Addressing the Functional Determinants of FAK during Ciliogenesis in Multiciliated Cells

Abstract

We previously identified focal adhesion kinase (FAK) as an important regulator of ciliogenesis in multiciliated cells. FAK and other focal adhesion (FA) proteins associate with the basal bodies and their striated rootlets and form complexes named ciliary adhesions (CAs). CAs display similarities with FAs but are established in an integrin independent fashion and are responsible for anchoring basal bodies to the actin cytoskeleton during ciliogenesis as well as in mature multiciliated cells. FAK down-regulation leads to aberrant ciliogenesis due to impaired association between the basal bodies and the actin cytoskeleton, suggesting that FAK is an important regulator of the CA complex. However, the mechanism through which FAK functions in the complex is not clear, and in this study we examined the role of this protein in both ciliogenesis and ciliary function. We show that localization of FAK at CAs depends on interactions taking place at the amino-terminal (FERM) and carboxyl-terminal (FAT) domains and that both domains are required for proper ciliogenesis and ciliary function. Furthermore, we show that an interaction with another CA protein, paxillin, is essential for correct localization of FAK in multiciliated cells. This interaction is indispensable for both ciliogenesis and ciliary function. Finally, we provide evidence that despite the fact that FAK is in the active, open conformation at CAs, its kinase activity is dispensable for ciliogenesis and ciliary function revealing that FAK plays a scaffolding role in multiciliated cells. Overall these data show that the role of FAK at CAs displays similarities but also important differences compared with its role at FAs.

Introduction

FAK² is a non-receptor-tyrosine kinase that becomes activated and functions at FA complexes in response to several stimuli, including activation of integrins after binding on substrates of the extracellular matrix and growth factor receptor activation. At FAs, FAK acts as a regulator of the actin cytoskeleton, cell adhesion, and migration as well as a regulator of cell proliferation and survival (1, 2). These functions of FAK are tightly linked to its ability to interact with and regulate a variety of proteins. When recruited at FAs FAK becomes auto-phosphorylated at tyrosine 397, which leads to the recruitment of Src and the phosphorylation of several downstream targets, including paxillin, p130Cas, and tensin (3–9). Most of FAK's interactions are mediated through its amino-terminal (FERM) and carboxyl-terminal (FAT) domains, which flank the central kinase domain, whereas the linkers that connect these structural domains contain three proline-rich regions (PRR1–3) and other sites also responsible for interactions with additional binding partners.

The FERM domain of FAK shares homology with the band-4.1/ERM family of proteins, which act as linkers between the membrane and the cytoskeleton (10, 11). It has been implicated in a plethora of interactions that regulate the activity of FAK and downstream signaling. Such interactions include binding to growth factor receptors and phosphatidylinositol 4,5-biphosphate (PIP₂) (2, 12–15). Moreover, the FERM domain has been shown to interact with the Arp2/3 complex and N-WASP, providing a link between integrin activation and remodeling of the actin cytoskeleton, which is critical for cell spreading and migration, whereas it has also been proven to be responsible for the interaction of FAK with the ERM protein ezrin (16–18). In vitro studies have also shown direct binding of FERM to the cytoplasmic tail of integrin β1, but this interaction has not been confirmed in vivo (19). Even though when exogenously expressed FERM fails to localize at FAs (20), studies suggest that it is involved in targeting FAK at FAs and that it has a crucial role in controlling the dynamics of FAK at these complexes (21). In addition, FERM is thought to be responsible for the localization of FAK at membrane structures and cell-cell junctions (22, 23). Furthermore, the FERM domain has long been proposed to regulate the enzymatic activity of FAK (24–26). This is mediated through intramolecular, autoinhibitory interactions of FERM through the F2 lobe, which binds directly to the C-lobe in the kinase domain and the FERM's F1 lobe, which interacts with the activation site Tyr-397. These interactions retain FAK in a closed-inactive state through steric inhibition of the access to the activation site and the catalytic cleft by activating proteins (25, 27). In addition, recent data by Brami-Cherrier et al. (28) suggest that the FERM domain mediates intermolecular interactions leading to dimerization of the protein (FERM-FERM interaction), which is a critical step for its activation. It was proposed that dimerization takes place specifically at FAs and that the dimers are stabilized through an additional interaction between the FERM and the FAT domain (FERM-FAT interaction). Interestingly, binding of paxillin on the FAT domain of FAK appears to further strengthen the FERM-FAT interaction and, therefore, the stabilization of the FAK dimmers (28).

The FAT domain is a highly conserved four-helix bundle with a hydrophobic core shown to be both necessary and sufficient to target FAK at FAs (29, 30). This is believed to be mediated through interactions with other FA proteins and mainly through an interaction with paxillin (31). Two hydrophobic patches (HP1 and HP2) of the FAT domain are responsible for binding paxillin, and each one engages one of the two paxillin LD motifs (leucine-rich motifs) responsible for the interaction with FAK (LD2 and LD4). Importantly, this requires the integrity of the four-helix bundle structure of the FAT domain (29, 32, 33). Mutations of conserved amino acids reveals that either of the two patches is sufficient for binding paxillin and that two mutations (I936E/I938E) are needed for this interaction to be completely abolished (29). Loss of paxillin interaction affects the localization of FAK at FAs and is thus believed to be critical for FA targeting; however it is not considered the sole determinant as some mutants of FAK that can not bind paxillin can still localize at FAs (34, 35). In addition, the interaction of FAK, through its FAT domain, with talin has also been implicated in FAK's FA localization (36). Specifically, the interaction between FAK and talin was shown to be important for the regulation of FAs, and loss of talin leads to impaired localization of FAK at mature FAs (36, 37). However, other data suggest a role for FAK in driving talin at nascent adhesions (20, 38, 39). Moreover, the FAT domain is responsible for mediating interactions with additional binding partners including p190RhoGEF and Grb2 (40–45).

In a previous study we identified FAK as an important regulator of ciliogenesis in multiciliated cells. We showed that FAK and other FA proteins, such as paxillin, talin and vinculin, co-localize and form a complex at a region close to the basal foot and at the ends of the ciliary rootlets of motile cilia. We named these complexes CAs and showed that they connect the basal bodies and their rootlets to the apical and subapical networks of actin, respectively. CAs form during basal body migration and connect the basal bodies to an internal actin network responsible for their transport to the apical cell surface (46). Although little is known about CA regulation, recent work has revealed that Cp110 is required for the normal formation of CA complexes in multiciliated cells and that Cp110 knockdown leads to aberrant ciliogenesis through defects in basal body transport (47). Basal body migration is a critical step in the differentiation of multiciliated cells during which multiple, newly formed centrioles/basal bodies travel from the cytoplasm, where they are formed, to the apical surface of the cells, where they dock (48–51). This process has been shown to be driven by directed vesicular trafficking and is believed to depend on an intact actin-myosin network (50–56). In addition to its role in the transport of the basal bodies, the actin cytoskeleton has also been shown to be involved in other aspects of ciliogenesis including basal body docking, spacing, tissue level, and rotational polarity (54, 56–59). Loss of FAK in the multiciliated cells of the Xenopus epidermis leads to aberrant ciliogenesis due to impaired basal body-actin association and consequently impaired actin-mediated processes, including defective basal body migration, docking, and spacing (46).

In this study we focused on an in-depth characterization of the role of FAK in multiciliated cells and attempted to identify the determinants for both the localization and function of the protein at CAs. We show that FAK's function during ciliogenesis depends on the ability of the protein to localize at CAs, suggesting that its involvement is direct and not through FAK's role in cell adhesion. Importantly, CA localization, unlike FA localization, requires both the amino- and carboxyl-terminal domains, FERM and FAT. We go on to show that the interaction with paxillin, through the FAT domain, is necessary for localization and function at CAs. Interestingly however, we show that the kinase activity of FAK is dispensable during both ciliogenesis as well as for proper ciliary function in the multiciliated epithelium of the Xenopus epidermis, suggesting that FAK acts as a scaffolding molecule at CAs.

