The erlin2 T65I mutation inhibits erlin1/2 complex–mediated inositol 1,4,5-trisphosphate receptor ubiquitination and phosphatidylinositol 3-phosphate binding

The erlin1/2 complex is a ∼2-MDa endoplasmic reticulum membrane–located ensemble of the ∼40-kDa type II membrane proteins erlin1 and erlin2. The best defined function of this complex is to mediate the ubiquitination of activated inositol 1,4,5-trisphosphate receptors (IP3Rs) and their subsequent degradation. However, it remains unclear how mutations of the erlin1/2 complex affect its cellular function and cause cellular dysfunction and diseases such as hereditary spastic paraplegia. Here, we used gene editing to ablate erlin1 or erlin2 expression to better define their individual roles in the cell and examined the functional effects of a spastic paraplegia–linked mutation to erlin2 (threonine to isoleucine at position 65; T65I). Our results revealed that erlin2 is the dominant player in mediating the interaction between the erlin1/2 complex and IP3Rs and that the T65I mutation dramatically inhibits this interaction and the ability of the erlin1/2 complex to promote IP3R ubiquitination and degradation. Remarkably, we also discovered that the erlin1/2 complex specifically binds to phosphatidylinositol 3-phosphate, that erlin2 binds this phospholipid much more strongly than does erlin1, that the binding is inhibited by T65I mutation of erlin2, and that multiple determinants within the erlin2 polypeptide comprise the phosphatidylinositol 3-phosphate–binding site. Overall, these results indicate that erlin2 is the primary mediator of the cellular roles of the erlin1/2 complex and that disease-linked mutations of erlin2 can affect both IP3R processing and lipid binding.

Upon stimulation of IP 3 /IP 3 R-dependent signaling, a fraction of IP 3 Rs are polyubiquitinated and thus targeted for proteasomal degradation (7) via the ubiquitin-proteasome pathway (UPP) (8). More specifically, degradation is mediated by the ER-associated degradation (ERAD) pathway, the facet of the UPP responsible for selective degradation of aberrant ER proteins (9,10). This results in a rapid reduction in the steady-state level of IP 3 Rs (t1 ⁄ 2 ϳ15-60 min), termed "IP 3 R down-regulation" (4,7), and may also be the route by which IP 3 Rs are turned over under basal conditions (7,11). Our studies on the mechanism of IP 3 R processing by the UPP have revealed that a complex formed from an assemblage of the ER membrane proteins erlin1 and erlin2 mediates IP 3 R polyubiquitination and degradation by binding to activated IP 3 Rs and recruiting the ubiquitin ligase RNF170 (7,(11)(12)(13)(14).
Erlin1 and -2 are ϳ40-kDa type II ER membrane proteins with an N-terminal transmembrane domain and the bulk of the polypeptide in the ER lumen (12,13,15). They oligomerize into a toroidal, ϳ2-MDa complex to which RNF170 is constitutively bound (7,13,14). By far the best defined function of erlins to date is to mediate IP 3 R processing (7), but other possible roles have been proposed, including binding cholesterol (16), regulation of lipid metabolism (16,17), and regulation of proteins that mediate SV40 exit from the ER (18). Interestingly, mutations to erlin1 and, more commonly, erlin2 have been linked to rare neurodegenerative diseases, e.g. hereditary spastic paraplegia (7, 19 -27). As yet, however, it has not been possible to define how the mutations affect the cell biological function of erlin2 or the erlin1/2 complex and cause cell dysfunction (7). This work was supported by National Institutes of Health Grants DK107944 and GM121621. The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. 1

cro ARTICLE
Here, we used gene editing to delete either erlin1 or -2 to better define the role of each protein in terms of IP 3 R processing. We found that erlin2 is the dominant partner within the erlin1/2 complex and that a novel hereditary spastic paraplegia-linked mutation to erlin2 (threonine to isoleucine at position 65; T65I) 4 dramatically inhibits the IP 3 R-processing function of the complex. Remarkably, we also show that the erlin1/2 complex binds specifically to phosphatidylinositol 3-phosphate (PI(3)P) in a manner inhibited by the T65I mutation to erlin2.
