The interaction of with the is essential for activation and formation

Krishna Chinthalapudia, Erumbi S. Rangarajana, and Tina Izarda,1

aCell Adhesion Laboratory, Department of Integrative Structural and Computational Biology, The Scripps Research Institute, Jupiter, FL 33458

Edited by Barry Honig, Howard Hughes Medical Institute and Columbia University, New York, NY, and approved August 23, 2018 (received for review April 11, 2018) Multicellular organisms have well-defined, tightly regulated mech- major phosphorylation sites, T144 and T150) (3), and the talin anisms for . Heterodimeric αβ integrin receptors play preceding F0 subdomain has a -like fold. The talin F3 central roles in this function and regulate processes for normal cell subdomain harbors the primary β-integrin–binding site (4, 5). The functions, including signaling, , and development, talin rod domain consists of 13 domains, R1–R13, composed of 62 binding to the , and senescence. They are in- amphipathic α-helices arranged into four-helix (R2, R3, R4, and volved in hemostasis and the immune response, participate in leu- R8) and five-helix (R1, R5, R6, R7, R9, R10, R11, R12, and R13) kocyte function, and have biological implications in angiogenesis bundle domains (6, 7) and a C-terminal dimerization domain (6). and cancer. Proper control of integrin activation for cellular commu- Each domain has unique properties, including binding to other nication with the external environment requires several physiolog- talin domains, to integrin, and to (8, 9). A secondary ical processes. Perturbation of these equilibria may lead to constitutive integrin-binding site in the rod domain (residues 1,974–2,293) of integrin activation that results in bleeding disorders. Furthermore, uncertain function has also been identified (10). There are three play key roles in cancer progression and metastasis in which -binding sites located on the head domain and R4–R8 and certain tumor types exhibit higher levels of various integrins. Thus, the R13 subdomains (6, 11, 12), and 13 vinculin-binding sites that are integrin-associated signaling complex is important for cancer therapy α – single amphipathic -helices (7, 13 20). BIOPHYSICS AND development. During inside-out signaling, the cytoskeletal talin The role of lipids in integrin activation remains unclear despite plays a key role in regulating integrin affinity whereby the talin head a large body of literature and the known functional importance of COMPUTATIONAL BIOLOGY domain activates integrin by binding to the cytoplasmic tail of talin attachment to the membrane (8). In the first stages of cell β-integrin and acidic membrane phospholipids. To understand the attachment, the talin F3 FERM domain binds to the NPxY motif mechanism of integrin activation by talin, we determined the crystal of the integrin cytoplasmic β tail, thereby inducing reorganization structure of the talin head domain bound to the acidic phospholipid of the integrin heterodimer and activating integrin (5, 21–24). phosphatidylinositol 4,5-bisphosphate (PIP2), allowing us to design a Talin attachment to the plasma membrane is enhanced by phos- – talin lipid-binding deficient talin mutant. Our confocal microscopy with phatidylinositol 4,5-bisphosphate (PIP2), which induces a confor- knockout cells suggests that the talin–cell membrane interaction seems mational change in talin to expose the integrin-binding site (22, essential for focal adhesion formation and stabilization. Basal integrin 25–28). The role of PIP2 in integrin activation is particularly in- activation in Chinese hamster ovary cells suggests that the lipid- teresting since PIP2 is a major phosphoinositide of the inner binding–deficient talin mutant inhibits integrin activation. Thus, mem- brane attachment of talin seems necessary for integrin activation and Significance focal adhesion formation. Vertebrate cell growth, division, locomotion, , angiogenesis | cell adhesion | integrin activation | phospholipids | and development rely on the dynamic interactions of cells with talin activation extracellular matrix components via cell surface complexes termed focal adhesions that are composed of heterodimeric αβ alin is a key player in integrin activation. Vertebrates express integrin receptors, associated signaling molecules, and the large Ttwo isoforms in which talin1 is ubiquitously expressed, while cytoskeletal protein talin. While it is known that talin activation talin2 is found primarily in striated muscle and in the brain. As a and binding to β-integrin requires interactions with lipids, little multidomain cytoskeletal protein, talin contains discrete binding is known regarding the structure and function of inactive vs. sites for acidic phospholipids, β-integrin, actin, and vinculin, as activated talin, and what is known is often disputed. Here we well as layilin, PIPK1γ90, and . Talin links report that talin binding to the cell membrane seems necessary to the cytoplasmic membrane at cell-extracellular matrix adhesion for integrin activation and focal adhesion formation, a finding sites. This process depends critically on talin. Talin consists of a that significantly advances our understanding of integrin acti- polypeptide chain of 2,541 amino acids and is often described as vation and might aid the development of novel integrin having an N-terminal FERM (four-point-one, , , therapeutic agents. ) domain connected by a linker (residues 401–482) that harbors a -II cleavage site to a large “rod” domain (resi- Author contributions: K.C., E.S.R., and T.I. designed research; K.C. and E.S.R. performed dues 483–2,541). The talin head domain is different from all other research; K.C., E.S.R., and T.I. analyzed data; and K.C., E.S.R., and T.I. wrote the paper. FERM domain-containing in that it has four subdomains, The authors declare no conflict of interest. F0–F3 (instead of the typical three, F1–F3), and they adopt an This article is a PNAS Direct Submission. extended structure (1) instead of the canonical cloverleaf con- Published under the PNAS license. formation seen in the ERM family of proteins (2). As seen in Data deposition: The atomic coordinates and structure factors have been deposited in the other FERM domain-containing proteins, the talin FERM sub- , (PDB ID code 6mfs). domains contain a ubiquitin-like F1, acyl-CoA–binding protein- 1To whom correspondence should be addressed. Email: [email protected]. like F2 and phosphotyrosine-binding–like F3 subdomain. Unlike This article contains supporting information online at all other F1 FERM subdomains, the talin F1 subdomain has an 1073/pnas.1806275115/-/DCSupplemental. unstructured insert (F1 loop, residues 133–165, harboring two PNAS Latest Articles | 1of6 Downloaded by guest on September 28, 2021 membrane (29, 30), and because talin regulates the local PIP2 concentration in the membrane by binding and activating PIPK1γ (31, 32). PIP2 regulates important processes, such as vesicular trafficking, platelet activation, organization (33–35), and focal adhesion turnover (25, 26, 36). This process evolves by targeting proteins to the membrane, often through induction of a or oligomerization (36, 37). In mammals, the heterodimeric integrin transmembrane re- ceptors are composed of 18 distinct α and β chains (38, 39). By responding to extracellular and intracellular stimuli, integrins connect the extracellular matrix to the cytoskeleton, and trans- duce signals across the plasma membrane in both directions, termed outside-in and inside-out signaling, respectively (39). Integrin activation is important in platelets and leukocytes as well as many tissues in which extracellular matrix remodeling, angiogenesis, and cell migration are involved. These processes require tightly controlled integrin activation mechanisms that involve conformational changes of these receptors. Thus, un- derstanding the molecular mechanisms of how talin activates integrin is fundamental for gaining insight into important path- ological states and recognizing how integrin activation might aid the development of novel integrin antagonists. Here we report the talin1 head/PIP2 complex crystal structure together with biochemical and functional data that answer im- portant questions, including how PIP2 activates talin. Our data provide several surprises and answers to longstanding mecha- nistic questions and suggest a mechanism in which on re- cruitment of cytosolic talin by PIPK1γ to the plasma membrane (32), PIP2 activates talin by severing the head–tail interaction, thereby exposing the integrin-binding site. Remarkably, our in vivo data suggest that the talin–PIP2 interaction is crucial for talin localization to the cell membrane, affects the scaffolding of cells, and thus is likely key for cell spreading and adhesion. We further find that disrupting talin binding to the membrane affects integrin activation, and that this talin–PIP2 interaction seems necessary for focal adhesion formation. Collectively, our study provides a major advance in our understanding of the dynamic control of focal adhesions by talin. Results

