Rho differentially regulates the Hippo pathway by modulating the interaction between Amot and Nf2 in the blastocyst

The first cell fate decision during embryogenesis results in the segregation of the inner cell mass (ICM) and the trophectoderm (TE) in the blastocyst (Cockburn and Rossant, 2010; Zernicka-Goetz et al., 2009; Chen et al., 2010). ICM cells further differentiate into the epiblast and the primitive endoderm (PE). The epiblast give rise to the fetus, while the PE, together with the TE, contributes to the placenta.

The Hippo pathway, the core components of which are the kinases Mst1/2 and Lats1/2, and the downstream effector Yap, is involved in a wide range of biological processes, such as cell proliferation, cell death, cell differentiation, organ size control, tissue homeostasis and cancer development (Pan, 2010; Yu et al., 2015; Yu and Guan, 2013). It has been demonstrated that the Hippo pathway also regulates the differentiation of the ICM and the TE (Nishioka et al., 2009; Lorthongpanich et al., 2013). In outside TE cells of the blastocyst, Hippo signaling is repressed, and the unphosphorylated downstream effector Yap is transported into the nucleus. Consequently, Yap cooperates with Tead4 to activate the expression of transcription factor genes essential for TE development, such as Cdx2 and Gata3 (Nishioka et al., 2009; Yagi et al., 2007; Nishioka et al., 2008; Ralston et al., 2010). In inside ICM cells, Hippo signaling is activated and phosphorylates Yap, resulting in cytoplasmic retention of Yap. Thus, Cdx2 cannot be activated by Tead4 in the ICM (Nishioka et al., 2009).

How the Hippo pathway is differentially regulated in TE and ICM cells of the blastocyst is a fundamental issue for understanding the first cell fate determination event. In other biological systems, cell polarity, cell adhesion, cell contact and mechanical cues have been shown to be upstream regulators for the Hippo pathway (Yu et al., 2015). Coincidently, TE cells in the blastocyst are polarized, whereas ICM cells are apolar (Thomas et al., 2004; Plusa et al., 2005; Zernicka-Goetz et al., 2009). Tight junctions are formed at the apicolateral cell contact region among polarized TE cells, whereas other cell contacts within the embryo are mediated by adherens junctions (AJs) (Sheth et al., 2000; Fleming et al., 1989; Sheth et al., 1997). The difference in cell polarity and cell adhesion between TE and ICM cells might contribute to the differentially regulated Hippo signaling. In support of this view, it has been shown that downregulation of polarity molecules, such as Par3, Par6 and atypical PKC (aPKC), leads to cytoplasmic distribution of Yap and reduced expression of Cdx2 in the TE, consequently suppressing TE formation and promoting ICM specification (Alarcon, 2010; Hirate et al., 2013; Plusa et al., 2005). Inhibition of Rho-ROCK signaling disrupts the apical-basal polarity and activates Hippo signaling (Kono et al., 2014). In addition, the junction-associated proteins angiomotin (Amot) and angiomotin-like 2 (Amotl2) mediate the signaling of cell polarity and AJs to modulate the Hippo pathway. In apolar ICM cells, the association of Amot with AJs activates the Hippo pathway. In contrast, in polarized TE cells, Amot is restricted to the apical region and absent in basolateral AJs, thereby failing to activate Hippo signaling. The phosphorylation of Amot at S176 (S175 for human AMOT) plays a crucial role in Amot binding to AJs and subsequent activation of Hippo signaling (Hirate et al., 2013; Leung and Zernicka-Goetz, 2013). Another upstream activator of the Hippo pathway, Nf2/Merlin, which is randomly distributed in the plasma membrane regions of both ICM and TE cells, is required for the activation of Hippo signaling in ICM cells (Cockburn et al., 2013), implying that suppression of Nf2 activity is necessary for TE cells to inactivate the Hippo pathway. Yet how Nf2 is repressed in TE cells, and how cell polarity, Rho-ROCK, Amot and Nf2 cooperate to regulate the Hippo pathway in the blastocyst remain to be explored.

To understand the regulatory mechanisms of the Hippo pathway in preimplantation embryos, we screened a mini-library of inhibitors and activators for signaling pathways, and identified Rho-ROCK signaling as a Hippo repressor in TE cells and PKA as a Hippo activator in ICM cells. We further demonstrated that the repressive effect of Rho on Hippo signaling is independent of cytoskeleton and cell polarity. Rather, active Rho prevents the phosphorylation of S176 in Amot, stabilizing the association between Amot and filamentous actin (F-actin). Meanwhile, Rho competes with Nf2 and binds to the coiled-coil (CC) domain of Amot. Consequently, sequestering Amot from Nf2 secures the inactive status of Hippo signaling in TE cells. Our results reveal a novel mechanism by which Rho can regulate Hippo signaling.

