ACCELERATED PUBLICATION EXPERIMENTAL PROCEDURES. * This work was supported by National Institutes of Health Grants HL55426

Transcription

1 ACCELERATED PUBLICATION This paper is available online at THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 281, NO. 30, pp , July 28, by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A. Orai1 and STIM Reconstitute Store-operated Calcium Channel Function * Received for publication, May 19, 2006, and in revised form, June 7, 2006 Published, JBC Papers in Press, June 9, 2006, DOI /jbc.C Jonathan Soboloff 1,2, Maria A. Spassova 1, Xiang D. Tang, Thamara Hewavitharana, Wen Xu, and Donald L. Gill 3 From the Department of Biochemistry and Molecular Biology, University of Maryland School of Medicine, Baltimore, Maryland The two membrane proteins, STIM1 and Orai1, have each been shown to be essential for the activation of store-operated channels (SOC). Yet, how these proteins functionally interact is not known. Here, we reveal that STIM1 and Orai1 expressed together reconstitute functional SOCs. Expressed alone, Orai1 strongly reduces store-operated Ca 2 entry (SOCE) in human embryonic kidney 293 cells and the Ca 2 release-activated Ca 2 current (I CRAC ) in rat basophilic leukemia cells. However, expressed along with the store-sensing STIM1 protein, Orai1 causes a massive increase in SOCE, enhancing the rate of Ca 2 entry by up to 103-fold. This entry is entirely storedependent since the same coexpression causes no measurable store-independent Ca 2 entry. The entry is completely blocked by the SOC blocker, 2-aminoethoxydiphenylborate. Orai1 and STIM1 coexpression also caused a large gain in CRAC channel function in rat basophilic leukemia cells. The close STIM1 homologue, STIM2, inhibited SOCE when expressed alone but coexpressed with Orai1 caused substantial constitutive (storeindependent) Ca 2 entry. STIM proteins are known to mediate Ca 2 store-sensing and endoplasmic reticulum-plasma membrane coupling with no intrinsic channel properties. Our results revealing a powerful gain in SOC function dependent on the presence of both Orai1 and STIM1 strongly suggest that Orai1 contributes the PM channel component responsible for Ca 2 entry. The suppression of SOC function by Orai1 overexpression likely reflects a required stoichiometry between STIM1 and Orai1. Ca 2 signals control many cellular functions ranging from short term responses such as contraction and secretion to longer term regulation of cell growth and proliferation (1, 2). * This work was supported by National Institutes of Health Grants HL55426 and AI and by the Interdisciplinary Training Program in Muscle Biology, University of Maryland School of Medicine. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 These authors contributed equally to the work. 2 To whom correspondence may be addressed: Dept. of Biochemistry and Molecular Biology, University of Maryland School of Medicine, 108 North Greene St., Baltimore, MD Tel.: (office) or (laboratory); Fax: ; jsobo001@umaryland.edu. 3 To whom correspondence may be addressed: Dept. of Biochemistry and Molecular Biology, University of Maryland School of Medicine, 108 North Greene St., Baltimore, MD Tel.: (office) or (laboratory); Fax: ; dgill@umaryland.edu. Receptor-induced Ca 2 signals involve two closely coupled components, rapid, inositol 1,4,5-trisphosphate-mediated Ca 2 release from ER 4 stores, followed by Ca 2 entry through store-operated channels (SOCs) (1, 3 6). The activation of SOCs is key to mediating longer term cytosolic Ca 2 signals and replenishing intracellular stores (4 6). In many cell types, including hematopoietic cells, SOCs carry a highly Ca 2 -selective, non-voltage-gated, inwardly rectifying current, termed the Ca 2 release-activated Ca 2 current or I CRAC (3, 5, 6). Molecular characterization of SOCs and the activation process for store-operated Ca 2 entry (SOCE) have remained elusive (5, 6). High through-put RNA interference screens