PIP2 Binding to Talin Allosterically Blocks the Integrin and Talin Tail- Binding Sites. We determined the crystal structures of the N- terminal talin head domain (residues 1–400) and the deletion mutant of that domain Δ139–168 (talin residues 1–400), both bound to PIP2 (Fig. 1 and SI Appendix, Tables S1 and S2). However, the full-length head domain structure did not show clear electron density for residues 139–168. The electron density map for the lipid was poorer compared with the PIP2-bound Δ139–168 talin structure. Fig. 1. PIP2 binding to talin allosterically blocks the integrin- and talin tail- – The talin FERM domain architecture is linear compared with binding sites. (A) Superposition of our talin/PIP2 structure (F2, residues 209 304, green; F3, residues 311–398, blue) onto the talin head domain (residues the classical cloverleaf structure. Our PIP2/talin (1-400; Δ139– 209–400; yellow) bound to the tail rod R9 subdomain (residues 1655–1824, 168) complex structure harbors one PIP2-binding site that differs from the classical phosphoinositide-binding mode of other cyan; PDB ID code 4F7G) (28). The red double arrow indicates how PIP2 binding displaces the tail domain. (B) Superposition of our talin/PIP2 structure known modules (40), including the ERM protein radixin, where (F2, green; F3, blue) onto the two talin2 F2-F3/integrin β1D (integrin, cyan; IP3 bound between the interface between F1 and F3, making the talin, yellow; PDB ID code 3G9W) (4) heterodimers in the asymmetric unit. The