To assess the assembly status of erlin1 and erlin2, lysates from control, E1KO, and E2KO cells were subjected to nondenaturing PAGE (Fig. 1B). This showed that both erlin1 alone (lane 3) and erlin2 alone (lane 5) were able to assemble into high-molecular-weight complexes of size similar to that of the native erlin1/2 complex (lanes 1 and 4). p97, which associates into an ϳ600-kDa hexamer (32), served as a loading control. Furthermore, to assess the association status of RNF170, anti-erlin1 and anti-erlin2 immunoprecipitates were examined (Fig.  1C). Interestingly, although RNF170 interacted strongly with immunoprecipitated native erlin1/2 complex (lanes 1 and 2), it bound only weakly with immunoprecipitated erlin2 or erlin1 alone (lanes 4 and 5), suggesting that only the native complex provides the ideal module for RNF170 binding.
Finally, the relative expression of erlin1 and erlin2 and the composition of the erlin1/2 complex were assessed with a newly developed erlin pan antibody, raised against a completely conserved region of erlin1 and -2. This antibody was not entirely specific and strongly cross-reacted with several erlin-unrelated proteins in cell lysates. 5 This difficulty was ameliorated by analyzing membrane preparations. The relative intensity of the erlin pan -immunoreactive bands at 43 (erlin2) and 41 kDa (erlin1) revealed that the ratio of erlin2 to erlin1 in ␣T3 cell membrane extracts was ϳ2:1 (Fig. 1D, lane 1), and about the same ratio was found in erlin1/2 complex immunopurified with either anti-erlin1 or anti-erlin2 (Fig. 1D, lanes 4 and 5) or complex associated with activated IP 3 R1 (lane 7), indicating that all cellular erlin1 and -2 exist in homogeneous complexes with an ϳ2:1 ratio. A similar ratio was obtained by mass spectral analysis of erlin1/2 complex immunopurified as in Fig. 1D, lane 5; the number of unique erlin2 and erlin1 peptides identified (an index of relative abundance) was 19 and 13, respectively.

Erlin1/2 complex-mediated ubiquitination and PI(3)P binding
and recruiting RNF170. 6 However, in these E1KO cells, IP 3 R1 ubiquitination and the association of erlin2 and RNF170 were substantially inhibited relative to control cells (Fig. 2, A, lanes 1-10, and B; ϳ67 Ϯ 2% (n ϭ 4) inhibition of peak ubiquitination), indicating that the complex formed by erlin2 and RNF170 functions relatively inefficiently. In contrast, analysis of immunoprecipitates from E2KO cells showed that these cells did not retain the capacity to substantially ubiquitinate activated IP 3 R1, and, importantly, neither erlin1 nor RNF170 coimmunoprecipitated ( Fig. 2A, lanes [11][12][13][14][15], suggesting that erlin1 alone cannot bind to activated IP 3 R1; IP 3 R1 ubiquitination was inhibited by ϳ94 Ϯ 2% (n ϭ 4) (Fig. 2B). Measurement of IP 3 R1 down-regulation, the decline in steady-state IP 3 R1 levels that is the culmination of IP 3 R1 ubiquitination and proteasomal degradation (7), revealed that the response was lost in E2KO cells, whereas it was considerably inhibited in E1KO cells (Fig. 2, C and D). This correlated with the extent of IP 3 R1 ubiquitination in E2KO and E1KO cells (Fig.  2, A and B). Finally, to check that the reductions in ubiquitination and down-regulation seen after erlin1 or erlin2 deletion are not due to impairment of the IP 3 /IP 3 R signaling pathway, we measured Ca 2ϩ mobilization in response to GnRH (Fig. 2E).
This revealed that changes in cytosolic free [Ca 2ϩ ] ([Ca 2ϩ ] c ) were unaffected by erlin1 or erlin2 KO, indicating that the IP 3 / IP 3 R signaling pathway is not impaired. Taken together, the data in Figs. 1 and 2 indicate that although both erlin1 and erlin2 alone can assemble into high-molecular-weight complexes and bind RNF170, only erlin2 contains the motif for interaction with activated IP 3 R1, thereby allowing for IP 3 R1 ubiquitination.

