MAP4K family kinases act in parallel to MST1/2 to activate LATS1/2 in the Hippo pathway

The Hippo pathway plays a central role in tissue homoeostasis, and its dysregulation contributes to tumorigenesis. Core components of the Hippo pathway include a kinase cascade of MST1/2 and LATS1/2 and the transcription co-activators YAP/TAZ. In response to stimulation, LATS1/2 phosphorylate and inhibit YAP/TAZ, the main effectors of the Hippo pathway. Accumulating evidence suggests that MST1/2 are not required for the regulation of YAP/TAZ. Here we show that deletion of LATS1/2 but not MST1/2 abolishes YAP/TAZ phosphorylation. We have identified MAP4K family members—Drosophila Happyhour homologues MAP4K1/2/3 and Misshapen homologues MAP4K4/6/7—as direct LATS1/2-activating kinases. Combined deletion of MAP4Ks and MST1/2, but neither alone, suppresses phosphorylation of LATS1/2 and YAP/TAZ in response to a wide range of signals. Our results demonstrate that MAP4Ks act in parallel to and are partially redundant with MST1/2 in the regulation of LATS1/2 and YAP/TAZ, and establish MAP4Ks as components of the expanded Hippo pathway.

T issue homoeostasis is maintained through the precise regulation of cell proliferation, apoptosis and differentiation; dysregulation of any of these processes can result in aberrant tissue growth and carcinogenesis. The Hippo pathway, which is conserved from Drosophila to mammals, has been recognized as a master regulator of cell fate, tissue homoeostasis and organ size [1][2][3][4][5] . Recent advances have rapidly expanded the understanding of the Hippo pathway, leading to the identification of more than 30 components of this pathway 6,7 . However, only five proteins are considered to comprise the core of the Hippo pathway in Drosophila: Ste20-like kinase Hippo (Hpo) with its adaptor protein Salvador (Sav), the NDR family kinase Warts (Wts) with its adaptor Mats, and the transcriptional effector Yokie (Yki). The Hpo-Sav complex phosphorylates and activates Wts-Mats, which in turn phosphorylates and inhibits Yki. Dysregulation of this Hpo/Wts kinase cascade leads to aberrant activation of Yki and uncontrolled growth in the Drosophila eye and wing [8][9][10][11][12][13][14] .
The core components of the Hippo pathway in mammals consist of Mammalian Ste20-like kinases 1/2 (MST1/2, homologues of Hpo) and their adaptor protein Sav family WW domain-containing protein 1 (SAV1, homologue of Sav), Large tumour suppressor 1/2 (LATS1/2, homologues of Wts) and their adaptor proteins MOB1A/1B (homologues of Mats), and the two Yki homologues Yes-associated protein (YAP) and transcriptional co-activator with PDZ-binding motif (TAZ). The physiological importance of the Hippo pathway in mammals has been revealed in many different genetically engineered mouse models. For example, deleting MST1/2 or overexpressing YAP in the liver leads to hepatomegaly and hepatocellular carcinogenesis in mice 12,15,16 . Deleting the core Hippo pathway components in mice leads to neoplasia in other organs as well 3 . Furthermore, mutations in the core and peripheral Hippo pathway components are associated with a number of human malignancies 17 .
A variety of signals have been reported to either activate or inhibit the Hippo kinase cascade. In epithelial cells, apical-basal polarity regulates activities of Wts/LATS through interactions between the upstream components and intercellular junctionassociated proteins 6 ; these interactions may also be responsible for initiating YAP/TAZ phosphorylation and degradation in response to cell-cell contact 18 . Recent reports found that serum deprivation or energy stress activates LATS1/2 and inhibits YAP/TAZ [19][20][21][22] . Extracellular hormones can modulate LATS1/2 kinase activity via G-protein-coupled receptors to regulate the Hippo pathway 19,23 . Changes in Rho-GTPase activities and cytoskeletal dynamics appear to be the major mediators for YAP/TAZ regulation by G-protein-coupled receptors, as well as cell detachment and mechanical forces [24][25][26][27] .
