n e ws a nd v i e ws
TCR signaling: the barrier within Michael L Dustin & Simon J Davis
TCR Membrane
Lat
b
Inactive
Primed
Active
Lck SH3
CD45
Lck
Csk
CD45
c
No perturbation CskAS inhibition CskAS inhibition + actin disruption (or costimulation)
SH2 K
P
T cell
P
C
sk
Active
45
T cell
D
45
136
Glycocalyx
a
D
Michael L. Dustin is with the Kennedy Institute of Rheumatology and Nuffield Department of Orthopedics and Musculoskeletal Sciences, The University of Oxford, Oxford, UK. Simon J. Davis is with the Medical Research Council Human Immunology Unit and Radcliffe Department of Clinical Medicine, Weatherall Institute of Molecular Medicine, The University of Oxford, Oxford, UK. e-mail: [email protected]
tyrosine kinase CD45 in the case of Lck, which favors activation. Unlike the SFKs, Csk lacks a membrane-recruitment domain and thus needs to bind through its SH2 domain to membrane-anchored adaptors such as Cskbinding protein (CBP). Weiss and colleagues ask what would happen if Csk could be rapidly turned off. Published studies have used overexpression of dominantnegative forms of Csk5, but approaches that involve stable gene expression are subject to compensation and other caveats. To acutely
extension that contains a tyrosine residue that can be phosphorylated. The tertiary structure of these kinases allows two inhibitory intramolecular interactions: interaction of the SH3 domain with the polyproline motif, and interaction of the SH2 domain with the phosphorylated C-terminal tyrosine. The aptly named ‘cytosolic C-terminal Src kinase’ (Csk) phosphorylates that C-terminal tyrosine and ‘encourages’ intramolecular binding of the SH2 domain and inhibition of the kinase. The phosphate is removed by the transmembrane
C
uch attention has been focused on the barrier to T cell activation posed by the ‘forest’ of bulky glycoproteins that form the glycocalyx of the cell surface, but there are even deeper potential barriers to signaling on the other side of the membrane that have only recently started coming into focus1. The cell cortex comprises a crosslinked network of filamentous actin (F-actin), myosin and other proteins that lie directly underneath the plasma membrane and form a barrier typically 100 nm in thickness2 (Fig. 1a). The cortex offers resistance to externally applied forces and has a key role in cell-shape change. In this issue of Nature Immunology, Weiss and colleagues test a new tool for ‘kick-starting’ T cell activation and discover a new F-actindependent barrier to signaling via the T cell antigen receptor (TCR) in thymocytes3. That barrier is breached at least in part by engagement of the coreceptor CD28 and thus has the potential to be important for the generation of CD28-dependent T cell lineages such as Foxp3+ regulatory T cells4. Src-family kinases (SFKs) such as Lck undergo allosteric regulation driven by tyrosine phosphorylation (Fig. 1b). SFKs all share a unique amino (N)-terminal domain that is involved in membrane interactions, a Srchomology 3 (SH3) domain, an SH2 domain, a linker with a polyproline motif, a kinase domain (SH1) and a carboxy (C)-terminal
C
M
k Lc
npg
© 2014 Nature America, Inc. All rights reserved.
Acute inhibition of the regulatory kinase Csk reveals additional checkpoints for full activation of thymocytes via the T cell antigen receptor.
rk p-E -g1 LC p-P
F-actin
P P
PLC-g1
Figure 1 Effects of Csk inhibition and actin disruption on TCR signaling. (a) The actin cortex forms a physical barrier between the membrane and the rest of the cytoplasm. Only the F-actin part of the cortex is presented here. The proteins are drawn to scale; the cytoplasmic domains of TCR-CD3 and Lat, even at their theoretical maximal length (~3.5Å per residue), are occluded by the actin cortex. Scale bar, 10 nm. (b) Role of Csk in the regulation of Lck. The inactive, primed and active isoforms of Lck interconvert through the phosphorylation of Csk and autophosphorylation of Lck and through CD45-mediated dephosphorylation. (c) Signaling thresholds and the effects of global perturbation of receptors. Arrows indicate signaling cascades initiated by Lck-mediated phosphorylation of individual TCRs in the resting state; arrow length corresponds to the depth of the signaling cascade reached (so for a given condition the lengths vary, reflective of stochastic effects): basal signaling (gray); signaling initiated by inhibition of CskAS (purple); or signaling initiated by inhibition of CskAS and disruption of actin (or costimulation via CD38; red). Blue and red dashed lines represent signaling thresholds that correspond to the phosphorylation (p-) of PLC-g1 and Erk. In the unperturbed cell, the signaling cascades reach neither threshold. Inhibition of CskAS probably globally enhances Lck activity and TCR phosphorylation and thus increases the number of signaling cascades and the probability that at least one cascade breaches the threshold required for the phosphorylation of PLC-g1. The phosphorylation of Erk also requires disruption of actin, which may enhance the access of PLC-g1 to substrates in the membrane.
