LFA-1 activates focal adhesion kinases FAK1/PYK2 to generate LAT-GRB2-SKAP1 complexes that terminate T-cell conjugate formation

Lymphocyte function-associated antigen 1 (LFA-1) affinity and avidity changes have been assumed to mediate adhesion to intercellular adhesion molecule-1 for T-cell conjugation to dendritic cells (DC). Although the T-cell receptor (TCR) and LFA-1 can generate intracellular signals, the immune cell adaptor protein linker for the activation of T cells (LAT) couples the TCR to downstream events. Here, we show that LFA-1 can mediate both adhesion and de-adhesion, dependent on receptor clustering. Although increased affinity mediates adhesion, LFA-1 cross-linking induced the association and activation of the protein-tyrosine kinases FAK1/PYK1 that phosphorylated LAT selectively on a single Y-171 site for the binding to adaptor complex GRB-2-SKAP1. LAT-GRB2-SKAP1 complexes were distinct from canonical LAT-GADs-SLP-76 complexes. LFA-1 cross-linking increased the presence of LAT-GRB2-SKAP1 complexes relative to LAT-GADs-SLP-76 complexes. LFA-1-FAK1 decreased T-cell-dendritic cell (DC) dwell times dependent on LAT-Y171, leading to reduced DO11.10 T cell binding to DCs and proliferation to OVA peptide. Overall, our findings outline a new model for LFA-1 in which the integrin can mediate both adhesion and de-adhesion events dependent on receptor cross-linking.

The integrin family is comprised of 18a and 8b subunits that form 24 hetero-dimeric integrins with different ligand specificities 22 . Lymphocyte function-associated antigen 1 (LFA-1) is expressed on T cells and binds to intercellular adhesion molecules (ICAM)-1 and -2 on major histocompatibility complex (MHC) bearing antigen-presenting cells (APCs) 23 . The integrin controls conjugate formation between T cells and antigenpresenting cells as well as the migration of T cells to sites of inflammation and within lymph nodes (LNs) 24 . LFA-1 can be activated by TCR and chemokine receptor ligation 25 . Increased affinity and clustering (that is, avidity) are thought to contribute to adhesion. Increased affinity is accomplished by the extension of the extracellular domain 26 . Disruption of conformational changes in the extracellular domain blocks adhesion 27 . Clustering has been thought to increase contact numbers and stabilize adhesion, although the full purpose of this event has yet to be proven. Previous studies have underscored the primacy of affinity over clustering in regulation of adhesion 27 .
Increased affinity is accomplished by the extension of the extracellular domain and separation of LFA-1 aL and b2 cytoplasmic subunits 28 . FERM-domain carrying talin and kindlins to the b-chain alters the conformation of the extracellular domains 29 . Adaptor protein RIAM (Rap1-GTPinteracting adapter molecule) facilitates talin localization 30 and interacts with the cytoskeleton via profiling and Ena/VASP proteins. GTP-binding protein Rap1 and associated RapL bind to the a-chain, whereas RapL-deficient lymphocytes exhibit impaired adhesion and migration 31 .
Members of the focal adhesion kinase family also regulate cell adhesion and motility 44 . The family includes FAK1 (Focal Adhesion Kinase 1) and PYK2 (proline-rich tyrosine kinase-2). FAKs are comprised of an N-terminal FERM (band 4.1, ezrin, radixin, moesin homology) domain, a 40 residue linker region, a kinase domain, a 200 residue proline-rich region, and a C-terminal focal adhesion targeting domain 45 . The FERM and kinase domains form an auto-inhibitory interaction 45 , which is released in focal adhesions 46 . Focal adhesion kinases regulate focal adhesion contacts, motility and cell survival 47 . Non-lymphoid cells from FAK-deficient mice show enhanced focal adhesion contact formation and reduced cell motility 48 . FAK auto-phosphorylation at Tyr-397 is needed for kinase activation and acts to bind to the SH2-domain of p60 Src kinase 49 . TCR engagement induces FAK and PYK2 phosphorylation and translocation to the immunological synapse (IS) [50][51][52] .
The observation that increased LFA-1 affinity for adhesion precedes the clustering of the integrin caused us to question whether the cross-linking of LFA-1 operated solely to increase LFA-1 adhesion, or whether it could also auto-regulate to terminate adhesion and the interaction between T cells and APCs. Given that LFA-1 generates co-signals, it is also of interest whether LFA-1 can interface with TCR signalling by altering the make-up of the LAT adaptor complex. Here, we show that while affinity changes increase LFA-1 mediated adhesion, the cross-linking of LFA-1 activated associated focal adhesion kinases FAK/PYK2 to phosphorylate LAT Y-171 for a reduction in the contact times of T cells with DCs. These findings support an 'auto-regulatory on-off model' for LFA-1 where the integrin can mediate both adhesion and de-adhesion, dependent on receptor cross-linking.