revealed stromalinteracting molecule (STIM1), is required for SOCE (7, 8) and conductance through CRAC channels (9). STIM1 is likely the sensor of Ca 2 within ER Ca 2 stores (8, 10), moving in response to store depletion into ER puncta close to the PM (8). Recently, SOCE and CRAC channels have also been shown to require the plasma membrane four-transmembrane spanning protein, Orai1 or CRACM1 (11, 12). Hence, a naturally occurring R91W mutation in Orai1 led to elimination of I CRAC (11, 12), as did Orai1 knockdown (11, 12). While expression of Orai1 wild-type (wt) restored CRAC to normal levels (11, 12), the exact role of Orai1 in CRAC activation was not established. Here we show that expression of STIM1 and Orai1 in combination results in an enormous gain in function of SOCE and CRAC channel activity, indicating that the PM Orai1 protein is likely the channel entity mediating SOC function. EXPERIMENTAL PROCEDURES DNA Expression Constructs, Mutagenesis, and Transfection wt human STIM1 and STIM2 were subcloned into piresneo (Clontech, Palo Alto, CA) as described previously (13). The human STIM1 E87A mutation was introduced using the QuikChange site-directed mutagenesis kit (Stratagene) and confirmed by sequencing. Orai1, expressed in pcmv6-xl4, was obtained from Origene (Rockville, MD). Constructs were introduced by electroporation using the Gene Pulser II Electroporation system (Bio- Rad) at 350 V, 960 microfarads, and infinite resistance, followed by 48 h in culture. Development of Stable Cell Lines Human embryonic kidney 293 (HEK293) were maintained as described previously (14). HEK293 stable cell lines were generated by electroporation of the above described wt human STIM1 or STIM2 pires constructs. After selection with appropriate antibiotics, cells were cloned and selected based upon expression of the genes and the amount of SOCE. Ca 2 Measurements Cells grown on coverslips were placed in cation-safe medium free of sulfate and phosphate anions (NaCl (107 mm), KCl (7.2 mm), MgCl 2 (1.2 mm), glucose ( The abbreviations used are: ER, endoplasmic reticulum; SOC, store-operated channel; SOCE, store-operated Ca 2 entry; I CRAC,Ca 2 release-activated Ca 2 current; fura-2/am, fura-2 acetoxymethylester; BAPTA, 1,2-bis(2- aminophenoxy)ethane-n,n,n,n -tetraacetic acid; F, farad; PM, plasma membrane; wt, wild-type; HEK, human embryonic kidney; RBL, rat basophilic leukemia; 2-APB, 2-aminoethoxydiphenyl borate. JULY 28, 2006 VOLUME 281 NUMBER 30 JOURNAL OF BIOLOGICAL CHEMISTRY 20661

2 mm), HEPES-NaOH (20 mm), ph 7.2) and loaded with fura-2/ acetoxymethylester (2 M) for 30 min at 20 C, as described elsewhere (15, 16). Cells were washed, and dye was allowed to de-esterify for a minimum of 30 min at 20 C. Approximately 95% of the dye was confined to the cytoplasm as determined by the signal remaining after saponin permeabilization (17). Ca 2 measurements were made using an InCyt dual-wavelength fluorescence imaging system (Intracellular Imaging Inc.). The concentration of intracellular free Ca 2 was calculated according to the following formula of Grynkiewcz et al. (18), [Ca 2 ] i K d F min /F max ) (R R min )/(R max R (Eq. 1) where R is the ratio of the fluorescence intensities measured at 340 and 380 nm during the experiments, and F is the fluorescence intensity measured at 505 nm. R min, R max, F min, and F max were determined from in situ calibration of unlysed cells using 40 M ionomycin in the absence (R min and F min ;10mM EGTA) and presence of (R max and F max )ofca 2. K d (135 nm) isthe dissociation constant for fura-2 at room temperature. All measurements shown are averages of cells and representative of a minimum of three experiments. Electrophysiology Studies were performed in rat basophilic leukemia (RBL) and HEK293 cells cotransfected with yellow fluorescent protein to select transfected cells. We used conventional whole cell recordings as described (9). Immediately after establishment of the whole-cell