The T65I mutation to erlin2 inhibits erlin1/2 complex binding to activated IP 3 Rs and IP 3 R ERAD
The hereditary spastic paraplegia-linked T65I mutation is in a region highly conserved between erlin2 and erlin1 (Fig. 3A). To explore how the T65I mutation affects erlin2 function, WT human erlin2HA (E2HA WT ) and T65I erlin2HA (E2HA T65I ) were stably expressed in E2KO ␣T3 cells (Fig. 3B). Two lines of each were examined (Fig. 3, B-I). E2HA T65I migrated slightly more rapidly that E2HA WT , was expressed at approximately the same levels as E2HA WT (indicating that the mutation does not affect stability), and did not alter the expression level of other relevant proteins (Fig. 3B); notably, however, only E2HA WT restored IP 3 R1 expression to control levels (compare lanes 3 and 5 with lane 1). Further characterizations showed that both E2HA WT and E2HA T65I were able to assemble into high-molecular-weight species similar to endogenous erlin1/2 complex (Fig. 3C) and, via coimmunoprecipitation, interact normally with endogenous erlin1 and RNF170 (Fig. 3D). Taken together, 6 It is intriguing that the amount of RNF170 that coimmunoprecipitates with activated IP 3 R1 via erlin2 in E1KO cells ( Fig. 2A, lanes 6 -10) is more than would be expected based on the low amount of RNF170 that coimmunoprecipitates when erlin2 is purified from E1KO cells (Fig. 1C, lane 4). This suggests that in stimulated E1KO cells ( Fig. 2A, lanes 6 -10) activated IP 3 R1 associates selectively with RNF170-bound erlin2 complexes. ] c in control, E1KO, and E2KO ␣T3 cells exposed to 100 nM GnRH (n Ն 9). Error bars represent S.E.

Erlin1/2 complex-mediated ubiquitination and PI(3)P binding
these data indicate that both E2HA WT and E2HA T65I bind to, and assemble with, their endogenous partners normally. Analysis of GnRH-induced IP 3 R1 processing showed that introduction of E2HA WT into E2KO cells restored IP 3 R1 ubiquitination, whereas, remarkably, E2HA T65I did not (Fig. 3, E and F). Further analysis of the immunoprecipitates showed that erlin1 and RNF170 coimmunoprecipitated with activated IP 3 R1 when E2HA WT was present (Fig. 3E, lanes 1-3 and 7-9), indicating that E2HA WT allows for the formation of a functional erlin1/2 complex-RNF170 module. In contrast, neither E2HA T65I , erlin1, nor RNF170 associated with activated IP 3 R1 when E2HA T65I was present (lanes 4 -6 and 10 -12), indicating that E2HA T65I is incapable of forming a functional erlin1/2 complex-RNF170 module. Consistent with these results, E2HA WT , but not E2HA T65I , was able to mediate IP 3 R1 downregulation in response to GnRH (Fig. 3, G and H). The lack of GnRH-induced IP 3 R1 ubiquitination and down-regulation in E2HA T65I -expressing cells was not due to an impairment of IP 3 R1 function or IP 3 -dependent signaling as GnRH-induced Ca 2ϩ mobilization was essentially identical in cells expressing . Amino acid identity is indicated by asterisks; colons and periods indicate strongly and weakly conservative differences, respectively. B, lysates from control and E2KO ␣T3 cells and E2KO cell lines stably expressing E2HA WT or E2HA T65I (two of each; lanes 3-6) were probed in immunoblots for ubiquitin, IP 3 R1, p97, erlin1, erlin2, HA, and RNF170. C, nondenaturing PAGE of cell lysates probed with anti-erlin1, anti-erlin2, or anti-p97 as a loading control. D, anti-erlin2 or anti-HA immunoprecipitates (IP) from cell lines stably expressing E2HA WT or E2HA T65I probed for the proteins indicated. E and F, cells were treated with 100 nM GnRH for 0, 5, or 10 min; anti-IP 3 R1 immunoprecipitates were probed in immunoblots for the proteins indicated; and ubiquitin immunoreactivity was quantitated and graphed (n ϭ 2). G and H, cells were treated with 100 nM GnRH for 0, 30, or 60 min; lysates were probed for the proteins indicated; and IP 3 R1 immunoreactivity was quantitated and graphed (n ϭ 7). I, [Ca 2ϩ ] c in cell lines stably expressing E2HA WT or E2HA T65I exposed to 100 nM GnRH (n ϭ 11). Error bars represent S.E.