To search for upstream regulators of the Hippo pathway in the morula and the blastocyst, a small-scale screening was carried out with a mini-library of inhibitors and activators for signaling pathways that are known Hippo regulators or crucial for preimplantation development. To avoid inhibitor- or activator-induced developmental defects, which might perturb proper Hippo regulation, a short-time (2 h) exposure to inhibitors or activators was performed with late stage morula and early stage blastocysts. To read out the status of the Hippo pathway, the subcellular distribution of Yap was examined in control and treated embryos by immunofluorescence staining. Okadaic acid (OA), an inhibitor for protein phosphatase 1 and 2A (PP1 and PP2A), was used as a control, because of its activation effect on the Hippo pathway (Hata et al., 2013). Similar to OA, inhibitors targeting the Rho-ROCK signaling, including C3 transferase (C3) and CCG1423, and Y27632 promote cytoplasmic localization of Yap in TE cells. Meanwhile, these treatments lead to blastocoel collapse (Fig. 1A-C). Conversely, Yap is localized to the nucleus of ICM cells upon PKA inhibition (Fig. 1A, Fig. S1F).

We were particularly interested in the regulatory effect of Rho-ROCK signaling on the Hippo pathway, because it is also important for early embryo development (Clayton et al., 1999). To ensure the regulatory effect of Rho on the Hippo pathway, we tested whether overexpression of constitutively active RhoA (caRhoA) rescues the mislocalization of Yap in C3-treated embryos. Glutamine 63 in the switch II region of RhoA is replaced with a leucine to mimic the deamidation status, thus leading to constitutive activation of RhoA and resistance to C3 transferase (Vogelsgesang et al., 2007). mRNAs encoding caRhoA and H2B-mCherry were injected into one blastomere of the two-cell embryo. At the late morula stage, embryos were treated with C3 for 2 h. Expression of caRhoA leads to the nuclear localization of YAP in ICM cells. More importantly, the nuclear localization of Yap is maintained in blastomeres expressing caRhoA, even after C3 treatment (Fig. 1D). We then asked whether the activity of Rho is differentially regulated in the blastocyst. As demonstrated by Rho-GTP affinity assay, active Rho (Rho-GTP) is enriched at the apical region of TE cells and is almost undetectable in the ICM (Fig. 1E). Consistently, apical enrichment of Rho-GTP is demonstrated by immunofluorescence staining with an antibody specifically recognizing active RhoA (Fig. 1F). These data indicate that Rho signaling suppresses the Hippo pathway in the TE, and that lack of Rho activity in the ICM leads to the activation of Hippo.

It has been proposed that Rho GTPases, modulated by G-protein coupled receptors (GPCRs) and mechanical cues, including cell contacts and cell attachment status, remodel the cytoskeleton to regulate Hippo signaling (Zhao et al., 2012, 2007; Yu et al., 2012, 2015). To test whether Rho inhibits Hippo signaling through modulating the organization of F-actin and/or microtubules, we first examined the organization of F-actin and microtubules after C3 treatment, and found that C3 treatment has a negligible effect on cytoskeleton organization in embryos (Fig. 2A,B). Next, we treated embryos with chemicals that altered actin or microtubule dynamics. To our surprise, actin polymerization inhibitors, cytochalasin D (CCD) and latrunculin B (LatB), do not activate the Hippo pathway in TE cells (Fig. 2C), despite the observation that LatB treatment leads to Hippo activation and cytoplasmic localization of Yap in 3T3 cells (Fig. S1B). Similarly, neither inhibition nor induction of microtubule polymerization by nocodazole and taxol, respectively, affects Hippo signaling (Fig. 2D). The effects of CCD, LatB, nocodazole and taxol on the cytoskeleton organization in embryos are demonstrated in Fig. S1C,D. These data suggest that, unlike in cultured cells, cytoskeleton remodeling does not affect the Hippo pathway in the blastocyst. Therefore, Rho regulates Hippo signaling independently of cytoskeleton remodeling in peri-implantation embryos.

Rho is also involved in cell polarity regulation (Etienne-Manneville and Hall, 2002) and it has been shown that inhibition of Rho-ROCK signaling promotes ICM specification by impairing the apical-basal polarization and activating the Hippo pathway in the blastocyst (Kono et al., 2014). We then addressed whether Rho represses Hippo signaling through regulating cell polarity. Embryos treated with C3 for 2 and 12 h were harvested at the late morula stage and the distribution of aPKC and E-cadherin (E-cad) were examined by immunofluorescence. C3 treatment for 2 h does not alter the apical distribution of aPKC and the basal-lateral localization of E-cad in a majority of embryos, while this treatment is sufficient to activate the Hippo pathway and leads to cytoplasmic localization of Yap in TE cells. Prolonged C3 treatment (12 h) compromises the polarized distribution of aPKC and E-cad, and activates Hippo signaling in TE cells (Fig. 2E,F). These data demonstrate that disruption of cell polarity is not required for the activation of Hippo by inhibition of Rho.