configuration, voltage ramps of 50-ms duration spanning the voltage range of 100 to 100 mv were delivered from a holding potential of 0 mv at a rate of 0.5 Hz. The intracellular solution contained (mm): 145 CsGlu, 10 HEPES, 10 BAPTA, 8 Na,5Mg 2, 2 Mg-ATP (total 8 mm Mg 2 ), ph mm Mg 2 and ATP were used to inhibit TRPM7 (19). The extracellular solutions contained (mm): 145 NaCl, 10 CaCl 2, 10 CsCl, 2 MgCl 2, 2.8 KCl, 10 HEPES, 10 glucose, ph 7.4. We applied 10 mv junction potential compensation. Materials Orai1 was from Origene (Rockville, MD). Thapsigargin was from EMD Biosciences (San Diego, CA). G418 was from Sigma. Fura-2/acetoxymethylester was from Molecular Probes (Eugene, OR). RESULTS AND DISCUSSION We examined the function of Orai1 by expressing it in HEK293 cells lines stably expressing STIM1 or the control vector-expressing HEK293 cells. Ca 2 entry in fura-2-loaded cells was measured after thapsigargin-induced store depletion in Ca 2 -free medium. Despite its necessity in SOCE (11, 12), Orai1 expressed in vector control cells strongly suppressed SOCE (Fig. 1A; see also Fig. 3A). In STIM1-expressing cells there was a modest enhancement of SOCE, as shown previously (14). Remarkably, Orai1 coexpressed in STIM1-expressing cells resulted in a massive and rapid increase in SOCE (Fig. 1A). No significant change in store content was observed under any of these expression conditions. To reveal the extent of Ca 2 entry, the fura-2 ratiometric data is expressed as free Ca 2 concentrations. Since the ratios reached in STIM1-Orai1-expressing cells (ratios approaching 10) were beyond the linear range of fura-2, the highest Ca 2 levels are only approximate. In vector-control FIGURE 1. Coexpression of STIM1 and Orai1 reconstitutes store-operated Ca 2 entry. Ca 2 concentration was monitored in HEK293 cells stably expressing wt human STIM1 or empty pires vector after transient transfection with Orai1 or empty vector. A, HEK293 cells were pretreated with thapsigargin (2 M) in the absence of extracellular Ca 2 (10 min) to deplete Ca 2 stores. 1 mm Ca 2 was added at the arrow to assess SOCE. B,1mM Ca 2 was added to store-replete cells that had briefly ( 5 min) been maintained in nominally Ca 2 -free medium. C, HEK293 cells expressing both STIM1 and Orai1 were pretreated with thapsigargin (2 M) in the absence of extracellular Ca 2 (10 min) to deplete Ca 2 stores. 1 mm Ca 2 was added at the arrow to assess SOCE. 5 M 2-APB followed by 50 M 2-APB were added as indicated by their respective arrows. cells, the maximal increase in cytosolic Ca 2 resulting from SOCE was 150 nm. This value decreased to nm in Orai1-expressing vector control cells. However, using STIM1- expressing cells, the expression of Orai1 resulted in a staggering fold increase in the maximal level of cytosolic Ca 2 mediated by store-emptying. In this experiment, Ca 2 rose to 3400 nm. Most significantly, the initial rate of Ca 2 entry into Orai1/STIM1-expressing cells was enormously increased in this experiment by 103-fold compared with vector-expressing cells. All of the increased Ca 2 entry due to Orai1/STIM1 expression was dependent on store depletion. Hence, no change in resting Ca 2 levels could be detected with any combination of expression of Orai1 and STIM1 (Fig. 1B). Moreover, the huge increase in SOCE was rapidly blocked by application of 50 M 2-aminoethoxydiphenyl borate (2-APB), typical of known SOCE and CRAC channel function (20, 21). Based on the dramatic gain in Ca 2 entry observed in the presence of coexpressed Orai1 and STIM1, all of which is storedependent, we considered it likely that the Orai1 protein was increasing the density of store-operated channels. To further assess the possible channel role, we examined the effects of Orai1 and STIM1 expression on the density of CRAC channels measured in the RBL mast cell-derived line. The time course of development of endogenous I CRAC in response to BAPTA-in JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 281 NUMBER 30 JULY 28, 2006