volume 15 number 2 february 2014 nature immunology
npg
© 2014 Nature America, Inc. All rights reserved.
n ews a n d views and specifically inhibit Csk, the Weiss laboratory teamed up with the Shokat laboratory to generate a bulky ATP analog–sensitive mutant of Csk (CskAS)6. In published work, they found that when a membrane-anchored form of CskAS was expressed in the Jurkat human T lymphocyte cell line, inhibition with the bulky analog resulted in the full activation of TCR signaling, including phosphorylation of the kinase Erk and an increase in cytoplasmic Ca2+ concentrations6. Now, Weiss and colleagues have introduced the gene encoding CskAS into mice via bacterial artificial chromosome transgenesis to gauge the effect of acute inhibition of Csk in primary cells3. When Csk is acutely inhibited in primary thymocytes, the early TCR activation cascade is triggered, including activating phosphorylation of Lck, the cytoplasmic kinase Zap70, phospholipase C-γ1 (PLC-γ1) and the PLC-γ1 adaptor Lat. These results indicate that the TCR-proximal elements of the tyrosine kinase cascade work as expected, with acute inhibition of Csk pushing the equilibrium toward greater activation of Lck, which drives assembly of the classical proximal signalosome without engagement of the TCR. Unexpectedly, however, whereas inhibition of Csk generates robust TCR-proximal signals, no activation of Erk or elevation in cytoplasmic Ca2+ ensues, which identifies a previously unsuspected signaling checkpoint comprising the cleavage of phosphatidylinositol-4,5-bisphosphate by PLC-γ1 (ref. 3). However, that checkpoint is overcome by disruption of F-actin with cytochalasin D or by engagement of CD28 in thymocytes with dendritic cells or beads coated with the CD28 ligand CD86, which results in increased recruitment of F-actin to the synapse. Mature T cells have a more complex response to manipulation of F-actin after inhibition of Csk, so the existing view that F-actin has a positive role in mature T cells still holds true7. Little is known about the actin cortex of lymphocytes. At the typical depth of 100 nm (and a mesh size of 15–30 nm), the cortex of T cells would be much deeper than the maximum theoretical reach of the TCR or Lat (Fig. 1a). That is why processes such as secretion need to generate holes in the actin cortex to allow vesicles to reach the plasma membrane8. Cortical actin is argued to be a diffusion barrier in signaling via the B cell antigen receptor, and the work of Weiss and colleagues now shows that cortical F-actin acts as a barrier that apparently constrains the access of otherwise active PLC-γ1 to phosphoinositide lipids3. Exactly how the barrier function is achieved is unclear, as there is at least some access of PLC-γ1 to the signalosome because it is phosphorylated after inhibition of Csk without
disruption of actin. Bringing CD28 into the equation as a remodeler of actin nevertheless has interesting functional implications, as a phosphorylated tyrosine residue in the cytoplasmic domain of CD28 may also directly modulate Lck activity through binding to Lck’s SH2 domain in competition with the terminal tyrosine phosphorylated by Csk9. During the thymic differentiation of regulatory T cells, therefore, CD28 may both overcome the actin barrier and contribute to Lck-activation mechanisms that are critical for additional downstream signaling. It is necessary, however, to bring just one note of caution to interpretation of these new data. This is because whereas the types of perturbation imposed by Weiss and colleagues3—that is, global disruption of actin and acute inhibition of Csk—will affect all the TCRs present simultaneously, the T cell is able to respond in a physiological setting to just a few TCR ligands10, which indicates that individual TCRs acting locally are capable of strong signaling. Perturbations that are imposed on very large numbers of TCRs simultaneously, whose individual effects might otherwise be relatively weak, will tend to increase the overall probability of signaling and elicit responses that might be similar in strength to those primed by ‘strong’ ligands acting through just a few TCRs, due to stochastic effects11 (Fig. 1c). Among the challenges that remain, it will be necessary to show that local remodeling of actin, perhaps mediated by costimulation, contributes substantially to local signaling responses to conventional ligands rather than contributing to other local processes. The classical feedback model for Csk is that active SFKs phosphorylate sites on adaptors at the membrane that then recruit Csk, which in turn inhibits the active SFKs by phos- phorylating their C-terminal tyrosine residues12. It has been shown, however, that the docking site for Csk in CBP is transiently dephosphorylated during triggering of the TCR5,13,14. Nevertheless, the role of CBP in controlling Csk localization seems to be somewhat redundant, as CBP-deficient mice have no alteration in T cell phenotype15. While Weiss and colleagues clearly show that inhibiting Csk is a potent means of triggering the TCR3, as the authors argue elsewhere, it is difficult to envisage how inhibition of Csk might be coopted in vivo in a way that initiates or promotes new signaling14. After all, Weiss and colleagues observe only a relatively slow decay in terminal tyrosine phosphorylation of Lck after inhibition of Csk3. On the other hand, as the authors also note, given the scale of the effects of Csk blockade in the absence of TCR engagement, it is difficult
nature immunology volume 15 number 2 february 2014
to envisage how conformational rearrangements, such as dissociation of the immunoreceptor tyrosine-based activation motifs of the invariant signaling protein CD3 from the membrane16, could be considered necessary elements of the TCR-triggering mechanism. In addition, ligand-dependent mechanical distortion17 is also largely out of the question. Mechanical feedback on T cell responses may instead be integrated through changes in the F-actin cortex. The elegant experiments of Weiss and colleagues have two important ‘take-home messages’. First, along with their previous studies of CskAS, the new data show that the TCR signalosome can be activated in primary thymocytes (and Jurkat cell lines) without engagement of the receptor simply by an acute increase in the activation of SFKs, which vividly illustrates the sensitivity of the TCR to changes in the balance of kinase and phosphatase activities. Second, cortical actin can form a functional barrier between active PLC-γ1 and its substrate. It is not so much the quantity of F-actin that matters, perhaps, but also its organization, as engagement of CD28 increases the accumulation of F-actin at the site of engagement while still overcoming the barrier to full signaling also breached by the disruption of F-actin. The work of Weiss and colleagues thus substantially expands the possibilities for how cortical F-actin could regulate signaling in cells of the T lineage. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. 1. Treanor, B. et al. Immunity 32, 187–199 (2010). 2. Clark, A.G., Dierkes, K. & Paluch, E.K. Biophys. J. 105, 570–580 (2013). 3. Tan, Y.X. et al. Nat. Immunol. 15, 186–194 (2014). 4. Zhang, R. et al. J. Clin. Invest. 123, 580–593 (2013). 5. Torgersen, K.M. et al. J. Biol. Chem. 276, 29313–29318 (2001). 6. Schoenborn, J.R., Tan, Y.X., Zhang, C., Shokat, K.M. & Weiss, A. Sci. Signal. 4, ra59 (2011). 7. Valitutti, S., Dessing, M., Aktories, K., Gallati, H. & Lanzavecchia, A. J. Exp. Med. 181, 577–584 (1995). 8. Brown, A.C. et al. PLoS Biol. 9, e1001152 (2011). 9. Kong, K.F. et al. Nat. Immunol. 12, 1105–1112 (2011). 10. Irvine, D.J., Purbhoo, M.A., Krogsgaard, M. & Davis, M.M. Nature 419, 845–849 (2002). 11. Aoki, K. et al. Mol. Cell 52, 529–540 (2013). 12. Kawabuchi, M. et al. Nature 404, 999–1003 (2000). 13. Brdicka, T. et al. J. Exp. Med. 191, 1591–1604 (2000). 14. Davidson, D., Bakinowski, M., Thomas, M.L., Horejsi, V. & Veillette, A. Mol. Cell. Biol. 23, 2017–2028 (2003). 15. Dobenecker, M.W., Schmedt, C., Okada, M. & Tarakhovsky, A. Mol. Cell. Biol. 25, 10533–10542 (2005). 16. Xu, C. et al. Cell 135, 702–713 (2008). 17. Ma, Z., Janmey, P.A. & Finkel, T.H. FASEB J. 22, 1002–1008 (2008).