Results
LFA-1 induces SKAP1 binding to LAT. The LAT binds to PLCg1, GADs-SLP-76 and GRB2 for the activation of T cells 2,10 . To assess whether the LAT complex could be altered by LFA-1, T cells were ligated with either anti-CD3 or anti-LFA-1 in an in situ proximity ligation assay (PLA) (Fig. 1a). Unless otherwise stated, both anti-CD3 and anti-LFA-1 were bivalent and therefore crosslink their respective receptors. Antibodies to LAT, SLP-76 and SKAP1 were employed using isotype-specific antibodies with the DuolinkTM detection system 53 . Anti-CD3 induced SLP-76-LAT proximity signals as shown by an increase in fluorescent dots (Fig. 1a, panel b, also right histogram). Anti-LFA-1 induced no SLP-76-LAT proximity signals (Fig. 1a, panel c), whereas the combination of anti-CD3/LFA-1 reduced the signal compared with anti-CD3 alone (Fig. 1a, panel d). Interestingly, by contrast, anti-LFA-1 induced a moderate PLA signal between LAT and SKAP1 (Fig. 1a, panel g; see right histogram), whereas anti-LFA-1 and anti-CD3 produced the strongest PLA signal between SKAP-1 and LAT (Fig. 1a, panel h). Anti-CD3 alone induced a relatively weak proximity signal between LAT and SKAP1 (Fig. 1a, panel f). These results showed that LFA-1 cross-linking increased the proximity of LAT and SKAP1 either alone or in conjunction with anti-CD3.
in response to anti-CD3 ligation 54,55 . We previously also reported that vesicular LAT interacts with surface SLP-76 (ref. 56). Jurkat T cells transfected with LAT-cherry and SKAP1-GFP were imaged on anti-CD3 or anti-CD3-ICAM1-Fc-coated cover slips and monitored as previously described [56][57][58] . Anti-CD3 induced LAT-cherry clusters by 2 min ligation in the central and peripheral contact regions (Fig. 1b,   Jurkat T-cells were ligated with antibodies as shown, were precipitated with anti-SKAP1 and subjected to blotting with anti-LAT, GADs, SLP-76 and ADAP (upper panels). The same cells were ligated precipitated as above but then subjected to blotting with anti-RapL (n ¼ 3).
clusters, although usually at slightly lower levels (Fig. 1b, left lower panel). By contrast, Although anti-CD3 induced a diffuse array of SKAP1-GFP around the peripheral region of the cell (Fig. 1b, middle right panels), co-ligation of LFA-1 by ICAM1 induced markedly larger and more discrete SKAP1 clusters that were concentrated in the peripheral contact band of the contact region Fig. 1b, left lower panels). LAT-cherry and SKAP1-GFP clusters overlapped in this region (Fig. 1b, see yellow dots). These data confirmed that LFA-1 cross-linking promoted SKAP1-LAT proximity or clusters, which appeared primarily in the peripheral contact band, an area known for the localization of LFA-1. We next assessed biochemically whether LAT could bind to SKAP1 in response to LFA-1 cross-linking (Fig. 1c). DC27.10 mouse T cells were cross-linked with antibody followed by precipitation with anti-SKAP1 or LAT. Membrane and cytosolic fractions were purified by cell disruption and centrifugation, as described in the Methods 37 . Anti-SKAP1 co-precipitated LAT in response to anti-LFA-1 alone and anti-CD3/LFA-1 ligation from membranes (Fig. 1c, upper panels, lanes 4 and 5). Anti-LFA-1 induced more SKAP1/LAT binding than anti-CD3 in over 8 experiments (Fig. 1c, lanes 4 versus 3). Anti-CD3/LFA-1 cooperated to produce the greatest SKAP1/LAT binding (Fig. 1c, lane 5), consistent with PLA analysis (Fig. 1a). No co-precipitated SLP-76 was observed in the anti-SKAP1 IPs. Similarly, anti-LAT co-precipitated more SKAP1 in response to anti-LFA-1/CD3, more than seen with anti-CD3 ligation from membranes ( Fig. 1c, lower panel, lane 6 versus 5).
Further, the cross-linking of LFA-1 by an incubation of cells with plate immobilized ICAM1-Fc induced LAT-SKAP1 complexes as seen by the co-precipitation of LAT with anti-SKAP1 from the membrane fraction (Fig. 1d, lanes 8,9). No association was seen in precipitates from the cytosolic fractions (Fig. 1d, lanes 3-5). Neither GADS nor SLP-76 was observed in the co-precipitates.
Plate immobilized ICAM1-Fc alone or in combination with soluble anti-CD3 also induced SKAP-1-LAT binding in peripheral blood human primary T cells (Fig. 1e). Anti-SKAP1 was observed to co-precipitate LAT from the membrane fraction of cells ligated by ICAM1-Fc over 5-20 min (Fig. 1e, upper panel, lanes 3-5) and by anti-CD3/LFA-1 (Fig. 1e, lanes 6-8). The increased level of LAT associated SKAP1 correlated with the increased presence of SKAP1 in the membrane fraction of cells. We previously showed that activation signals induce the translocation of SKAP1 to the membranes of T cells 37,38 . Neither GADs nor SLP-76 was observed in the co-precipitates. Collectively, these data showed that the cross-linking of LFA-1 alone or together with anti-CD3 induced the association of SKAP1 with LAT in mouse and human T cells.
