, we and others (11) have highlighted a specific role of ␤1 integrins in YAP nuclear localization. This ques-

Cell adhesion control of YAP activation
tion is even more intriguing knowing that both ␤1 and ␤3 integrins can regulate FAK and SRC (36). Clearly, additional work focusing on these early signaling events will be required to address this question. Actin cytoskeleton was proposed to be critical for controlling YAP nuclear translocation, but the identification of a clear mechanism was elusive. Actin cytoskeleton remodeling may modulate YAP activity through the Rho GTPases family (23,24). Our present data support a critical role for RAC1 rather than RHOA or CDC42 in good agreement with recent reports showing that Arl4c triggers YAP nuclear translocation via the up-regulation of RAC1 while blocking RHOA activity (37). Although our data do not highlight any role for RHOA in YAP nuclear translocation downstream of ␤1-dependent cell adhesion, we cannot rule out that this small GTPase as an important regulator of actin networks in turn may control membrane tar-geting of important players of the Hippo pathway under specific conditions. It is noteworthy that during cell spreading, integrin engagement inhibits RHOA activity to dynamically regulate actin cytoskeleton re-organization (38). This decrease in RHOA activity corresponds in a timely manner to YAP nuclear translocation; therefore, a direct role of the latter GTPase in adhesion-dependent YAP nuclear translocation is very unlikely. Similarly, suspended cells display an elevated level of RHOA, and yet YAP is sequestered in the cytoplasm.
In the future, it will be important to gain a better insight into how those GTPases cross-talk in space and time regarding their activation and recruitment of their downstream effectors. The outcome of such a cross-talk may impact actin remodeling and YAP nuclear localization.
Here, we propose an integrated view of how ␤1 integrins regulate YAP nuclear translocation (Fig. 10D). Indeed, using Note that the Ser-518 mutation decreases YAP interaction, although the residue does not belong to the mapped binding site. B, Western blot analysis of merlin/NF2 (NF2) and YAP from ␤1 f/f and ␤1 Ϫ/Ϫ stably expressing FLAG-YAP. Cells were fractionated to isolate total cell membranes from the cytoplasm; then YAP was immunoprecipitated using the FLAG epitope. The presence of merlin was visualized in the different fractions. Merlin was also immunoblotted from the input fraction. C, Western blot analysis of merlin/NF2 (NF2) and YAP from ␤1 f/f and IPA3-treated cells stably expressing FLAG-YAP. Cells were fractionated to isolate the total cell membranes and the cytoplasm, and then YAP was immunoprecipitated using the FLAG epitope. The presence of merlin was visualized in different fractions. Merlin was also immunoblotted from the input fraction. D, summary of the ␤1 integrin control on YAP/merlin activation.

Cell adhesion control of YAP activation
the expression of an activated form of RAC1 and PAK1 or a mutant of merlin (loss of function), we rescued the defective YAP localization that characterizes ␤1 deficient cells. In contrast, blocking RAC1 and PAK activity in control cells impaired YAP nuclear localization. Together these data, with the observation that RAC1 and its effector PAK1 did not accumulate at the plasma membrane in ␤1-deficient cells, strongly support a picture in which ␤1 integrins regulate RAC1 delivery to locally activate PAK1 that in turn modulates YAP in a merlin-dependent manner. Fitting with this view, merlin, a well-known PAK substrate, is underphosphorylated in ␤1-deficient cells, a posttranslational modification that favors its active state. Activated merlin promotes its interaction with LATS and YAP thereby blocking YAP nuclear translocation.