To elucidate the mechanism by which Rho suppresses the Hippo pathway in TE cells, we then investigated the relationship of Rho and known Hippo regulators in the blastocyst: Amot and Nf2. Nf2 siRNA or control siRNA, together with H2B-mCherry mRNA, were injected into one blastomere of the two-cell embryo. At the late morula stage, embryos were treated with or without C3 for 2 h. C3 treatment leads to cytoplasmic distribution of Yap in TE cells of the embryos injected with control siRNA. In contrast, progeny cells derived from the two-cell blastomere injected with Nf2 siRNA are resistant to C3 treatment, and the nuclear localization of Yap is retained in these cells after C3 treatment (Fig. 3A). Consistently, suppression of Nf2 activity by dominant-negative Nf2 (dnNf2) also promotes nuclear distribution of Yap, and renders the nuclear localization of Yap insensitive to C3 treatment (Fig. 3B). Similarly, downregulation of Amot by siRNA or double-stranded RNA (dsRNA) induces C3 resistance (Fig. 3C,D), even though a smaller fraction of cells with downregulated Amot maintain the nuclear localization of Yap after C3 treatment when compared with cells expressing dnNf2. This might be due to incomplete knockdown of Amot and the redundancy of Amotl2. When Amot and Amotl2 are simultaneously knocked down, more cells display nuclear localization of Yap and resistance to C3 treatment, suggesting the redundancy of Amot and Amotl2 in mediating the effect of Rho on Hippo (Fig. S2C). Altogether, these data indicate that both Nf2 and Amot are essential for the activation of Hippo signaling upon inhibition of Rho.

It is noteworthy that Amot is restricted to the apical domain of TE cells, whereas it is evenly distributed in the plasma membrane of ICM cells (Hirate et al., 2013; Leung and Zernicka-Goetz, 2013). Even though we have demonstrated that disruption of cell polarity is not required for Hippo activation by inhibition of Rho, it remains possible that Rho regulates the polarized distribution of Amot in TE cells. Indeed, treating embryos with ROCK inhibitor Y27632 from the two-cell stage leads to altered localization of Amot (Mihajlović and Bruce, 2016). Thus, we examined the localization of Amot in morula before and after C3 treatment, and found that Amot spreads out to the basolateral region in TE cells after 2 h of C3 treatment (Fig. 4A).

The next question we asked is to which polarized molecule in TE cells does Amot anchor? Amot is known as an F-actin-binding protein (Ernkvist et al., 2006). F-actin is apically distributed in TE cells (Liu et al., 2013). We then tested whether Amot is associated with F-actin in TE cells. First, ectopically expressed wild-type Amot colocalizes with F-actin in TE cells (Fig. 4B). Phosphorylation of S176 in Amot leads to dissociation of Amot from F-actin (Hirate et al., 2013). Replacing S176 with alanine (Amot-SA) does not change the colocalization of Amot and F-actin, whereas the mutation of S176 to an aspartate (Amot-SD), mimicking a phosphorylated residue, results in diffuse cytoplasmic distribution of Amot-SD (Fig. 4B). These data imply the association of Amot with F-actin in TE cells. To further validate this conclusion, embryos were treated with LatB to disrupt F-actin organization. Apical F-actin becomes fragmented. The distribution of both wild-type Amot and Amot-SA still overlaps with F-actin, suggesting Amot indeed binds to F-actin in TE cells (Fig. 4B).

To demonstrate that Rho regulates the binding of Amot to F-actin, Flag-tagged Amot, together with caRhoA or dominant-negative RhoA (dnRhoA), were expressed in MCF-7 cells. As expected, Amot-SD diffuses in the cytoplasm, not overlapping with F-actin, regardless of whether caRhoA or dnRhoA is expressed. In contrast, Amot-WT and Amot-SA colocalize with F-actin, with or without caRhoA expression. In the presence of dnRhoA, the colocalization of Amot-WT and F-actin is disrupted. However, the colocalization of Amot-SA and F-actin is resistant to dnRhoA, implying that phosphorylation of Amot S176 plays a pivot role in regulating the association between Amot and F-actin by Rho (Fig. 4C and Fig. S3B). In the blastocyst, a similar observation was made. In TE cells, inhibition of Rho by C3 leads to Amot-WT spreading out from the apical region to the basolateral region, whereas Amot-SA is restricted at the apical region and undetectable at the lateral region, even after C3 treatment (Fig. 4D). Moreover, western blot showed that caRhoA expression indeed decreases the level of phosphorylated Amot at S176 (p-Amot), and that the level of p-Amot is elevated upon dnRhoA expression in HEK293T cells (Fig. 4E). Consistently, C3 treatment of embryos also enhances the level of p-Amot (Fig. 4F). These data indicate that Rho prevents the phosphorylation of Amot S176 to stabilize the interaction between Amot and F-actin.