molecular events for membrane-mediated spatiotemporal regu- red double arrow indicates how PIP2 binding prevents the membrane- lation of talin inhibition and activation unique. The PIP2-binding proximal integrin binding. (C) Superposition of our talin/PIP2 structure (col- site is lined by residues from the F2 (K272) and F3 (K316, K324, ored spectrally: F0, residues 4–83, orange; F1, 85–195, yellow; F2, 209–304, E342, and K343) FERM subdomains, in which the 4′-phosphate green; F3, 311–398, blue; 139-171, disordered) onto the unbound talin1 structure (residues 1–400, Δ139–168, gray; PDB ID code 3IVF) (1) highlights the group of PIP2 interacts with talin residues K272, E342, and K343 and the 5′-phosphate group interacts with E342 and K316. In relative F0-F1 domain movements of approximately 7° and 3 Å. addition, K272 interacts with the hydroxyl moiety of the inositol, while one of the carbonyls of the diacylglycerol moiety is within hydrogen-bonding distance to K324. binding to the talin rod R9 subdomain as well as to integrin, but γ A – Superposition of our lipid-bound structure onto the talin head/ not to PIPK1 (Fig. 1 ). In our lipid-bound structure, the 318 2 tail (F2-F3/R9) structure [Protein Data Bank (PDB) ID code 325 main chain has temperature factors of approximately 81 Å , 4F7G] (28) reveals a large movement of approximately 10 Å of with the nearest crystal contacts occurring >4 Å between N323 the F3 loop (residues 318–325) that is extensively involved in and the symmetry-related T354. In the head/tail structure, N323

2of6 | Chinthalapudi et al. Downloaded by guest on September 28, 2021 engages in hydrogen-bonding interactions with T1767 and D1809. Strikingly, K324 is approximately 6 Å closer to the lipid- binding site in the talin/PIP2 structure compared with its position in the head/tail structure. This causes the 318–325 loop to move, whereby K320 releases the talin R9 rod subdomain by steric hindrance. Collectively, the structures show the molecular basis of how the binding of PIP2 and the talin R9 rod subdomain seem mutually exclusive, particularly since the lipid-binding residues are not involved in crystal contacts (SI Appendix, Fig. S1). This finding is supported by solution studies showing that PIP2 vesi- cles compete effectively with talin R9 binding to the talin F2F3 subdomains (28). Similarly, superposition of our talin/PIP2 structure onto the talin2 F2-F3/integrin β1D structure (PDB ID code 3G9W) shows that the 318–325 loop adopts a similar conformer in the integrin- bound state and the R9-bound state (Fig. 1B). Notably, the talin L325R abolishes its binding to the integrin membrane- proximal region (4). The position of the loop in our lipid-bound structure causes K321 to occupy the integrin membrane- proximal binding site. In addition, the integrin and the talin R9 rod domain-binding sites overlap on the talin head domain. These findings are consistent with studies showing that the integrin membrane-distal site is necessary for talin-induced integrin activation (41). In our lipid-bound structure, N323 en- gages only in water-mediated crystal contacts with a symmetry- related T354. Thus, it seems unlikely that the distinct loop conformation is caused by crystal contacts, and more likely that it

is caused by lipid binding (SI Appendix, Fig. S2). BIOPHYSICS AND

The unbound structure (PDB ID code 3IVF) (1) is iso- COMPUTATIONAL BIOLOGY morphous to and almost identical to our PIP2-bound structure except for relative domain movements. Significantly, superposi- tion of the respective F2F3 domains shows relative F0F1 domain movements of approximately 7° and 3 Å (Fig. 1C). Notably, the 318–325 loop is in the PIP2-bound conformer in the apo struc- ture. PIP2 also engages in crystal contacts with K334 and E335, and the E335 side chain moves to make room for the lipid (SI Appendix, Fig. S3). Collectively, this suggests that integrin or R9 binding causes the distinct 318–325 loop conformation. The interpretation of comparisons with the PIPK1γ/talin struc- tures (PDB ID code 2G35) (42) is less obvious, perhaps because these structures are from either the talin2 isoform or a talin1- PIPK1γ chimera. Furthermore, PIPK1γ is not in contact with the loop region that binds the integrin membrane-proximal site, but instead overlaps with the integrin membrane-distal site on talin.

Fig. 2. The talin head domain harbors one PIP2-binding site. (A) Structure- The Talin Head Domain Harbors Only One PIP2-Binding Site. The F1 – based mutagenesis confirms our talin1 lipid-binding sites by lipid cosedi- loop (residues 133 170) has been shown to be required for mentation assay. The 3 M mutant (Y377F, R358Q, K357Q) binds lipids as seen integrin activation but not for integrin binding. It can form an for wild type (WT) talin, while the minimal lipid-binding–deficient mutant α-helix and as an isolated peptide, interacts with lipids. To de- 5M (K272Q, K316Q, K324Q, E342Q, K343Q) and the lipid-binding-deficient termine if in the context of the talin head domain there is an- mutant 8M (K272Q, K316Q, K324Q, E342Q, K343Q, Y377F-K357Q-R358Q) other lipid-binding site, we mutated K272Q, K316Q, K324Q, show insignificant binding to PIP2/PC vesicles. P, pellet. S, supernatant. (B) E342Q, and K343Q, residues that we had identified as lipid- FRET experiment showing the fluorescence emission spectra of 1.5 μM CFP- binding amino acids in our complex crystal structure. We con- talin (donor) and 4 μM YFP-talin (acceptor) in the absence and presence of firmed the integrity of our mutant proteins by thermal denaturation increasing concentrations of PIP2 micelles on excitation at 414 nm. The or- SI Appendix dinate shows the relative fluorescence, and the horizontal axis shows the ( ,Fig.S4). This mutant talin exhibited similar melting μ ± wavelength (nm) of the fluorescence emission scan plot. 0 and 200 MPIP2 temperatures (52.02 0.27 °C) as was seen for wild-type talin are shown in black and red traces, respectively; traces for the other con- (53.94 ± 0.27 °C). Thus, the do not seem to affect the centrations are colored spectrally. (Top) Wild type was predominantly mo-