Results

The FAT Domain Is Necessary for the Recruitment of FAK at CAs and for Its Function during Ciliogenesis

The domains of FAK have been extensively studied with respect to their role in the context of FAs. Despite many parallels and similarities between FAs and CAs, the absence of integrins from CAs suggests that the two complexes are established through different mechanisms. In an effort to identify the domains of FAK that are important for CA localization and function during ciliogenesis, we initially focused on FAK's carboxyl-terminal domain. The carboxyl-terminal domain of FAK (FA targeting, FAT domain) has been shown to be both necessary and sufficient for FAK's localization at FAs (30). To examine a possible role of the FAT domain in targeting FAK to CAs, we initially used a deletion mutant that lacks the FAT domain (FAK Δ FAT) (60) and compared its localization to wild type (WT) FAK. We co-injected the mRNA of GFP FAK Δ FAT with that of centrin RFP (a basal body marker) at the two ventral blastomeres of four-cell stage embryos. Live imaging of multiciliated cells from stage 32 embryos revealed that FAK lacking the FAT domain fails to localize at CAs of differentiated multiciliated cells, unlike WT FAK, which localizes adjacent to the basal bodies in a polarized manner (Fig. 1, a, a′ and b, b′). In addition, as shown in Fig. 2, unlike FAK, FAK Δ FAT does not associate with the migrating basal bodies of stage 18–20 embryos, stages when ciliated cells are intercalating into the epidermis (Fig. 2, e, e′ and f, f′). These data suggest that the FAT domain is necessary for the recruitment of FAK at CAs. We next wanted to examine if the FAT domain is sufficient for CA localization. We thus generated a GFP FAT construct and examined its localization in the multiciliated cells of the epidermis of live embryos. As shown in Fig. 1, c and c′, and Fig. 2, g and g′, GFP FAT surprisingly fails to localize at CAs as efficiently as WT FAK, suggesting that the FAT domain is not sufficient to recruit FAK at these complexes.

FIGURE 1.

FAK is recruited at CAs through its FERM and FAT domains. Merged and single channel (for GFP) maximum intensity projections of multiciliated cells, expressing centrin RFP (red in merged images) and GFP FAK or GFP-fused FAK deletion mutants. a and a′, full-length FAK displays strong localization at CAs, and each basal body is associated with a punctum of GFP FAK. CA complexes appear as bright puncta throughout the cell, displaying an anteroposterior intensity gradient. b and b′, FAK Δ FAT does not associate with the basal bodies as efficiently as FAK, and there are no bright CA puncta, showing that the FAT domain is necessary for proper localization at CA complexes. c and c′, the FAT domain is not sufficient to target FAK at CAs, and only few faint CA-puncta are present in the cell. d and d′, GFP FAK Δ FERM localizes strongly on the rootlets, whereas the apical CA localization displayed by full-length FAK (a and a′) is absent. e and e′, the FERM domain is not sufficient to drive FAK at CAs, and only few faint puncta, corresponding to CAs, are present. f and f′, GFP FF displays strong localization at CAs similarly to full-length FAK showing that combination of the FERM and FAT domains is sufficient for proper localization of FAK. Scale bars: 5 μm.

FIGURE 2.

Interaction with paxillin is not the sole determinant for FAK's CA localization. Merged and single channel (for GFP) maximum intensity projections of multiciliated cells expressing centrin RFP (red in merged images) and GFP FAK or GFP-fused FAK mutants. a and a′, full-length FAK displays strong localization at CAs. b and b′, talin binding-deficient FAK (E1015A) displays strong localization at CAs, indicating that interaction with talin is not necessary for correct localization. Bright puncta in close association with basal bodies are present throughout the cell. c and c′, FAK L1034S displays strong localization at CAs, whereas FAK I936E/I998E (d and d′) fails to associate with the basal bodies. e–j, Optical sections of intercalating multiciliated cells expressing centrin RFP and GFP-fused FAK mutants. e and e′, wild type FAK localizes at CAs during basal body migration. CA-localized FAK appears as bright puncta associated with the basal bodies. Neither GFP FAK Δ FAT (f and f′) nor GFP FAT (g and g′) associate with migrating basal bodies, suggesting that neither the FAT domain nor the FERM domain (present in GFP FAK Δ FAT) are individually sufficient for localization at CAs. h and h′, GFP FAK E1015A localizes at CAs similarly to WT FAK. GFP FAK L1034S (i and i′) associates with the migrating basal bodies, whereas GFP FAK I936E/I998E (j and j′) fails to do so. Scale bars: 5 μm.

The above results suggest that the FAT domain has an important role in recruiting FAK at CAs, but it is not sufficient for correct localization. We went on to examine if the FAT domain is important for the function of FAK during ciliogenesis. To do so, we tested if the FAK Δ FAT mutant is able to rescue the ciliogenesis defects elicited by FAK down-regulation. We thus used a previously characterized FAK morpholino (MO) to down-regulate FAK in multiciliated cells as previously described (21, 46, 61). As shown in Fig. 3b, morphant cells have fewer cilia projecting from their surface compared with control cells (Fig. 3a), with the majority (80.5 ± 2.2%, mean ± S.E.) projecting very few or even no cilia (severe ciliogenesis defects) (Fig. 4a). Moreover, morphant embryos presented impaired directional flow along their epidermis when compared with control embryos, which generated robust flow along their anterior-posterior axis (Fig. 5, a and b). Co-injection of the mRNA encoding WT FAK partially rescued the ciliogenesis phenotype as 39.3 ± 2.96% of the cells projected cilia normally, whereas 41 ± 1.84% of them presented moderate defects (Figs. 3c and 4a). In contrast, despite similar expression levels (Fig. 3k) co-injection of the mRNA encoding FAK Δ FAT with the FAK MO failed to rescue the phenotype, and the majority of cells (90.9 ± 6.9%) displayed severe ciliogenesis defects (Figs. 3d and 4a), whereas the flow along these embryos remained impaired (Fig. 5d) unlike that of WT FAK expressing embryos (Fig. 5c). These results suggest that CA localization of FAK is important for its function in multiciliated cells and that the FAT domain is necessary for proper ciliogenesis and ciliary function either by recruiting FAK at CAs and/or by mediating important interactions.

FIGURE 3.

FAK's function during ciliogenesis depends on the FERM and FAT domains and interaction with paxillin but is kinase-independent. Comparison of the ability of different mutants to rescue the FAK MO elicited ciliogenesis defects. Surface view (maximum intensity projections) of the multiciliated epidermis of Xenopus tadpoles, immunostained against acetylated tubulin, which stains the cilia (red). Tadpoles have been injected with a FAK MO together with centrin RFP (gray) and different FAK constructs. Targeted cells are identified by expression of centrin RFP. a, in control embryos, cells appear normal with numerous cilia projecting from their apical surface. b, in morphant embryos, cells display severe ciliogenesis defects and have fewer or even no cilia. c, expression of WT FAK rescues the phenotype and cells project cilia normally. d, expression of FAK Δ FAT fails to rescue the phenotype showing that the FAT domain is necessary for proper function of FAK during ciliogenesis. e, expression of FAK E1015A partially rescues the phenotype and the majority of cells project cilia normally, showing that interaction of FAK with talin may only play a minor role during ciliogenesis. f and g, expression of two paxillin binding-deficient mutants cannot rescue the ciliogenesis defects, as the majority of cells have very few or no cilia, indicating that interaction with paxillin is critical for proper ciliogenesis. h, expression of FAK Δ FERM does not rescue the defects, showing that the FERM domain of FAK has an important role for its function in ciliogenesis. i, expression of FF leads to partial rescue of the phenotype. j, expression of kinase inactive (KD) FAK rescues similarly to wild type FAK showing that kinase activity is dispensable during ciliogenesis. k, expression levels of the FAK constructs used in rescue experiments analyzed in Western blots. All constructs are stable and at the expected molecular weight. Scale bars: 10 μm.

FIGURE 4.