3 FIGURE 2. Coexpression of STIM1 and Orai1 reconstitutes I CRAC in RBL cells. A, representative time course of the activation of CRAC channel activity recorded at 80 mv with overexpression of STIM1 (blue), Orai1 (green), or STIM1 together with Orai1 (red). CRAC channel activation in control cell is shown for comparison (black). B, representative I-V curves at the time of maximal activation of the CRAC channel activity. C, average maximal current at 100 mv in RBL cells overexpressing empty vector (control), STIM1, Orai1, or STIM1 Orai1. FIGURE 3. STIM2 and Orai1 coexpression results in constitutive Ca 2 entry. Ca 2 concentration was monitored in HEK293 cells stably expressing wt human STIM2 or empty pires vector after transient transfection with Orai1 or empty vector. A, HEK293 cells were pretreated with thapsigargin (2 M)in the absence of extracellular Ca 2 (10 min) to deplete Ca 2 stores. 1 mm Ca 2 was added at the arrow to assess SOCE. B, the average increase in Ca 2 entry observed after the addition of 1 mm Ca 2 to thapsigargin-pretreated cells overexpressing empty vector, STIM2, Orai1, or STIM2 Orai1. C,1mM Ca 2 was added to store-replete cells that had briefly ( 5 min) been maintained in nominally Ca 2 -free medium. D, the average increase in Ca 2 entry observed after the addition of 1 mm Ca 2 to untreated cells overexpressing both STIM2 and Orai1 as compared with STIM2-expressing control cells. N represents the number of separate experiments, each including individual cells. * indicates p 0.05; ** indicates p 0.01; *** indicates p duced store depletion is shown in Fig. 2A. Compared with control cells, the expression of Orai1 was strongly inhibitory, in agreement with the results for SOCE in HEK293 cells shown in ACCELERATED PUBLICATION: Orai, STIM, and SOC Channels Fig. 1. Expression of STIM1 caused a modest enhancement of I CRAC, whereas coexpression of both Orai1 together with STIM1 resulted in a substantial increase in CRAC channel activity. Quantitation of maximal current from multiple experiments (Fig. 2B) revealed that the current density in vector-transfected cells was pa/pf (n 3). This value decreased to pa/pf (n 3) with expression of Orai1 alone. STIM1 expressed alone approximately doubled current density to pa/pf (n 3), whereas coexpression of Orai1 with STIM1 increased current some 9-fold to a total of pa/pf (n 5). As shown in Fig. 2C, the I/V profile in the absence or presence of Orai1 and STIM1 reveals inward rectification and a reversal potential of approximately 50 mv, typical of the highly Ca 2 -selective CRAC channel (3, 6). Hence, the results provide compelling evidence that channel density is increased by Orai1/STIM1 expression. The larger relative fold increase in SOCE in HEK293 cells likely reflects both lower basal CRAC levels relative to RBL cells (12) and greater transfection efficiency. STIM1 and STIM2 are closely related in structure, but STIM2 has an opposing inhibitory action on SOCE (14). We utilized stable STIM2-expressing HEK293 cells to examine the functional interaction between Orai1 and STIM2 (Fig. 3). The inhibitory actions of either STIM2 (14) or Orai1 expression on SOCE are clear. Unexpectedly, combined expression resulted in elevated SOCE, especially at later time points (Fig. 3, A and B). In store-replete cells, a substantial increase in constitutive Ca 2 entry was observed with Orai1/STIM2 expression but not with STIM2 expression alone (Fig. 3, C and D). STIM1 and STIM2 are functionally distinct. STIM1 accumulates in near-pm puncta of the ER upon store depletion (8, 14) where it is believed to activate Ca 2 entry channels. In contrast, STIM2 expressed alone does not change its ER distribution in response to store depletion (14). Puncta are considered preexisting PMjuxtaposed regions of the contiguous ER (8). We suspect that a manyfold increase in STIM2 expressed throughout the ER might lead to an increase in STIM2 in puncta, independently of altered luminal Ca 2. Thus, at high levels of STIM2 expression, while it is normally an inhibitor of SOCs, STIM2 may mimic the action of STIM1 and, through interaction with overexpressed Orai1, result in significant constitutive SOC activation. As seen in Fig. 3C, the STIM2-mediated constitutive Ca 2 entry was enhanced by 50 M 2-APB. This enhancing effect was consistent with our previous observation in store-depleted STIM2-transfected RBL cells (22). 