Angiomotins stimulate LATS kinase autophosphorylation and act as scaffolds that promote Hippo signaling

The Hippo pathway controls cell proliferation, differentiation, and survival by regulating the Yes-associated protein (YAP) transcriptional coactivator in response to various stimuli, including the mechanical environment. The major YAP regulators are the LATS1/2 kinases, which phosphorylate and inhibit YAP. LATS1/2 are activated by phosphorylation on a hydrophobic motif (HM) outside of the kinase domain by MST1/2 and other kinases. Phosphorylation of the HM motif then triggers autophosphorylation of the kinase in the activation loop to fully activate the kinase, a process facilitated by MOB1. The angiomotin family of proteins (AMOT, AMOTL1, and AMOTL2) bind LATS1/2 and promote its kinase activity and YAP phosphorylation through an unknown mechanism. Here we show that angiomotins increase Hippo signaling through multiple mechanisms. We found that, by binding LATS1/2, SAV1, and YAP, angiomotins function as a scaffold that connects LATS1/2 to both its activator SAV1–MST1 and its target YAP. Deletion of all three angiomotins reduced the association of LATS1 with SAV1–MST1 and decreased MST1/2-mediated LATS1/2-HM phosphorylation. Angiomotin deletion also reduced LATS1/2's ability to associate with and phosphorylate YAP. In addition, we found that angiomotins have an unexpected function along with MOB1 to promote autophosphorylation of LATS1/2 on the activation loop motif independent of HM phosphorylation. These results indicate that angiomotins enhance Hippo signaling by stimulating LATS1/2 autophosphorylation and by connecting LATS1/2 with both its activator SAV1–MST1/2 and its substrate YAP.

A major regulator of Hippo signaling is the mechanical environment, which includes factors such as cell density, substrate stiffness, and cell stretch or mechanical tension. How these mechanical stimuli act through the core Hippo regulators to control YAP activity is not clear. Several studies have shown that the actin cytoskeleton, which responds dynamically to mechanical changes, is a major regulator of Hippo signaling (16 -20). Reduction in F-actin activates LATS1/2. How F-actin controls LATS1/2 activity is not certain but may involve angiomotins (AMOT, AMOTL1, and AMOTL2) (21)(22)(23). Angiomotins bind to both LATS1/2 and YAP and can inhibit YAP through two mechanisms: binding and retention of YAP in the cytoplasm/plasma membrane and activation of LATS1/2 (21, 24 -27). Other studies have shown that AMOT can control YAP nuclear/cytoplasmic localization, depending on its phosphorylation state (28). Angiomotins also bind the LATS1/2 activator NF2 (29,30), and a study using proximity labeling found that SAV1 and MOB1 may interact with AMOT (31). We and others have proposed that angiomotins may act as scaffolds to promote LATS1/2 activation. A conserved N-terminal region in the long form of AMOT (AMOT130, hereafter referred to as AMOT) contains an F-actin binding motif and flanking L/PPXY sites that bind WW-containing proteins (such as YAP). We have shown previously that F-actin and YAP com-pete for binding to angiomotins (22), which may provide one mechanism for how F-actin can influence YAP activity; when F-actin levels go down, angiomotins are free to bind and inhibit YAP in the cytoplasm and possibly activate LATS1/2. Angiomotins are required for relocalization of YAP from the nucleus to the cytoplasm when F-actin is disrupted (22). In addition, elimination of angiomotin regulation of YAP causes patterning defects in early mouse embryos (26,27), transformation of Madin-Darby canine kidney cells (25), and epithelial to mesenchymal transition in MCF10A cells (32). How angiomotins activate LATS1/2 is not known. Here we show that the angiomotin protein AMOT interacts with multiple core Hippo pathway regulators to stimulate LATS1/2 to phosphorylate YAP via three distinct mechanisms. First, AMOT promotes interaction between LATS1/2 and its activators MST1/2 and SAV1. Second, AMOT enhances LATS1/2 phosphorylation of YAP by promoting interaction between the two proteins. And third, AMOT collaborates with MOB1 to enhance LATS1/2-AL autophosphorylation, even in the absence LATS-HM site phosphorylation.

