Ezrin activation by LOK phosphorylation involves a PIP2-dependent wedge mechanism

How cells specify morphologically distinct plasma membrane domains is poorly understood. Prior work has shown that restriction of microvilli to the apical aspect of epithelial cells requires the localized activation of the membrane-F-actin linking protein ezrin. Using an in vitro system, we now define a multi-step process whereby the kinase LOK specifically phosphorylates ezrin to activate it. Binding of PIP2 to ezrin induces a conformational change permitting the insertion of the LOK C-terminal domain to wedge apart the membrane and F-actin-binding domains of ezrin. The N-terminal LOK kinase domain can then access a site 40 residues distal from the consensus sequence that collectively direct phosphorylation of the appropriate threonine residue. We suggest that this elaborate mechanism ensures that ezrin is only phosphorylated at the plasma membrane, and with high specificity by the apically localized kinase LOK. DOI: http://dx.doi.org/10.7554/eLife.22759.001


Introduction
All nucleated cells can polarize to generate morphologically and biochemically distinct regions at the cell surface. For example, the apical and basolateral domains of epithelial cells have distinct protein and lipid compositions, and microvilli are restricted to the apical domain. How cells maintain morphologically distinct regions of their cell surface is not clear.
We have been addressing this issue by examining how microvilli are assembled and specifically localized to the apical surface of epithelial cells (Sauvanet et al., 2015). A critical component of epithelial microvilli is ezrin, the founding member of the closely-related ezrin/radixin/moesin (ERM) protein family, that serves as a regulated membrane-F-actin linking protein (Bretscher, 1983(Bretscher, , 1989Fehon et al., 2010). Genetic knockout of ezrin in the mouse results in enterocytes with shorter and disorganized microvilli. In the fruit fly, loss of the single ERM protein is lethal, but when selectively knocked out in photoreceptor cells, microvilli are lost (Karagiosis and Ready, 2004;Saotome et al., 2004;Speck et al., 2003). Thus, ERM proteins provide a critical function in polarized morphogenesis. ERM proteins are regulated by a reversible head-to-tail interaction ( Figure 1A). Like all ERMs, ezrin contains an N-terminal FERM domain that binds the plasma membrane and a C-terminal F-actin-binding domain (ezrin-CTD) that can attach to the underlying actin filaments that make up the core of microvilli (Gould et al., 1989;Turunen et al., 1994). In the closed inactive state, the FERM domain is tightly associated with the~80 residues of ezrin-CTD, masking the membrane association and F-actin-binding sites (Gary and Bretscher, 1995;Pearson et al., 2000;Reczek and Bretscher, 1998). Linking these two regions is a~150 residue a-helical region that folds into an anti- parallel coiled coil hairpin in the closed structure (Li et al., 2007). In the open structure, the unmasked FERM domain binds the plasma membrane, the unmasked C-terminal domain binds F-actin, and the a-helical region presumably unravels. The unmasked FERM domain can bind many proteins, including the NHERF family adaptor proteins EBP50 and E3KARP, and the trans-membrane proteins CD44, ICAMs, and b-dystroglycan (Mori et al., 2008;Reczek et al., 1997).
The transition between closed and open ERMs requires phosphorylation of a specific threonine residue (T567 in ezrin; T564 in radixin and T558 in moesin) (Nakamura et al., 1995;Simons et al., 1998). Additionally, in vivo phosphorylation was found to require binding of the FERM domain to the plasma membrane phospholipid phosphatidylinositol 4,5-bisphosphate (PIP 2 ) (Fievet et al., 2004;Hao et al., 2009). Structural analysis of the radixin FERM domain in the presence or absence of IP 3 led to the suggestion that PIP 2 binding to FERM directly changes the conformation so as to partially release the C-terminal domain to make the phosphorylation site accessible (Hamada et al., 2000). Consistent with this model, in the closed configuration, the T567 that becomes phosphorylated is buried within the interface between the FERM and the ezrin-CTD (Li et al., 2007;Pearson et al., 2000).