structure of the proteins. nomeric up to approximately 20 μMPIP2 and dimeric for all other four μ We determined lipid binding via a lipid cosedimentation assay higher PIP2 concentrations (50, 75, 100, and 200 M). (Bottom) The lipid- (Fig. 2A), which we used previously to detect micromolar lipid binding–deficient mutant (K357Q, R358Q, Y377F, K272Q, K316Q, K324Q, μ binding (43). The mutant showed approximately 12-fold less E342Q, K343Q) does not dimerize at PIP2 concentrations up to 200 M. binding (as assessed using ImageJ) to the lipid vesicles compared with the wild-type talin. We found a strong electron-dense feature near the side chain We initially interpreted this feature as a second PIP2-binding μ of talin residue R358 that is part of the membrane-distal binding site, since two PIP2-binding sites (with affinities of 0.4 M and μ site and has been identified as involved in the interaction with 5 M for PIP2diC8) were observed by isothermal titration calo- integrin by NMR studies (4). This feature is also near N285 and rimetry (28). However, another study using a phospholipid Q288 from a symmetry-related molecule (SI Appendix, Fig. S5). bilayer that contained 10% PIP2 and was immobilized on a

Chinthalapudi et al. PNAS Latest Articles | 3of6 Downloaded by guest on September 28, 2021 Biacore L1 chip found that the second, weaker site resided on fibers and much smaller cells, comparable to the talin-null cells. F0F1 but did not affect the stoichiometry of the interaction of talin Importantly, large pools of GFP-tagged lipid-binding–deficient with acidic bilayers, and that the contribution of the second site mutant talin (5M or 8M) accumulated in the compared with was therefore negligible (26). Furthermore, our talin K358Q wild-type or mutant (3M) expressing cells. When we previously mutation resulted in similar binding to lipid vesicles as seen for mutated our vinculin or metavinculin lipid-binding residues (9, 36, wild-type talin (Fig. 2A). Thus, K272Q, K316Q, K324Q, E342Q, 44), focal adhesions were never completely disassembled; however, K343Q is a bonafide lipid-binding–deficient mutant. the focal adhesions were largely disrupted for cells expressing We modeled two phosphates into this electron density. The the lipid-binding–deficient talin mutant. Thus, lipid binding seems only non-backbone interaction of the phosphate groups occurs necessary for talin localization to the focal adhesion membrane sites with the guanidinium group of R358. The phosphates are also and for talin regulation of the scaffolding effects. Furthermore, within hydrogen-bonding distance to the carbonyl of N285 and talin-null cells expressing lipid-binding–deficient talin mutants the amide of Q288 of a symmetry-related molecule (SI Appendix, were approximately fivefold smaller (with a chaotic and diffused Fig. S5). With respect to the two reported binding constants, it is actin network) compared with talin-null cells expressing wild- interesting to note that the crystallographic twofold generates a type talin, in which a pronounced F-actin is visible, as the talin dimer. A dimer is also detected by fluorescence resonance cellular integrity is maintained by intact talin–membrane inter- energy transfer of CFP (donor) and YFP (acceptor) fused wild- actions and proper cytoskeletal rearrangements. type talin proteins, but the same assay does not detect di- Next, to determine the effects of talin binding to the plasma merization by our lipid-binding–deficient mutant talin (Fig. 2B). membrane, we assessed the PIP2-mediated integrin activation in Chinese hamster ovary (CHO) cells that stably expressed integrin Talin–PIP2 Interactions Are Essential for Focal Adhesion Formation. (αIIbβ3). When we transiently transfected talin1, we obtained too To elucidate how PIP2 binding to the talin head domain contrib- few cells expressing the protein to perform our integrin activation utes to focal adhesion formation, we generated mutant and wild- assays in triplicate. We overcame this by generating stable CHO type constructs that were tagged with GFP at the N-terminus. cells expressing full-length talin1 tagged with EGFP. For each These constructs were expressed in talin knockout cells that lack construct, we measured the mean fluorescence intensity (MFI) of endogenous talin and do not adhere to the extracellular matrix the bound ligand in at least two independent experiments. The (12). Exogenous expression of GFP-tagged full-length wild-type PAC1 antibody, which recognizes only activated αIIbβ3receptors, talin or the “3M” mutant that resides near the membrane-distal bound to integrin αIIbβ3 with higher affinity in cells expressing wild- integrin-binding site (K357Q, R358Q, Y377F) rescued focal ad- type talin1, with an MFI of 13.5% compared with the cell-expressing hesion formation, cell spreading, and cells clearly displaying focal mutant talin1 (MFI of 1.57%) or the lipid-binding–deficient mu- adhesions connected to prominent actin stress fibers (Fig. 3A). tants(MFIof0.98%for5Mand≤0.26% for 8M) (Fig. 3 B and C Surprisingly, the minimal lipid-binding–deficient mutant “5M” and SI Appendix,Fig.S6). Collectively, our data show that dis- (K272Q, K316Q, K324Q, E342Q, K343Q) and the lipid-binding– rupting talin binding to the membrane affects integrin activation. deficient mutant “8M” (K272Q, K316Q, K324Q, E342Q, K343Q, Since focal adhesion formation was significantly affected by our K357Q, R358Q, Y377F) disrupted focal adhesion formation. The talin lipid-binding–deficient mutant, we assessed the mobility of structure of both mutants seemed to be preserved, as judged by talin by FRAP in the talin-null cell background (Fig. 4A). These thermal denaturation (SI Appendix,Fig.S4). Furthermore, cells experiments were not possible with the lipid-binding–deficient expressing these constructs showed diffused and chaotic actin stress talin mutants that lack distinct focal adhesions in which the