␤1 integrins regulate the formation of a YAP inhibitory complex
It appears that several key players such as NF2/merlin and LATS that negatively regulate YAP are also concentrated in plasma membrane extensions. This observation fits with previous data showing that LATS is recruited and activated by NF2/ merlin at the cell membrane (31) and with the localization of NF2/merlin in membrane ruffles (39). Our data extended the picture and showed that ␤1 integrins and PAK1 negatively regulated YAP-merlin interactions at the plasma membrane. The reduced YAP-merlin interaction is likely due to the capacity of PAK1 to phosphorylate merlin at the Ser-518 residue to limit YAP and LATS access. Recently, it has been reported that merlin phosphorylation at Ser-518 reduces its interaction with AMOT family members (a YAP interacting partners) (33). However, AMOT recruitment to merlin induces and/or stabilizes merlin open conformation and in turn allows LATS binding on the FERM domain of merlin. Our data would favor such a model, in which merlin YAP and LATS belong to a membrane-associated inhibitory complex that may be dissociated upon PAK1 phosphorylation. Indeed, we observed that LATS and YAP interact with non-overlapping domains of merlin. Although LATS is recruited on merlin using its FERM domain, YAP interacts with the C-terminal part of merlin. Although our data did not establish whether the YAP-merlin interaction is direct or via AMOT, they clearly indicate that merlin acts as a scaffold to bring in close proximity LATS with YAP. The ␤1-dependent regulation of PAK1 increases merlin phosphorylation and thereby decreases YAP and LATS recruitment on merlin. Therefore, we proposed that YAP and LATS are recruited in an inhibitory complex at the plasma membrane orchestrated by merlin. Upon cell adhesion, RAC1 and PAK1 are locally activated and induce merlin phosphorylation to disrupt merlin, YAP, and LATS complex, a prerequisite for YAP nuclear translocation (Fig. 10D).
Although our data provide insights into how YAP is controlled by ␤1 integrins upon cell adhesion, the mechanism of RAC1 targeting to the plasma membrane by ␤1 engagement is still elusive albeit extensively described. Recently, FAK, PI3K, and SRC were shown to regulate YAP nuclear translocation (11), but at the molecular level how those proteins control YAP via LATS was not investigated. It is noteworthy that RAC1 activity is modulated by FAK (40) as well as by PI3K/SRC (41).
Our data add to these findings, showing that actually SRC belongs to the ␤1 integrin-signaling pathways that controls YAP localization. Therefore, an open possibility is that FAKand SRC-dependent regulation of YAP nuclear translocation also relies on a mechanism that converges on the release of YAP from merlin upon RAC/PAK1 activation. Our data highlighted that the loss of ␤1 integrins specifically affects RAC1 at the plasma membrane, and thus we could speculate that SRC/FAK could be important for RAC1 activation/localization and that ␤1 integrins would specifically regulate RAC1 coupling to its effector PAK. Indeed, similarly to RAC1, SRC and FAK were also shown to be activated on endosomes upon growth factor stimulation. Once activated, RAC1 is then translocated to the plasma membrane in a microtubule-and cell adhesion-dependent manner (42). Cell adhesion is then important to regulate microtubule targeting to the plasma membrane (43,44).

Mouse genetics
Mouse strain with floxed alleles of ␤1 integrin (Itgb1 tm1Ref ) have been described previously (45) and were kindly provided by Dr. R. Fässler (Max Planck Institute, Martinsried, Germany). The Osx1-GFP:Cre deletor mouse was described previously (46) and was kindly provided by Dr. A. McMahon. Conditional knock-in mice bearing the G12D mutation at the K-Ras locus (Kras tm4Tyj ) were obtained from the NCI mouse repository and originally generated by Dr. T. Jacks (47). Mice were kept under regular conditions of husbandry accordingly to the European rules and approved by the University Ethical committee.

Cell lines and mesenchymal stem cell culture
Primary MEFs were isolated at embryonic day 14.5 (E14.5) from K-Ras G12D ␤1 f/f or K-Ras G12D ␤1 ϩ/f embryos using a standard procedure. Cells were immortalized with the large SV40 T antigen. Immortalized K-Ras G12D ;␤1 f/f and K-Ras G12D ␤1 ϩ/f cells were infected with an adenoviral supernatant encoding the Cre recombinase for 1 h in PBS supplemented with 2% FCS and 1 mM MgCl 2 . All other cell lines were generated upon retrovirus transduction, and transgene expression was verified by Western blotting and immunostaining.
Primary mesenchymal stem cells were isolated from WT and ␤1 Ost-ko bone marrow and selected on their capacity to adhere on plastic (48). The differentiation process was visualized by alkaline phosphatase staining described previously (49), and the number of alkaline phosphatase colonies having a diameter higher than 0.5 mm was evaluated using a stereomicroscope (Olympus SZX10).