137
TCR signaling: the barrier within Michael L Dustin & Simon J Davis
TCR Membrane
Lat
b
Inactive
Primed
Active
Lck SH3
CD45
Lck
Csk
CD45
c
No perturbation CskAS inhibition CskAS inhibition + actin disruption (or costimulation)
SH2 K
P
T cell
P
C
sk
Active
45
T cell
D
45
136
Glycocalyx
a
D
Michael L. Dustin is with the Kennedy Institute of Rheumatology and Nuffield Department of Orthopedics and Musculoskeletal Sciences, The University of Oxford, Oxford, UK. Simon J. Davis is with the Medical Research Council Human Immunology Unit and Radcliffe Department of Clinical Medicine, Weatherall Institute of Molecular Medicine, The University of Oxford, Oxford, UK. e-mail: [email protected]
tyrosine kinase CD45 in the case of Lck, which favors activation. Unlike the SFKs, Csk lacks a membrane-recruitment domain and thus needs to bind through its SH2 domain to membrane-anchored adaptors such as Cskbinding protein (CBP). Weiss and colleagues ask what would happen if Csk could be rapidly turned off. Published studies have used overexpression of dominantnegative forms of Csk5, but approaches that involve stable gene expression are subject to compensation and other caveats. To acutely
extension that contains a tyrosine residue that can be phosphorylated. The tertiary structure of these kinases allows two inhibitory intramolecular interactions: interaction of the SH3 domain with the polyproline motif, and interaction of the SH2 domain with the phosphorylated C-terminal tyrosine. The aptly named ‘cytosolic C-terminal Src kinase’ (Csk) phosphorylates that C-terminal tyrosine and ‘encourages’ intramolecular binding of the SH2 domain and inhibition of the kinase. The phosphate is removed by the transmembrane
C
uch attention has been focused on the barrier to T cell activation posed by the ‘forest’ of bulky glycoproteins that form the glycocalyx of the cell surface, but there are even deeper potential barriers to signaling on the other side of the membrane that have only recently started coming into focus1. The cell cortex comprises a crosslinked network of filamentous actin (F-actin), myosin and other proteins that lie directly underneath the plasma membrane and form a barrier typically 100 nm in thickness2 (Fig. 1a). The cortex offers resistance to externally applied forces and has a key role in cell-shape change. In this issue of Nature Immunology, Weiss and colleagues test a new tool for ‘kick-starting’ T cell activation and discover a new F-actindependent barrier to signaling via the T cell antigen receptor (TCR) in thymocytes3. That barrier is breached at least in part by engagement of the coreceptor CD28 and thus has the potential to be important for the generation of CD28-dependent T cell lineages such as Foxp3+ regulatory T cells4. Src-family kinases (SFKs) such as Lck undergo allosteric regulation driven by tyrosine phosphorylation (Fig. 1b). SFKs all share a unique amino (N)-terminal domain that is involved in membrane interactions, a Srchomology 3 (SH3) domain, an SH2 domain, a linker with a polyproline motif, a kinase domain (SH1) and a carboxy (C)-terminal
C
M
k Lc
npg
© 2014 Nature America, Inc. All rights reserved.
Acute inhibition of the regulatory kinase Csk reveals additional checkpoints for full activation of thymocytes via the T cell antigen receptor.
rk p-E -g1 LC p-P
F-actin
P P
PLC-g1
Figure 1 Effects of Csk inhibition and actin disruption on TCR signaling. (a) The actin cortex forms a physical barrier between the membrane and the rest of the cytoplasm. Only the F-actin part of the cortex is presented here. The proteins are drawn to scale; the cytoplasmic domains of TCR-CD3 and Lat, even at their theoretical maximal length (~3.5Å per residue), are occluded by the actin cortex. Scale bar, 10 nm. (b) Role of Csk in the regulation of Lck. The inactive, primed and active isoforms of Lck interconvert through the phosphorylation of Csk and autophosphorylation of Lck and through CD45-mediated dephosphorylation. (c) Signaling thresholds and the effects of global perturbation of receptors. Arrows indicate signaling cascades initiated by Lck-mediated phosphorylation of individual TCRs in the resting state; arrow length corresponds to the depth of the signaling cascade reached (so for a given condition the lengths vary, reflective of stochastic effects): basal signaling (gray); signaling initiated by inhibition of CskAS (purple); or signaling initiated by inhibition of CskAS and disruption of actin (or costimulation via CD38; red). Blue and red dashed lines represent signaling thresholds that correspond to the phosphorylation (p-) of PLC-g1 and Erk. In the unperturbed cell, the signaling cascades reach neither threshold. Inhibition of CskAS probably globally enhances Lck activity and TCR phosphorylation and thus increases the number of signaling cascades and the probability that at least one cascade breaches the threshold required for the phosphorylation of PLC-g1. The phosphorylation of Erk also requires disruption of actin, which may enhance the access of PLC-g1 to substrates in the membrane.