Further, depletion analysis of cell lysates was also conducted (Fig. 2c). Anti-SKAP1 depleted SKAP1 from lysates by carrying out serial precipitations of the lysate (5 Â ) prior to precipitating with anti-SLP-76. Anti-SKAP1 depletion reduced the presence of SKAP1 in the cell lysate (Fig. 2c, upper panel). Anti-SLP-76 co-precipitated LAT from depleted and non-depleted lysates of cells stimulated with anti-CD3 (Fig. 2c, lanes 2 and 5). Serial precipitations (6 Â ) with anti-SLP-76 were carried out to deplete lysates of SLP-76 (Fig. 2d, upper panel). We observed that anti-SKAP1 co-precipitated equal levels of LAT from SLP-76 non-depleted and depleted lysates from cells ligated with anti-LFA-1 (Fig. 2d, lower panel, lane 3 and 6). Anti-SKAP1 co-precipitated more LAT from anti-LFA-1 than anti-CD3 crosslinked cells. Taken together, these data indicated that there are distinct complexes of LAT-SLP-76 and LAT-SKAP1 in T cells.
Previous studies have shown that ant-CD3 induction of phospholipase Cg-1 (PLCg1) binding to pY-132 on LAT is key to the generation of activation signals by the adaptor 59 . However, anti-SKAP1 failed to co-precipitate PLCg1 as seen in lysates from cells that been ligated for 10 min with anti-LFA-1 or anti-CD3 and blotted with anti-PLCg1 ( Adaptor GRB2 binds to SKAP1. We next assessed whether SKAP1 binds to the LAT binding partners, GRB-2, GADs and PLCg1 (Fig. 3, Supplementary Fig. 1). Flag-tagged SKAP1 was co-expressed with GFP-GRB-2 in 293T cells followed by anti-SKAP1 precipitation and blotting. SKAP1 precipitated GRB-2 as detected by anti-GFP blotting (Fig. 3a, lane 6). Western blotting of lysates with anti-Flag confirmed the expression of SKAP1 (Fig. 3a, lower panel, lanes 1-3). By contrast, Flag-tagged SKAP1 failed to co-precipitate either GADS or PLCg1 (Supplementary Fig. 1a, lane 6). These data showed that GRB-2, but neither GADs nor PLCg1, bind to SKAP1.
At last, for confirmation of the interaction in Jurkat T cells, GST-SKAP1 fusion proteins were used in pull-down assays to precipitate antigen from anti-LFA-1 ligated Jurkat T cells. GST SKAP1 WT precipitated LAT, GRB-2 and ADAP from lysates (Fig. 3d, lane 3). Neither PLCg1 nor SLP-76 was co-precipitated. As a control, the GST-SKAP1 SH3 domain failed to precipitate GRB2, but did precipitate ADAP (Fig. 3d, lane 4), as we and others have reported 35,36 .
LFA-1 recruits FAK-1 to selectively phosphorylate LAT Y-171. Next, we assessed whether LFA-1 cross-linking could induce LAT phosphorylation (Fig. 4a). DC27.10 T cells were incubated with anti-CD3 or anti-LFA-1 for 5 min followed by membrane purification and blotting with anti-phospho-specific LAT antibodies. As previously described 59  The level of Y-171 phosphorylation was similar, or greater, than seen with anti-CD3 (n ¼ 7). Neither Y-191 or Y-132 was phosphorylated. As a control, anti-LAT blotting showed the expression of equal levels of protein in the solubilized membrane fractions. These data showed the novel finding that LFA-1 phosphorylated LAT at a restricted Y-171 residue.
This phosphorylation of a single site suggested that LFA-1 uses a kinase distinct from ZAP-70. FAK1 and PYK2 can associate with focal adhesions where auto-phosphorylation is an essential step for activation 47 . We therefore assessed whether these kinases associated with LFA-1 and mediated phosphorylation (Fig. 4b).
We next used a cold in vitro kinase assay to assess whether FAK1 directly phosphorylated Y-171, (Fig. 4c). 293T cells were transfected with various mutants of Flag-tagged LAT. Anti-Flag was used to precipitate LAT followed by an in vitro kinase assay in the presence of exogenous added recombinant FAK1 and non-radioactive ATP followed by blotting with an anti-phosphotyrosine (4G10). In this, wild-type LAT, Y-191F and the Y-132F Vector + Flag-SKAP1 GFP-GRB-2+ Flag-SKAP1 GFP-GRB-2  mutants were phosphorylated by FAK1 (Fig. 4c, lanes 1, 3 and 5, respectively). However, Y-171F and Y-171/191-F mutants showed a markedly reduced signal (Fig. 4c, lanes 2, 4). Anti-Flag blotting confirmed the equal expression and precipitation of WT LAT and mutants. These data showed that Y-171 was the preferred phosphorylation site of FAK1 in an in vitro kinase assay. We also co-transfected 293T cells with Myc-tagged LAT and either Flag-tagged FAK1, or PYK2, followed by precipitation and blotting with anti-phospho-LAT specific antibodies (Fig. 4d). Remarkably, again, FAK1 phosphorylated LAT on Y-171, but not on Y-191, Y-132 or Y-226 (Fig. 4d, lane 2 versus 1). The expression of related Flag-PYK2 also phosphorylated LAT on Y-171 but not on the other sites (Fig. 4d, lane 4 versus 3).