Although the role of MST1/2 in the Hippo pathway has been firmly established, accumulating evidence indicates that MST1/2 are not essential for LATS1/2 activation under various conditions 19,23,24,28 , suggesting the existence of MST1/2independent LATS-kinases. In this study, we screened a human kinome library by in vitro kinase assay with LATS1 as the substrate and identified several mitogen-activated protein kinase kinase kinase kinase (MAP4K) family members, including Hematopoietic progenitor kinase 1 (HPK1/MAP4K1), Germinal centre kinase (GCK/MAP4K2), Germinal centre kinase-like kinase (GLK/MAP4K3), HPK/GCK-like kinase (HGK/MAP4K4), Misshapen-like kinase 1 (MINK1/MAP4K6) and TRAF2 and NCK interacting kinase (TNIK/MAP4K7), as potent LATS1/2activating kinases. We further demonstrate that MAP4Ks are physiological regulators of LATS1/2 and YAP/TAZ in response to a number of upstream signals. Our results indicate that MAP4Ks are components of the Hippo pathway by directly phosphorylating and activating the LATS1/2 kinases.
We tested the roles of LATS1/2 and MST1/2 in YAP/TAZ phosphorylation in response to a variety of signals. High cell density induced YAP phosphorylation in wild-type (WT) HEK293A cells, as shown by YAP-S127 phosphorylation and mobility shift in phos-tag gels (Fig.1b). TAZ contains two phosphodegrons and its degradation is strongly enhanced by LATS-dependent phosphorylation. As anticipated, high cell density reduced TAZ protein levels (Fig. 1b). However, these effects were abolished in LATS1/2-dKO cells, in which high cell density failed to induce YAP phosphorylation or reduce TAZ protein levels. Surprisingly, MST1/2 deletion did not compromise YAP phosphorylation or TAZ protein reduction. This observation was confirmed in MST1/2-dKO U2OS cells ( Supplementary  Fig. 1a), in which high density-induced YAP phosphorylation was only slightly reduced ( Supplementary Fig. 1b).
Considering the indispensable role of LATS1/2 in YAP/TAZ regulation, identification of the MST1/2-independent LATS1/2 kinases is crucial for understanding the Hippo pathway signal transduction. We, therefore, screened the human kinome using in vitro kinase assays to identify candidates that can directly phosphorylate the LATS hydrophobic motif. A truncated human LATS1 (638-1,130) was expressed and purified from Eschericia coli, and used as a substrate for the in vitro kinase assays. We screened a human kinome library 31 supplemented with kinase constructs available in our lab (Fig. 2d), which covered 354 of the 518 putative protein kinases in the human kinome 32 . By this approach, we identified MST1/2 and six additional kinases that can efficiently phosphorylate LATS1-HM (Fig. 2e). These include MAP4K2/4/6, NIMA-related kinase 9 (NEK9), serine/threonine kinase 32B (STK32B) and eukaryotic elongation factor-2 kinase (eEF2K). MAP4K2/4/6 displayed kinase activity comparable with that of MST2 (Fig. 2f). In contrast, MST3, which was thought to be more evolutionally related to MST1/2, displayed no significant kinase activities towards LATS.
MAP4K2/4/6 and MST1/2 both belong to the STE20-like kinase family, and their kinase domains are highly homologous to one another 33 (Supplementary Fig. 2a). Both MAP4K4/6 and MST1/2 possess coiled-coil structures ( Supplementary Fig. 2b), which are important for facilitating kinase-substrate interactions, although MAP4K2/4/6 do not contain Sarah domains that are present in MST1/2. Notably, MAP4K2/4/6 possess a Citron domain in their C-terminal regions. The Citron domain is known for binding of Rac and RhoA 34 , which are crucial regulators of the Hippo pathway. For these reasons, the MAP4K family kinases were the most appealing candidates. Among the MAP4K2/4/6, MAP4K4 and MAP4K6 are closely related. Therefore, we initially focused our efforts on MAP4K4.
To verify that MAP4K4, but not an associated kinase in the MAP4K4 immunoprecipitate, was responsible for LATS1-HM phosphorylation, we generated a MAP4K4 kinase-inactive mutant (MAP4K4-K54R). This mutation abolished MAP4K4's ability to phosphorylate LATS1 and LATS2 (Fig. 2g). We next tested whether MAP4K4 could phosphorylate MOB1-T35, a known MST1/2-specific site, and found that it did not have significant kinase activity towards MOB1-T35 ( Supplementary  Fig. 3a). This is consistent with the observation that MOB1 phosphorylation is completely absent in the MST1/2-dKO cells (Fig. 1a). Furthermore, MAP4K4 physically interacts with both LATS1 and LATS2 ( Supplementary Fig. 3b,c). We confirmed the interaction by co-immunoprecipitation of endogenous proteins ( Supplementary Fig. 3d   ARTICLE individually induced LATS-HM phosphorylation, while expressing MST3 did not (Fig. 2h). Furthermore, MAP4K2/4/6 were as potent as, if not more potent than, MST2 in promoting LATS-HM phosphorylation.