Rescue efficiency of FAK mutants in Xenopus multiciliated cells. Quantification of the MO-induced ciliogenesis defects and rescue efficiency for the constructs shown in Fig. 3. Bar charts show the percentage of cells that are normal with respect to the number-density of cilia, present moderate (∼50% of normal cilia number) or severe defects (very few or no cilia). Results were analyzed with two tailed unpaired t tests. The error bars represent S.E. The density of cilia was demonstrated by the density of the acetylated tubulin signal over each cell. a, in control embryos 96.25 ± 0.252% of cells project cilia normally (n = 8 embryos), whereas in morphant embryos only 1.4 ± 0.43% of the cells appear to be normal (n = 9 embryos) (***, p < 0.0001). Expression of WT FAK leads to partial rescue of the phenotype (n = 9 embryos) (***, p < 0.0001 when compared with morphants). FAK Δ FAT fails to rescue, and only 0.95 ± 0.25 of the cells are normal (n = 10 embryos) (***, p < 0.0001 when compared WT FAK expressors). b, expression of FAK E1015A improves the phenotype. Only 20.45 ± 4.85% of the cells present severe defects (n = 14 embryos) (***, p < 0.0001 when compared with morphants or FAK Δ FAT expressors) while 24.85 ± 1.15% of the cells project cilia normally (**, p = 0.001 when compared to morphant embryos). However, the rescue is not as efficient as that with WT FAK (*, p = 0.049). c, expression of FAK L1034S fails to rescue (n = 12 embryos), and similar results are obtained by expression of FAK I936E/I998E (n = 10 embryos) (***, p < 0.0001 when compared with embryos expressing WT FAK). d, expression of FAK Δ FERM also fails to rescue, and the majority of cells present severe defects, whereas only 2.45 ± 2.03 of the cells are normal (n = 10 embryos) (***, p < 0.0001 when compared with embryos expressing WT FAK). e, expression of FF partially rescues the phenotype observed by loss of FAK (n = 9 embryos) (***, p < 0.0001 when compared with morphants) but not as efficiently as WT FAK (**, p = 0.0046 when comparing the amount of normal cells). f, expression of FAK KD rescues the phenotype, and only 18.73 ± 1.73% of the cells presented severe defects (n = 10 embryos) (***, p < 0.0001 when compared with morphants), whereas 35.5 ± 7.4% of the cells project cilia normally. ns, not sigficant.

FIGURE 5.

Interaction with paxillin is necessary, whereas FAK kinase activity is dispensable for the generation of ciliary flow. a–f, representation of the flow of QDs as it was tracked over the skin of tadpoles. a, control embryos display robust and directional flow. FAK morphants (b) display impaired flow, whereas expression of WT FAK (c) partially rescues the phenotype leading to improved flow. Expression of FAK Δ FAT (d) does not improve the flow, whereas in FAK I936E/998E-expressing tadpoles (e) the flow is still impaired. f, expression or kinase inactive FAK leads to improved flow along the tadpoles showing that the kinase activity of FAK is dispensable for proper function in cilia.

FAK's Interaction with Paxillin Is Necessary for Both Localization and Function during Ciliogenesis

The FAT domain has been shown to mediate interactions with various proteins at FAs, including GRB2, p190 RhoGEF, talin, and paxillin, and such interactions are important for the regulation of FA complexes (20, 29, 40, 44, 62). To dissect the role of the FAT domain and determine how it regulates FAK's function during ciliogenesis, we focused on interactions that have been previously shown to be important for FAK's FA localization and examined if these interactions are also important for recruitment to CAs. Importantly, two of the FAT-binding proteins shown to be involved in the recruitment of FAK to FAs, talin and paxillin, are also components of the CA complex (29, 30, 36, 37, 46, 63).

The interaction between FAK and talin has been shown to be important for the regulation of FAs, and loss of talin leads to impaired localization of FAK at mature FAs, whereas FAK was reported to be important for the recruitment of talin to nascent adhesions (20, 37, 64). To address whether this interaction is important for the localization of FAK at CAs, we used a previously characterized point mutant of FAK that abolishes interaction with talin (FAK E1015A; supplemental Fig. 1 and Ref. 20). The localization of FAK E1015A in relation to the basal bodies was initially examined in ciliated cells undergoing radial intercalation (stage 18–20). As shown in Fig. 2, h and h′, FAK E1015A associates with migrating basal bodies, similar to the WT protein (Fig. 2, e and e′). Imaging of stage 32-embryos expressing GFP FAK E1015A and centrin RFP revealed that the localization of GFP FAK E1015A at CAs in mature ciliated cells is almost identical to that of GFP FAK (Fig. 2, a, a′ and b, b′). These results suggest that the interaction with talin is not required for the recruitment of FAK at CAs.

We then used FAK E1015A in rescue experiments to address a possible functional role of this interaction during ciliogenesis. Expression of the talin binding mutant partially rescued the ciliogenesis defects elicited through MO-mediated down-regulation of FAK (Fig. 3e and 4b). As shown in Fig. 4b, in embryos injected with the FAK MO together with the mRNA of FAK E1015A the percentage of severely affected cells dropped dramatically (20.45 ± 4.85% compared with 80.5 ± 2.2% for MO injected embryos) to similar levels as in embryos rescued with WT FAK (19.73 ± 3.69%). The majority of cells (54.7 ± 3.7%) presented moderate defects, whereas the percentage of normal cells in FAK E1015A-expressing embryos increased significantly compared with that in morphant embryos (24.85 ± 1.15% and 1.4 ± 0.43%, respectively). This suggests that the FAK-talin interaction is not critical during ciliogenesis; however, a minor role cannot be ruled out given that FAK E1015A, despite similar expression levels to WT FAK (Fig. 3k), rescues slightly less efficiently than WT FAK. Specifically the percentage of normal multiciliated cells in FAK E1015A-expressing embryos is lower than in FAK expressing embryos, and the difference is statistically significant (24.85 ± 1.15% and 39.3 ± 2.96% respectively; *, p = 0.0491). However, the difference in the percentage of defective cells between embryos expressing FAK E1015A and FAK Δ FAT (20.45 ± 4.85% and 90.0 ± 6.9% respectively; ***, p < 0.0001) shows that many aspects of the phenotype are rescued in the absence of FAK-talin interaction, suggesting that other FAT mediated interactions are more important (Fig. 4b).

As mentioned above, another interaction mediated through the FAT domain shown to be important for the recruitment of FAK at FAs as well as for proper regulation of FA dynamics is that with paxillin (34, 35, 62). This interaction is mediated through the binding of two paxillin LD motifs (LD2 and LD4) on two hydrophobic patches (HP1 and HP2) of the four-helix bundle that makes up the FAT domain (29). To address a possible role of this interaction in the recruitment of FAK to the CAs we initially used a point-mutant (FAK L1034S; supplemental Fig. 1) shown to disrupt this interaction in immunoprecipitation experiments (65). Co-injection of the mRNA of GFP FAK L1034S with that of centrin RFP revealed that this mutant can associate with basal bodies during their migration toward the apical surface (Fig. 2, i and i′) and after they have apically docked in a similar fashion as WT FAK (Fig. 2, c and c′). We then examined if this mutant, which abolishes paxillin binding but still localizes at CAs, could rescue the ciliogenesis defects elicited by FAK down-regulation. As shown, FAK L1034S failed to rescue the ciliogenesis defects (Fig. 3f), with the majority of cells (60.6 ± 3.4%) presenting severe defects, projecting very few or even no cilia (no significant difference when compared with MO-injected, p = 0.058) (Fig. 4c). The inability of the FAK L1034S mutant to rescue, when expressed at similar levels to WT FAK (Fig. 3k), indicates that localization of FAK at CA complexes is not sufficient for proper ciliogenesis and that interaction with paxillin is necessary.

The fact that FAK L1034S was able to localize at CAs raised the possibility that FAK may be responsible or involved in driving paxillin to CAs. To explore this possibility, we examined the localization of paxillin in multiciliated cells from embryos injected with the FAK MO. As shown in Fig. 6, a, a′ and b, b′, the localization of GFP paxillin was not affected by the loss of FAK indicating that its recruitment at CA complexes is not mediated by FAK. To confirm this, we used a mutant of paxillin (paxillinC), which lacks the amino-terminal LD motifs including the FAK-binding sites of the protein (LD2 and LD4) (46), and compared its localization to that of full-length paxillin in multiciliated cells. As shown in Fig. 6, c and c′, GFP paxillinC associates with the basal bodies in an identical manner as GFP paxillin, confirming that the localization of paxillin at CAs is independent of its interaction with FAK.