Despite the findings above, how ezrin activation is restricted to the apical membrane was not resolved. At steady state, about one half of the ezrin in an epithelial cell is phosphorylated. Moreover, in vivo phosphorylation of ezrin turns over on the time-scale of minutes, similar to the lifetime of microvilli. These observations strongly suggest a morphogenetic principle: a dynamic system of local ezrin activation by phosphorylation coupled with delocalized deactivation by dephosphorylation. Lymphocyte-oriented-kinase (LOK) and its close paralog Ste20-like-kinase (SLK) were then found to be the major ezrin kinases in epithelial cells. Remarkably, LOK and SLK are currently the only kinases known to selectively localize to the apical membrane of epithelial cells . LOK had previously been identified as the kinase that phosphorylates ezrin in lymphocytes (Belkina et al., 2009).
LOK and SLK belong to the germinal center-like kinase (GCK) -V subfamily of kinases. They consist of a conserved N-terminal kinase domain, a less conserved intermediate region, and a moderately conserved C-terminal domain ( Figure 1A). A single ortholog exists in Drosophila melanogaster (Slik) and Caenorhabditis elegans (GCK4). In the fly, Slik has been identified as the sole kinase that phosphorylates the single ERM ortholog, moesin (Carreno et al., 2008;Hipfner et al., 2004;Kunda et al., 2008). It is notable that LOK is highly selective for members of the ERM family, as the kinase's target sequences requires a conserved tyrosine at position À2 relative to the substrate threonine (Belkina et al., 2009).
Here, we describe an in vitro system to reveal why ezrin has to bind PIP 2 to become a substrate for phosphorylation by LOK. As far as we are aware, this is the first example of a protein that has to bind a phosphoinositide lipid to serve as a kinase substrate. Our results indicate that PIP 2 binding to the ezrin FERM domain transmits a conformational change through the a-helical region to weaken the FERM/ezrin-CTD association. PIP 2 -priming of ezrin permits the LOK C-terminal domain to act as  Figure 1-source data 1), ****p<0.0001. (D) In vitro kinase assay showing that 18 mM ezrin is specifically primed by 90 mM of PIP 2 micelles and not by IP 3 or other phospholipids at 90 mM concentrations. Blots are derived from same membrane. Data are presented as mean ± SE, n = 3, two-way ANOVA (See also Figure 1-source data 1), ****p<0.0001. (E) In vitro kinase assay showing that unilamellar liposomes DOPC:PIP 2 (90 mol% DOPC, 10% PIP 2 ) or DOPC:DOPS:PIP 2 (80 mol% DOPC, 10 mol% DOPS, 10 mol% PIP 2 ) promote phosphorylation of 18 mM ezrin by 10 nM LOK, whereas DOPC (100 mol % DOPC) or DOPC:DOPS (70 mol% DOPC, 30 mol% DOPS) fail to promote LOK-mediated ezrin phosphorylation. (F-G) 10 nM LOK-N phosphorylates 18 mM ezrin-CTD but not full-length ezrin in presence of 90 mM PIP 2 . Data are presented as mean ± SE, n = 3, two-way ANOVA (See also Figure 1source data 1), ***p<0.0002, ****p<0.0001. Total ezrin is shown in red and phosphorylation of T567 in green in dual color Western blots. DOI: 10.7554/eLife.22759.002 The following source data and figure supplement are available for figure 1: Source data 1. Experimental replicates for Figure 1B a wedge between the FERM and ezrin-CTD to allow the kinase domain to gain access and phosphorylate T567. In support of this model, we design a chimeric protein and show that it can recapitulate the activity of LOK in vitro and ex vivo. Further, we define a distal site in ezrin about 40 residues from T567 that is also necessary for phosphorylation by the kinase domain. In addition to being required for ezrin phosphorylation, the LOK C-terminal domain negatively regulates the kinase activity of LOK. Thus, the PIP 2 -dependent mechanism of ezrin phosphorylation by LOK involves several distinct steps, thereby ensuring the specificity of the reaction.