volume 15 number 2 february 2014 nature immunology
npg
© 2014 Nature America, Inc. All rights reserved.
n ews a n d views and specifically inhibit Csk, the Weiss laboratory teamed up with the Shokat laboratory to generate a bulky ATP analog–sensitive mutant of Csk (CskAS)6. In published work, they found that when a membrane-anchored form of CskAS was expressed in the Jurkat human T lymphocyte cell line, inhibition with the bulky analog resulted in the full activation of TCR signaling, including phosphorylation of the kinase Erk and an increase in cytoplasmic Ca2+ concentrations6. Now, Weiss and colleagues have introduced the gene encoding CskAS into mice via bacterial artificial chromosome transgenesis to gauge the effect of acute inhibition of Csk in primary cells3. When Csk is acutely inhibited in primary thymocytes, the early TCR activation cascade is triggered, including activating phosphorylation of Lck, the cytoplasmic kinase Zap70, phospholipase C-γ1 (PLC-γ1) and the PLC-γ1 adaptor Lat. These results indicate that the TCR-proximal elements of the tyrosine kinase cascade work as expected, with acute inhibition of Csk pushing the equilibrium toward greater activation of Lck, which drives assembly of the classical proximal signalosome without engagement of the TCR. Unexpectedly, however, whereas inhibition of Csk generates robust TCR-proximal signals, no activation of Erk or elevation in cytoplasmic Ca2+ ensues, which identifies a previously unsuspected signaling checkpoint comprising the cleavage of phosphatidylinositol-4,5-bisphosphate by PLC-γ1 (ref. 3). However, that checkpoint is overcome by disruption of F-actin with cytochalasin D or by engagement of CD28 in thymocytes with dendritic cells or beads coated with the CD28 ligand CD86, which results in increased recruitment of F-actin to the synapse. Mature T cells have a more complex response to manipulation of F-actin after inhibition of Csk, so the existing view that F-actin has a positive role in mature T cells still holds true7. Little is known about the actin cortex of lymphocytes. At the typical depth of 100 nm (and a mesh size of 15–30 nm), the cortex of T cells would be much deeper than the maximum theoretical reach of the TCR or Lat (Fig. 1a). That is why processes such as secretion need to generate holes in the actin cortex to allow vesicles to reach the plasma membrane8. Cortical actin is argued to be a diffusion barrier in signaling via the B cell antigen receptor, and the work of Weiss and colleagues now shows that cortical F-actin acts as a barrier that apparently constrains the access of otherwise active PLC-γ1 to phosphoinositide lipids3. Exactly how the barrier function is achieved is unclear, as there is at least some access of PLC-γ1 to the signalosome because it is phosphorylated after inhibition of Csk without
disruption of actin. Bringing CD28 into the equation as a remodeler of actin nevertheless has interesting functional implications, as a phosphorylated tyrosine residue in the cytoplasmic domain of CD28 may also directly modulate Lck activity through binding to Lck’s SH2 domain in competition with the terminal tyrosine phosphorylated by Csk9. During the thymic differentiation of regulatory T cells, therefore, CD28 may both overcome the actin barrier and contribute to Lck-activation mechanisms that are critical for additional downstream signaling. It is necessary, however, to bring just one note of caution to interpretation of these new data. This is because whereas the types of perturbation imposed by Weiss and colleagues3—that is, global disruption of actin and acute inhibition of Csk—will affect all the TCRs present simultaneously, the T cell is able to respond in a physiological setting to just a few TCR ligands10, which indicates that individual TCRs acting locally are capable of strong signaling. Perturbations that are imposed on very large numbers of TCRs simultaneously, whose individual effects might otherwise be relatively weak, will tend to increase the overall probability of signaling and elicit responses that might be similar in strength to those primed by ‘strong’ ligands