5 Dramatically, overexpression of Orai1 in STIM2-expressing HEK293 cells increased the 2-APB-induced Ca 2 entry to levels approaching Orai1/STIM1-expressing store- 5 M. A. Spassova and D. L. Gill, unpublished observations. JULY 28, 2006 VOLUME 281 NUMBER 30 JOURNAL OF BIOLOGICAL CHEMISTRY 20663

4 FIGURE 4. Model to depict functional coupling between STIM1, the ER luminal Ca 2 sensor, and Orai1, the PM Ca 2 entry channel. The STIM1 protein contains an intraluminal Ca 2 -binding EF-hand domain and sterile- -motif (SAM), a single transmembrane segment, two cytoplasmic coiled-coil domains (CC), and a proline-rich region (P). The Orai1 protein has four predicted transmembrane helices, a proline-rich segment near its cytoplasmic N terminus. The arginine (R) mutated in hereditary immunodeficiency (11) likely fixes the first transmembrane sequence longitudinally within the bilayer by interaction with negative lipid phosphate charges. The glutamate (E) may perform a function in transmitting Ca 2 across the membrane. depleted cells. Electrophysiological measurements in Orai1/ STIM2-transfected HEK293 cells revealed constitutive but otherwise typical CRAC-like current (with I/V relationship similar to Fig. 2C)of pa/pf (n 4), which increased to pa/pf (n 4) after 50 M 2-APB. In Orai1/STIM1-expressing HEK293 cells, we measured thapsigargin-induced CRAC current of pa/pf (n 3). These remarkable CRAC current values are enormously larger than basal values of 1 pa/pf in HEK293 cells (12). The action of 2-APB on SOCE is biphasic, lower 2-APB ( 10 M) enhances, while higher levels of 2-APB ( M) are strongly inhibitory (20, 21). Likely, this reflects two distinct sites of action of high and low affinity, respectively (21). As suggested previously (9), the inhibitory (low affinity) target of 2-APB on physiological SOC activation may be the STIM1 protein itself. Hence, when STIM2 rather than STIM1 couples to the channel, 2-APB does not inhibit. Instead, the high-affinity 2-APB site dominates, leading to enhanced Ca 2 entry. Notwithstanding the intricacies of the mechanism of action of 2-APB, the high level of Ca 2 entry observed in the presence of overexpressed Orai1 is the predominant effect, consistent again with Orai1 providing the channel component. STIM1 can be mutated to prevent its Ca 2 sensing function resulting in constitutive Ca 2 entry and CRAC channel function (7 9, 14), thereby circumventing the entire store depletion process. We examined whether Orai1 could be directly activated in this manner. Using HEK293 cells stably expressing the STIM1-D76A- E87A EF-hand mutant, we found that Orai1 expression again dramatically enhanced Ca 2 entry (data not shown). Similarly, in RBL cells expressing the STIM1-E87A EF-hand mutant together with Orai1 resulted in an 15-fold increase in CRAC channel function (data not shown). Thus, the store-independent STIM1 mutant, together with Orai1, reconstitutes both functional Ca 2 entry and CRAC channel activity. One question is why overexpression of Orai1 results in substantially lower SOCE in HEK293 cells (Figs. 1A and 3, A and B) or decreased CRAC channel activity in RBL cells (Fig. 2). We would explain this by assuming the coupling stoichiometry between channel and sensor is not unity, as predicted by Putney (23). Thus, assuming more than one sensor must interact with each channel, the predominance of channels reduces the probability of successful coupling. While we observe a powerful inhibitory action of