We have shown that Amot is anchored to apical F-actin in TE cells. Upon C3 treatment, Amot becomes phosphorylated at S176, is dissociated from F-actin and spreads out to the basolateral membrane of TE cells, consequently activating the Hippo pathway. Which protein(s) facilitate the basolateral membrane localization of Amot? How does basolaterally localized Amot activate Hippo signaling? Nf2 has been shown to be essential for the activation of Hippo signaling in the blastocyst (Cockburn et al., 2013), as well as for the activation of Hippo by Rho inhibition in TE cells (Fig. 3A,B). In addition, Nf2 activates the Hippo pathway by recruiting Lats to the plasma membrane (Yin et al., 2013). And Nf2 is evenly distributed in the plasma membrane of both TE and ICM cells (Cockburn et al., 2013). To test whether Nf2 is able to recruit Amot to the plasma membrane, Amot-GFP, with or without Nf2-mCherry, was expressed in HEK293T cells. In the absence of Nf2-mCherry, Amot-GFP is located in the cytoplasm, likely associated with F-actin. When Nf2-mCherry is expressed, Amot-GFP colocalizes with Nf2-mCherry at the plasma membrane (Fig. 5A). Experiments repeated in HeLa cells (Fig. S4A) also support the conclusion that Nf2 recruits Amot to the plasma membrane. Moreover, knockdown of Nf2 disrupts the membrane localization of Amot in ICM cells. In contrast, Amot distribution at the apical region of TE cells is maintained upon Nf2 knockdown (Fig. 5B). These data suggest that Amot is recruited to the membrane by Nf2 in ICM cells, whereas the apical distribution of Amot in TE cells depends on the interaction between Amot and F-actin.

It has been shown that Amot interacts with Nf2 (Hirate et al., 2013; Yi et al., 2011; Li et al., 2015). With co-immunoprecipitation, we also detected the interaction between Amot and Nf2, which is regulated by Rho activity. caRhoA attenuates the association of Amot and Nf2, whereas dnRhoA slightly enhances the interaction between Amot and Nf2 (Fig. 5C). More importantly, Amot and Nf2 collaboratively activate the Hippo pathway, elevating the level of p-Lats1/2 and p-Yap (Fig. 5D). In addition, caRhoA blocks the plasma membrane recruitment of Amot by Nf2 (Fig. 5E), suggesting that Rho might regulate the Hippo pathway through repressing the interaction between Amot and Nf2.

Given the important role of Amot S176 phosphorylation in regulating the interaction between Amot and F-actin, we suspected that phosphorylation of Amot S176 may affect the association of Amot and Nf2. To test this hypothesis, Amot-WT, SD and SA were co-expressed with Nf2 in HEK293T cells. Nf2 is able to recruit Amot-WT and Amot-SD, but not Amot-SA to the plasma membrane (Fig. S4B), suggesting that Amot has to be phosphorylated to allow its interaction with Nf2. We then asked whether Rho attenuates the interaction between Amot and Nf2 through blocking the phosphorylation of Amot S176. Expression of caRhoA impairs the recruitment of both Amot-WT and Amot-SD to the plasma membrane by Nf2 (Fig. 5E), whereas the plasma membrane localization of Nf2 is not affected by caRhoA (Fig. S4C), implying that Rho may suppress the interaction between Amot and Nf2 independently of Amot S176 phosphorylation. Nevertheless, compared with Amot-WT, Amot-SD is more resistant to caRhoA-induced dissociation from the plasma membrane. Upon caRhoA overexpression, ∼45% cells retain the plasma membrane distribution of Amot-SD, whereas the plasma membrane distribution of Amot-WT persists in only a few cells (∼10%) (Fig. 5E). These data suggest that dephosphorylation of Amot S176 contributes to the attenuated interaction between Amot and Nf2 by Rho. Moreover, co-immunoprecipitation experiments showed that caRhoA is able to weaken the association of Nf2 with Amot, regardless of the phosphorylation status of Amot S176 (Fig. 5F). The Nf2-interacting site on Amot has been mapped to the CC domain (Li et al., 2015). The binding between Nf2 and the CC domain of Amot is also reduced by caRhoA, even though the N-terminal domain of Amot, including S176, is absent (Fig. 5G). These data suggest that additional mechanism(s), other than regulating Amot S176 phosphorylation, are employed by Rho to suppress the binding between Amot and Nf2.