Fig. 3. Talin–PIP2 interactions seem to be essential for focal adhesion formation. (A) Talin-null papillary collecting duct cells (PCDs) engineered to express full-length wild-type GFP-talin1 or mutant GFP-talin1 fusion proteins were analyzed by confocal laser scan- ning microscopy. Representative images, which define the localization of GFP-talin1 (green) at FAs decorating F-actin (red), along with the merged channels, are shown along with the nuclei stained with DAPI. Data shown are representative of five independent experi- ments. (Scale bars: 2 and 5 μm as indicated.) (B and C) Flow cytometry analyses of (B) PAC1 binding to wild- type talin1 and (C) the minimal lipid-binding–deficient mutant 5M (K272Q, K316Q, K324Q, E342Q, K343Q) represented as FACS dot plots. Abscissas, PAC1 binding; ordinates, full-length talin1 expression. Each experi- ment was repeated twice. Stably transfected wild-type talin1 increases integrin activation (correlated to PAC1 binding), but the lipid-binding–deficient mutant sig- nificantly reduces integrin activation.

4of6 | Chinthalapudi et al. Downloaded by guest on September 28, 2021 the side chain of R358 to stack with integrin W775, while K357 is in electrostatic interaction with integrin E779 at the membrane- distal region. In contrast, the lipid-binding–deficient mutant af- fected both focal adhesion formation and integrin activation (Fig. 3). Notably, the distal integrin membrane interaction with talin has been shown to provide the initial linkage between talin and integrin, while strong activation arises from the subsequent binding of talin to the integrin membrane-proximal region (5). Our studies suggest that both integrin-binding sites are involved in integrin activation, while only the membrane-proximal site is involved in focal adhesion formation. Fig. 4. Talin1 influences focal adhesion dynamics at the plasma membrane. The majority of previous in vivo studies have used the tran- (A) Representative images of FRAP recovery of EGFP-tagged wild-type full- siently transfected talin head domain (residues 1–400) or just the length talin1. Focal adhesions are indicated before and after photo- – – −/− F2F3 (residues 203 400) or F3 (residues 309 400) talin FERM bleaching (arrows). (B) FRAP recovery curve of talin1 in talin PCD cells. A subdomain (5, 24), which mimic the activated talin form. In double-exponential model was used to fit these normalized fluorescence curves. The red line is the calculated curve that fits the experimental data contrast, we used full-length inactive talin to measure integrin and is the best fit of a nonlinear regression analysis with >99% confidence. activation by stably expressing talin proteins and selective sorting The results represent the mean ± SEM of 20 independent measurements. via two rounds of fluorescent-activated cell sorting (FACS) be- Error bars are shown in the form of bands (light gray) to represent the SEM. fore performing the integrin activation assays. These stable pools (Scale bar: 5 μm.) of talin-expressing cells enabled us to reliably and reproducibly measure integrin activation rates using A5 CHO cells. PIP2 activates talin by severing the head–tail interaction, major pool of GFP-talin proteins are present in the cytosol. The thereby exposing the integrin-binding site, although simultaneous > fitting of our full-length talin FRAP data with 99% confidence binding of talin to the integrin membrane-proximal site and to PIP2 was possible only with the double exponential (Fig. 4B). The seem mutually exclusive. Mutating talin residues involved in binding resulting recovery curve revealed that the fluorescence recovery to the integrin membrane-distal site did not affect focal adhesion ± was biphasic with an initial fast-phase t1/2 of 1.7 0.15 s and a slow- formation (Fig. 3A), while integrin activation was reduced by ap-