Angiomotins promote LATS1/2 activation
Phosphorylation at the LATS2-HM site is presumed to be the primary regulatory site for LATS1/2 because, when phosphorylated, it promotes autophosphorylation in the AL of the kinase to make it fully active (6,10,11), allowing it to phosphorylate YAP and other substrates. We tested the effect of angiomotins on phosphorylation at the LATS1/2-HM site. Overexpression of all three angiomotins (AMOT, AMOTL1, and AMOTL2) promoted LATS2-HM phosphorylation (Fig. 1A). (Note that in this study all experiments utilize the long form of AMOT (AMOT130)). Consistent with these observations, deletion of all three angiomotin genes (Amot-3KO) in HEK293 cells using CRISPR (Fig. S1, A and B) resulted in reduced activating phosphorylation of LATS1/2 at the HM and AL sites after F-actin disruption by Latrunculin B (Lat B) treatment (Fig. 1B). Similar results were observed in HeLa cells with all three angiomotins knocked down using siRNA (Fig. S1C). Consistent with our findings, other studies showed that knockdown of angiomotins reduces LATS1/2 activity, as judged by reduced YAP Ser 127 phosphorylation in HEK293T cells, Madin-Darby canine kidney cells, and MCF10A cells (21,25,32). Together, these findings show that angiomotins can activate LATS1/2 by promoting phosphorylation at the HM site in response to F-actin perturbation.
To better understand how angiomotins activate LATS1/2, we tested which core Hippo components are required for angiomotins to promote HM phosphorylation. We found that, in HEK293 cells, MST1/2 are the primary LATS1/2-HM kinases because deletion of MST1/2 (Fig. S1D) causes a major decrease in HM phosphorylation in unperturbed and Lat B-and okadaic acid-treated cells (Fig. S1, E and F). Consistent with this observation, the increase LATS2-HM phosphorylation after angiomotin overexpression was largely lost in MST1/2-deleted cells (Fig. 1A), suggesting that angiomotin-stimulated LATS2-HM phosphorylation works primarily through MST1/2 in HEK293 cells. Similar results were observed in the absence of SAV1 and NF2 ( Fig. 1C and Figure S1D), the MST1/2 and LATS1/2 binding partners, respectively. Note that not even the phosphomimetic (active) form of AMOT (AMOT-175E) is able to activate LATS1/2 in the absence of SAV1 or NF2. These results are consistent with the idea that angiomotins may act together with the two kinase modules (MST-SAV1 and LATS1/2-NF2) to promote LATS1/2 activation.

MST1/2 and SAV1 stimulate AMOT-LATS2 binding
While examining how MST1/2, SAV1, and NF2 contribute to angiomotin activation of LATS1/2, we discovered a potential positive feedback loop whereby MST1/2 and SAV1 promote AMOT-LATS2 binding. When we tested whether MST1/2, SAV1, and NF2 affect AMOT-LATS2 binding, we observed that deletion of MST1/2 or SAV1 greatly reduced binding of AMOT to LATS2 (NF2 deletion reduced binding, but the effect did not reach the level of significance) (Fig. 1, D and E; see Fig. S1G for further antibody controls). One explanation for these results could be that AMOT-LATS2 binding requires LATS2 activity and that MST1/2 and SAV1 are required for LATS2 activation. This would be consistent with our previous study showing that AMOTL2 bound better to WT than kinase-dead LATS2 or LATS2 with both sites of activating phosphorylation mutated to alanine residues (21). We tested whether AMOT had similar binding preferences. This experiment showed that AMOT, like AMOTL2, bound preferentially to WT LATS2 compared with LATS2 with both sites of activating phosphorylation mutated ( Fig. 2A). Because LATS2 is known to phosphorylate AMOT on Ser 175 (22,26,(33)(34)(35), these results could be explained if either LATS2 phosphorylation of AMOT promotes AMOT-LATS2 binding or if AMOT preferentially binds to the active form of LATS2. Consistent with the first model, we observed, as seen previously (26), that a phosphomimetic version of AMOT (AMOT-175E) binds to (Fig. 2B) and activates LATS2 (Fig. S2A) better than the nonphosphorylatable form of AMOT (AMOT-175A). However, because expression of AMOT-175E did not rescue the AMOT-LATS2 binding defect in MST1/2-KO cells, additional factors must contribute to the AMOT-LATS2 binding defect in MST1/2-KO cells (Fig. 2C). Therefore, we tested the second model; namely, that angiomotins bind preferentially to active LATS2. If correct, then expression of a phosphomimetic version of LATS2 (LATS2-1041E) should rescue the defect in AMOT-LATS2 interaction in MST1/2-deleted cells. Indeed, this turned out to be the case (Fig. 2D). These results show that MST1/2 and SAV1 activation of LATS1/2 stimulates a positive feedback mechanism by promoting AMOT-LATS1/2 binding.