FIGURE 6.

The FAK-paxillin interaction modulates FAK's affinity for the CA complex. a and b, merged and single channel (for GFP) images of the localization of GFP paxillin and centrin RFP in multiciliated cells of control (a and a′) and FAK MO injected (b and b′) tadpoles. GFP paxillin associates with the basal bodies both in the FAK-morphant cells (b and b′) as well as in control cells (a and a′). c and c′, multiciliated cell expressing GFP paxillinC and centrin RFP. GFP paxillinC localizes at the posterior area of the basal bodies like full-length paxillin (a and a′). d–d″, localization of mKate FAK and GFP FAK in multiciliated cells. Both proteins display strong localization at CAs. e–e″, localization of mKate FAK and GFP FAK L1034S in multiciliated cells. The FAK L1034S mutant displays reduced localization at CAs (e″) compared with the WT protein (e′). f, quantification of the mKate/GFP intensity ratio at CAs, from two tailed unpaired t test. Error bars represent S.E. In control cells (expressing GFP FAK) the mean ratio is 1.2 ± 0.02 (mean ± S.E.) (n = 78 CAs from 7 embryos), whereas in cells expressing GFP FAK L1034S the ratio is 2.1 ± 0.04 (n = 85 CAs from 6 embryos). g, quantification of the CA/cytosolic GFP intensity ratio, from two tailed unpaired t test. Error bars represent S.E. The ratio in cells expressing GFP FAK is higher (1.8 ± 0.03, mean ± S.E., n = 78 CAs and cytosolic regions from 7 embryos) when compare with that in GFP FAK L1034S (1.4 ± 0.02, n = 85 CAs and cytosolic regions from 6 embryos) indicating lower affinity of the L1034S mutant for CAs. Scale bars: 5 μm.

Because the FAK L1034S mutation has been shown to not only disrupt the FAK-paxillin interaction but also to affect the structure of the FAT domain and thus disrupt additional interactions mediated by this domain (40), including an interaction with RhoGEF, we decided to use a second FAK mutant (FAK I936E/I998E, supplemental Fig. 1) that specifically abolishes interaction with paxillin by compromising the two HPs without affecting the structure of the four helix bundle (29). When co-injected with the FAK MO at the ventral animal pole of eight-cell stage embryos, FAK I936E/I998E, expressed at similar levels to WT FAK (Fig. 3k), failed to rescue the ciliogenesis defects, similarly to FAK Δ FAT and FAK L1034S (Fig. 3g), or the impaired flow along the anterior-posterior axis of the tadpoles (Fig. 5e) confirming that a direct interaction with paxillin is required for FAK's role during ciliogenesis. As shown in Fig. 4c, only 3.14 ± 0.64% of the multiciliated cells appeared to be normal, and 15.04 ± 1.29% showed moderate defects in embryos expressing FAK I936E/I998E, compared with 39.3 ± 2.96% and 41 ± 1.82%, respectively, in embryos expressing WT FAK. Interestingly, GFP FAK I936E/I998E, unlike FAK L1034S, failed to localize at CAs in intercalating ciliated cells of stage 18–20 embryos (Fig. 2, j and j′) and in mature multiciliated cells (Fig. 2, d and d′). In both cases no association with the basal bodies was observed, and the localization of the paxillin binding mutant resembled that of FAK Δ FAT (Figs. 1, b and b′ and 2, f and f′).

The striking difference in localization of the FAK L1034S and FAK I936E/I998E mutants raised questions with respect to the precise role of the FAK-paxillin interaction in the recruitment of FAK to CAs. Interestingly, despite the apparent loss of FAK-paxillin interaction in immunoprecipitation experiments, the L1034S mutant still localizes at FAs, and the same is true for CAs (Fig. 2, a and c, and Fig. 7, a and b) (65). On the other hand, the I936E/I998E mutant fails to localize at both FAs and CAs (Fig. 2, a and d, and Fig. 7, a and c) (35). These data suggest that FAK's interaction with paxillin is not the sole determinant for either FA or CA localization of FAK and that the I936E/I998E mutation may disrupt interactions with additional proteins (29). Alternatively, although the L1034S mutation abolishes the FAK-paxillin interaction in immunoprecipitation experiments, the two proteins may still interact in the cell albeit with lower affinity, explaining the correct localization of this mutant when overexpressed in both cells and embryos (29, 31). We initially tested the possibility that the L1034S mutant retains some capacity to interact with paxillin. As shown, neither mutant can immunoprecipitate either endogenous paxillin or exogenous overexpressed paxillin, despite a robust ability of the WT protein to do so (Fig. 7d). These data show that both mutants have a dramatically reduced affinity for paxillin, yet their localization at both CAs and FAs differs, providing additional evidence that paxillin binding is not the sole determinant for FAK's localization. To address the possibility that this interaction modulates FAK's affinity for the CA complex, we overexpressed both WT FAK and FAK L1034S in the same cells to determine if the WT protein has a higher affinity for CAs. As shown in Fig. 6, e and e″, when co-expressed with mKate FAK, GFP FAK L1034S displays visibly reduced localization at CAs compared to WT FAK (mKate FAK), suggesting that the WT protein competes the mutant off the complex. In contrast, GFP FAK co-expressed with mKate FAK displays strong CA localization similarly to mKate FAK (Fig. 6, d and d″). To confirm this result, we quantified the mKate-to-GFP intensity ratio at CAs from multiciliated cells expressing mKate FAK together with GFP FAK or GFP FAK L1034S. As shown in Fig. 6f, in cells expressing mKate FAK and GFP FAK, the mean mKate/GFP ratio is 1.2 ± 0.02 (mean ± S.E.), suggesting that the two constructs display similar affinity for CAs. On the other hand, in cells expressing mKate FAK and GFP FAK L1034S, the mean mKate/GFP ratio is 2.1 ± 0.04 (mean ± S.E.), showing that mKate FAK associates with the basal bodies more efficiently than GFP FAK L1034S. In addition, quantification of the GFP intensity at CAs in relation to the cytosolic signal revealed a higher CA-to-cytosolic ratio in cells expressing GFP FAK (1.8 ± 0.03, mean ± S.E.) than in cells expressing GFP FAK L1034S (1.4 ± 0.02), also supporting a reduced ability of this mutant to localize at CAs (Fig. 6g).

FIGURE 7.

The interaction of FAK with paxillin is not the sole determinant for FAK's FA localization. a–c, HeLa cells expressing RFP paxillin (FA marker) and GFP FAK (a and a′), GFP FAK L1034S (b and b′), or GFP FAK I936E/I998E (c and c′). FAK L1034S localizes at FAs, whereas FAK I936E/I998E does not. d, representative blots showing immunoprecipitated GFP FAK, GFP FAK L1034S, and GFP FAK I936E/I998E and blotted for GFP and paxillin. WT FAK co-precipitates paxillin, whereas none of the paxillin binding mutants does. e and e′, paxillin −/− and reconstituted (marked with a yellow asterisk) cells immunostained against FAK. In both types of cells FAK is recruited at FAs, showing that interaction with paxillin is not the only determinant for localization. Scale bars: 5 μm.

Overall these data suggest that the interaction with paxillin does play a role in the recruitment of FAK to CAs and FAs; however, it does not provide direct evidence given the possibility that additional proteins may also bind the HPs on the FAT domain (29).