Results
In vitro phosphorylation of ezrin by LOK requires PIP 2 and the LOK C-terminal domain We first established an in vitro phosphorylation assay for purified LOK using the isolated ezrin-CTD as substrate. LOK phosphorylated the ezrin-CTD as detected using pT567 antibody ( Figure 1B), but failed to phosphorylate full-length ezrin ( Figure 1C). We therefore explored the possible role of phospholipid phosphatidylinositol 4,5-bisphosphate (PIP 2 ) as previous reports have suggested a regulatory role for this phospholipid (Fievet et al., 2004;Hao et al., 2009). In the presence of PIP 2 micelles, LOK was readily able to phosphorylate ezrin ( Figure 1C), but PIP 2 did not influence phosphorylation of the ezrin-CTD by LOK ( Figure 1-figure supplement 1A). The requirement for PIP 2 was specific, as other phosphoinositides, or soluble IP 3 , failed to substitute for PIP 2 in phosphorylation of ezrin ( Figure 1D). The requirement of PIP 2 for ezrin phosphorylation was not simply a negative charge effect as PIP 2 -dependency was reproduced in 100-nm-sized synthetic unilamellar liposomes of DOPC or DOPC:DOPS containing 10 mol% PIP 2 , which represent a more physiological system ( Figure 1E).
We next explored which domains of LOK are required for ezrin phosphorylation. LOK consists of an N-terminal kinase domain (LOK-N) and a C-terminal region (LOK-C) containing putative polo kinase kinase (PKK) domains of unknown function ( Figure 1A). In vitro, both full-length LOK and LOK-N can phosphorylate ezrin-CTD, although the isolated kinase domain was less active ( Figure 1F). Whereas intact LOK phosphorylated ezrin in a PIP 2 -dependent manner, LOK-N was unable to phosphorylate ezrin, even in the presence of PIP 2 micelles and unilamellar DOPC:DOPS liposomes containing 10 mol% PIP 2 ( Figure 1G and Figure 1-figure supplement 1B). We conclude that in vitro phosphorylation of full-length ezrin requires both PIP 2 and the C-terminal domain of LOK.

In vitro phosphorylation requires active participation by the central ahelical domain of ezrin
The FERM domain of ezrin is linked to the ezrin-CTD by a~150 residue a-helical coiled coil hairpin, as represented in the structure of the close homolog Sfmoesin ( Figure 2A) (Li et al., 2007). This region consists of three a-helices that connect the FERM domain to the ezrin-CTD via an antiparallel coiled coil (Li et al., 2007). In the absence of the a-helical region, the individual FERM and ezrin-CTD domains bind avidly together (Pearson et al., 2000). To explore whether binding of PIP 2 to the FERM domain simply weakens the association with the ezrin-CTD as proposed (Hamada et al., 2000), we asked whether a complex of isolated GST-ezrin-CTD and FERM could be phosphorylated by LOK or LOK-N in the presence of PIP 2 . Remarkably, while GST-ezrin-CTD was phosphorylated by LOK and LOK-N, we found that the isolated GST-ezrin-CTD+FERM complex could not be phosphorylated by LOK or LOK-N in the presence of PIP 2 ( Figure 2B). Furthermore, LOK failed to phosphorylate GST-ezrin-CTD in complex with FERM in presence of DOPC:DOPS:PIP 2 unilamellar liposomes (Figure 2-figure supplement 1). We conclude that the central a-helical region is required for translating PIP 2 -binding to the FERM domain into a weakened association between FERM and ezrin-CTD domains, thus allowing in vitro phosphorylation of ezrin by LOK.