Jurkat T cells were also transfected with Flag-PYK2 followed by cross-linking with anti-LFA-1 in the presence or absence of the FAK1/PYK2 inhibitor PF 431396. Anti-LFA-1 cross-linking induced Y-171 phosphorylation in mock-transfected cells (Fig. 4e, lane 2 versus 1), and this was further increased by PYK2 transfection (Fig. 4e, lane 4). PF431396 inhibited the anti-LFA-1 induced phosphorylation of Y-171 (Fig. 4e, lane 6). These data showed that PYK2 expression co-operated with LFA-1 cross-linking to induce the phosphorylation of the Y-171 site.
To further determine whether LAT and the LAT pY-171 site was needed for anti-SKAP1 co-precipitation of the kinases, various combinations of Flag-FAK, Flag-PYK2, SKAP1 and LAT were co-expressed in 293T cells and assessed for binding (Fig. 4h). Anti-SKAP1 co-precipitated LAT and Flag-FAK1 or Flag-PYK2 from cells co-transfected with WT LAT, SKAP1 and Flag-PYK2 or Flag-FAK1 (Fig. 4h, lanes 5 and 7). By contrast, greatly reduced when LAT-pY171 and Flag-PYK-2 or Flag-FAK was co-precipitated by anti-SKAP1 (Fig. 4h, lanes 6 and 8). These observations showed that SKAP1 binding to FAK1 and PYK2 requires the presence of binding sites on LAT at Y-171. LFA-1 cross-linking is needed to activate FAK1 and LAT pY-171. We next assessed whether increased LFA-1 affinity mediated binding could activate FAK1/PYK2 and LAT -pY-171, or whether receptor cross-linking was required. The combination of 5 mM MgCl 2 þ and 1 mM EGTA buffer is known to induce conformational changes for high-affinity LFA-1 without receptor cross-linking 22,60 . To confirm the effectiveness of MgCl 2 þ / EGTA, Jurkat cells were incubated with the reagents for 15 min and assess for adhesion to immobilized ICAM-Fc on plates (Fig. 5a). This showed a marked increase in the number of cells bound to plates (that is, 1,800 versus 10 cells per cm À 2 ), thereby confirming affinity activation.
FAK1 limits T-cell dwell times with DCs. We next assessed whether FAK1 and LAT-Y171 influenced the dwell times of T cells with DCs. DO11.10 CD4 þ T cells were transfected with FAK1 siRNA followed by a measure of the dwell times with DCs, as described 37,61 (Fig. 6). Mature DCs were labelled with SNARF-1 and pre-incubated with OVA peptide (DC-OVA) prior to incubation with T cells 61 . FITC-siRNA electroporation showed an 80% siRNA uptake by T cells and reduced expression of FAK1 by anti-FAK1 blotting (Fig. 6a, right panels). These T cells showed significantly longer contact times with DCs than T cells transfected with the siRNA scrambled control. FAK1 siRNA increased the mean contact time of T cells from 675 to 814 s in the presence of OVA peptide, and from 334 to 482 s in the absence of peptide (Fig. 6a, left and middle upper panels). Antigen independent conjugation between T cells and DCs has been previously described. Concurrently, we observed a reduction in T-cell motility and displacement owing to FAK1 siRNA (Fig. 6a, lower right and left panels). In the presence of OVA peptide, T cells with FAK siRNA showed a mean velocity of 6 mm min À 1 compared with 8.6 mm min À 1 for scrambled, whereas velocity in the absence of antigen increased from 10 mm min À 1 relative to 13 mm min À 1 for the scrambled control. Similarly, displacement (distance from the site of origin) decreased with FAK1 siRNA from 48 to 32 mm in the presence of peptide and from 80 to 54 mm in its absence. Consistent with this, FAK1 siRNA T cells were also more rounded and less polarized (that is, less amoeboid) in the presence and absence of OVA peptide relative to cells expressing control siRNA (as defined as 1.4/1.0 length versus width) ( Fig. 6b; right panel and images).
Conversely, FAK1-GFP transfection of DO11.10 CD4 þ T cells reduced contact times with DCs (Fig. 6c, left panel). In the presence of OVA peptide, we found that FAK1-GFP transfected cells showed a mean contact time of 486.3 relative to 664.2 s for vector-GFP transfected control cells. In the absence of antigen, the cells showed a mean reduced dwell time of 275 compared with 389 s for vector-GFP. At the same time, mean motility was significantly increased from 8.9-11.7 s in the presence of antigen and 14-16.4 s in its absence of OVA peptide (Fig. 6c, middle  panel). Displacement was also increased from 41 to 65 mm in the presence of antigen and 88 to 117 mm in the absence of OVA peptide (Fig. 6c, right panel). These observations show that FAK1 in T cells reduces contact times between T cells and DCs.

FAK limitation on T-cell dwell times depends on LAT Y-171.
Given the phosphorylation of LAT Y-171 by FAK1, we next asked whether the FAK1 induced reduction in T-cell conjugation was dependent on LAT Y-171 (Fig. 7). In this case, DO11.10 CD4 þ T cells were co-transfected with FAK1-GFP and wild-type LAT or LAT-Y171F followed by a measurement of dwell times. In the presence of OVA peptide, the combination of FAK-GFP and LAT reduced the mean contact times from 751.1 to 583.3 s, and in the absence of antigen, from 396 to 275 s (Fig. 7a). Conversely, the combination of FAK1-GFP and LAT increased motility and displacement (Fig. 7b,c).