± BIOPHYSICS AND phase t1/2 of 9.2 0.36 s. Thus, talin is recruited to focal adhesions, proximately 8.5-fold compared with wild type (Fig. 3B and SI Ap-

and this localization is maintained in cells moving in a persistent pendix,Fig.S6). In contrast, the lipid-binding–deficient mutant COMPUTATIONAL BIOLOGY manner, allowing determination of the dynamics for the association affected both focal adhesion formation and integrin activation. of talin with the membrane. The initial fast rate of recovery pos- This suggests that the integrin membrane-proximal region plays sibly accounts for the bulk of recovery and most likely reflects this a role in integrin activation and focal adhesions, whereas the rebinding of cytosolic talin to focal adhesions at the plasma membrane-distal region impacts only integrin activation. membrane. A relatively small amount of recovery also may have Talin recruitment to the membrane leads to integrin activa- occurred by lateral diffusion from the adjacent membrane, which tion. These sequential events are closely linked because integrin would be consistent with the second observed, slower t1/2. Collec- activation requires the talin head domain to be positioned close tively, talin binding to the membrane is a dynamic process impor- to the integrin tail on the cytoplasmic face of the membrane. tant for its scaffolding effects at the focal adhesion membrane sites. Talin-mediated integrin activation requires that the auto- inhibitory interactions between the talin head and rod domains Discussion are released by PIP2. The autoinhibitory talin F2F3/R9 structure We determined the talin head/PIP2 complex crystal structure and identified the head–tail interface and suggested a negatively confirmed the lipid-binding site seen in the crystal biochemically. charged surface to repel the membrane, although the orientation Based on sequence similarity and mutagenesis, one of the bona- of the lipid was unknown. However, our talin/PIP2 structure fide residues involved in PIP2 binding, K272, has previously been shows that this negatively charged surface is not planar, but “ ” identified as the membrane orientation patch. Other postulated rather is almost perpendicular to the PIP2-binding site (SI Ap- lipid-binding residues (K256, K274, and R277) (4) are not part of pendix, Fig. S7). It remains to be seen what surfaces on full- the lipid-binding site. Our structural and biochemical identifica- length talin are actually solvent-exposed. tion of K324 as another bonafide PIP2-binding residue also agrees The unique F1 loop has been suggested to become helical – with NMR studies revealing a perturbation in the 268 278 and when in contact with PIP2-rich microdomains to decrease the 318–324 regions on binding to IP3 (28). Surface plasmon reso- distance between talin and the plasma membrane. Furthermore, nance data showed that the K324A mutant bound PIP2 sixfold two of the talin phosphorylation sites that have been mapped weaker compared with wild type. Molecular dynamics simulations from activated human platelets (3) are located on this loop predicted that K324 might be in hydrogen-bonding interactions (T144 and T150). This suggests that phosphorylation of these with PIP2 in addition to being in contact with residue R995 from sites might prevent talin–membrane interactions. However, our the integrin α subunit and thus releasing the electrostatic in- lipid cosedimentation results showed that the lipid-binding– teraction with D723 from the integrin β subunit. However, in the deficient talin, which has an intact F1 loop, does not bind to absence of our high-resolution talin/PIP2 structure, functional lipids. This suggests that in the context of the entire talin head studies had been difficult to interpret. Here we mutated the five domain (residues 1–400), this loop does not interact with the lipid-binding residues that we confirmed structurally and bio- membrane. It remains to be seen if the second integrin-binding chemically to be involved in PIP2 binding: K272Q, K316Q, site located in the talin tail domain (residues 1984–2,113) allows K324Q, E342Q, and K343Q. We found that the talin–lipid in- for simultaneous binding of full-length talin to the membrane via teraction seems to be essential for focal adhesion formation and F2F3 and to integrin via the second integrin-binding site. stabilization, and that this interaction increases integrin activation. While the talin–integrin interaction was shown to be enhanced As shown in our high-resolution confocal imaging studies, the by PIP2 (31), binding of integrin via its membrane-proximal site K357Q, R358Q, and Y377F talin mutant that targets the integrin and PIP2 seem to be mutually exclusive (Fig. 1B). In agreement membrane-distal binding site does not affect focal adhesion with the earlier studies, we show how PIP2 severs the talin head– formation (Fig. 3A). Comparison of our lipid-bound structure tail interaction to expose the cryptic integrin-binding site. Col- with the integrin-bound talin structure revealed a movement of lectively, these findings indicate that on talin recruitment to the