PIP 2 primes ezrin to serve as an efficient substrate for phosphorylation by LOK Having established the broad features of the PIP 2 -dependent phosphorylation of ezrin by LOK, we explored the kinetic parameters of the phosphorylation reactions in order to gain more insight into the mechanism (Table 1 and Figure 1-figure supplement 1). Full-length LOK and just the kinase domain, LOK-N, had about the same affinity for the ezrin-CTD (K m 252 and 176 mM, respectively), but the maximal turnover rate for LOK-N was lower (k cat of 19.8 min À1 for LOK versus 1.4 min À1 for LOK-N). In the presence of PIP 2 , LOK phosphorylated ezrin with a K m of 5 mM and displayed a lower maximal turnover (k cat of 8.5 min À1 ). In the absence of PIP 2 , LOK hardly phosphorylated ezrin, and the K m was difficult to measure, but was of the order of 800 mM. Thus, PIP 2 enhances the affinity of  Table 1. A comparison of the kinetic constants for kinase-substrate pairs based on 32 P incorporation. Concentration of kinase in each experiment was 10 nM. Data are represented as mean ± SE, n = 3 (See also Figure 1-figure supplement 1C-D and Table 1-source data 1).  LOK for ezrin by about 160 fold. LOK-N phosphorylated intact ezrin so poorly that it was not possible to derive a K m value, emphasizing the importance of LOK C-terminal domain. Our data reveal that the catalytic efficiency (k cat /K m ) of LOK for ezrin and ezrin-CTD is 1.56 and 0.08 mM À1 min À1 , respectively. Thus, as a substrate for LOK, ezrin is about twenty-fold better than the ezrin-CTD. Since the affinity of LOK for ezrin is greatly enhanced by PIP 2 , and this affinity is much higher than LOK for the ezrin-CTD, the LOK C-terminal domain (LOK-C) must actively engage full-length ezrin in a PIP 2dependent manner.
The LOK C-terminal domain serves as a wedge between the FERM and ezrin-CTD domains To explore if LOK-C binds specific domains of ezrin, equilibrium-binding curves were generated. We measured the depletion of LOK-C from the supernatant of reactions containing sedimented beads with either immobilized ezrin-CTD or FERM domain ( Figure 3A and B). Measurements revealed that the LOK-C binds both the free FERM and ezrin-CTD domains with a K d of about 10 mM and 2 mM, respectively. We also attempted to measure the association of the LOK-C with full-length ezrin in the absence and presence of PIP 2 , but minimal binding could be detected (data not shown). Thus, priming of ezrin by PIP 2 does not unmask binding sites sufficiently to enable the determination of a binding constant. The combination of the kinetics and binding experiments lead to a model for the basic mechanism that allows the LOK kinase domain to gain access to T567 ( Figure 3C). Ezrin binds PIP 2 and this induces a conformational change that is transmitted through the a-helical linker domain to transiently loosen the association between the FERM and ezrin-CTD. The loosening of FERM-ezrin-CTD interaction allows LOK-C to enter between the FERM and ezrin-CTD -working like a wedge -to render T567 accessible to the kinase domain.
The PIP 2 -dependent wedge mechanism of ezrin phosphorylation can be recapitulated with a chimeric kinase Our model implies that site(s) in ezrin for binding LOK-C are masked in the closed, inactive state. This is also true of the binding site for ezrin-binding protein of 50 kD (EBP50). EBP50 consists of two PDZ domains and a C-terminal tail region. The latter binds tightly to a surface on the FERM domain that is masked in closed full-length ezrin Reczek and Bretscher, 1998). Structural work has shown that the C-terminal tail of EBP50 (EBP50t) binds to the same surface of the FERM domain that is occupied by a C-terminal a-helix belonging to ezrin-CTD, thereby explaining why the binding site for EBP50t is masked in the full-length inactive protein (Finnerty et al., 2004). We therefore tested if EBP50t could compensate functionally for the LOK C-terminal domain in phosphorylation of PIP 2 -primed ezrin. We generated and purified a LOK-N-EBP50t chimera ( Figure 4A) and used it in in vitro phosphorylation assays ( Figure 4B). As expected, both LOK-N-EBP50t and LOK could phosphorylate the ezrin-CTD. When presented with full-length ezrin, neither kinase could phosphorylate ezrin in the absence of PIP 2 , while both LOK and LOK-N-EBP50t could phosphorylate ezrin robustly when PIP 2 was included. Thus, the chimeric protein supports the PIP 2dependent wedge model, by interaction of C-terminal EBP50t with ezrin FERM, and consequently compensating for LOK-C, when ezrin is primed by PIP 2 .