Significantly, by contrast, the co-expression of the LAT-Y171F mutant completely reversed the reduced dwell effect of FAK1-GFP on conjugation times (Fig. 7a, upper left panels). It even increased the mean contact from 715 to 851 s in the presence of OVA and from 396 to 544 s in the absence of OVA peptide (Fig. 7a, upper left panels). This increase in contact time was even higher than seen in the GFP vector control cells. Similarly, the expression of LAT-Y171F blocked the ability of FAK1 to increase the mean velocity and displacement of cells (Fig. 7b,c, right and  lower panels). These observations showed that the ability of FAK1 to limit T-cell contact with DCs was dependent on LAT Y-171, the site phosphorylated by FAK1.
Similar results were obtained when LAT-Y171F was expressed in T cells in the absence of co-transfected FAK1 (Fig. 7d). In the absence of overexpressed FAK1, the effect of the wild-type LAT was less apparent, however, the expression of LAT-Y171F increased contact times (620-850 s) (left panel) concurrent with a reduced velocity (7.7-4.5 mm min À 1 ) (Fig. 7d, middle panel) and displacement (27-12 mm) (Fig. 7d, right panel). These observations showed that LAT-Y171F was effective in modulating conjugation in the presence of endogenous FAK1.
This modulation of dwell times by FAK1 and LAT-Y171F was also reflected in the proliferative response of DO11.10T cells to OVA peptide (Fig. 7e). DO11.10T cells were labelled with carboxyfluorescein diacetate succinimidyl ester, incubated with DCs plus OVA peptide and assessed for cell division by flow cytometry at day 6. Although vector-transfected cells under cell division can be seen as the presence of M2-M6 cell divisions, transfection of FAK1 and WT LAT inhibited the proliferation (that is, mostly M2-M3). By contrast, the combination of FAK1 and LAT-Y171F restored to proliferation (that is, M4-M6) to that seen in the vector transfected cells (Fig. 7e, see right histogram). These observations showed that LAT Y-171 reversal of the dissociative effects of FAK1 on T-cell/DC conjugation was reflected in a restoration of T-cell proliferation.
At last, JCAM.2 T cells lacking LAT were transfected with wild-type LAT versus Y-171 and assessed for random motility on ICAM-1-Fc coated plates (Fig. 7f). Although wild-type LAT supported extensive random migration as seen by the presence of extensive tracks, Y-171 mutant arrested motility as shown by the absence of tracks (Fig. 7f, right panels). Blotting showed equal levels of expressed wild-type and mutant (Fig. 7f, left panel). These observations indicated, using another model, that the LAT Y-171 site is needed for T-cell motility, in this instance, random motility.

Discussion
Although LFA-1 binding to ICAM-1 initiates T-cell adhesion to antigen-presenting cells, the identity of signals that limit or terminate conjugate is poorly understood. In this context, LFA-1 undergoes changes in affinity and avidity (that is, clustering), events that have been assumed to increase adhesion for conjugate formation. LFA-1 generates 'outside-in' co-signals, although whether these events affect TCR signalling has been unclear. Here, we show that although increased affinity initiates adhesion, LFA-1 cross-linking recruits and activates FAK1 and PYK2 to phosphorylate LAT selectively on a single Y-171 site that binds to the GRB2-SKAP1 complex and limits dwell times with DCs. Our findings made two novel points, firstly that LFA-1-FAK-1 co-signals can intersect with TCR signalling by re-structuring the LAT complex in T cells, and second, that LFA-1 is an 'auto-regulatory on-off' receptor that can mediate the opposing roles of adhesion and deadhesion dependent on affinity versus clustering.
Previous studies underscored the primacy of affinity over clustering in regulation of LFA-1 adhesiveness 27 . The current model is that increased affinity cooperates with increased avidity to increase adhesion. The fact an increased LFA-1 affinity precedes clustering caused us to question whether cross-linking might generate co-signals that also terminate conjugation (that is, auto-regulatory model). In agreement, we found that the crosslinking of LFA-1 recruited and activated focal adhesion kinases FAK1/PYK2 to phosphorylate LAT Y-171 and reduce contact   3). P values were calculated by one-way analysis of variance (ANOVA) between groups followed by a series of ttests. times between T cells and DCs. In our model, increased LFA-1 affinity to monovalent ligand ICAM1 mediates contact with DCs which is then followed by clustering and cross-linking of the receptor, which reduces or terminates adhesion. This occurs via the FAK1 activation, and its mono-phosphorylation of LAT Y-171. Since LFA-1 clusters in the pSMAC, it seems reasonable that some degree of receptor cross-linking would occur. Importantly, FAK1/PYK2 were recruited and activated by LFA-1 cross-linking, and not by affinity activation of LFA-1 with MgCl2 þ /EGTA, or the monovalent binding of anti-LFA-1 Fab' or in combination. FAK1/PYK2 auto-phosphorylation or LAT Y-171 phosphorylation depended on cross-linking with a bivalent antibody. These general effects were seen in DO11.10, Jurkat and primary T cells. LFA-1 is, therefore, similar to receptors such as c d ARTICLE CD4 and CD8, which we previously showed to recruit the kinase p56 lck to activate signalling pathways in cells 4,6,7 . In the case of LFA-1, associated FAK1 interferes or terminates conjugate formation initiated by affinity induced LFA-1 adhesion. The pathway could also have roles during leukocyte endothelial transmigration.