We hypothesized that Rho might compete with Nf2 in binding to Amot. With co-immunoprecipitation experiments, the interaction between caRhoA and Amot was detected (Fig. 5H). We further demonstrated that caRhoA also interacts with the CC domain of Amot, which also mediates the interaction with Nf2 (Fig. 5H). Conversely, Nf2 overexpression also reduces the interaction between caRhoA and Amot (Fig. S4D). Thus, competitive binding at the CC domain of Amot allows RhoA to suppress the interaction between Nf2 and Amot. Both caRhoA and dnRhoA binds to Amot, and the binding between caRhoA and Amot is slightly stronger than that of dnRhoA and Amot (Fig. 5I), implying that Rho activity is not essential for the binding between RhoA and Amot.

Another possible mechanism by which Rho could reduce the association of Nf2 and Amot is post-translational modifications (PTMs) of these proteins. The Amot CC domain was expressed without or with caRhoA. The resulting protein lysates were resolved in regular SDS-PAGE and Phos-tag gels. No difference in the electrophoretic mobility of Amot CC domain was detected in either gels (Fig. S4E,F), implying that caRhoA might not alter the PTM status of the CC domain. Nevertheless, some PTMs do not change the gel mobility of proteins. Therefore, we cannot rule out the possibility that Rho regulates the binding of Amot to Nf2 through PTMs.

So far, we have characterized two mechanisms by which Rho can prevent the binding of Amot and Nf2 (Fig. 6A). First, active Rho blocks the phosphorylation of Amot S176, thus stabilizing the association between unphosphorylated Amot and F-actin, and reducing the concentration of free Amot. Second, active Rho competes with Nf2 in binding to the CC domain of Amot. The second mechanism might be complementary to the first one in suppressing the interaction between Nf2 and Amot transiently dissociated from F-actin. The cooperation of these two mechanisms might be crucial for the suppression of the Hippo pathway in TE cells, in which Amot and Nf2 are colocalized in the apical region.

In this study, we have identified Rho as a negative regulator of the Hippo pathway in TE cells, and have demonstrated that Rho suppresses the Hippo pathway through modulating the interaction between Amot and Nf2. In TE cells of the blastocyst, Rho is active. Amot is unphosphorylated and binds to apical F-actin. Moreover, Rho occupies the CC domain of Amot, and prevents the binding between Nf2 and Amot that is transiently dissociated from apical F-actin. Without the formation of Amot and the Nf2 complex, Lats1/2 cannot be activated. Thus, the Hippo pathway is inactive, and Yap translocates into the nucleus to promote TE differentiation. In contrast, the activity of Rho is absent in ICM cells. Amot becomes phosphorylated and dissociated from F-actin, which allows the interaction between Nf2 and Amot. A complex formed with Amot and Nf2, as well as with other proteins, leads to the activation of Hippo signaling and cytoplasmic retention of Yap, allowing the specification of ICM cell fate (Fig. 6B).

What features of the blastocyst lead to the differential status of Rho in TE and ICM cells? Previous results and our data support that cell contacts are essential for the differential regulation of Rho in the blastocyst. First, when cell contacts are impaired in embryos lacking both maternal and zygotic E-cadherin, nuclear Yap localization and Cdx2 expression in TE cells are not affected. Rather, more Cdx2 positive cells are present in E-Cadherin knockout embryos (Stephenson et al., 2010). Second, when blastomeres of the 32-cell embryo are dissociated into single cells, the Hippo pathway is repressed and Yap becomes localized to the nucleus, regardless of the original inside or outside positions of the cells (Hirate et al., 2013). Moreover, when dissociated blastomeres were treated with C3 to inhibit Rho, Yap is distributed in the cytoplasm in all cells (Fig. S1E). These data imply that cell contacts inactivate Rho, and subsequently activate the Hippo pathway in ICM cells. It has been suggested that cell polarity of TE cells is required for inactivating the Hippo pathway, and that difference in cell contractility between polar and apolar cells regulates the Hippo pathway and cell fate (Hirate et al., 2013; Cao et al., 2015; Anani et al., 2014; Maître et al., 2016). Moreover, isolated eight-cell blastomeres become polarized, and the apical domain is required and sufficient for the specification of TE cell fate (Korotkevich et al., 2017). Cell contacts, cell polarity and cell contractility are tightly co-regulated. It is very likely that cell contacts might regulate Rho through affecting cell polarity and cell contractility in the blastocyst. Cell contacts inhibit cell polarity in ICM cells by preventing the formation of polarized apical domain, resulting in high cell contractility. In contrast, cell contacts stabilize the cell polarity of TE cells by separating the apical and basolateral regions, thus maintaining low cell contractility. When cell contacts are disrupted either by E-Cadherin knockout or blastomere dissociation, inside cells might initiate the formation of apical domain and reduce contractility, whereas the polarity of outside cells is compromised, but still maintained. The polarity and low contractility of cells consequently leads to Rho activation, Hippo suppression and nuclear localization of Yap.