To directly address the role of paxillin on FAK's FA localization, we carried immunofluorescence experiments in paxillin null and reconstituted cells, which revealed that FAK localizes at FAs in both cases. FAK staining at FAs of reconstituted cells does appear stronger; however, all FAs of paxillin null cells contain FAK, suggesting that paxillin is not required for FAK's FA localization but rather increases its affinity for this complex (Fig. 7, e and e′). On the other hand, interpretation of these results is complicated by the expression of the paxillin homologs Hic-5 and leupaxin, which are localized at FAs and bind FAK through their respective LD motifs, compensating for the loss of paxillin (66–68). Given the prominent expression of Hic-5 in epithelial tissues (69), down-regulation of paxillin in the embryo to determine its effect on FAK's CA localization would not be very informative. We thus decided to take advantage of the strong CA localization of GFP paxillinC and the fact that it lacks the FAK binding LD motifs. We postulated that because this mutant does localize at CAs, its overexpression would lead to partial displacement of endogenous paxillin from the complex. If the interaction with paxillin is important for FAK's localization at CA's, displacement of endogenous paxillin would be expected to lead to defective CA localization of FAK, as paxillinC is unable to interact with FAK (32, 70). We thus co-injected the mRNA of mKate FAK with that of GFP paxillinC and centrin CFP, whereas as a control we injected embryos with the mRNAs of mKate FAK, GFP paxillin, and centrin CFP. As shown in Fig. 8, b and b″, in cells overexpressing GFP paxillinC, mKate FAK was still able to localize at CAs (Fig. 8, a and a″). However, quantification of the localization efficiency of mKate FAK revealed that CA recruitment in paxillinC-expressing cells was impaired. As shown in Fig. 8c, the GFP-to-mKate ratio (GFP/mKate) in control cells, expressing GFP paxillin, is 1.2 ± 0.026 (mean ratio, ± S.E.), whereas in cells expressing GFP paxillinC the mean ratio is 2.8 ± 0.052 (***, p < 0.0001). This suggests that the localization of paxillinC at CAs leads to reduced ability of FAK to localize and suggests that the interaction with paxillin modulates FAK's affinity for the CA complex.

FIGURE 8.

The FAK-paxillin interaction is partially responsible for FAK's CA localization. a–a″, multiciliated cell expressing centrin CFP, mKate FAK, and GFP paxillin. FAK co-localizes with paxillin at CAs. Each basal body displays strong localization of both paxillin and FAK. b–b″, multiciliated cell expressing centrin CFP, mKate FAK, and GFP paxillinC. FAK can still localize at CAs together with paxillinC but less efficiently. c, quantification of the ratio of GFP/mKate signal on separate CAs from two-tailed unpaired t test. Error bars represent S.E. Although in control cells (expressing GFP paxillin) the mean ratio is 1.2 ± 0.026 (mean ± S.E.) (n = 226 CAs from 9 embryos) in cells expressing GFP paxillinC, the mean ratio is increased more than 2-fold to 2.8 ± 0.052 (mean ± S.E.) (n = 207 CAs from 11 embryos) (***, p < 0.0001). Scale bars: 2 μm.

The FERM Domain Is Necessary for the Recruitment of FAK at CAs and for Its Function during Ciliogenesis

The above data show that the FAT domain and interactions with paxillin and other partners at the carboxyl terminus are not sufficient to drive FAK to the complexes, suggesting that additional interactions are required to achieve correct localization. The amino-terminal domain of FAK (FERM domain) is known to mediate multiple interactions, including interactions with growth factor receptors, phosphoinositides, ezrin, and the Arp2/3 complex (2, 12, 13, 17, 18). Moreover, it mediates intramolecular, interactions important for the regulation and dynamics of FAK at FAs (21, 25, 28).

We have previously shown that FAK lacking the FERM domain is unable to rescue the ciliogenesis defects elicited by FAK down-regulation (46). We repeated this experiment under the conditions used in this manuscript. As shown in Fig. 3h, FAK Δ FERM fails to rescue the ciliogenesis defects elicited by FAK down-regulation. Specifically, the majority of cells (85.41 ± 6.1%) present severe ciliogenesis defects similarly to morphant embryos (80.5 ± 2.2%). Only 2.45 ± 2.02% of the cells project cilia normally, and 12.14 ± 4.16% of them have moderate defects compared with 39.3 ± 2.96% and 41 ± 1.82%, respectively, in embryos expressing WT FAK, confirming that the FERM domain is required for proper ciliogenesis (Fig. 4d). We, therefore, went on to examine a possible involvement of the FERM domain in the localization of FAK at CA complexes. We co-injected GFP FAK Δ FERM with centrin RFP in the ventral animal blastomeres of 8-cell-stage embryos and examined the localization of the proteins in the ciliated cells of the epidermis of stage 32 embryos. As shown in Fig. 1, d and d′, FAK Δ FERM exhibits a dramatically different localization compared with WT FAK, displaying strong rootlet localization. This localization partially corresponds to the subapical CAs, which form at the end of the striated rootlets and have been suggested to link basal bodies to the subapical actin network (46); however, it is clear that deletion of the FERM domain affects the localization of the protein, as the strong apical enrichment (apical CAs) displayed by WT FAK is compromised (Fig. 1, a and a′), whereas the rootlet localization is enhanced, showing that the FERM domain is necessary for correct localization. Given the requirement of an intact FERM domain for correct localization to CAs, we then asked if it is sufficient to target FAK at the apical complex. To address this we injected embryos with the mRNA of GFP FERM together with the mRNA of centrin RFP and examined the localization in multiciliated cells. As shown, despite a few basal bodies showing apical FERM accumulation (arrows), the GFP signal is mainly cytoplasmic, showing that the FERM domain is not sufficient for CA localization (Fig. 1, e and e′).

The above data suggest that interactions mediated by the FERM domain are important for FAK's localization at CAs. Because such interactions have been shown to relieve the autoinhibitory binding of the FERM domain on the kinase domain, leading to FAK's activation, we went on and examined the status of FAK with respect to its conformation at CAs. We used a previously characterized intramolecular FRET biosensor of FAK, which has YFP (acceptor) fused at the amino terminus and CFP (donor) inserted into the linker between the FERM and the kinase domains. This sensor responds to conformational changes of the FERM domain upon binding to a target and release of the autoinhibitory interactions with increased FRET (71). We thus injected the mRNA of the FAK biosensor into the ventral animal poles of 8-cell stage embryos and recorded CFP and YFP fluorescence using spectral imaging (lambda mode, λ scan). Quantification of the intensity of YFP in relation to CFP revealed higher YFP/CFP ratios specifically at CAs, suggesting elevated FRET at these areas (Fig. 9a). Specifically, the YFP/CFP ratio is 17 ± 0.01% higher at CAs in relation to the mean YFP/CFP ratio within the cytosol. This is visualized in images where the YFP/CFP ratio is generated after processing and is presented color-coded as described by Kardash et al. (72) in Fig. 9a′. The increased FRET at CAs suggests that conformational changes take place specifically at these sites, leading to the release of the FERM domain from the kinase domain. Because this release is triggered by the binding of the FERM domain to proteins and/or lipids, the open conformation of FAK at CAs suggests that the FERM domain specifically interacts with a partner in this complex.

FIGURE 9.

The FERM-kinase interaction is disrupted at CAs in a kinase activity-independent manner. a–c, quantification of the YFP to CFP intensity ratio of the WT (a) and the kinase inactive (K454R) FAK biosensor (b and c) in control (b) and Y15-treated embryos (c), at CAs and in the cytosol. Results were analyzed with a two-tailed unpaired t test. Error bars represent S.E. a, the YFP/CFP ratio at CAs is higher than in the cytosol (***, p < 0.0001, n = 37 CAs, and n = 33 cytosolic regions from 3 embryos) suggesting that FRET is specifically elevated at CAs. b and c, the ratio remains higher in the absence of any enzymatic activity showing that the conformational change at CAs is kinase activity-independent (***, p < 0.0001, n = 40 CAs and n = 30 cytosolic regions from 5 embryos in b and ***, p < 0,0001, n = 28 CAs, and n = 24 cytosolic regions from 3 embryos in c). a′, b′, and c′). Color-coded images of the YFP/CFP intensity ratio of the FAK biosensors showing higher ratio at CAs and suggesting that FAK is in the open conformation at these sites. d and e, epidermal region of embryos immunostained for pY576 FAK and acetylated tubulin. d, FAK is phosphorylated at cell-cell contacts, and dense cilia project from the apical surface of multiciliated cells in control embryos. e, treatment with the Y15 inhibitor blocks phosphorylation on tyrosine 576, and no signal is detected at cell-cell contacts, whereas cilia project normally. Treatment of embryos with the inhibitor during ciliogenesis stages did not affect the density of multiciliated cells along the epidermis (g) or the docking and spacing of basal bodies (i) compared with control embryos (f and h). j and k, tracking of the QD flow over the skin of control or inhibitor-treated tadpoles. Like controls (j), embryos treated with the Y15 inhibitor (k) produced robust posterior flow, showing that inhibition of FAK's kinase activity does did affect either ciliogenesis or the function of cilia and their ability to generate flow. Scale bars: 2 μm in a′, b′, and c′; 10 μm in d and e; 5 μm in h and i.