We next tested whether MAP4Ks could directly activate LATS1/2 kinase by performing a sequential kinase assay of MAP4K-LATS-YAP in vitro. Both MST2 and MAP4K4 enhanced LATS2 autophosphorylation, as determined by LATS activation loop phosphorylation (pLATS-AL), as well as the ability of LATS2 to phosphorylate YAP (Fig. 3a). Therefore, the in vitro reconstitution experiments with purified proteins demonstrate that MAP4K4 can directly activate LATS. Collectively, our data strongly suggest MAP4K4 as a direct LATS-activating kinase.
To investigate the role of MAP4K4 in regulating the Hippo pathway, we co-transfected MAP4K4 and YAP into WT, LATS1/2-dKO and MST1/2-dKO cells. Ectopic MAP4K4 expression strongly induced YAP/TAZ phosphorylation in a LATS1/2-dependent but MST1/2-independent manner (Fig. 3b)  phosphorylation of endogenous YAP and LATS, whereas expression of the MAP4K4 kinase-dead mutant (MAP4K4-KR) suppressed the phosphorylation of LATS and YAP (Fig. 3c). Therefore, the MAP4K4-KR likely acted in a dominant-negative manner. We thus overexpressed the WT and the kinase-dead MAP4K4 in HEK293A cells, and tested YAP phosphorylation in response to various signals. As expected, the kinasedead MAP4K4-KR, but not WT MAP4K4, antagonized YAP phosphorylation induced by high cell density, serum deprivation or LatB treatment ( Supplementary Fig. 4), suggesting a role of MAP4K4 in YAP regulation by different signals. YAP/TAZ interact with TEA domain family (TEAD) transcription factors to promote gene transcription. Phosphorylation of YAP leads to its cytoplasmic localization and reduces its interaction with TEAD. As expected, MAP4K4 overexpression decreased YAP-TEAD4 association (Fig. 3d). Consistently, the activity of a TEAD luciferase reporter and the expression of YAP target genes (Connective tissue growth factor CTGF and Cysteine-rich angiogenic inducer 61 CYR61) were also significantly decreased by MAP4K4 overexpression (Fig. 3e,f). In contrast, the kinase-dead MAP4K4 promoted the transcription of CTGF and CYR61. Furthermore, we tested whether MAP4K4 could inhibit cell proliferation through LATS1/2. Overexpression of MAP4K4, but not the kinase-dead mutant, suppressed the proliferation of WT cells. Importantly, MAP4K4 did not inhibit the proliferation of LATS1/2-dKO cells (Fig. 3g). Collectively, these results suggest a model where MAP4K4 acts through LATS to inhibit YAP and cell proliferation.
Deletion of MAP4K4/6/7 reduces YAP phosphorylation. To determine the in vivo functions of MAP4K4, we deleted MAP4K4 by CRISPR in HEK293A cells ( Supplementary Fig. 5a,b). However, deletion of MAP4K4 alone had no obvious effects on YAP phosphorylation under the conditions tested ( Supplementary  Fig. 5c,d,e). We speculated that this lack of effect could be due to functional redundancy with the kinases related to MAP4K4, such as MAP4K6 and MAP4K7, which are homologous to the Drosophila Misshapen (Msn) and also expressed in HEK293A cells ( Supplementary Fig. 2b, Supplementary Fig. 5a) 35 . Therefore, we generated HEK293 cell lines with the MAP4K4/ 6/7 triple deletion (tKO; Fig. 4a). MAP4K4/6/7 deletion modestly decreased the basal level of YAP phosphorylation based on phos-tag gel analysis, although it did not have significant impacts on density-induced YAP phosphorylation (Fig. 4b). Moreover, the MAP4K4/6/7-tKO cells showed decreased or delayed phosphorylation of LATS1/2 and YAP in response to energy stress and serum deprivation (Fig. 4c,d). LATS phosphorylation and YAP mobility shift induced by LatB were also slightly reduced in MAP4K4/6/7-tKO cells (Fig. 4e).