GPCR signaling has been shown to regulate the Hippo pathway through Rho GTPases (Yu et al., 2012, 2013). Our data rule out the possibility that Rho is activated by GPCRs coupled to Gα_12/13 or Gα_q/11 in the TE. Neither the ligand for GPCRs coupled to Gα_12/13 or Gα_q/11, lysophosphatidic acid (LPA), nor the LPA antagonist Ki16425 alters Hippo signaling in the blastocyst (Fig. 1A). PKA, which mediates the signaling from Gα_s-coupled GPCRs to suppress Rho and subsequently activate Hippo signaling in breast cancer cells (Yu et al., 2013), was identified as an activator for the Hippo pathway in ICM cells in our initial screen. It raises the possibility that Gα_s-coupled GPCRs activate PKA, which in turn suppresses Rho in the ICM. However, when embryos were treated with PKA inhibitor H89 and Rho inhibitor C3 simultaneously, the Hippo pathway is inactivated and Yap becomes nuclear localized in all blastomeres, regardless of their position in the embryo (Fig. S1F). The data indicate that Rho is not downstream of PKA or Gα_s-coupled GPCRs in regulating the Hippo pathway in the blastocyst. The relationship between PKA and Rho in modulating Hippo signaling in the blastocyst needs further investigation.

It has been shown that Rho GTPases repress the Hippo pathway through promoting F-actin formation in cultured cells (Zhao et al., 2012; Mo et al., 2012; Yu et al., 2012; Feng et al., 2014). However, in TE cells, cytoskeleton remodeling is uncoupled from the Hippo pathway in the blastocyst. Drugs affecting cytoskeleton dynamics, including CCD, LatB, nocodazole and taxol, do not affect Hippo signaling in the blastocyst (Fig. 2C,D). It is very likely that F-actin and microtubule lose their mechanosensor function in the blastocyst. In support of this, both F-actin and microtubule are enriched at the apical region of the blastocyst, a distinct distribution pattern compared with that in cultured cells. Nevertheless, F-actin is still involved in regulating the Hippo pathway in the blastocyst, through sequestering Amot. Inhibitors of actin polymerization, such as CCD and LatB, disrupt the integrity of F-actin network, but do not interfere with the binding of Amot to the fragmented F-actin. Therefore, neither CCD nor LatB perturbs the Hippo pathway in the blastocyst. Only Rho inhibition by C3, which induces the dissociation of Amot from F-actin, leads to the activation of Hippo signaling in TE cells.

Regulating the phosphorylation status of Amot S176 by Rho signaling is a crucial step in modulating the interaction between Amot and F-actin, and activating the Hippo pathway. Yet how Rho prevents the phosphorylation of Amot S176 is unknown. Lats kinases have been shown to phosphorylate Amot S176 (Hirate et al., 2013). However, our data showed that upon Amot knockdown by siRNA or dsRNA, the Hippo pathway is not activated in TE cells after C3 treatment, suggesting that Amot is upstream of Lats kinases, and downstream of Rho. It seems that Amot is phosphorylated by a kinase other than Lats in TE cells upon Rho inhibition. Yet it remains possible that Amot and Nf2 collaboratively activate Lats kinases, and activated Lats kinases phosphorylate Amot. This positive feedback circuit formed by Amot/Nf2 and Lats may be essential for quick and efficient activation of the Hippo pathway. Disruption of any component in the circuit leads to a failure in full activation of the Hippo pathway.

In TE cells, Amot is restricted to the apical region by binding to F-actin (Fig. 4A). How does Amot anchor to the plasma membrane in ICM cells and C3-treated TE cells? We have shown that, in HEK293T and HeLa cells, Nf2 recruits Amot to the plasma membrane (Fig. 5A and Fig. S4A). Amot dissociates from the membrane of ICM cells upon Nf2 knockdown (Fig. 5B), further confirming that Amot is recruited to the basolateral membrane by Nf2. In addition, other components of adherens junction might contribute to the membrane distribution of Amot in the blastocyst. The interaction between E-cadherin and Amot has been demonstrated by co-immunoprecipitation experiments (Hirate et al., 2013).

Taken together, our results demonstrated the crucial role of Rho in the differential regulation of the Hippo pathway in the TE and the ICM. In contrast to the prevailing view that Rho represses Hippo signaling by modulating the actin cytoskeleton, Rho inhibits the interaction between Amot and Nf2 to inactivate the Hippo pathway in TE cells. These two mechanisms are not exclusive, and might cooperate to suppress the Hippo pathway in other biological systems.

All animal experiments were carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health (China). The use of mice for this research is approved by Nankai Animal Care and Use Committee.