To see if LOK-N-EBP50t could mimic LOK ex vivo, we explored its localization in cultured placental choriocarinoma JEG3 cells in which LOK has been shown to selectively localize to the apical domain . Upon transient expression, both LOK and LOK-N-EBP50t localized to microvilli, whereas LOK-N did not ( Figure 4C). To see if LOK-N-EBP50t could compensate functionally for LOK or SLK, using CRISPR/Cas9, we generated a JEG-3 cell line lacking LOK and knocked down endogenous SLK by siRNA expression. As with double siRNA knockdown , kinase suppression reduced ezrin T567 phosphorylation by over 80% and resulted in loss of microvilli (Figure 4-figure supplement 1A and B). Introduction of LOK or LOK-N-EBP50t into these cells restored microvilli, whereas LOK-N failed to restore microvilli ( Figure 4D). LOK-N-EBP50t also restored ezrin phosphorylation in kinase-suppressed cells (Figure 4-figure supplement 1C). Thus, LOK-N-EBP50t is able to recapitulate both the enzymatic properties of LOK as well as its function in microvilli formation, via restricting activated ezrin to the apical membrane ( Figure 4E).

The LOK C-terminal domain is involved in auto-inhibition
High expression of LOK-C in JEG-3 cells inhibits ezrin phosphorylation and results in the loss of microvilli. However, in low-level expressing cells, LOK-C is localized to regions of individual microvilli with local loss of ezrin . Thus in cells, LOK-C has the properties of a negative regulatory domain.
We therefore examined the effect of LOK-C in the phosphorylation of ezrin or ezrin-CTD in vitro. In these assays, we employed LOK or LOK-N-EBP50t at 10 nM, or LOK-N at 30 nM and the ezrin or ezrin-CTD substrates at 18 mM. In the presence of 90 mM PIP 2 , LOK-C inhibited the ability of LOK or LOK-N-EBP50t to phosphorylate ezrin with an IC 50 of 1.9 ± 0.1 mM and 1.6 ± 0.1 mM, respectively ( Figure 5A and C). Similarly, the inhibition curves with ezrin-CTD as substrate showed an IC 50 for LOK of 1.9 ± 0.2 mM and for LOK-N of 1.2 ± 0.2 mM ( Figure 5B and C). As these IC 50 values were all much lower than the 18 mM substrate concentration, inhibition by LOK-C is due to a direct effect on the LOK kinase activity rather than affecting the substrate. These similar IC 50 values also suggest that the mechanism of inhibition is similar in all cases, namely competitive inhibition. Moreover, since

LOK-N requires a docking site distal to the T567 consensus sequence
Kinases require a well-defined consensus sequence immediately flanking the phosphorylation site. In the case of LOK, there is a strong preference for a tyrosine at P-2 (Y565), which is conserved in all ERM proteins (Belkina et al., 2009). In order to explore the presence of additional regions that contribute to specific LOK phosphorylation of ezrin, we used N-terminal truncations of the ezrin-CTD to assess their ability to serve as substrates for LOK and LOK-N. As shown earlier, LOK-N has a slightly lower activity than LOK for GST-ezrin-CTD ( Figure 6A). Robust phosphorylation was seen up to GST-ezrin-520-585. The shorter constructs, GST-ezrin-530-585 and GST-ezrin-555-585, were very poor substrates for LOK and LOK-N. As this suggests that the region 520-530 is important for recognition by LOK-N, we made a construct, GST-ezrin-490-(D520-530)À585, lacking this region. This truncated substrate was a significantly poorer substrate than GST-ezrin-490-585. Thus, the ezrin-CTD has an additional distal docking site approximately 40 residues N-terminal to the kinase consensus sequence.