Overall these results uncover an important role of the FERM domain in both the localization and function of FAK in multiciliated cells and suggest that in addition to the interaction with paxillin, interactions through the FERM domain are necessary for the recruitment and function of FAK at CAs.

The Amino and Carboxyl Termini of FAK Cooperate to Target It at CAs

Since the FAT and FERM domains are both necessary for proper localization at CAs but neither is sufficient, we decided to examine whether the two domains could cooperate to correctly target FAK at the complex. We thus took advantage of a construct, called FF, which combines the two domains. This construct, which lacks the kinase domain, is composed of the amino terminus of FAK, which contains the FERM domain, and the carboxyl terminus, which contains the FAT domain, and has been shown to mimic the localization of WT FAK both at FAs and in the embryo (21). As shown, unlike GFP FERM and GFP FAT, GFP FF displays strong CA localization that is indistinguishable from that of WT FAK (Fig. 1, f and f′). This confirms that unlike FAs, where the FAT domain is both necessary and sufficient for FAK's localization, in CAs the FERM domain is also necessary and interactions through this domain are critical for correct localization. The localization of FF and all mutants examined in this study in relation to the basal bodies and rootlets is summarized in the schematic of Fig. 10.

FIGURE 10.

Localization summary for FAK mutants. A schematic showing the basic structure of cilia (gray, axoneme; red, basal body; yellow, rootlet) and the regions where CAs (green) form. The efficiency of deletion mutants of FAK to localize at CAs is presented color-coded. WT FAK displays strong localization apically, next to the basal bodies and subapically, at the end of ciliary rootlets. Deletion of the FAT domain leads to loss of subapical localization, whereas only weak apical localization remains. A similar localization is displayed by the FERM and FAT domains when expressed alone. Deletion of the FERM domain leads to near loss of apical localization and pronounced localization at the rootlet. FF displays a localization that is almost identical to that of WT FAK.

Since FF combines the two necessary domains and displays strong CA localization, in an effort to determine if FAK's role in multiciliated cells requires the kinase domain, we decided to use FF to rescue the ciliogenesis defects elicited by FAK down-regulation. As shown in Figs. 3i and 4e, expression of FF in embryos injected with the FAK MO partially rescued the phenotype, albeit with reduced efficiency than WT FAK. Specifically, 18.25 ± 2.65% of the cells projected cilia normally in FF-expressing embryos compared with 1.4 ± 0.43% of normal cells in morphant embryos and 39.3 ± 2.96% in embryos expressing WT FAK. The amount of severely affected cells dropped significantly to 33.5 ± 2.2% in FF expressors compared with 80.5 ± 2.2% in morphant embryos, and the majority of cells (48.25 ± 0.45%) presented moderate defects (Fig. 4e). Despite FF's reduced ability to rescue compared with WT FAK, FF expression led to significantly better rescue compared with expression of FAK Δ FERM or FAK Δ FAT. As shown in Fig. 4e, in FF-expressing embryos 18.25 ± 2.65% of cells project cilia normally, whereas in FAK Δ FERM and FAK Δ FAT the amount of cells with normal cilia is 2.45 ± 2.03% and 0.95 ± 0.25%, respectively. Taken together, these results confirm that the amino and carboxyl termini of FAK cooperate to target FAK at CAs and suggest that FAK plays important kinase-independent, scaffolding roles at these complexes.

The Kinase Activity of FAK Is Dispensable for Its Role in Ciliogenesis

The inability of FF to rescue as efficiently as WT FAK suggests that the kinase domain of FAK is important during ciliogenesis. To directly address a possible requirement of FAK's kinase activity during ciliogenesis, we used the kinase-inactive FAK K454R (supplemental Fig. 1) mutant in rescue experiments (73). We co-injected the mRNA of FAK K454R with the FAK MO and then scored the tadpoles for defects in ciliogenesis. As shown, the “kinase dead” mutant could rescue with the same efficiency as WT FAK (Figs. 3j and 4f). As shown in Fig. 4f, 35.5 ± 7.4% of the cells were normal with respect to the density of the cilia projecting from their surface, which is not significantly different from 39.65 ± 4.9% for WT FAK-expressing embryos. Moreover, the percentage of cells with severe ciliogenesis defects dropped significantly (18.73 ± 1.73%) when compared with that in morphant embryos (80.5 ± 2.2%). These results suggest that FAK acts as a scaffold, promoting interactions between CA proteins during ciliogenesis, rather than mediating signal transduction through its catalytic activity. To examine whether the kinase activity is also dispensable for proper ciliary function, we went on and examined embryos rescued with FAK KD for their ability to generate directional flow. As shown in Fig. 5f, robust flow is generated along the anterior-posterior axis of the tadpole. These data indicate that the kinase activity of FAK is dispensable for both ciliogenesis and ciliary function, suggesting that FAK fulfils a purely scaffolding role in this context.

Given the critical role of FAK's kinase activity with respect to the regulation of FAs, we wanted to confirm the above findings. We thus went on to use a previously characterized FAK inhibitor (Y15) that prevents autophosphorylation at the activation site, tyrosine 397, and blocks FAK's kinase-mediated downstream signaling (74, 75). Embryos were treated with 20 μm concentrations of the inhibitor at neurula stages (stage15) when ciliated cells begin to intercalate and basal bodies migrate toward the apical surface (76). Embryos were subsequently allowed to develop in the presence of the inhibitor to stage 32. At this stage the multiciliated epidermis is fully differentiated, and directional flow is present along the long axis of the rescued embryo. The effectiveness of the inhibitor was confirmed via staining with an anti-phospho-FAK (Tyr(P)-576) antibody that recognizes Tyr(P)-576, which is in the activation loop of FAK's kinase domain and regulates its kinase activity (Fig. 9, d and e). The tadpoles were then scored for ciliogenesis defects and examined for defects in ciliary flow. As shown, treatment with the inhibitor did not affect ciliogenesis (Fig. 9, f–i) or the cilia-driven flow, which was robust and directional along the anterior-posterior axis of the tadpole (Fig. 9, j and k). These data confirm that FAK's role in ciliogenesis is kinase-independent. Given the FAK conformational FRET sensor data, showing that FAK is in the open active conformation at CAs, we decided to introduce a kinase-inactive version of the FRET FAK biosensor (FAK sensor K454R) into the embryos to determine if the conformational change of FAK at CAs is kinase-dependent. As shown in Fig. 9, b and b′, quantification of the YFP/CFP intensity ratio in multiciliated cells revealed the presence of increased FRET, specifically at CAs. However, due to the presence of endogenous WT FAK in these cells, trans-phosphorylation of the sensor by the endogenous protein cannot be excluded, so we went ahead and repeated these experiments in the presence of the Y15 FAK inhibitor. As shown, FRET is still elevated in the presence of the FAK inhibitor, confirming that the conformational changes of FAK taking place at CAs are kinase activity-independent (Fig. 9, c and c′). Together, these results suggest that although FAK is, at least partially, in the open conformation at CAs, its role in these complexes is kinase-independent, and the release of the FERM-kinase domain interaction is likely elicited through the interaction of the FERM domain with binding partners, specifically enriched at CAs.