Erlin1/2 complex-mediated ubiquitination and PI(3)P binding
E2HA WT and E2HA T65I (Fig. 3I). Thus, the T65I mutation to erlin2 appears to greatly impair the ability of the erlin1/2 complex to associate with activated IP 3 Rs and mediate their ERAD.

The erlin1/2 complex specifically binds PI(3)P, and the T65I mutation disrupts this binding
The erlin1/2 complex has been suggested to exist in lipid raft-like, or detergent-resistant, membrane domains within the ER and thus has been suggested to have some lipid-binding capacity, such as binding cholesterol (15,16). Thus, endogenous erlin1/2 complex was immunopurified from ␣T3 cells with anti-erlin1, and its binding to a panel of immobilized lipids (33,34) was examined by probing for erlin2 (Fig. 4). Surprisingly, no specific binding to cholesterol was discovered, but rather binding to phosphatidylinositol 4-phosphate (PI(4)P) was detected (Fig. 4A). Further analysis showed that erlin1/2 complex interacted selectively with monophosphorylated phosphoinositides with the strongest binding to PI(3)P, ϳ5-fold greater than that to PI(4)P and phosphatidylinositol 5-phosphate (PI(5)P) (Fig. 4, B and C). This binding appears to be direct as immunopurified erlin1/2 complex is largely free of contaminants (13) with the exception of RNF170 (14), and A and B, erlin1/2 complex was immunopurified from control ␣T3 cells, and binding to lipid arrays was assessed by probing with anti-erlin2. TG, triglyceride; DAG, diacylglycerol; PA, phosphatidic acid; PS, phosphatidylserine; PE, phosphatidylethanolamine; PC, phosphatidylcholine; PI, phosphatidylinositol, PI(x)Py, various PI phosphates; LPA, lysophosphatidic acid; LPC, lysophosphocholine; S1P, sphingosine 1-phosphate. C, erlin1/2 complex was immunopurified from control and RNF170KO ␣T3 cells (11), and binding of equal amounts of complex to lipid arrays was assessed by probing with anti-erlin2. In the repeat experiment with similar results, RNF170 was depleted using siRNA (14). D, mouse E2FLAGHA WT and E1FLAGHA WT were expressed transiently in ␣T3 E1/E2KO cells and immunopurified with anti-HA, and binding of equal amounts to lipid arrays was assessed with anti-FLAG. The lowest panel shows an immunoblot of the material added to the lipid arrays probed with anti-FLAG, demonstrating that equal amounts of E1FLAGHA WT and E2FLAGHA WT (migrating at ϳ 47 and 48 kDa, respectively) were used. E, WT and chimeric erlin1 and erlin2 constructs either FLAGHA-tagged or HA-tagged. The switches were made between erlin2 and erlin1 at amino acid positions 31 and 33 and positions 299 and 301. Amino acid homology (identity) between mouse erlin2 and -1 in the N-terminal (NT), intervening, and C-terminal (CT) regions is 33, 86, and 25%, respectively. F, the constructs shown in E were expressed transiently in ␣T3 E1/E2KO cells and immunopurified with anti-HA, and binding of equal amounts of protein to PI(3)P (800 or 400 pmol) was assessed with anti-FLAG (lanes 1-4), anti-erlin2 (lanes 5 and 6), or anti-erlin1 (lanes 7 and 8). The lower panels show immunoblots of the material added to the PI(3)P arrays probed with the corresponding antibodies, demonstrating that matched amounts of the exogenous proteins were used. The epitopes for anti-erlin1 and anti-erlin2 are located in the C-terminal regions of erlin1 and -2 (13).