Embryo manipulation experiments were carried out as described previously (Liu et al., 2013). Female ICR mice (4-6 weeks) were induced to superovulate by intraperitoneal injections of 5 IU of pregnant mare serum gonadotropin (PMSG, Calbiochem) and, 48 h later, 5 IU human chorionic gonadotropin (hCG, Sigma). Females were then paired with ICR males overnight and checked for vaginal plugs next morning. Zygotes were flushed out from oviducts at 12 h post-hCG, and two-cell embryos were flushed out at 42-48 h post-hCG. Embryos were cultured in groups of 20-30 in a 50 µl droplet of potassium simplex optimization medium (KSOM) with amino acids (Millipore) covered by mineral oil (Sigma) in a 37°C incubator with 6.5% CO₂. The concentrations of inhibitors or activators to treat late morula and early blastocysts are listed in Table S1. All experiments were performed with groups of more than 10 embryos and repeated three times.

For the injection of two-cell embryos, 10 mM siRNA, 100 ng/μl Amot dsRNA, 10 ng/μl caRhoA mRNA or 100 ng/μl dnNf2 mRNA, together with 100 ng/μl H2B-mCherry mRNA, were injected into one blastomere of the late two-cell stage embryo. For control embryos, 200 ng/μl H2B-mCherry mRNA was injected into one blastomere of the late two-cell stage embryo. For zygotic injection, 200 ng/μl Amot-Flag (WT, SA or SD) mRNA was injected into the cytoplasm of fertilized eggs in M2 medium. We demonstrated that Yap distribution in embryos is not affected when 200 ng/µl Amot-Flag mRNA is injected (Fig. S3A). After injection, embryos were cultured in KSOM medium until the late morula stage.

HEK293T, HeLa and MCF-7 cells were cultured in DMEM (Invitrogen) supplemented with 10% fetal bovine serum, penicillin (100 U/ml) and streptomycin (100 mg/ml) under an atmosphere of 5% CO₂ at 37°C. Transfection of plasmid DNA was performed with lipofectamine 3000 (Invitrogen) according to the manufacturer's instructions. HEK293T and HeLa cells were obtained from George Daley's laboratory at Harvard Medical School (Boston, MA, USA). MCF-7 cells were purchased from ATCC. These cell lines were recently authenticated using Short Tandem Repeat DNA profiling analysis.

Embryos were fixed in 4% paraformaldehyde for 20 min, and then permeabilized with 0.2% Triton X-100 for 30 min. After being blocked with 5% goat serum for 2 h, embryos were incubated with primary antibodies for 4-6 h at room temperature or overnight at 4°C. Embryos were then washed and incubated with secondary antibodies and/or rhodamine-phalloidin (Molecular Probes). Alexa Fluor 488 anti-mouse, Alexa Fluor 488 anti-rabbit and Alexa Fluor 594 anti-rabbit were used as secondary antibodies (Molecular Probes), and Hoechst 33342 (Sigma) was used for nuclei staining. Epifluorescent images were taken using Olympus IX81 microscope. Confocal images were captured using Leica TCS SP5 confocal microscope.

In some experiments, localization of aPKC, E-cadherin and Amot proteins along the apical-basal or apical-lateral cell polarity was assessed with the ImageJ software. After conversion to TIFF format, images were opened in ImageJ, a line was drawn across the apical and basal/lateral sides of an outer cell, and pixel intensities along this line were examined using the Plot Profile Tool.

Analysis of active Rho protein was conducted as previously described (Berdeaux et al., 2004). Briefly, embryos were fixed with freshly prepared 4% paraformaldehyde in PBS for 30 min at room temperature. After washing, embryos were permeabilized in PBS containing 3% BSA, 0.1 M glycine and 0.05% Triton X-100, blocked in the same buffer containing 5% goat serum, and then incubated with 50 μg/ml GST-tagged Rhotekin-Rho-binding domain (GST-RBD) (RT01A, Cytoskelton) for 12 h at 4°C. Anti-GST primary antibody and Alexa Fluor 488-conjugated anti-rabbit antibody were used to visualize the GST-RBD and Rho-GTP complex. Confocal images were captured using Leica TCS SP5 confocal microscope. The specificity of Rho-GTP affinity assay is demonstrated by a negative control (C3 treated embryos) and a positive control (caRhoA-overexpressing embryos) (Fig. S1A).