Discussion

FAK is one of the best studied focal adhesion proteins serving both signaling and scaffolding functions and regulating a variety of cellular processes (1, 2). We previously identified a novel role of FAK, as a member and regulator of CA complexes in multiciliated cells and showed that loss of FAK in these cells leads to aberrant ciliogenesis due to impaired association of the basal bodies with the actin cytoskeleton (46). In this study we address the determinants for FAK's localization and function at the CA complexes in multiciliated cells. Our findings suggest that efficient localization of FAK at CAs requires the cooperation between the amino- and carboxyl-terminal regions of the protein and specifically the FERM and the FAT domains (Fig 10). Deletion of either domain prevented correct recruitment of FAK to the basal bodies, whereas combining the two was sufficient for correct CA localization. In addition, none of the mutants that failed to localize correctly could rescue the ciliogenesis defects elicited by down-regulation of endogenous FAK, indicating that the localization of FAK at the base of the cilia is critical for FAK's role in ciliogenesis.

The localization determinants for FAK at CAs are in contrast to those for FA localization. Although initial work suggested that the carboxyl terminus of FAK was sufficient for CA localization, upon closer examination it became clear that this is only the case for the subapical complex found associated with the ciliary rootlets. In contrast, the apical CA localization is compromised in the absence of the amino-terminal FERM domain. Although FA localization of FAK strictly depends on the FAT domain, which is both necessary and sufficient to target FAK at FA complexes, both the FAT and the FERM domains are important for localization at CAs (30, 31, 62). Specifically, our results show that CA localization of FAK strongly depends on the FAT domain and the interaction with paxillin, similarly to FAK's FA localization, but in addition requires an intact FERM domain. The FAK-paxillin interaction is also critical for FAK's function during ciliogenesis given the fact that the L1034S mutant, which localizes at CAs, fails to rescue the ciliogenesis defects. Unlike FAK on the other hand, we show that paxillin localizes at CAs through its carboxyl-terminal LIM domains, which are also sufficient for targeting the protein at FAs, even though the mechanism and binding partners that drive this localization are still unknown (32).

The important role of the FERM domain is also supported by evidence showing that CA-localized FAK is in an open state, as a result of conformational changes triggered from the release of the FERM-kinase domain interaction. Importantly, such changes have been suggested to occur as a consequence of the interaction between FERM and binding partners at FAs, suggesting that such FERM-mediated interactions take place specifically at CAs (71). Thus, in addition to the interaction with paxillin, additional interactions through the FERM domain are important for targeting, whereas others could be necessary for stabilizing FAK at CAs. Interestingly, even though the FAT domain is necessary and sufficient for FA localization, the FERM domain has been shown to be involved in regulating the dynamics of FAK at FAs and to be responsible for localization at the plasma membrane and at cell-cell junctions (21, 22). Moreover, in addition to interactions with other proteins, the FERM domain has been shown to mediate the dimerization of FAK (through FERM-FERM and FERM-FAT interactions) specifically at FAs, raising the possibility that it may regulates FAK's localization and function at CAs through dimerization (28). However, given the qualitative differences in the localization of full-length FAK and the Δ FERM mutant, it is unlikely that the FERM domain's role is solely to mediate dimerization at CAs. Additional work will be required to elucidate the mechanism through which the FERM domain is involved in this process and to identify potential binding partners at CAs. It would for example be interesting, given the fact that the FERM domain has been reported to interact with PIP₂ and given the close proximity of CAs to the plasma membrane, to examine if the FERM domain is directly interacting with PIP₂ at the plasma membrane.

Our results suggest that even though dispensable for FAK's localization at CAs, an interaction with talin has a minor role during ciliogenesis. It would be interesting to further investigate the importance of this interaction and examine the possibility that FAK is involved in driving talin to CAs, considering that a similar role has been reported for the recruitment of talin at nascent adhesions in adherent cells (20, 38, 39). Examination of the mechanism of talin's localization at CAs would be intriguing, also given the absence of integrins from CA complexes and considering that talin directly binds to integrin cytoplasmic domains, an interaction that is thought to be responsible for its FA localization (77–79).

Finally, our results suggest that the role of FAK in multiciliated cells is kinase activity-independent, and this was shown using a kinase inactive mutant of FAK as well as through pharmacological inhibition of kinase-mediated downstream signaling of FAK during ciliogenesis and ciliary function. Kinase activity-independent functions of FAK have been reported in other contexts as well. Recent work from our group revealed that FAK is involved in the sensing of forces that guide the mitotic spindle and that the kinase activity is dispensable in this context (60). Moreover, a kinase-independent role of FAK in cell proliferation and survival has also been reported (80–82). We now show that the catalytic activity of FAK is dispensable for its function in multiciliated cells, where it acts as a scaffolding protein. However, a role for the kinase domain of FAK cannot be ruled out given the inability of FF (which lacks the kinase domain) to rescue as efficiently as WT or kinase-inactive FAK. This suggests that even though the enzymatic activity is not required, the kinase domain of FAK may still serve scaffolding or regulatory roles important for FAK's function in ciliated cells. Even though a binding partner for this domain has not been yet identified, it plays a crucial role in the conformation of the protein through its interaction with the FERM domain. Given the FRET data, suggesting that at least some of the protein at CAs is in the open conformation, the correct regulation of FAK through the FERM-kinase interaction may be important, and future work will focus on this role through the use of mutants that disrupt the FERM kinase interaction.

Experimental Procedures

Embryo Manipulations, Microinjections, and Lysates

Female adult Xenopus laevis were ovulated by injection of human chorionic gonadotropin. Eggs were fertilized in vitro, de-jellied in 2% cysteine (pH 7.8), and embryos were maintained in 0.1× Marc's modified ringers (MMR) and staged according to Nieuwkoop and Faber (83). Microinjections were performed in 4% Ficoll in 0.33× MMR, and injected embryos were reared for 2 h or until stage 8 in 4% Ficoll and then washed and maintained in 0.1× MMR. Microinjections were made at the ventral blastomeres of 4- or 8-cell-stage embryos to target the epidermis. For all experiments we injected 12 ng of the FAK morpholino per blastomere (at 8-cell stage embryos) and mRNAs at various amounts. Embryos were allowed to develop to the appropriate stage and then imaged live or fixed in MEMFA (1 m MOPS, 20 mm EGTA, 10 mm MgSO₄, 37% Formaldehyde) for 1–2 h at room temperature. For live imaging, embryos were anesthetized in 0.01% benzocaine in 0.1× MMR. Fixed embryos were used immediately.

Protein lysates were prepared by homogenizing embryos in ice-cold MK's modified lysis buffer (50 mm Tris (pH 8.0), 150 mm NaCl, 0.5% Nonidet P-40, 0.5% Triton X-100, 100 mm EGTA, 5 mm NaF) supplemented with protease and phosphatase inhibitors. Homogenates were cleared by centrifugation at 15,000 × g for 30 min at 4 °C. ¾-1 embryo equivalent was used for Western blot analysis of protein expression levels.

All procedures were carried out according to the University of Cyprus Animal Ethics Guidelines under licenses granted by the Animal Health and Welfare Division of the Veterinary Services department of the Ministry of Agriculture, Rural Development, and Environment of the Republic of Cyprus.

FAK Inhibitor Treatment

Embryos were incubated at room temperature with 20 μm FAK Y15 inhibitor (Santa Cruz Biotechnologies) diluted in 2% DMSO in 0.1× MMR. Embryos were incubated at stage 15 and were allowed to develop in the presence of the inhibitor to stage 32. Control embryos of the same stage were incubated with 2% DMSO in 0.1− MMR.