Transfections and infections
HEK293 GP cells (Clontech, St Germain en Laye, France) were transfected with plasmid DNA using ExGen500 Transfection reagent (Euromedex, Souffelweyersheim, France) according to the manufacturer's instructions. Osteoblast retroviral infections were performed as described previously (43).

Histomorphometric analysis
Tibiae were fixed and embedded in methyl methacrylate. Sections were deplasticized and stained for Masson-Goldner with hematoxylin (Gill II), acid fuchsin/Ponceau xylidine, and phosphomolybdic acid/orange G to stain the cells and osteoid and light green to stain the mineralized matrix (51). The total absolute number of osteoblasts in the area extending from 150 m below the growth plate down 2 mm was evaluated and reported.

TUNEL and BrdU in vivo staining assay
Fluorescein "In Situ" Cell Death Detection Kit (Roche Diagnostics, Meylan, France) was used for TUNEL staining. Briefly, bone sections were deparaffinized and hydrated. Antigen retrieval and endogenous peroxidase quenching were performed and then TUNEL staining was achieved according to the manufacturer's instructions. The TUNEL-positive cells and total cells (DAPI positive) in five areas of periosteum and trabecular bone from each of the mice in the experiments were counted under a ϫ20 objective microscope lens.
For BrdU staining, mice were sacrificed 2 h after being injected with BrdU (150 g/g). Following deparaffinization and hydration, sections were treated 20 min with 4 N HCl, and then antigen retrieval was performed using trypsin 10 min at 37°C. Finally, bone sections were immunostained for BrdU as described under "Immunofluorescence staining."

Cell fractionation
All the operations were carried out at 4°C. Cells from four 10-cm Petri dishes were washed twice with PBS and then scraped into 1 ml of PBS with a rubber policeman. They were centrifuged at 1500 rpm for 5 min. The pellet was resuspended in a hypotonic buffer made of 10 mM HEPES, pH 7.4, 1 mM EDTA, and a mixture of protease inhibitors (Complete, Roche Diagnostics, Meylan, France) and phosphatase inhibitors (Sigma, l'Isle d'Abeau, France) and incubated for 10 min on ice. The cells were broken with a Dounce homogenizer (Piston B, 25 stokes). Unbroken cells and nuclei were eliminated by centrifugation at 2500 rpm for 5 min. Mitochondria were further removed by a 6000 rpm centrifugation for 15 min.
For cytoplasm and total membrane recovery, the supernatant was centrifuged at 120,000 rpm at 4°C for 20 min in a fixed angle AT120 rotor in a Hitachi micro-ultracentrifuge. The whole-membrane fraction was recovered from the pellet fraction after solubilizing in 50 mM Tris, pH 7.4, 1% Triton X-100, 150 mM NaCl, 1 mM EDTA for 20 min at 4°C. Cytoplasmic proteins were recovered from the previous supernatant.

RHOA and Rac1 activity
GST-Rhotekin and GST-PAK-Crib based pulldown assays were carried out as reported previously (52).

Pharmacological inhibition of Rac1 and group I PAK
Cells were resuspended in DMEM and preincubated in suspension for 30 min at 37°C. Then either ETH1864 (53) or IPA3 (54) was added at the concentration of 50 and 10 M, respectively, and the incubation was pursued for another 30 min. The cells were then plated in the presence of the inhibitors for 1 h before fixation with paraformaldehyde and immunostaining.

Immunoblotting
Cells were lysed using RIPA lysis buffer containing protease and phosphatase inhibitors. Cell lysates were centrifuged at 15,000 rpm for 30 min at 4°C, and supernatants were used for immunoblotting using standard protocol.