Serum deprivation rapidly induces YAP/TAZ phosphorylation and cytoplasmic localization in 30 min, which was, as expected, abolished in the LATS1/2-dKO cells (Fig. 4f), confirming the essential role of LATS1/2 in YAP/TAZ regulation by serum. In contrast, MST1/2 deletion caused only marginal defect on serum starvation-induced YAP/TAZ cytoplasmic localization. On the other hand, MAP4K4/6/7 deletion more significantly suppressed the serum deprivation-induced YAP/TAZ cytoplasmic localization (Fig. 4f,g). However, YAP was still translocated into nucleus on longer serum starvation. Taken together, our observations support a physiological role of MAP4K4/6/7 in Hippo pathway regulation.
Consistently, TAZ protein levels were also significantly elevated. We further found that, though high cell density still induced LATS and YAP/TAZ phosphorylation in MM-5KO cells, the levels of LATS and YAP phosphorylation were significantly lower than the WT cells (Fig. 5b, Supplementary Fig. 7b). The 2-DG-induced phosphorylation of LATS and YAP/TAZ was also abolished in the MM-5KO cells, similar to that observed in the LATS1/2-dKO cells (Fig. 5c). Serum deprivation-induced phosphorylation of LATS and YAP/TAZ was largely blocked in MM-5KO cells (Fig. 5d, Supplementary Fig 7c). Furthermore, a minimal induction of LATS and YAP/TAZ phosphorylation was observed on LatB treatment (Fig. 5e, Supplementary Fig. 7d). To evaluate the synergistic effects of MST1/2 and MAP4K4/6/7, we serum-starved cells for a longer time (60 min). Consistent with the YAP phosphorylation, MM-5KO cells showed significantly reduced YAP cytoplasmic localization in response to serum deprivation compared with MST1/2-dKO and MAP4K4/6/7-tKO cells (Fig. 5f,g). The effects on both YAP phosphorylation and localization by deleting the five kinases were much more dramatic than deleting either MST1/2 or MAP4K4/6/7 alone, suggesting that MST1/2 and MAP4K4/6/7 function in a partially redundant manner to activate LATS1/2 and, therefore, lead to YAP/TAZ phosphorylation and inactivation.
Phosphorylation of YAP and TAZ by LATS inhibits their transcriptional activity. We examined the effect of MAP4K4/6/7 on the expression of the YAP/TAZ target genes CTGF and CYR61. As expected, deletion of LATS1/2 strongly induced the CTGF and CYR61 expression. Deletion of either MST1/2 or MAP4K4/6/7 significantly increased the expression of CTGF and CYR61 (Fig. 6a). Moreover, the combined deletion of MST1/2 and MAP4K4/6/7 further elevated the expression of CTGF and CYR61, which, however, was still lower than that in the LATS1/2-dKO cells, indicating that additional LATS-activating kinases are involved in regulating YAP/TAZ transcriptional activity. Nevertheless, these data demonstrate that MST1/2 and MAP4K4/ 6/7 inhibit YAP/TAZ transcriptional activity in a partially redundant manner.
In contrast, MAP4K5 did not increase YAP phosphorylation ( Supplementary Fig. 9c). Consistently, the kinase activities of MAP4K2/3 were required for their ability to induce YAP phosphorylation (Fig. 7b). To examine the role of endogenous MAP4K1/2/3 in the Hippo pathway signal transduction, we deleted MAP4K1/2/3 in the MST1/2 and MAP4K4/6/7 five gene knockout (KO) cells (MM-5KO) to generate the knockout lines with deletion of the eight kinases (combined deletion of MST1/2 and MAP4K1/2/3/4/6/7, MM-8KO), which were confirmed by sequencing of the genomic DNAs ( Supplementary Fig. 9d-f Supplementary Fig. 10a-c)   ARTICLE clones when compared with the MM-5KO cells (Fig. 7c,  Supplementary Fig. 10d). However, it is worth noting that residual YAP phosphorylation was still observed in the MM-8KO cells, indicating the existence of additional LATS-activating kinase(s). Together, our data demonstrate that the MAP4K family kinases are physiological LATS-activating kinases and emphasize the complexity of the Hippo pathway (Fig. 7d).