DNA Constructs and Morpholino Oligonucleotide

All plasmids were constructed using standard molecular biology techniques and verified by sequencing. The GFP-tagged and HA-tagged versions of chicken FAK, FAK Δ FAT, FAK L1034S, FAK E1015A, FAK I936E/I998E, FERM, FAK K454R, FF, paxillin, and paxillinC in addition to centrin CFP and mKate FAK have been described elsewhere (21, 22, 46, 60). The sequence of chicken FAK with the mutation sites indicated is provided in Supplemental Fig. 1. The pCS108 GFP FAT construct was generated in two steps using the primers F/FAT, (5′-aagtcgacatcaagccacaggaaatcagccctcct-3′) R/FAT, (5′-tttctcgagttagtggggcctggactggctgatcatttt-3′), F/GFP, (5′-aaagcggccgcatggtgagcaagggcgaggagctg-3′), R/GFP, (5′-gftgacttacttgtacagctcgtccatgccga-3′), and pCS2++ GFP FAK as a template. The pCS2++ GFP FAK Δ FERM construct was generated by inserting the GFP sequence in the pCS2++ HA FAK Δ FERM plasmid. The pCS108 FAK sensor and the pCS108 FAK K454R sensor plasmids were generated by subcloning from the YFP-FAK153-CFP (wild type and K454R mutant) plasmids (kindly provided by Gertrude Bunt's laboratory). The pCS2++ FusionRed paxillin and pCS2++ RFP paxillin constructs were generated by replacing the sequence of GFP with that of FusionRed or RFP in the pCS2++ GFP paxillin construct (46). The pCS2+ Centrin4 RFP plasmid was kindly provided by Brian J. Mitchell's laboratory. All plasmids were transcribed into mRNA and used for microinjections. The expression of all HA-tagged constructs was verified after immunostaining against HA in tadpoles. The sequence of FAK MO used in our experiments is TTGGGTCCAGGTAAGCCGCAGCCA (21, 46, 61).

Immunostaining

Paxillin −/− cells plated on glass coverslips were washed with PBS and fixed for 10 min in 4% paraformaldehyde solution in PBS. Fixation was followed by the addition of 50 mm glycine solution in PBS, and then cells were permeabilized using 0.2% Triton X-100 solution in PBS for 10 min. Cells were blocked using 10% normal donkey or goat serum (Jackson ImmunoResearch) in PBS for 30 min. Cells were incubated with a primary antibody against FAK (1:500, Millipore) for 1.5 h, washed several times in PBS, and then incubated with a Cy3 anti-mouse secondary antibody (Jackson ImmunoResearch) for 1 h.

For whole mount immunostaining, fixed embryos were rehydrated by serial washes. Embryos were then permeabilized in PBDT (1× PBS + 0.5% Triton X-100 + 1% DMSO) for several hours at room temperature and blocked in PBDT + 10% normal goat or donkey serum for 1 h at room temperature. Primary antibodies were then added (in block solution). Primary antibodies used were: FAK pY576 (1:200, Invitrogen), acetylated α-tubulin (6–11B-1) (1:500, Santa Cruz), and HA (Y-11) (1:200, Santa Cruz). The embryos were incubated in the antibody solution for 4 h at room temperature or overnight at 4 °C. The next day embryos were washed in PBDT. Embryos were then incubated in secondary antibodies diluted in the blocking solution at room temperature for 1–2 h. Secondary antibodies used were: Alexa 488 (1:500, Invitrogen), Cy3 (1:500, Jackson ImmunoResearch), and IgG-CFL 647 (1:500, Santa Cruz Biotechnologies). Embryos were washed in PBDT, post-fixed in MEMFA for 15–30 min at room temperature, washed in 1× PBS, and imaged immediately.

Cell Culture, Transfection, Lysis, and Immunoprecipitation

HEK 293T and HeLa cells (ATTC) were maintained in DMEM supplemented with 10% FBS, and paxillin −/− cells (gift of Dr. Scott Vande Pol) were maintained in DMEM supplemented with 10% FBS, 1 mm sodium pyruvate, and nonessential amino acids. Transient transfections were performed using calcium phosphate or Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol.

For immunoprecipitation, HEK 293T cells were co-transfected with pCS2++ FusionRed paxillin and either pCS2++ GFP FAK, pCS2++ GFP FAK L1034S, or pCS2++ GFP FAK I936E/I998E. 48 h after transfection cells were rinsed with ice-cold PBS and lysed with ice-cold lysis buffer (130 mm NaCl, 20 mm HEPES (pH 7.2), 3 mm EDTA, 0.3% Triton X-100) supplemented with protease and phosphatase inhibitors. Cell lysates were cleared by centrifugation at 15,000 × g at 4 °C for 10 min and then incubated with the GFP-G1 antibody (Developmental Studies Hybridoma Bank) for 2 h at 4 °C under continuous shaking. Protein G plus agarose beads (Santa Cruz Biotechnologies) were added, and the mixture was incubated for 2 h at 4 °C under continuous shaking. Beads were collected by centrifugation at 2500 × g, washed with ice-cold lysis buffer, and resuspended in Laemmli buffer. Samples were used for Western blot analysis.

Western Blot Analysis

Proteins were run on 7.5% SDS-polyacrylamide gels and blotted onto nitrocellulose membranes with the WesternC ladder (Bio-Rad). Blots were blocked in 5% milk in TBST (1× TBS and 0.1% Tween 20) and then incubated with the primary antibodies in 5% milk overnight at 4 °C. The primary antibodies used were: actin (1:1000, Santa Cruz), HA (1:500, Santa Cruz), GFP (1:2000, Abcam), and paxillin (1:7000 BD Biosciences). Visualization was performed using HRP-conjugated antibodies (Santa Cruz) after 1 h of incubation at room temperature, and a signal was detected using LumiSensor (GeneScript) on a UVP BioSpectrum Imaging System.

Fluid Flow Assay

For the examination of the cilia-driven flow over the epidermis, stage-32 tadpoles were anesthetized in 0.01% benzocaine in 0.01× MMR and placed in thick silicone grease wells on glass slides. 655- or 700-nm QDs (Invitrogen) were then added into the media, and a glass coverslip was placed on top of the wells without touching the tadpoles' skin. To image the flow of QDs, we carried out time-lapsed microscopy using the Zeiss Axio Imager Z1 microscope equipped with a Zeiss Axiocam MR3 and the Axiovision software 4.7. Tracking of the flow was done manually using ImageJ.

Imaging

Imaging was done on a Zeiss LSM 710 laser scanning confocal microscope with the Zen 2010 software.

Intramolecular FRET of the FAK Biosensor

FRET experiments were accomplished using a laser scanning confocal microscope (Zeiss LSM 710) with a Plan-Apochromat 63×/1.40 Oil differential interference contrast M27 objective lens (Zeiss). Tadpoles expressing the FAK biosensor (wild type or the kinase dead mutant) were anesthetized in 0.01% benzocaine in MMR and immobilized in silicone grease wells on glass slides. A 458-nm laser was used for excitation (CFP excitation), and emission was recorded using the lambda mode for spectral imaging (λ scan). Spectral unmixing of the acquired images revealed one emission peak at 478 nm (maximum emission of CFP) and a second one at 527 nm (maximum emission of YFP). YFP/CFP intensity ratios were quantified as follows. The YFP/CFP ratio for each selected region of a cell (ciliary adhesion or cytosolic region) was divided by the mean cytosolic YFP/CFP ratio of the cell, in order to normalize any differences at the expression level of individual cells and embryos.

Quantification

All quantification and statistical analysis was done using GraphPad Prism.

Quantification of the ciliogenesis defects was done based of the number of cilia projecting from the apical surface of individual cells. Cells have been denoted as normal, with moderate defects (having ∼50% of their cilia) or with severe defects (having near complete loss of their cilia). 130–150 cells from each embryo were used for quantification, and the experiments were repeated 3 times (at least) for each FAK construct used for rescue. The number of embryos for each experiment is denoted in the figure legends and under “Results.” Due to the fact that wild type FAK partially rescued the ciliogenesis defects, the ability of FAK mutants to rescue was compared with that of wild type FAK.

For the quantification of the efficiency of FAK L1034S to localize at ciliary adhesions we quantified the mKate/GFP intensity ratio on ciliary adhesions in cells expressing mKate FAK and GFP FAK (control) or mKate FAK and GFP FAK L1034S. To normalize any differences regarding the expression levels or the total ratios of the proteins between individual cells, the mKate/GFP intensity ratio for each ciliary adhesion was divided by the total mKate/GFP intensity ratio of the cell.

Author Contributions

I. A. participated in the design of the experiments and carried out most of them, analyzed the data, and wrote the manuscript. P. S. helped to carry out some of the localization experiments. N. C. carried out the experiments using the FAK biosensor. P. A. S. participated in the design of the experiments and wrote the manuscript. All authors read and approved the final manuscript.