Chinthalapudi et al. PNAS Latest Articles | 5of6 Downloaded by guest on September 28, 2021 cell membrane, lipid binding to the talin head domain releases proteins were transfected in A5 CHO cells that stably expressed integrin the interaction of the talin head with the talin rod domains. (αIIbβ3) receptors, and basal integrin activation assays were performed by Activated talin can then activate integrin, which releases the FACS using A5 CHO cells. interaction of talin with the membrane. It remains to be seen if – the integrin membrane-proximal binding site is solvent-exposed X-Ray Crystallography. Lipid-bound talin1 head domain (residues 1 400; Δ139–168) crystals were obtained by hanging-drop vapor diffusion by op- in the full-length talin structure. timizing the PIP2diC8-to-talin ratio. X-ray diffraction data were collected at the Stanford Synchrotron Radiation Lightsource, beamline 12–2, and the Experimental Procedures X-ray diffraction data were indexed, integrated, and scaled using XDS and DNA Constructs and Protein Preparation. All bacterial expression plasmids and AIMLESS as implemented in autoPROC (45). The unbound talin1 structure mammalian expression plasmids of talin1 used in this study were cloned using (PDB ID code 3IVF) (1) was used to obtain phases for our talin/PIP2 complex, mouse full-length talin1 as a template. Site-directed mutagenesis was per- and crystallographic refinement was performed using autoBuster (46). formed to generate the talin1 mutants, and all DNA constructs were sequence-verified. Proteins were expressed in Rosetta 2 or BL21-CodonPlus ACKNOWLEDGMENTS. We are indebted to the staff of the Stanford (DE3)-RIL host cells and purified by nickel affinity and size exclusion Synchrotron Radiation Lightsource for synchrotron support. We thank the chromatography. Max Planck Florida Light Microscopy facility and Florida Atlantic University, Nikon Center of Excellence for imaging facilities, Dr. Roy Zent (Vanderbilt In Vitro and in Vivo Functional Assays and Confocal Microscopy. The talin1 head Center for Kidney Disease) for talin-null cells, and Dr. Mark Ginsberg (Uni- domain proteins were used for in vitro FRET assays as CFP and YFP FRET donor versity of California, San Diego) for A5 CHO cells. We thank Marina Primi [The Scripps Research Institute (TSRI)] and Charmane Gabriel (Oxbridge Acad- and acceptor pairs. In brief, FRET measurements were performed for 1.5 μM emy) for protein expression and Louis Shane (Palm Beach Gardens, FL) and μ CFP-talin1 and 4 M YFP-talin1 wild-type and mutant proteins in the absence Douglas Bingham (TSRI) for helpful discussions regarding the manuscript. T.I. is – μ andpresenceofincreasingconcentrationsofPIP2 micelles (0 200 M). supported by grants from the National Institute of Health and the Department High-resolution confocal microscopy and FRAP experiments were of Defense, and by startup funds provided to TSRI from the State of Florida. performed with talin-null epithelial cells. EGFP-tagged full-length talin1 This is publication 29675 from The Scripps Research Institute.