overexpressed Orai1 on SOCE and I CRAC, the recent report of Vig et al. (12) showed little effect of overexpressed Orai1 on CRAC channel activity in HEK293 cells. We would suggest that the extremely low level of endogenous CRAC activity in these cells simply precluded an observable decrease. Based on our results and those previously showing the requirement for STIM1 and Orai1 (7 12,14), we can now suggest that these two proteins are both necessary and sufficient to mediate the process of store-operated channel function. After submission of this manuscript, a paper from Peinelt et al. (24) described a similar synergism between Orai1 and STIM1. The scheme shown in Fig. 4, likely portrays the coupled function of these two proteins. The ER and PM location of STIM1 and Orai1, respectively, are consistent with their roles as ER Ca 2 sensor and PM Ca 2 channel, respectively. It should also be considered that STIM1 has a PM location (14), and it is therefore possible that its functional coupling to Orai1 may be within the PM. Acknowledgment We thank Marie Dziadek for invaluable scientific input. REFERENCES 1. Berridge, M. J., Lipp, P., and Bootman, M. D. (2000) Nat. Rev. Mol. Cell. Biol. 1, Berridge, M. J., Bootman, M. D., and Roderick, H. L. (2003) Nat. Rev. Mol. Cell. Biol. 4, Parekh, A. B., and Penner, R. (1997) Physiol. Rev. 77, Putney, J. W., Jr., Broad, L. M., Braun, F. J., Lievremont, J. P., and Bird, G. S. (2001) J. Cell Sci. 114, Venkatachalam, K., van Rossum, D. B., Patterson, R. L., Ma, H. T., and Gill, D. L. (2002) Nat. Cell Biol. 4, E263 E Parekh, A. B., and Putney, J. W., Jr. (2005) Physiol. Rev. 85, Roos, J., DiGregorio, P. J., Yeromin, A. V., Ohlsen, K., Lioudyno, M., Zhang, S., Safrina, O., Kozak, J. A., Wagner, S. L., Cahalan, M. D., Velicelebi, G., and Stauderman, K. A. (2005) J. Cell Biol. 169, Liou, J., Kim, M. L., Heo, W. D., Jones, J. T., Myers, J. W., Ferrell, J. E., and Meyer, T. (2005) Curr. Biol. 15, Spassova, M. A., Soboloff, J., He, L. P., Xu, W., Dziadek, M. A., and Gill, D. L. (2006) Proc. Natl. Acad. Sci. U. S. A. 103, Zhang, S. L., Yu, Y., Roos, J., Kozak, J. A., Deerinck, T. J., Ellisman, M. H., Stauderman, K. A., and Cahalan, M. D. (2005) Nature 437, Feske, S., Gwack, Y., Prakriya, M., Srikanth, S., Puppel, S. H., Tanasa, B., Hogan, P. G., Lewis, R. S., Daly, M., and Rao, A. (2006) Nature 441, Vig, M., Peinelt, C., Beck, A., Koomoa, D. L., Rabah, D., Koblan-Huberson, M., Kraft, S., Turner, H., Fleig, A., Penner, R., and Kinet, J. P. (2006) Science 312, Williams, R. T., Manji, S. S., Parker, N. J., Hancock, M. S., Van Stekelen JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 281 NUMBER 30 JULY 28, 2006

5 burg, L., Eid, J. P., Senior, P. V., Kazenwadel, J. S., Shandala, T., Saint, R., Smith, P. J., and Dziadek, M. A. (2001) Biochem. J. 357, Soboloff, J., Spassova, M. A., He, L.-H., Xu, W., Johnstone, L. J., Dziadek, M. A., and Gill, D. L. (2006) Curr. Biol., in press 15. Ma, H.-T., Patterson, R. L., van Rossum, D. B., Birnbaumer, L., Mikoshiba, K., and Gill, D. L. (2000) Science 287, Patterson, R. L., van Rossum, D. B., Ford, D. L., Hurt, K. J., Bae, S. S., Suh, P. G., Kurosaki, T., Snyder, S. H., and Gill, D. L. (2002) Cell 111, Venkatachalam, K., Ma, H. T., Ford, D. L., and Gill, D. L. (2001) J. Biol. Chem. 276, Bian, J., Ghosh, T. K., Wang, J. C., and Gill, D. L. (1991) J. Biol. Chem. 266, Prakriya, M., and Lewis, R. S. (2002) J. Gen. Physiol. 119, Ma, H. T., Venkatachalam, K., Parys, J. B., and Gill, D. L. (2002) J. Biol. Chem. 277, Prakriya, M., and Lewis, R. S. (2001) J. Physiol. (Lond.) 536, Spassova, M. A., Soboloff, J., Xu, W., Dziadek, M., and Gill, D. L. (2006) Biophys. J. 90, (abstr.) 23. Putney, J. W., Jr. (2004) Trends Cell Biol. 14, Peinelt, C., Vig, M., Koomoa, D. L., Beck, A., Nadler, M. J., Koblan-Huberson, M., Lis, A., Fleig, A., Penner, R., and Kinet, J. P. (2006) Nat. Cell Biol., in press JULY 28, 2006 VOLUME 281 NUMBER 30 JOURNAL OF BIOLOGICAL CHEMISTRY 20665