acting through just a few TCRs, due to stochastic effects11 (Fig. 1c). Among the challenges that remain, it will be necessary to show that local remodeling of actin, perhaps mediated by costimulation, contributes substantially to local signaling responses to conventional ligands rather than contributing to other local processes. The classical feedback model for Csk is that active SFKs phosphorylate sites on adaptors at the membrane that then recruit Csk, which in turn inhibits the active SFKs by phos- phorylating their C-terminal tyrosine residues12. It has been shown, however, that the docking site for Csk in CBP is transiently dephosphorylated during triggering of the TCR5,13,14. Nevertheless, the role of CBP in controlling Csk localization seems to be somewhat redundant, as CBP-deficient mice have no alteration in T cell phenotype15. While Weiss and colleagues clearly show that inhibiting Csk is a potent means of triggering the TCR3, as the authors argue elsewhere, it is difficult to envisage how inhibition of Csk might be coopted in vivo in a way that initiates or promotes new signaling14. After all, Weiss and colleagues observe only a relatively slow decay in terminal tyrosine phosphorylation of Lck after inhibition of Csk3. On the other hand, as the authors also note, given the scale of the effects of Csk blockade in the absence of TCR engagement, it is difficult
nature immunology volume 15 number 2 february 2014
to envisage how conformational rearrangements, such as dissociation of the immunoreceptor tyrosine-based activation motifs of the invariant signaling protein CD3 from the membrane16, could be considered necessary elements of the TCR-triggering mechanism. In addition, ligand-dependent mechanical distortion17 is also largely out of the question. Mechanical feedback on T cell responses may instead be integrated through changes in the F-actin cortex. The elegant experiments of Weiss and colleagues have two important ‘take-home messages’. First, along with their previous studies of CskAS, the new data show that the TCR signalosome can be activated in primary thymocytes (and Jurkat cell lines) without engagement of the receptor simply by an acute increase in the activation of SFKs, which vividly illustrates the sensitivity of the TCR to changes in the balance of kinase and phosphatase activities. Second, cortical actin can form a functional barrier between active PLC-γ1 and its substrate. It is not so much the quantity of F-actin that matters, perhaps, but also its organization, as engagement of CD28 increases the accumulation of F-actin at the site of engagement while still overcoming the barrier to full signaling also breached by the disruption of F-actin. The work of Weiss and colleagues thus substantially expands the possibilities for how cortical F-actin could regulate signaling in cells of the T lineage. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. 1. Treanor, B. et al. Immunity 32, 187–199 (2010). 2. Clark, A.G., Dierkes, K. & Paluch, E.K. Biophys. J. 105, 570–580 (2013). 3. Tan, Y.X. et al. Nat. Immunol. 15, 186–194 (2014). 4. Zhang, R. et al. J. Clin. Invest. 123, 580–593 (2013). 5. Torgersen, K.M. et al. J. Biol. Chem. 276, 29313–29318 (2001). 6. Schoenborn, J.R., Tan, Y.X., Zhang, C., Shokat, K.M. & Weiss, A. Sci. Signal. 4, ra59 (2011). 7. Valitutti, S., Dessing, M., Aktories, K., Gallati, H. & Lanzavecchia, A. J. Exp. Med. 181, 577–584 (1995). 8. Brown, A.C. et al. PLoS Biol. 9, e1001152 (2011). 9. Kong, K.F. et al. Nat. Immunol. 12, 1105–1112 (2011). 10. Irvine, D.J., Purbhoo, M.A., Krogsgaard, M. & Davis, M.M. Nature 419, 845–849 (2002). 11. Aoki, K. et al. Mol. Cell 52, 529–540 (2013). 12. Kawabuchi, M. et al. Nature 404, 999–1003 (2000). 13. Brdicka, T. et al. J. Exp. Med. 191, 1591–1604 (2000). 14. Davidson, D., Bakinowski, M., Thomas, M.L., Horejsi, V. & Veillette, A. Mol. Cell. Biol. 23, 2017–2028 (2003). 15. Dobenecker, M.W., Schmedt, C., Okada, M. & Tarakhovsky, A. Mol. Cell. Biol. 25, 10533–10542 (2005). 16. Xu, C. et al. Cell 135, 702–713 (2008). 17. Ma, Z., Janmey, P.A. & Finkel, T.H. FASEB J. 22, 1002–1008 (2008).
137