Previous studies have focused on the central role played by LAT in the mediation of TCR signals 2 . Our study has identified two new kinases, FAK1 and PYK2, which can phosphorylate and interface with the LAT adaptor. Intriguingly, unlike ZAP-70 which phosphorylates multiple sites on LAT (that is, Y-132, Y-191, Y-171 and Y-226), FAK1 and PYK-2 phosphorylated only a single site on LAT at Y-171. This remarkable fidelity was seen in overexpression and in in vitro kinase assays, as well as with FAK1 drug inhibition following LFA-1 ligation. LFA-1-FAK1 therefore skewed the nature of LAT regulation from poly-to mono-phosphorylation at Y-171. FAK1 and PYK2 showed that same pattern consistent with the conserved nature of their kinase domains. LFA-and FAK1 in turn facilitated recruitment of a novel GRB-2-SKAP1 complex. As with GADs binding to SLP-76, we found that the SH3 domains of GRB-2 bound to SKAP1, whereas the SH2 domain of the adaptor bound to LAT. In this manner, GRB-2 brings SKAP1 to the LAT complex. FAK1 and PYK2 were also part of the complex by unexplained mechanisms.
Surprisingly, we found that LAT-GRB-2-SKAP1 complexes were distinct from LAT-GADS-SLP-76 complexes. At no point did anti-SKAP1 precipitates show the presence of either SLP-76 or GADs in response to anti-LFA-1 or anti-LFA-1/CD3, despite co-precipitating LAT, or vice versa. This new observation indicates that different receptors produce distinct LAT complexes for the regulation of function. As a possible explanation, anti-LFA-1 induced LAT-SKAP1 micro-cluster formation primarily in the pSMAC-like region enriched with LFA-1. This contrasted with anti-CD3 induced LAT-GADs-SLP-76 complexes that are seen in both the centre and periphery of cells 54,62 . This concentration of LAT-FAK1 might topographically allow for FAK1 mediated phosphorylation separate from ZAP-70. However, it was also noteworthy that TCR ligation cooperated strongly with anti-LFA-1 with increased LAT-SKAP1 proximity as seen by in situ proximity analysis and biochemically by the presence of co-precipitated SKAP1. This basis for this cooperation is not clear, but might be due to the well-known ability of the TCR to promote LFA-1 clustering and to independently induce FAK1 and PYK2 auto-phosphorylation that could synergize with anti-LFA-1.
Functionally, we found that FAK1 reduced T-cell contact times with DCs as observed by siRNA downregulation of FAK1 (that is, which increased dwell times), or by FAK1 transfection (that is, which reduced contact times). Increased dwell times in turn correlated with reduced motility, and conversely, reduced contact with increased motility. Intriguingly, co-expression of the LAT-Y171F mutant reversed the ability of FAK1-GFP to shorten conjugation times. In fact, in the absence of co-transfected FAK1, it even increased mean contact times relative to the vector transfected cells, presumably due to an ability to complete with endogenous wild-type LAT, resulting in increased contact times and enhanced proliferation at wild-type levels. These findings are therefore consistent with a model where LAT Y-171 residue negatively regulates T-cell conjugation formation and activation when phosphorylated by FAK1. Whether PYK2 operates the same way is uncertain. Reconstitution studies showed that PYK2 failed to replace FAK1 to enhance migration 63 .
Further downstream, the full range of signalling events responsible for the reduced conjugation of T cells remains to be determined. We previously showed that SKAP1 regulates LFA-1 mediated adhesion, conjugation and motility 32,34,37,39 . In this context, SKAP1 expression is needed for TCR induced Rap1-RapL complex formation, whereas RapL mutations that abrogate SKAP1 binding reverse its ability to reduce T-cell motility (that is, stop signal) in LNs 37 . We also previously showed that the expression of the SKAP1 binding partner, ADAP promoted the motility of T cells on ICAM1 coated plates 19 . At the same time, FAK1 promotes the turnover of cell contacts in non-lymphoid cells lacking ADAP and SKAP1 (refs 48,64,65). FAK1-deficient cells are less motile with larger focal adhesion plaques 48 . Downstream regulators may therefore also include other FAK1 targets such as a focal adhesion associated protein, paxillin or RhoA 66 . LAT-GRB-2 also binds to THEMIS, a mediator of cell survival 67 .
LAT has multiple potential sites for GADs and GRB-2 binding as well-established by the Samelson lab 2 . GRB-2 and GADs use their SH2 domains to bind LAT and their SH3 domains to bind to their respective binding partners. We found that the N-terminal SH3-1 domain was most effective in binding to the SK region of SKAP1 (between the SH3 and PH domains). The GADs SH3 domain binding is unusual in binding to an atypical motif on SLP-76 (ref. 17), that is not present in SKAP-1.