Amot- and Nf2-coding regions were amplified from cDNA of V6.5 mouse embryonic stem cells (ESCs), and inserted into the RN3P vector (a gift from Dr Na Jie). Amot was fused with 3×Flag. dnNf2 was constructed by replacing alanine for a seven amino acid stretch in the FERM domain (Johnson et al., 2002). Using SfiI-cut RN3P-Amot-Flag or RN3P-dnNf2 plasmids as DNA templates, mRNA was synthesized using the mMESSAGE mMACHINE T7 transcription Kit (Life Technologies).

dsRNA was prepared as previously described (Leung and Zernicka-Goetz, 2013). The DNA template for in vitro transcription of Amot dsRNA was amplified from V6.5 mouse ESC cDNA, with the following primers, 5′-TAATACGACTCACTATAGGGTGTGTTTGGGGAGAAAAGGA-3′ and 5′-TAATACGACTCACTATAGGGAAGTCCAGGAAAAGGCCTGA-3′. Amot dsRNA was synthesized with the DNA template using the T7 mMESSAGE mMACHINE Kit. The dsRNA was annealed by heating to 70°C for 5 min and cooling at room temperature for 2 h. The resulting RNA sample was treated with RNase A and Proteinase K to remove ssRNA and proteins.

siRNAs were synthesized by GenePharma Corporation (Shanghai, China). Amot siRNA, 5′-CAGGAGAAGCCUACUCAGCUA-3′; Nf2 siRNA, 5′-GGUGUUGGAUCAUGAUGUUTT-3′; negative control siRNA, 5′-UUCUCCGAACGUGUCACGU-3′. Stealth RNAi siRNA for Amotl2 (Amotl2 MSS226047) was purchased from Thermo Fisher. The knockdown efficiency of Amotl2 siRNA has been demonstrated previously (Hirate et al., 2013).

Expression plasmids were transfected into HEK293T cells. One day after transfection, cells were harvested and cell extracts were prepared in lysis buffer [20 mM Tris-HCl (pH 8.0), 137 mM NaCl, 10% glycerol, 1% NP-40, and 2 mM EDTA] with protease inhibitor (Roche) on ice for 30 min. After centrifugation at 12,000 g for 20 min, the supernatant was collected and incubated with anti-Flag M2 magnetic beads (Sigma) at 4°C overnight. The beads were washed three times with lysis buffer, and the bound proteins were released from the beads by boiling the beads in 2× SDS loading buffer for 5 min. Western blot was performed to detect the proteins present in immunoprecipitate samples.

Cells were lysed and total protein concentration was measured using a BCA Protein Assay Kit (Beyotime) to ensure equal loading. Samples were resolved by SDS-PAGE followed by transferring onto a PVDF membrane (Millipore). Membranes were probed with primary antibodies. Bound primary antibodies were recognized by HRP-linked secondary antibodies (GE Healthcare). HRP activity was detected by ECL Plus (Beyotime). Digital images were taken by the automatic chemiluminescence imaging analysis system (Tanon).

Trizol reagent (Roche) was used for RNA purification from cultured cells. cDNA was reverse transcribed using the Transcriptor First-Strand cDNA Synthesis Kit (Roche) with random primers according to the manufacturer's instruction. Real-time PCR was carried out with FastStart Universal SYBR Green Master (Roche) in a Bio-Rad IQ5 system. mRNA relative abundance was calculated by normalizing to β-actin mRNA. The following primers were used for real-time PCR: β-actin, 5′-CAGAAGGAGATTACTGCTCTGGCT-3′ and 5′-CAGAAGGAGATTACTGCTCTGGCT-3′; Nf2, 5′-CTCTTGGCGTCATATGCTGT-3′ and 5′-GAGCAATTCCTCTTGGGCTA-3′; Amot, 5′-GATGTGCAACCCAGATAAGCC-3′ and 5′-TCTCTGCATCAGGCTCTTGC-3′.

Primary antibodies used in this study were YAP [Santa Cruz, SC-101199; 1:1000 for western blot (WB), 1:200 for immunofluorescence (IF)], aPKC (Santa Cruz, SC-216; 1:200 for IF), E-cadherin (Abcam, ab15148; 1:200 for IF), pYAP (CST, 4911; 1:1000 for WB), LATS2 (Abcam, ab84158; 1:1000 for WB), pLATS1/2 (CST, 8654; 1:1000 for WB), Amot (Santa Cruz, sc-82491; 1:1000 for WB), pAmot (Millipore, ABS1045; 1:1000 for WB), Myc (Santa Cruz, SC-40; 1:5000 for WB), Flag (Sigma, F1804; 1:5000 for WB, 1:200 for IF), GST (Abcam, ab19256; 1:200 for IF), active RhoA-GTP (NewEast Biosciences, 26904; 1:100 for IF) and β-Tubulin (Huada, AbM59005-37B-PU; 1:10,000 for WB, 1:1000 for IF).

All data were analyzed using Student's t-test. Statistically significant P values are indicated in figures as follows: ***P<0.001, **P<0.01, *P<0.05.

We thank Professor Shian Wu for his critical reading of the manuscript, and the core facility at the College of Life Sciences, Nankai University, Tianjin, China.