1. Elliott PR, et al. (2010) The structure of the talin head reveals a novel extended 24. Vinogradova O, et al. (2002) A structural mechanism of integrin α(IIb)β(3) “inside-out” conformation of the FERM domain. Structure 18:1289–1299. activation as regulated by its cytoplasmic face. Cell 110:587–597. 2. Bretscher A, Chambers D, Nguyen R, Reczek D (2000) ERM- and EBP50 protein 25. Saltel F, et al. (2009) New PI(4,5)P2- and membrane-proximal integrin-binding motifs families in plasma membrane organization and function. Annu Rev Cell Dev Biol 16: in the talin head control beta3-integrin clustering. J Cell Biol 187:715–731. 113–143. 26. Moore DT, et al. (2012) Affinity of talin-1 for the β3-integrin cytosolic domain is 3. Ratnikov B, et al. (2005) Talin phosphorylation sites mapped by mass spectrometry. modulated by its phospholipid bilayer environment. Proc Natl Acad Sci USA 109: J Cell Sci 118:4921–4923. 793–798. 4. Anthis NJ, et al. (2009) The structure of an integrin/talin complex reveals the basis of 27. Goksoy E, et al. (2008) Structural basis for the autoinhibition of talin in regulating inside-out signal transduction. EMBO J 28:3623–3632. integrin activation. Mol Cell 31:124–133. 5. Wegener KL, et al. (2007) Structural basis of integrin activation by talin. Cell 128: 28. Song X, et al. (2012) A novel membrane-dependent on/off switch mechanism of talin – 171–182. FERM domain at sites of cell adhesion. Cell Res 22:1533 1545. 6. Gingras AR, et al. (2008) The structure of the C-terminal actin-binding domain of talin. 29. McLaughlin S, Wang J, Gambhir A, Murray D (2002) PIP(2) and proteins: Interactions, – EMBO J 27:458–469. organization, and information flow. Annu Rev Biophys Biomol Struct 31:151 175. 7. Papagrigoriou E, et al. (2004) Activation of a vinculin-binding site in the talin rod 30. Xu C, Watras J, Loew LM (2003) Kinetic analysis of -activated phosphoinosi- – involves rearrangement of a five-helix bundle. EMBO J 23:2942–2951. tide turnover. J Cell Biol 161:779 791. γ 8. Brown DT, Izard T (2015) Vinculin-cell membrane interactions. Oncotarget 6: 31. Ling K, Doughman RL, Firestone AJ, Bunce MW, Anderson RA (2002) Type I phos- 34043–34044. phatidylinositol phosphate targets and regulates focal adhesions. Nature 420: – 9. Izard T, Brown DT (2016) Mechanisms and functions of vinculin interactions with 89 93. phospholipids at cell adhesion sites. J Biol Chem 291:2548–2555. 32. Di Paolo G, et al. (2002) Recruitment and regulation of phosphatidylinositol phos- γ – 10. Moes M, et al. (2007) The integrin binding site 2 (IBS2) in the talin rod domain is phate kinase type 1 by the FERM domain of talin. Nature 420:85 89. 33. Sun Y, Thapa N, Hedman AC, Anderson RA (2013) Phosphatidylinositol 4,5- essential for linking integrin β subunits to the cytoskeleton. J Biol Chem 282: bisphosphate: Targeted production and signaling. BioEssays 35:513–522. 17280–17288. 34. Kwiatkowska K (2010) One lipid, multiple functions: How various pools of PI(4,5)P(2) 11. Hemmings L, et al. (1996) Talin contains three actin-binding sites each of which is are created in the plasma membrane. Cell Mol Life Sci 67:3927–3946. adjacent to a vinculin-binding site. J Cell Sci 109:2715–2726. 35. van den Bout I, Divecha N (2009) PIP5K-driven PtdIns(4,5)P2 synthesis: Regulation and 12. Atherton P, et al. (2015) Vinculin controls talin engagement with the actomyosin cellular functions. J Cell Sci 122:3837–3850. machinery. Nat Commun 6:10038. 36. Chinthalapudi K, et al. (2014) Lipid binding promotes oligomerization and focal ad- 13. Bois PR, Borgon RA, Vonrhein C, Izard T (2005) Structural dynamics of α-– hesion activity of vinculin. J Cell Biol 207:643–656. vinculin interactions. Mol Cell Biol 25:6112–6122. 37. Hirao M, et al. (1996) Regulation mechanism of ERM (ezrin/radixin/moesin) protein/ 14. Fillingham I, et al. (2005) A vinculin binding domain from the talin rod unfolds to plasma membrane association: Possible involvement of phosphatidylinositol turnover form a complex with the vinculin head. Structure 13:65–74. and Rho-dependent signaling pathway. J Cell Biol 135:37–51. 15. Izard T, Vonrhein C (2004) Structural basis for amplifying vinculin activation by talin. 38. Anthis NJ, Campbell ID (2011) The tail of integrin activation. Trends Biochem Sci 36: – J Biol Chem 279:27667 27678. 191–198. 16. Nhieu GT, Izard T (2007) Vinculin binding in its closed conformation by a helix ad- 39. Hynes RO (2002) Integrins: Bidirectional, allosteric signaling machines. Cell 110: – dition mechanism. EMBO J 26:4588 4596. 673–687. 17. Yogesha SD, Rangarajan ES, Vonrhein C, Bricogne G, Izard T (2012) Crystal structure of 40. Kutateladze TG (2010) Translation of the phosphoinositide code by PI effectors. Nat vinculin in complex with vinculin binding site 50 (VBS50), the integrin binding site 2 Chem Biol 6:507–513. – (IBS2) of talin. Protein Sci 21:583 588. 41. Liu J, Wang Z, Thinn AM, Ma YQ, Zhu J (2015) The dual structural roles of the 18. Yogesha SD, Sharff A, Bricogne G, Izard T (2011) Intermolecular versus intramolecular membrane-distal region of the α-integrin cytoplasmic tail during integrin inside-out – interactions of the vinculin binding site 33 of talin. Protein Sci 20:1471 1476. activation. J Cell Sci 128:1718–1731. 19. Izard T, et al. (2004) Vinculin activation by talin through helical bundle conversion. 42. Kong X, Wang X, Misra S, Qin J (2006) Structural basis for the phosphorylation- Nature 427:171–175. regulated focal adhesion targeting of type Igamma phosphatidylinositol phosphate 20. Bois PR, O’Hara BP, Nietlispach D, Kirkpatrick J, Izard T (2006) The vinculin binding kinase (PIPKIgamma) by talin. J Mol Biol 359:47–54. sites of talin and α-actinin are sufficient to activate vinculin. J Biol Chem 281: 43. Chinthalapudi K, et al. (2018) Lipid binding promotes the open conformation and 7228–7236. tumor-suppressive activity of neurofibromin 2. Nat Commun 9:1338. 21. Tadokoro S, et al. (2003) Talin binding to integrin β tails: A final common step in 44. Chinthalapudi K, Rangarajan ES, Brown DT, Izard T (2016) Differential lipid binding of integrin activation. Science 302:103–106. vinculin isoforms promotes quasi-equivalent dimerization. Proc Natl Acad Sci USA 22. Critchley DR (2009) Biochemical and structural properties of the integrin-associated 113:9539–9544. cytoskeletal protein talin. Annu Rev Biophys 38:235–254. 45. Vonrhein C, et al. (2011) Data processing and analysis with the autoPROC toolbox. 23. García-Alvarez B, et al. (2003) Structural determinants of integrin recognition by talin. Acta Crystallogr D Biol Crystallogr 67:293–302. Mol Cell 11:49–58. 46. Bricogne G, et al. (2011) BUSTER version 2.9 (Global Phasing Ltd, Cambridge, UK).

6of6 | Chinthalapudi et al. Downloaded by guest on September 28, 2021