Discussion
In order to maintain the morphological polarity of a cell, specific morphogenetic components have to be controlled locally. In the case of microvilli on epithelial cells, ezrin is locally activated through T567 phosphorylation mediated by the closely related LOK and SLK kinases. Here, we define a novel multi-step mechanism for specific phosphorylation of ezrin by LOK ( Figure 6B). In the first step, binding to PIP 2 induces a conformational change in ezrin. In the second step -that has to occur coincidentally with the first -the C-terminal domain of LOK (LOK-C) binds as a wedge between the FERM and ezrin-CTD domains. The wedge mechanism allows for the third step in which the kinase domain binds to a distal docking site in the ezrin-CTD upstream of the T567 phosphorylation site. The final step can now proceed, namely the phosphorylation of T567 within the appropriate consensus sequence setting. LOK itself is negatively regulated by LOK-C, and binding ezrin appears to relieve this inhibition. In the following narrative, we expand on each of these steps. The evidence for the first step is that LOK requires PIP 2 to phosphorylate ezrin ( Figure 6B, step 1). Since PIP 2 is a regulatory lipid largely confined to the plasma membrane, the highly specific priming of ezrin by PIP 2 ensures that ezrin phosphorylation is restricted to this membrane compartment. Our in vitro studies show that both phosphatidylinositol 4,5-bisphosphate (PIP 2 ) micelles and PIP 2containing unilamellar liposomes prime ezrin for LOK-mediated phosphorylation, while other phospholipids and more importantly, IP 3 , the head group in direct physical contact with ezrin, fail to prime ezrin for phosphorylation. From these data, we conclude that priming of ezrin for phosphorylation requires an appropriate charge distribution provided by phosphate groups in the head group of the phospholipid phosphatidylinositol 4,5-bisphosphate, and a high local concentration of the head group that can only be achieved in vitro using micelles or liposomes. Earlier studies have Source data 1. Source data for quantification of microvilli in Figure 4D and shown that PIP 2 binds the FERM domain and it has been suggested that this interaction results in a conformational change to loosen the associated FERM-ezrin-CTD (Ben-Aissa et al., 2012;Hamada et al., 2000). Therefore, we were surprised to find that in the presence of PIP 2 , LOK is unable to phosphorylate a complex of FERM associated with the ezrin-CTD. This revealed two points: First, PIP 2 acts through ezrin activation, and not by LOK activation, and second, the central The following source data is available for figure 5: Source data 1. Data summary and analysis for Figure 5C, based on experimental replicates represented in Figure 5A-B. DOI: 10.7554/eLife.22759.015 a-helical coiled coil hairpin connecting the FERM to the ezrin-CTD is necessary for priming of intact ezrin for phosphorylation. The detailed function of the central coiled coil hairpin is unknown, but one attractive possibility is that the a-helical coiled coil region acts as a spring counteracting the strong association between the FERM and ezrin-CTD, and upon binding PIP 2 , the strength of the spring is enhanced. The proposed spring model has similarities in other proteins. In the ERM-related protein merlin, the central a-helical hairpin undergoes an unfolding event, like a 'switchblade', during opening of the protein (Li et al., 2007). Another example occurs within the influenza virus hemagglutinin protein. This protein is a membrane-fusion glycoprotein with a rod-shaped a-helical ectodomain that undergoes a dramatic 'switchblade-like' unfolding event at low pH (Skehel and Wiley, 2000). The active involvement of LOK-C ( Figure 6B, step 2) is shown by the finding that the isolated LOK kinase domain cannot phosphorylate ezrin under any condition, whereas it can phosphorylate the isolated ezrin-CTD. Moreover, in the presence of PIP 2 , the K m of full-length LOK for ezrin is about 45 times lower than for the ezrin-CTD, indicating that the combined action of PIP 2 and the FERM domain greatly enhances specificity. Consistent with the kinetics of PIP 2 -primed full-length ezrin, we found that LOK-C can bind with low affinity to both the isolated FERM and ezrin-CTD. Thus, PIP 2 priming of ezrin loosens the association between the FERM and ezrin-CTD, allowing LOK-C to pry the domains apart, acting like a wedge, and thus permitting the kinase domain access to T567 in the ezrin-CTD. Indeed, using a cross-linking approach ex vivo, we have shown that LOK is recovered more efficiently with open ezrin rather than with closed ezrin-T567A that cannot be phosphorylated, suggesting that LOK has an affinity for sites masked in closed ezrin ( Figure 6B, step 5) (Viswanatha et al., 2013). Moreover, in support of this model, we have shown that a chimeric protein consisting of the LOK kinase domain fused to the domain of EBP50 that binds to a site on the FERM domain normally masked in the closed protein, can recapitulate the PIP 2 -dependent phosphorylation of ezrin in vitro and also partially ex vivo.