Cell adhesion control of YAP activation RNA extraction, reverse transcription, qPCR, and transcriptomic analyses
Total RNA was isolated using TRIzol reagent (Thermo Fisher Scientific, Waltham, MA) and RNeasy kit (Qiagen, Courtaboeuf, France) following the manufacturer's instructions. Total RNA quantification was performed using the Nanodrop ND-1000 spectrophotometer (Thermo Fisher Scientific). RNA was reverse-transcribed with the iScript Reverse Transcription Supermix (Bio-Rad). Real-time qPCR analysis was performed using iTaq Universal SYBR Green Supermix (Bio-Rad) on Bio-Rad CFX96. The integrity of the extracted RNAs was assessed with the Bioanalyzer 2100 and the RNA6000 nano kit (Agilent Technologies Inc., Santa Clara, CA). An RNA integrity number greater or equal to 7.00 was achieved for all samples. No sign of DNA contamination was detected in any of the samples analyzed. The starting amount of total RNA used for the reactions was 400 ng per sample for all samples. The Illumina Total Prep RNA Amplification kit (Applied Biosystems/Ambion, Austin, TX) was used to generate biotinylated, amplified cRNA according to the manufacturer's recommendations. Hybridization, staining, and detection of cRNAs on Illumina Mouse WG-6 version 2 Expression BeadChips were performed according to the manufacturer's protocol. The Mouse WG-6 version 2.0 BeadChip profiles more than 45,200 transcripts derived from the National Center for Biotechnology Information Reference Sequence (NCBI RefSeq) database (Build 36, Release 22), the Mouse Exonic Evidence-based Oligonucleotide (MEEBO) set, and from exemplar protein-coding sequences described in the RIKEN FANTOM2 database. The Illumina I-Scan system was used to scan all Expression BeadChips, according to Illumina recommendations.
Using the Gene Expression Module 1.9.0 of GenomeStudio version 2011.1 software (Illumina), the Quantile normalization method was applied to the primary probe data. Processed probe data were then filtered according to the following criteria: minimal signal intensity fold change of 1.50 across all samples and minimal probe signal intensity absolute change of 150 across all samples. Filtered data were then log2-transformed, and the expression values were compared between the b1 Ϫ/Ϫ cells and wild-type ␤1 f/f samples using Omics Explorer 3.2 (Qlucore, Sweden). Genes were considered differentially expressed when their expression level satisfied two criteria: the adjusted p value (q value) was Ͻ 0.01(which corresponded to an ͉R͉ Ͼ0.96 ii) the absolute fold change between the mean expression value in the samples from mutant cells compared with that in controls was Ͼ1.5. Two-dimensional hierarchical clustering analysis was performed using Omics Explorer 3.2 software on normalized data (mean ϭ 0, variance ϭ 1) with the average linkage option.

Cell proliferation assay
Cells were treated with BrdU or alternatively with EdU (10 M, Sigma) during 1 h or 30 min for osteoblasts and MEFs, respectively. For BrdU staining, cells were fixed with Carnoy's fixative (75% methanol, 25% glacial acetic acid) 20 min at Ϫ20°C and then denatured using 2 M HCl for 1 h at 37°C. Cells were then immunostained for BrdU as described earlier. BrdU-positive cells were counted under Axioimager microscope (Carl Zeiss, Inc.). For EdU staining, manufacturer's protocol was used, after an incubation of cells with EdU for 30 min.

Quantification of YAP nuclear localization
Cells were immunostained with an anti-YAP, and immunomicroscopy was carried out with a confocal laser-scanning microscope (Zeiss LSM510) equipped with a ϫ63 planapo oil immersion objective (n.a. 1.4) and a pinhole set to 1 Airy. On each cell image, a ROI was defined positioned either within the nuclei or in the cytoplasmic area next to the nuclei envelope. Because the thickness of the two ROI positions were likely identical, the average fluorescence intensity is likely proportional to YAP concentration and was estimated using ImageJ public software. Within the same cell, the ratio of both fluorescence intensities reflects YAP concentration ratio in both compartments. This ratio was represented under a logarithmic scale to have an identical range for positive and negative ratios. Measurements were performed with n Ն 50, and statistical significance was estimated with Student's t test. Boxplots were performed using R public software.

Colocalization microscopy
Confocal images were taken using LSM510 Zeiss microscope. Visualization and quantification of colocalized pixels were carries out using Wright cell imaging (15) facility plugins of ImageJ.