AMOT is a Hippo pathway scaffold protein
AMOT could activate LATS1/2 either by stimulating MST1/2 activation or by promoting the ability of MST1/2 to phosphorylate LATS1/2. Our results support the later possibility because AMOT overexpression did not affect the levels of MST1/2-activating phosphorylation (Fig. S2B). (Note that, to remove potential issues with feedback from LATS1/2, we conducted these and subsequent experiments with AMOT-175E.) We next investigated whether AMOT might function as a scaf- fold to bring LATS1/2 together with its activator MST1/2 and substrate YAP. This and previous studies showed that AMOT binds LATS1/2 (21,36). Thus, one potential mechanism of action for AMOT could be that it connects LATS1/2 with the MST1/2-SAV1 complex and YAP. Indeed, when we analyzed endogenous AMOT complexes using chromosomally tagged sfGFP-MAP-AMOT (37) in HEK293 cells (created using CRISPR-mediated genome modification), we could detect LATS1, SAV1, MST1, and YAP coming down with sfGFP-MAP-AMOT isolated using streptavidin beads (note that part of the sfGFP-MAP tag is a streptavidin-binding peptide) (Fig.  3A), consistent with endogenous AMOT associating with core Hippo pathway proteins. We next examined whether AMOT might contribute to association of core Hippo pathway proteins by overexpressing AMOT with different combinations of Hippo pathway proteins followed by immunoprecipitation. Because AMOT bound LATS1/2, we tested whether AMOT could stimulate LATS2 activation by also binding MST1. Although we were unable to observe binding between AMOT and MST1 when both proteins were co-expressed ( Fig.  3B), AMOT did bind to SAV1 in a manner that depended on the SAV1 WW domains and the 3 L/PPXY motifs (WW domain ligands) in AMOT (Fig. 3C). SAV1 expression allowed MST1 to come down in AMOT immune complexes (Fig. 3B), indicating that SAV1 could bridge interaction between AMOT and MST1. These results support the notion that AMOT could connect LATS2 to SAV1-MST1. However, as observed previously in Drosophila (14), SAV1 was also able to bind LATS2 (Fig. 3D) (presumably by interaction of its WW domain with the PPXY motif in LATS2), showing that it can act as a scaffold for MST1 and LATS2 without AMOT (at least when overexpressed). Interestingly, the amount of SAV1 that came down in LATS2 immune complexes was enhanced by expression of AMOT ( Fig. 3D). Similar experiments were also done to examine how the interaction of MST1 with LATS2 might depend on SAV1 and AMOT (Fig. 3E). MST1 and LATS2 were found to not interact unless SAV1 was co-expressed (Fig. 3E). As with the SAV1-LATS2 interaction (Fig. 3D), AMOT alone could not promote interaction between LATS2 and MST1 but was able to enhance the interaction of MST1 and SAV1 with LATS2 (Fig.  3E). Together, these results suggest that AMOT may enhance the assembly/stabilization of a complex between SAV1-MST1 Activation of LATS1/2 by AMOT and LATS2. In addition, AMOT may also act as a scaffold to bring LATS1/2 in proximity with its substrate YAP. We observed that, when YAP and LATS2 were co-expressed, YAP was only poorly recovered in LATS2 immune complexes (Fig.  3F) unless AMOT was also expressed. The ability of AMOT to enhance LATS2-YAP binding depended on the L/PPXY motifs in AMOT (Fig. 3F). This result suggested that YAP binding by AMOT might not just inhibit YAP by sequestering it in the cytoplasm as reported previously but might also bring it together with its inhibitory kinase LATS1/2. Consistent with this model, Amot-3KO cells show defects in YAP phosphorylation (Fig. S3A) and are defective in keeping YAP out of the nucleus after growth to high density or serum deprivation (Fig.  S3B). We also saw defects in association of endogenous core Hippo pathway proteins in the absence of angiomotins. Although immunoprecipitates of endogenous LATS1 from HEK293 cells contained SAV1 and YAP, both proteins were almost completely absent in LATS1 immunoprecipitates from Amot-3KO cells (Fig. 4A). Similarly, when we immunoprecipitated endogenous SAV1 from HEK293 cells, we could detect MST1, LATS1, AMOT, and YAP in the immune complexes. However, SAV1 immune complexes from Amot-3KO cells contained MST1 but almost no YAP and LATS1 (Fig. 4B). This indicates that, at endogenous protein levels, SAV1 alone cannot efficiently scaffold LATS1/2 with MST1/2. Collectively, these results show that AMOT may regulate LATS1/2 and YAP at least in part by acting as a scaffold to connect LATS1/2 to both its activator SAV1-MST1/2 as well as its target YAP.

Angiomotins promote LATS1/2-AL autophosphorylation
The ability of AMOT to promote a complex containing SAV1, MST1, and LATS2 could explain why LATS1/2-activating phosphorylation at the HM site (the site phosphorylated by MST1/2) is reduced in Amot-3KO cells' response to Lat B treatment (Fig. 1B). Similarly, we observed that LATS1/2-HM phosphorylation is reduced in Amot-3KO cells compared with WT when MST1 is overexpressed (Fig. 5A). The defect in LATS2-HM phosphorylation in Amot-3KO cells in response to MST1 overexpression could be rescued by co-overexpression of SAV1 (Fig. 5B), consistent with SAV1 being able to replace the scaffolding function of angiomotins, at least when overex-  A, HEK293 (WT) and Amot-3KO HEK293 cells were grown to high density and serum-starved for 4 h. Immunoprecipitation (IP) was performed on cell lysates with either anti-LATS1 or control (IgG) antibodies, and immune complexes and cell lysates were analyzed by Western blotting for AMOT, SAV1, LATS1, and YAP levels. LATS1 bg, residual LATS1 signal after stripping. B, HEK293 (WT) and Amot-3KO HEK293 cells were grown as in A. Immunoprecipitations were performed on cell lysates with either anti-SAV1 or control (IgG) antibodies, and immune complexes and cell lysates were analyzed by Western blotting for AMOT, MST1, SAV1, LATS1, and YAP levels.