To begin to examine the determinants of PI(3)P binding, E1FLAGHA WT or E2FLAGHA WT (Fig. 4E) were expressed in ␣T3 cells lacking both erlin1 and erlin2 (E1/2KO cells), immunopurified using anti-HA, and probed with anti-FLAG (Fig.  4D). This revealed that E2FLAGHA WT bound much more strongly than E1FLAGHA WT , suggesting that, within the erlin1/2 complex, erlin2 is the primary mediator of binding to PI(3)P. To further explore the PI(3)P-binding determinants in erlin2, we created chimeras in which the N and C termini of erlin2 were replaced with the corresponding regions of erlin1 because the termini are the only regions that differ substantially between erlin1 and -2 (Figs. 4E and 3A) (13). Switching the C termini showed that the C-terminal region of erlin2 contributes to PI(3)P binding because E2FLAGHA bound more strongly than E2(E1CT)FLAGHA (Fig. 4F, lanes 1 and 2), and E1FLAGHA bound less strongly than E1(E2CT)FLAGHA (lanes 3 and 4). Switching the N termini showed that the N-terminal region of erlin2 also contributes to PI(3)P binding because E2HA bound much more strongly than E2(E1NT)HA (Fig. 4F, lanes 5 and 6), and E1HA bound less strongly than E1(E2NT)HA (lanes 7 and 8). Overall, these data indicate that the PI(3)P-binding site in erlin2 is composed of an ensemble of polypeptide regions.
Remarkably, examination of the lipid-binding properties of E2HA T65I showed that it had significantly reduced binding capacity for PI(3)P (Fig. 5). This was the case with material immunopurified from ␣T3 E2KO cells stably transfected with E2HA WT and E2HA T65I (Fig. 5A), which contain some erlin1 (Fig. 5A, lowest panel), or material immunopurified from transiently transfected ␣T3 E1/2KO cells, which contain only the exogenous erlin2 constructs (Fig. 5B). Thus, mutation of the Thr-65 region inhibits the ability of complexes containing erlin2 to bind to PI(3)P.

Discussion
We have used gene editing to ablate erlin1 and erlin2 expression from ␣T3 cells to define the roles of these proteins in the functions of the erlin1/2 complex. We were able to show that although both erlin1 alone and erlin2 alone can form highmolecular-weight complexes and associate with RNF170, only erlin2 can bind efficiently to activated IP 3 Rs and mediate their ubiquitination and degradation. This indicates that the molecular determinants for the erlin1/2 complex interaction with activated IP 3 Rs lie within erlin2.
By far the best characterized cellular role of the erlin1/2 complex is to recruit RNF170 to activated IP 3 Rs and mediate IP 3 R ERAD (7). Thus, it was important to examine whether diseaselinked erlin2 mutations affect this role because a defect in IP 3 R processing might help explain the pathology. Remarkably, the previously uncharacterized T65I mutation to erlin2 4 completely blocked the ability of the erlin1/2 complex to associate with activated IP 3 Rs and mediate their ERAD. This was not because the erlin1/2 complex-RNF170 module was grossly perturbed as T65I erlin2 was fully capable of forming highmolecular-weight complexes and associating with erlin1 and RNF170. That the T65I-containing module fails to associate with activated IP 3 Rs suggests that the region of erlin2 that binds to activated IP 3 Rs is in the vicinity of Thr-65 and that the T65I mutation perturbs the binding interface. Clearly, additional investigation of this region may reveal more about how the erlin1/2 complex interacts with its primary target. Unfortunately, however, it has been impossible, so far, to recapitulate the interaction between IP 3 Rs and erlins in vitro using purified proteins. This has impeded the rapid screening of mutants and indicates that the interaction requires conformational changes in IP 3 Rs (2) and membrane localization of IP 3 Rs and erlins only seen in living cells.