Msn regulates Yki in parallel to the Hpo kinase. To examine the function of MAP4K4/6/7 in YAP/TAZ regulation in vivo, we tested whether Msn and Hpo regulate Yki in parallel, as only one homologue of each group of the kinases exists in Drosophila. We thus generated hpo mutant clones throughout wing imaginal discs and knocked down msn in the posterior compartment using the hh-Gal4 driver. Yki activity was assayed by examining the expression of the expanded-lacZ (ex-Z) reporter 39 and Cubitus interruptus expression was used to identify the anterior compartment (Fig. 8a-f). As previously reported, hpo null mutant clones, marked by lack of GFP, grew large and displayed increased ex-Z expression compared with GFPpositive WT cells (Fig. 8b,f). While knockdown of msn alone had no observable effect on ex-Z expression, hpo msn double mutant clones in the posterior compartment displayed a further increase of ex-Z expression (Fig. 8b,c,e arrowheads) compared with the single hpo clones in the anterior compartment (Fig. 8b,c,e asterisks). These data support a model in which Msn contributes to Yki inactivation in parallel to the Hpo kinase.

Discussion
Although MST1/2 are firmly established as the initiating kinases of the Hippo kinase cascade in mammals, it has been observed that MST1/2 are not absolutely required for YAP/TAZ regulation by a number of upstream signals. For example, RNA interference (RNAi) knockdown of MST1/2 does not affect LATS or YAP phosphorylation induced by serum depletion or high cell confluence 23,24 . MST1/2-dKO mouse livers showed only slightly decreased LATS phosphorylation 15 , and MST1/2-dKO murine embryonic fibroblasts displayed significant YAP phosphorylation under serum deprivation 19 . In this report, we identified MAP4K family members, including MAP4K1 (HPK1), MAP4K2 (GCK), MAP4K3 (GLK), MAP4K4 (HGK), MAP4K6 (MINK1) and MAP4K7 (TNIK), as important physiological LATS-activating kinases. Notably, a previous study implicated that some MAP4Ks inhibit YAP reporter activity, although the mechanism was not revealed 40 . Mechanistically, MAP4Ks directly phosphorylate the LATS-HM motif and activate LATS in vitro. Overexpression or deletion of MAP4Ks affects the phosphorylation and activity of LATS and YAP/TAZ. By acting in a LATS-dependent, but MSTindependent manner, MAP4Ks restrict the activity of YAP/TAZ by promoting their phosphorylation and inhibiting target gene expression. We show that MAP4Ks act in parallel to and display partial functional redundancy with MST in LATS and YAP/TAZ regulation. Furthermore, MAP4Ks, together with MST1/2, constitute the majority of LATS and YAP/TAZ regulatory activity in response to many upstream signals (Fig. 7d). It is north noting that for some upstream signals, such as serum deprivation, MAP4Ks are more important than MST1/2 in the regulation of LATS and YAP/TAZ. The Hippo pathway was named after the Drosophila Hpo gene (homologue of MST1/2), and the major signal output of the Hippo pathway is the inhibition of YAP/TAZ. This study, together with other reports, shows that the central regulatory molecule of the Hippo pathway in mammals is LATS because it is essential for YAP/TAZ regulation by virtually all signals tested. Surprisingly, the Hpo homologue MST1/2 are largely dispensable for YAP/TAZ regulation. We would like to suggest a broad and new definition of the 'Hippo pathway'. Proteins that specifically influence LATS kinase activity and the functional output of YAP/TAZ should be considered as Hippo pathway components. Even though MAP4Ks act independently of MST1/2, we propose that MAP4Ks are components of the Hippo pathway based on their direct role in LATS1/2 phosphorylation and activation.