Activation of LATS1/2 by AMOT
pressed. Interestingly, although Amot-3KO cells overexpressing MST1 and SAV1 had HM phosphorylation levels similar to WT controls expressing MST1 and SAV1, they still had a defect in LATS2 phosphorylation at the AL site (Fig. 5B). These results raised the possibility that angiomotins may promote LATS2 autophosphorylation at the LATS2-AL site, independent of their effect on MST1/2 phosphorylation at the LATS2-HM site. Note that the defect in LATS2-AL phosphorylation in Amot-3KO cells could be rescued by re-expression of the three angiomotin proteins (Fig. S1B). To distinguish between direct effects of AMOT on LATS2-AL phosphorylation and indirect effects via LATS2-HM phosphorylation, we tested whether AMOT could promote AL phosphorylation of a version of LATS2 that mimicked the HM-phosphorylated form (LATS2-1041E) but could not be regulated by MST1/2 (Fig. S4A) or other upstream HM kinases. We used LATS2-1041E because we assumed that AMOT alone might not be able promote LATS2-AL phosphorylation without HM phosphorylation. We found that AMOT expression enhanced AL phosphorylation of LATS2-1041E (Fig. 5C). Thus, AMOT can promote LATS2 autophosphorylation at the AL site independent of any effects it has on LATS2-HM phosphorylation. The only other protein known to stimulate LATS1/2 autophosphorylation is MOB1 (10,12). We observed that, when co-expressed (Fig. S4B), MOB1A could be detected in AMOT immune complexes, showing that the two proteins might interact. Therefore, we investigated the relationship between MOB1 and AMOT in promoting LATS2-AL phosphorylation. Several lines of evidence suggested that MOB1 may act independent of angiomotins and that the effects of AMOT on LATS1/2-AL phosphorylation may depend on MOB1. For example, MOB1A binding to LATS2 was not affected by deletion of all three angiomotins (Fig. S4C), and overexpression of AMOT was not able to significantly enhance AL phosphorylation of LATS2-1041E, which had a point mutation in the MOB-binding domain (MBD, R657A) (11), rendering it unable to bind MOB1 (Fig.  5D). In addition, strong overexpression of MOB1A could stimulate similar levels of AL phosphorylation of LATS2-1041E in angiomotin-deleted cells compared with WT cells (Figs. 5E). Interestingly, despite having the same levels of LATS-AL phos-phorylation, the angiomotin-deleted cells had a significant defect in YAP phosphorylation (Fig. 5E), consistent with our earlier results suggesting a scaffolding role for angiomotins to promote LATS1/2 phosphorylation of YAP. Other evidence suggests that MOB1A and AMOT may have additive or synergistic effects on LATS1/2-AL phosphorylation. When both MOB1A and AMOT were moderately overexpressed (at levels where each alone barely enhanced LATS2-AL phosphorylation), we observed that co-expression of both proteins significantly increased AL phosphorylation of LATS2 (Fig. 5F) beyond that observed for either protein alone. These results prompted us to test whether AMOT and MOB1A could promote AL phosphorylation even in the absence of HM phosphorylation (using LATS2-1041A). We found that, although moderately expressed AMOT and MOB1A alone did not increase AL phosphorylation, expression