The conditions that determine the specificity of SH2 binding to LAT are not fully understood. GRB-2 binding requires an N in the plus 2 position relative to the Y that are found in sites, Y-110 (YENE), Y171 (YVNV), Y191 (YVNV) and Y-226 (YENL) 68 . These motifs correspond to a class of hybrid ligand motifs that are permissive and less selective for GRB2 binding as seen with the CD28 pYMNM motif that binds to both GRB-2 and PI 3-kinase 69,70 . The putative sites are not equally bound or accessible to GRB-2 (or GADs) 59 . With LFA-1, we found that GRB-2-SKAP1 bound only to the Y-171 site, the site phosphorylated by FAK1/PYK2. Further, FAK1 and PYK-2 were also part of the complex as seen in co-precipitation experiments. It is possible that the presence of these other components promotes the preferential binding of GRB-2 for Y-171 motif.  3). P values were calculated by one-way analysis of variance (ANOVA) between groups followed by a series of t-tests.
Overall, our findings fit a new model (that is, auto-regulatory) where LFA-1 mediates both adhesion and de-adhesion. LFA-1 co-signals intersect with TCR signals at the level of the LAT complex. In our model, adhesion is mediated by increased affinity, whereas the subsequent clustering and cross-linking of LFA-1 leads to FAK1/PYK2 activation and the phosphorylation of LAT to limit or terminate T-cell/DC conjugation in adaptive immune responses. Enhanced TCR ligation at the nascent IS would activate ZAP-70 to induce classic activating LAT-PLCg1/SLP-76 complexes. ITK, RLK and ACK (Cdc42-associated kinase 1) phosphorylate SLP-76 (refs 71-73), whereas PLCg1 and GADs-SLP-76 mobilize calcium and activate the ERK pathway 9,11 . However, with the maturation of the IS and subsequent clustering, the cross-linking of LFA-1 (either alone or in conjunction with the TCR) would become increasing dominant with increased numbers of LAT-GRB-2 SKAP1 complexes at the expense of activating LAT-GADs-SLP-76-complexes. Further, by activating FAK1, LFA-1 would induce a pathway that leads to reduced conjugation and the dissociation of T cells from DCs. Our findings suggest the possible use of FAK1 inhibitors to increase T-cell conjugate formation in vaccine development and anti-tumour immunotherapy.

Methods
Cell culture. Jurkat cells (American Type Culture Collection (ATCC)) were grown in RPMI meeium 1640 with 10% fetal calf serum (FCS), 2 mM L-glutamine, penicillin and streptomycin. J14 cells (SLP-76-deficient Jurkat cells) were kindly provided by Dr A. Weiss (UCSF). DC27.10 cells were obtained from Dr Oreste Acuto (University of Oxford), whereas isolated mouse splenocytes were cultured after removal of red blood cells with hypotonic buffer (0.15 M NH 4 Cl, 1 mM NaHCO 3 , 0.1 mM EDTA, PH 7.25) in RPMI containing 10% FCS, 5% glutamine, 5% penicillin/streptomycin and 2-mercaptoethanol at 5 Â 10 À 5 M. To generate T-cell blasts, isolated splenocytes were cultured with Con A (2.5 mg ml À 1 ) for 2 days, and then cultured in growth medium with interleukin-2 (IL-2) (20 ng ml À 1 ) for 2-3 days. For experiments, T cells were washed and rested in growth medium for 2 days in the absence of growth factor. CD4 þ T cells were purified using Dynabeads mouse CD4 (11415D-Thermo Fisher Scientific) according to the manufacturer's instructions (Invitrogen). All experiments abided by the ethical review standards (Research Ethics Committee and the Animal Welfare and Ethical Review Body) of Cambridge University.
Human T cells were prepared from buffy coats isolated by leukophoresis (Addenbrookes Hospital Cambridge with donor consent). Mononuclear cells were isolated by Ficoll density gradient centrifugation. After being washed, cells were stimulated for 24 h at 37°C with 5 mg ml À 1 of phytohemagglutinin. After two washes, cells were maintained for 5-6 days in exponential growth phase in RPMI medium plus 10% FCS supplemented with 20 ng ml À 1 of recombinant IL-2, followed by washing and IL-2 starvation for 24 h. Cells were re-suspended at 10 8 cells per ml in 250 ml of complete medium containing 100 mg of SLP 76 EYFP expression construct.
Transfections were performed in 4 mm gap-cuvettes with the use of a BTX ECM 830 electroporator with a single pulse of 385 V and 6 ms. Cells were immediately transferred to pre-warmed complete medium and allowed to recover for 24 h prior to analysis. siRNAs specific for FAK1 (not PYK2) and control scrambled siRNAs were synthesized by Cell Signaling Technology. FAK siRNA purchased from Santa Cruz Biotechnology, Cat No: sc-35353. No effect of FAK-1 siRNA was noted on PYK2 expression.