In this study we reveal that LOK-C is a negative regulator of the kinase domain as it inhibits the activity of both LOK and LOK kinase domains toward ezrin or ezrin-CTD. Since LOK phosphorylates ezrin-CTD better than the isolated kinase domain does, binding the ezrin-CTD likely relieves the inhibition of full-length LOK by LOK-C in cis. LOK-C might have additional functional implications for kinase activity. The C-terminal domain of SLK shares 56% sequence identity with LOK-C and undergoes homodimerization, a process that enhances kinase activity in vitro and ex vivo (Delarosa et al., 2011;Sabourin and Rudnicki, 1999). Potential dimerization and autoactivation of LOK remains to be studied.
The specificity of kinases depends on local peptide consensus sequences, which in the case of ezrin phosphorylation by LOK involves the hydrophobic residue Y565 positioned -2 residues from T567 (Belkina et al., 2009). As is seen with some other serine/threonine kinases (Biondi and Nebreda, 2003;Sharrocks et al., 2000;Ubersax and Ferrell, 2007), we have also identified a distal docking site within residues 520-530 in the ezrin-CTD, required for the efficient phosphorylation of T567 by either LOK or the isolated kinase domain ( Figure 6B, steps 3-4), thereby adding another level of specificity to the kinase reaction. Binding of the kinase to this site may also relieve the inhibition imposed by LOK-C in the context of the full-length LOK. It is interesting to note that the residues equivalent to ezrin 520-530 are relatively conserved, and Y565 and T567 perfectly conserved, in all ERM proteins, but not in the closely related merlin that also undergoes a similar FERM/merlin-CTD interaction, where the threonine equivalent to ezrin-T567 is not phosphorylated (Nguyen et al., 2001;Sher et al., 2012).
To maintain morphological polarity, there are multiple levels of selectivity that ensure the specific phosphorylation of ezrin by LOK. At the genetic level, cells lacking LOK exhibit greatly reduced ezrin phosphorylation (Figure 4-figure supplement 1A). On the subcellular level, LOK-C targets the kinase to the apical membrane . At the molecular level, we have uncovered the simultaneous requirement for PIP 2 and LOK in an elaborate multiple-step mechanism to ensure specific and localized phosphorylation of ezrin. While substrate specificity of protein kinases has been the subject of many studies (Ubersax and Ferrell, 2007), fewer cases of conditional specificity have been examined in detail, with LOK-mediated ezrin phosphorylation being the first example of a specific requirement for a phosphoinositide binding to the substrate. As complex cell behaviors rely on the proper timing of phosphorylation events, such intimate co-regulation between substrates and kinases as described here may emerge as a general theme.

In vitro kinase assay and inhibition assay
Unilamellar liposomes were generated according the following recipes: DOPC (100 mol% DOPC), DOPC:PIP 2 (90 mol% DOPC, 10% PIP 2 ), DOPC:DOPS (70 mol% DOPC, 30 mol% DOPS), DOPC: DOPS:PIP 2 (80 mol% DOPC, 10 mol% DOPS, 10 mol% PIP 2 ), with 1 mol% DiR near-infrared dye (Avanti Polar Lipids, Alabaster, AL) to aid in visualization. Lipids were mixed, vacuum dried and hydrated in Kinase assay buffer (20 mM Tris pH 7.4, 140 mM NaCl, 1 mM EGTA, 1 mM DTT), followed by extrusion through 100 nm filters to generate liposomes. For micelles, phospholipids were vacuum dried and hydrated in Kinase assay buffer and sonicated in water bath prior to kinase assay. In general, IP 3 or phospholipids were added at 90 mM final concentration or at a 1:5 (substrate:phospholipid) molar ratio, and at a maximum concentration of 180 mM due to limited solubility in Kinase assay buffer. Unilamellar liposomes were used at a final concentration of 1 mM resulting in a final concentration of~100 mM PIP 2 . Kinase assays were performed in Kinase assay buffer, 500 mM MgCl 2 and 200 mM ATP at 37˚C for 15 min. 10 nM of purified kinase was used to phosphorylate 18 mM of ezrin, GST-ezrin-CTD or GST-ezin-CTD truncations. In the case of GST-ezrin-CTD+FERM complex, 18 mM GST-ezrin-CTD was pre-incubated with 18 mM FERM for 10 min on ice prior to kinase assay. In the inhibition assay, purified LOK-C was added to individual reaction at concentrations of 0-65 mM. 20% of each kinase reaction was analyzed by SDS-PAGE and immunoblotting.