of both proteins significantly increased AL phosphorylation of LATS2-1041A, albeit to lower levels than with WT LATS2 (Fig. 5F). We think that the AL phosphorylation of LATS2-1041A was due to autophosphorylation and not phosphorylation by another kinase or endogenous LATS1/2 because a LATS2 kinase-dead mutant did not show AL phosphorylation after MOB1A and AMOT expression (Fig. 5F). Together, these experiments indicate that AMOT and MOB1A can promote LATS2 autophosphorylation on the AL site independent of HM phosphorylation by upstream kinases.

Discussion
Previous studies showed that angiomotins promote LATS1/2 activity and YAP phosphorylation, but the mechanism was not known (21,25,32). We show here that angiomotins carry out these functions through multiple modes of action (Fig. 6). Several lines of evidence are consistent with AMOT acting as a scaffolding protein. Previous studies have shown that angiomotins bind LATS1/2, NF2, and YAP (21, 22, 24 -27, 29, 30), and here we show that they can also bind SAV1-MST1/2. Thus, they have the ability to bring LATS1/2 together with both their activator and substrate. Interestingly, the SAV1 protein has also been proposed to have a scaffolding function because it binds LATS1/2 (15) and MST1 (38 -41). However, we found that cells lacking angiomotins show severe defects in binding between endogenous SAV1-MST1 and LATS1/2, suggesting that, at physiological levels, angiomotins are important for SAV/MST-LATS complex stability. This raises the question of how Hippo signaling complexes are assembled in animals like Drosophila that lack angiomotins. One possible explanation is a rearrangement in the PPXY motif ligands for the SAV WW domains. In Drosophila, SAV binds to the LATS1/2 homolog WARTS through interaction between its WW domains and the five PPXY motifs in WARTS (14). Although a similar mechanism could operate in humans, it may not be as robust because LATS1 and LATS2 contain fewer PPXY motifs than Drosophila WARTS (LATS1 and LATS2 have two and one PPXY motifs, respectively). AMOT could act to recruit SAV1-MST1 to AMOT-LATS2 complexes via its three L/PPXY motifs (two PPXY motifs and one LPXY motif), which could supply additional binding sites for the WW domains of SAV1. Thus, Drosophila may not need angiomotins to recruit WW domain proteins to WARTS because WARTS has five PPXY motifs instead of the two and one in LATS1 and LATS2, respectively. Utilization of the PPXY domains of AMOT to recruit WW domain proteins to an AMOT-LATS1/2 complex could provide another level of regulation because we have previously shown that the L/PPXY motifs of AMOT are masked in the presence of F-actin. Here we showed that the phosphomimetic point mutant (AMOT-175E) that disrupts F-actin binding by AMOT (22,26,33,34) (thereby making the L/PPXY motifs of AMOT available for binding to WW domains) is better than the WT at promoting LATS2-HM phosphorylation. Together, this would create a mechanism whereby a reduction in F-actin frees AMOT to bind the WW domains of both SAV1 and YAP, allowing it to promote LATS1/2 activation and YAP phosphorylation by bringing LATS1/2 together with its activator SAV1-MST1/2 and its target YAP.