Constructs and transfection. Constructs of SKAP1 and GRB-2 were inserted into a pGEX5x-3 (GE Healthcare), into a 3xFlag-tagged and EGFP-tagged pcDNA3.1-Hygro (Invitrogen) vector. LAT and various mutants were cloned into a myc and 3xFlag-tagged pcDNA3.1 vector. PYK2 and FAK were cloned in a 3xFlag-tagged vector. SKAP1 and PLCgamma1 were cloned in pSRalpha expression vector in-frame with HA-tag. SLP-76 cDNA was sub-cloned in the pEYFP-N1 vector (Clontech). Mutants were generated by site-directed mutagenesis using the Quick Change protocol and Pfu Ultra II Fusion HS DNA Polymerase (Stratagene). All constructs were confirmed by sequencing. Transfections were performed in 4mm gap-cuvettes with the use of a BTX ECM 830 electroporator with a single pulse of 385 V and 6 ms. Cells were immediately transferred to pre-warmed complete medium and allowed to recover for 24 h prior to analysis. FAK and PYK inhibitor PF 431396 and PF 573228 were purchased from Tocris.
In situ PLA. In situ PLA was achieved using DuolinkTM in situ PLA reagents. Cells on slides were blocked with Duolink Blocking stock followed by the application of two PLA probes in 1x Antibody Diluent. Wash the slides in a wash buffer (1 Â TBS-T) for 5 min, twice and carry out a hybridization using Duolink Hybridization stock 1:5 in high-purity water and mix followed by incubation for 15 min at þ 37°C. Duolink Ligation with ligase and incubation of the slides in a pre-heated humidity chamber for 15 min at þ 37°C. Amplification was then achieved using Duolink Amplification stock and polymerase with incubation of the slides in a pre-heated humidity chamber for 90 min at þ 37°C. The quantification was based on the counts of five different fields in each experiment (n ¼ 4).
Immunoblotting. Precipitation was conducted by incubation of the lysate with the antibody for 1 h at 4°C, followed by incubation with 30 ml of protein G-Sepharose beads (10% w/v) for 1 h at 4°C as described 8,37,38 . Immunoprecipitates were washed three times with ice-cold lysis buffer and subjected to SDS-PAGE. For blotting, precipitates were separated by SDS-PAGE and transferred onto nitrocellulose filters (Schleicher and Schuell). Bound antibody was revealed with horseradish peroxidase-conjugated rabbit anti-mouse antibody using enhanced chemiluminescence (Amersham Biosciences). For purification of membrane fractions, Jurkat, DC27.10 or primary T cells were sheared in hypotonic buffer and the nuclei removed by low-speed centrifugation (1,500 r.p.m., 10 min), and the supernatant was recentrifuged at high speed (25,000 r.p.m.) for 1 h. The cytosolic fraction comprised the supernatant, whereas membranes remained in the pellet. The full length versions of the immunoblots in the paper is shown in Supplementary Fig. 2.
GST pull-down assay. The expression of recombinant GST-proteins was induced in Escherichia coli BL21 cells at 37 C for 2 h by the addition of 1 mM IPTG. GST-fused proteins were purified with the Cell Lytic B protocol (Sigma #B7435). Cell lysates were incubated with GST fusion proteins for 3 h followed by analysis with SDS-PAGE and western blotting.
Integrin adhesion assay. Jurkat cells (1-2 Â 10 5 cells per well) were incubated with 5 mM MgCl 2 þ and 1 mM EGTA for 15 min, were washed and assessed for adhesion to immobilized ICAM-1-Fc-coated plates by incubating cells on plates for 30 min at 37°C. Flat-bottomed 96-well plates had been coated previously with 4 mg ml À 1 murine ICAM-1 human Fc in PBS overnight at 4°C, washed with RPMI 1640 medium, and blocked with 2.5% BSA in PBS for 1 h at 37°C. Plates with adhered cells were then gently washed twice with RPMI 1640 medium followed by the counting of cells and photography of the plates 74 .
Live cell imaging. Jurkat T cells were transfected with LAT-cherry and SKAP1-GFP, and were imaged on anti-CD3 or anti-CD3-ICAM1-Fc coated cover slips. All imaging assays were performed in Poly-L-lysine-treated chambered glass culture slides (Lab-tek) as described 56,57 . Plates were coated with 5 mg ml À 1 anti-CD3 (OKT3) and 2 mg ml À 1 ICAM1-Fc overnight at 4°C. Cells were imaged at the interface using a Zeiss LSM 510 confocal microscope using excitation wavelengths of 514 nm for EYFP and 594 nm for mcherry and a Â 63 oil immersion objective (NA ¼ 1.2). Images were collected at 10 s intervals. Single Z sections were captured over time to improve the rate of image acquisition. All images were processed by Volocity software (Improvision). Standard deviations and standard errors were calculated with the use of Microsoft Excel and GraphPad Prism. Differences between means were tested using unpaired Student's t-test.
Statistical analysis. Pearson's correlation coefficients (Rr) were calculated by intensity correlation analysis with ImageJ. Column statistics were performed with GraphPad software (Prism). Paired and unpaired t-tests were performed to analyse the data where appropriate. In certain instances, one-way analysis of variance between groups followed by a series of t-tests. Mean values are shown and error bars represent the s.e.m. In the statistical analysis, P values 40.05 are indicated as nonsignificant (ns), P values between 0.01 and 0.001 are indicated by double asterisks (**), and P values smaller than 0.001 are indicated by triple asterisks (***).
Data availability. Data supporting the findings in this study are available within this article and from the corresponding author upon reasonable request.