Enzyme kinetics
Kinetic assays were initiated by the addition of 500 mM MgCl 2 and 200 mM ATP supplemented with 0.1 mL of 3000 Ci/mmol g-32 P ATP (PerkinElmer Inc., Waltham, MA) and incubated at 37˚C for 15 min. Reactions were stopped by the addition of 12.5 ml of 10% phosphoric acid and processed as previously described (Hastie et al., 2006). Briefly, samples were spotted on P81 cellulose paper (GE Healthcare, Pittsburg, PA) and washed three times for 10 min in 75 mM phosphoric acid to remove unincorporated g-32 P ATP. Filter papers were allowed to dry and then placed into scintillation vials containing 2.5 mL of Ecoscint fluid (National Diagnostics, Atlanta, GA). 32 P incorporation was measured using PerkinElmer Liquid Scintillation Counter, Tri-Carb 2810TR (PerkinElmer Inc., Waltham, MA).

Equilibrium binding assay
200 nM of purified LOK-C was diluted in 20 mM Tris pH 7.4, 150 NaCl and incubated with either 0-15 mM GST-ezrin-CTD immobilized on glutathione agarose or 0-18 mM CNBr-FERM for 15 min at RT. Samples were centrifuged at 5000Âg for 10 min, supernatants corresponding to 15% of total volume were removed, mixed with sample buffer and loaded on a 10% acrylamide gel for SDS-PAGE. Gels were stained with Coomassie blue and destained in 50% methanol and 10% acetic acid. Densitometric quantification was performed in ImageStudio 5.2.5 (LI-COR Biosciences, Lincoln, NB) by quantifying Coomassie signals after background correction.

Western blotting and densitometry
Western blots were performed as described . For analysis of endogenous ezrin phosphorylation, cells were rapidly boiled in 5X reducing sample buffer in kinase buffer, diluted 3fold, vortexed vigorously, and cleared by centrifugation prior to SDS-PAGE. For densitometric analysis of phospho-ezrin in relation to total ezrin, mouse monoclonal ezrin antibody and rabbit polyclonal pT567 antibody were detected with infrared fluorescent secondary antibodies donkey anti-mouse Alexa Fluor 647 (Thermo Fisher Scientific Cat# A-31571 RRID:AB_162542) or IRDye 800CW goat anti-rabbit (LI-COR Biosciences Cat# 827-08365 RRID:AB_10796098). Membranes were imaged using Odyssey CLx imager (Odyssey CLx , RRID:SCR_014579). Densitometric quantification was performed in ImageStudio 5.2.5 (LI-COR Biosciences, Lincoln, NB) by measuring pT567 and total ezrin after background correction. Total GST-ezrin-CTD was quantified from Coomassie stains. For SLK, The turnover number, k cat , was calculated using the following equation, For calculation of IC 50 , data were normalized to LOK/ezrin+PIP 2 or LOK/ezrin-CTD data sets in which phosphorylation signal at 0 mM LOK-C, defined as 'Top', was set to 100%. Data were fitted in GraphPad Prism using the four parameter logistic equation, Data are represented as means ± standard error (S.E.) from three independent experiments. Twoway ANOVA followed by Tukey's multiple comparisons test was performed using GraphPad Prism.