This model may also be relevant to our observation showing weak co-immunoprecipitation between YAP and LATS2, except for when AMOT was co-expressed, and reduced interaction between endogenous LATS1 and YAP in Amot-3KO cells. Here again, AMOT might be acting to provide additional L/PPXY sites for the WW domains of YAP to bind. Previous reports have given differing results regarding interaction between LATS2 and YAP, with some showing that LATS2 interacts with YAP via the first WW domain of YAP (42)(43)(44), whereas another report failed to observe co-immunoprecipitation between YAP and LATS2 (45). These studies could be consistent with the interaction between YAP and LATS2 being weak but observable when high enough levels of each protein are expressed. We showed previously that a competition for binding to AMOT between F-actin and YAP caused a reduction in F-actin levels to trigger AMOT to bind YAP and sequester it in the cytoplasm (22). Our current results suggest that the increased binding of AMOT to YAP caused by reduction in F-actin levels may help connect LATS1/2-AMOT complexes to YAP to promote its phosphorylation and cytoplasmic sequestration.
The other major discovery of this study was that AMOT can stimulate LATS2 to autophosphorylate on its activation loop. Previously, MOB1 was the only protein known to enhance LATS1/2-AL phosphorylation (10,12). We observed that coexpression of AMOT with MOB1A greatly enhanced LATS2-AL phosphorylation more than expression of either protein alone. Because both AMOT and MOB1A promote LATS1/2-HM site phosphorylation, and this phosphorylation can trigger LATS1/2 to autophosphorylate at the AL site, we needed to be able to analyze LATS2-AL phosphorylation in isolation. To do this, we used LATS2 mutants that had the HM phosphorylation site mutated to either nonphosphorylatable alanine or phosphomimetic glutamate. These results showed that activation or increased levels of AMOT and MOB1 can activate LATS2 even without activation by upstream kinases. The physiological relevance is at this point not clear. It is possible that AMOT and MOB1 always act in conjunction with LATS1/2-HM phosphorylation to promote activation of LATS1/2 through AL phosphorylation, or, alternatively, they might function under some circumstances to activate LATS2 independent of HM phosphorylation. Further studies will be required to address these possibilities. In addition, we show that both MOB1 and AMOT enhance LATS-AL phosphorylation independent of any effects they have on HM phosphorylation. D, the ability of fully active LATS to phosphorylate YAP may be enhanced by formation of a LATS-AMOT-YAP complex. Note that the model is speculative and its purpose is to illustrate the primary steps involved in AMOT activation of Hippo signaling. The Hippo pathway is regulated at many other levels, which could be occurring at the same time but are not shown for simplicity. It is presently not clear whether complexes exist in cells as pictured, with all components simultaneously bound to each other.