Genetic evidence in Drosophila also supports Hpo-independent regulation of Wts. Loss of function of the Hpo kinase causes strong phenotypes in imaginal discs and in enterocytes [8][9][10]14,[41][42][43] , the differentiated intestinal cells, but not in other cells such as the more progenitor-like intestinal enteroblasts, although loss of Wts causes strong phenotypes in all of these cell types 8,44 . In addition, myristylated Wts can partially bypass Hpo to rescue wing size in hpo mutants, indicating the existence of another Wts-activating kinase in Drosophila 36 . Drosophila has two MAP4K homologues, Msn (corresponding to MAP4K4/6/7) and Happyhour (corresponding to MAP4K1/2/3/5). Hpo is a master regulator of Drosophila midgut homoeostasis by controlling intestinal stem cell proliferation through suppressing Yki-induced Upd1/2/3 synthesis and secretion from enterocytes [41][42][43][44][45] . A recent study by Li et al. showed that Msn is also important for the midgut homoeostasis through controlling Upd3 secretion from another type of midgut cells, enteroblasts, by inactivating Yki through Wts independently of Hpo 46 . In addition, overexpression of MAP4K4 increased the phosphorylation of LATS and YAP in mammalian cells. Li et al. proposed that Msn and Hpo activate Wts in a non-redundant manner 46 . However, the molecular basis of LATS activation by MAP4K4 was not revealed in the previous study. Our results are consistent with the observations described by Li and colleagues that MAP4Ks act upstream of LATS. Moreover, our study reveals a mechanistic insight into LATS activation by MAP4Ks, which directly phosphorylate the LATS-HM, thereby activating the LATS kinase. We also observed partially redundant functions of Msn and Hpo as msn knockdown slightly enhanced the hpo loss of function phenotype in imaginal disc clones (Fig. 8). Furthermore, we demonstrate that MST1/2 and MAP4Ks have partially redundant function in mammalian cells to control LATS and YAP/TAZ activity.
MAP4Ks and MST1/2 clearly have distinct functions although both can activate LATS and inhibit YAP/TAZ. First, based on the analyses of MST1/2-dKO cells, MST1/2 are the major kinases responsible for phosphorylating MOB-T35 (Fig. 1a), and (a-f) Confocal images of a third instar wing imaginal disc stained for expanded-lacZ (ex-Z) expression to assay Yki activity and for Cubitus interruptus (Ci) to mark the anterior compartment. This disc contains hpo null mutant clones, which are marked by lack of GFP expression, and has RNAi driven knockdown of msn induced specifically in the posterior compartment using the Hh-Gal4 driver. Double mutant hpo, msn clones had increased ex-Z expression (arrowheads) compared to hpo single mutant clones (asterisks). Picture in e uses a 'spectrum' colour lookup table to better show differences in ex-Z expression.
MAP4K4 appears to be ineffective in phosphorylating MOB1 ( Supplementary Fig. 3a). Second, regarding upstream signals, MAP4K4/6/7 are more important than MST1/2 in serum deprivation-induced YAP phosphorylation (Fig. 4d,f). In contrast, MST1/2 appear to play a more prominent role in the phosphorylation of LATS and YAP induced by LatB (Fig. 1e,  Fig. 4e). Third, the relationship to other Hippo pathway components is not identical between MST1/2 and MAP4K4/6/7. For instance, MST1/2 are required for KIBRA-induced YAP phosphorylation but MAP4K4/6/7 are not. In contrast, both MST1/2 and MAP4K4/6/7 require NF2 to effectively induce YAP phosphorylation. Moreover, the relationship with SAV1 is different in which SAV1 is more important for MST than MAP4Ks to induce YAP phosphorylation. Consistently, unlike MST1/2, MAP4K4 does not interact with SAV1. We propose that MST1/2 and MAP4Ks have shared, as well as distinct functions, in relaying upstream signals to the LATS kinases (Fig. 7d). The Drosophila experiment with knockdown of Msn in Hpo mutants provides in vivo data supporting the above model. We further propose that the relative contribution of MST1/2 and MAP4Ks to the Hippo pathway is dependent on the nature of upstream signals as well as cell types.