Cell transfection, immunoprecipitation, and Western blotting
For the experiment in Fig. S1B, AMOT, AMOTL1, and AMOTL2 were transfected at low levels into HEK293 Amot-3KO cells with FuGENE HD (Promega) according to the manufacturer's protocol. Rescued A3KO cells were trypsinized 8 h after transfection, replated, and incubated overnight. The next morning, cells were treated with Lat B as indicated. Lipofectamine 2000 (Invitrogen) was the transfection reagent of choice for all other experiments. Because the Hippo pathway is activated by cell resuspension, all protein was collected from cells snap-frozen with liquid nitrogen in the plates on which they were grown. For the experiments shown in Figs. 3A and 4, A and B, a lysis buffer containing glycerol was used (10% glycerol, 20 M Tris HCl (pH 7), 137 mM NaCl, 2 mM EDTA, and 1% NP40). The lysis buffer for all other experiments contained 1% NP-40, 150 mM NaCl, 2 mM EDTA, 6 mM Na 2 HPO 4 , and 4 mM NaH 2 PO 4 . Both lysis buffers contained phosphatase and protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 1 mM Na 3 VO 4 , and mammalian protease inhibitor mixture (Sigma)). Lysis buffers were added to frozen cells, and cells were solubilized with the help of a cell scraper and back-and-forth pipetting. Lysates were cleared by centrifugation at 15,000 ϫ g for 5 min at 4°C. Dynabeads (Invitrogen) were used for all immunoprecipitation experiments according to the manufacturer's protocol. Nonspecific IgG antibodies were typically used as a control for immunoprecipitation experiments. For experiments involving coimmunoprecipitation of proteins with different epitope tags, control experiments were done to show that antibodies against the tag on the protein being immunoprecipitated did not bind to the co-expressed protein being tested for co-immunoprecipitation (Fig. S1G).

CRISPR-mediated KO cell line development
Target sequences were selected using the web tool developed by the Zhang lab at the Massachusetts Institute of Technology (http://crispr.mit.edu). 3 Complementary oligos (below) containing the target sequence (uppercase) and appropriate overhangs (lowercase) were annealed and cloned into a variant of the px330 plasmid with puromycin resistance (46): AMOT, caccGCCATACACCAGCAAGCCAC and aaacGTGGCTTG-CTGGTGTATGGC; AMOTL1, caccgCAAGTTCATGT-TCTCGGTTG and aaacCAACCGAGAACATGAACTTGc; AMOTL2, caccGCAGCGTGCGCGTCTCAGTC and aaacG-ACTGAGACGCGCACGCTGC; MST1, caccGGATCGTTA-TGGAGTACTGT and aaacACAGTACTCCATAACGATCC; MST2, caccGTTATGGAGTACTGTGGCGC and aaacGCG-CCACAGTACTCCATAAC; SAV1, caccGTTGGAATTGTT-GGACCATGc and aaacGCATGGTCCAACAATTCCAAc; NF2, caccgGAACTCCATCTCGGCGTCCA and aaacTGGA-CGCCGAGATGGAGTTC. KO cell lines were generated by transfecting 500 ng of the px330-based plasmid containing the CRISPR target sequence into HEK293 cells plated in 12-well plates using Lipofectamine 2000 (Invitrogen). The next day, cells were placed under selection for 48 h with 1 g/ml of puromycin (Gibco). Puromycin-resistant cells were then heavily diluted and plated on 10-cm plates for colony isolation. Clonal lines were then expanded, and the expression of the target gene was determined by Western blotting. Clonal lines that lacked expression of the target gene were reisolated from single cells and tested by Western blotting to ensure clonality. The sfGFP-MAP-AMOT-expressing cells were created similarly to the KO cell lines, but instead of px330 we used MLM3636 (Addgene, 43860) to generate the guide RNA (oligo sequences acac-cGAATTTCTCATCTCTATTGCg and aaaacGCAATAGAG-ATGAGAAATTCc), which cuts near the start codon for AMOT; JDS246 (Addgene, 43861) to express Cas9; and a rescue plasmid containing sfGFP-MAP (37) flanked by ϳ700 bp of genomic sequence immediately upstream and downstream of the AMOT start codon.
Pooled puromycin-resistant cells were used for further analysis. Western blotting using AMOT and GFP antibodies was used to confirm that the cells expressed full-length sfGFP-MAP-AMOT.