It should be noted that YAP phosphorylation was largely, but not completely, abolished in the MST1/2-MAP4K1/2/3/4/6/7-8KO (MM-8KO), indicating the existence of other LATS-activating kinases. Serum deprivation-induced YAP phosphorylation was blocked in the MM-5KO cells while high cell-density-induced YAP phosphorylation was still present, though much reduced, in the MM-8KO cells. On the other hand, the 2-DG-induced phosphorylation of LATS and YAP/TAZ was completely abolished in the MM-5KO cells. These data suggest that different LATS-activating kinases may preferentially participate in relaying different upstream signals. However, it is also apparent that there is no simple one-to-one linear relationship between upstream signals and the different subgroups of MSTs/MAP4Ks to LATS activation. Our kinome screen identified additional kinases that could phosphorylate LATS directly in vitro. Future studies are required to elucidate the function of NEK9, STK32B, and EEF2K in the Hippo pathway. This study uncovers the extraordinary complexity of the Hippo pathway, as numerous kinases relay upstream signals to LATS to regulate this pathway. This is not entirely surprising given the fact that the Hippo pathway responds to a wide range of signals, ranging from cell contact, extracellular hormones, intracellular energy status, cytoskeletal integrity, to mechanotransduction. The large numbers of LATSactivating kinases may serve to relay different upstream signals and ensure the proper regulation of this important signalling pathway, therefore to maintain precise control of cell, tissue, and organ growth and homoeostasis.

Methods
CRISPR. CRISPR genomic editing technology was used for the deletion of genes of interest 29 . The guide RNA sequences were cloned into the plasmids px459 (addgene 48319), a gift from Dr Feng Zhang 47 . The constructed plasmids were transfected into HEK293A or U2OS cells. 24 h after transfection, the transfected cells were enriched by 1 mg ml À 1 puromycin selection for 3 days and then were sorted onto 96-well plates with only one cell in each well. The clones were screened by Western blot with gene-specific antibodies and at least two independent clones for each gene deletion were used for each experiment described. The details of guide RNA sequences are provided in Supplementary Table 1.
Kinome screening. The flag-tagged human kinome constructs were transfected into HEK293A cells individually. The kinases were purified from the transfected cells by immunoprecipitation with anti-flag antibodies and protein A/G magnetic beads with a high stringency buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 50 mM NaF, 1% Triton X-100, 0.05% SDS, 0.25% Sodium Deoxycholate, 1 mM EDTA and 1 mM EGTA). The beads were subsequently washed with a buffer containing 50 mM Tris-HCl pH 7.5 and 150 mM NaCl, and then used for kinase assays with a truncated form of human LATS1 (amino acids 638-1,130) at 30°C for 30 min. A standard kinase assay buffer (0.5 mM ATP, 50 mM Tris-HCl pH 7.5, 10 mM MgCl 2 , 2 mM MnCl 2 , 0.1 mM EDTA, 2 mM DTT, 0.01% Brij 35) was used for most of the kinases except for CAMK family kinases, for which 1.2 mM calmodulin and 2.0 mM calcium were supplied. The immunoblot with an antibody detecting phosphorylated LATS hydrophobic motif was performed to detect LATS-HM phosphorylation signals. The detailed information of the tested kinases and the results are shown in Supplementary Figs 11-18.
Sequential kinase assay. Full-length GST-tagged mouse LATS2 proteins were purified from HEK293A cells treated with 10% fetal bovine serum for 45 min. The LATS2 proteins were purified by glutathione agarose slurry and eluted with glutathione. The proteins were dialyzed in 50 mM Tris-HCl pH 7.5, 150 mM NaCl, and then used in the coupled sequential kinase reactions with MST2 or MAP4K4, which was immunopurified from transfected HEK293A cells in a mild lysis buffer (20 mM Tris-HCl pH 7.5, 100 mM NaCl, 50 mM NaF, 2 mM EDTA, 1% NP40 substitute). 30 min after kinase reaction, recombinant GST-tagged YAP purified from E coli was added to the reaction as a substrate for LATS2. The second step of the kinase reaction was also performed at 30°C for 30 min.
For tetracycline-inducible gene expression, pRetroX-Tet-on-advanced and pRetrox-tight-puro plasmids (Clontech) were used to generate the stable cells according to the manufacturer's instructions. 500 ng ml À 1 Doxycycline was used to induce the gene expression for 24 h.
Immunoblot. Western blot was performed following standard methods. 7.5% phos-tag gel was used to resolve the phosphor-YAP proteins. The detailed information of the antibodies is provided in Supplementary Table 2. Uncropped blots are shown in Supplementary Fig. 19.
Immunofluorescence. HEK293A cells were sparsely seeded on fibronectin-coated plates. 24 h later, culture medium was replaced with fresh medium for 90 min before the serum deprivation. An antibody targeting YAP/TAZ (sc-101199, San Cruz Biotechnology, 1:200 dilution) was used for detection. The results were quantified in three randomly chosen fields for each sample.