We have also made the surprising discovery that the erlin1/2 complex binds specifically to monophosphorylated phosphoinositides and PI(3)P in particular. These lipids play various roles within the cell; e.g. PI(4)P is fairly widely distributed and has a well-established role in controlling cargo exit from the Golgi, whereas PI(5)P distribution is much more restricted and may play a limited role in modulating nuclear functions (33,34). Much attention has been paid to PI(3)P, which plays at least two major roles: by recruiting a range of adapter proteins to membranes it is critical both to the endocytic pathway and to autophagosome biogenesis at the ER (33)(34)(35)(36)(37)(38)(39). Given that the erlin1/2 complex seems to be localized exclusively to the ER (13,15,16) and that the erlin1/2 complex binds specifically to PI(3)P, it appears likely that the erlin1/2 complex-PI(3)P inter- Figure 5. Effects of the T65I mutation on lipid binding. A, erlin complex was immunopurified with anti-HA from ␣T3 E2KO cell lines stably expressing E2HA WT or E2HA T65I , and binding to lipid arrays was assessed by probing with anti-erlin2. The lowest panels show immunoblots of material added to the lipid arrays probed with anti-erlin2 and anti-erlin1, demonstrating that approximately equal amounts of E1HA WT and E2HA T65I were used and that erlin1 coimmunoprecipitates with E2HA WT and E2HA T65I . B, E2HA WT and E2HA T65I were expressed transiently in ␣T3 E1/E2KO cells and immunopurified with anti-HA, and binding of equal amounts to lipid arrays was assessed with anti-erlin2. The lowest panel shows an immunoblot of material added to the lipid arrays probed with anti-erlin2, demonstrating that approximately equal amounts of E2HA WT and E2HA T65I were used.

Erlin1/2 complex-mediated ubiquitination and PI(3)P binding
action has a role in the ER. A major puzzle that remains, however, is that PI(3)P is created and assumed to reside in the cytosol-facing leaflet of the ER membrane (33,40), although available data indicate that erlin1 and 2 are oriented into the ER lumen by their N-terminal transmembrane domains with only a few residues protruding into the cytosol (13). Thus, it is not currently obvious how PI(3)P and the erlins can be arranged to interact. As with the interaction between the erlin1/2 complex and IP 3 Rs, the interaction of the erlin1/2 complex with PI(3)P is mediated primarily by erlin2. Analysis of erlin1/2 chimeras suggests that the PI(3)P-binding site in erlin2 is composed of various parts of the polypeptide and, thus, that structural analysis will be required to gain clear insight into how erlin2 and PI(3)P interact in living cells.
Interestingly, PI(3)P binding to erlin2 appears to be mediated, at least in part, by the Thr-65 region because the T65I mutation inhibits PI(3)P binding. Thus, some of the determinants of PI(3)P binding and activated IP 3 R binding are located in the same region of erlin2, raising the possibility that PI(3)P could be a cofactor for the recognition of activated IP 3 Rs by the erlin1/2 complex or even contribute to the mysterious events that allow for the retrotranslocation of activated, ubiquitinated IP 3 Rs from the ER membrane (7). Finally, it is noteworthy that erlin1 and erlin2 are members of the SPFH domain-containing protein family that includes stomatins, prohibitins, and flotillins (15). The association of these proteins with lipid raft-like membrane domains (15) has led to speculation that these proteins bind lipids, which has been borne out by studies showing that, for example, stomatin binds cholesterol (41), stomatinlike protein 2 binds cardiolipin (42), and prohibitin binds PI(3,4,5)P 3 (43). Thus, an interaction of the erlin1/2 complex with PI(3)P fits with this familial trait.
In summary, we have shown that erlin2 is the dominant partner in the erlin1/2 complex and contains the determinants for binding to activated IP 3 Rs and PI(3)P. We have also identified two possible mechanisms by which the T65I mutation to erlin2 could cause hereditary spastic paraplegia. First, as T65I erlin2 is unable to interact with activated IP 3 Rs, IP 3 R ERAD is impaired. The expected chronic perturbation of IP 3 R levels and Ca 2ϩ handling could contribute to neurodegeneration (44). This logic parallels the effects of the R199C mutation to RNF170 that perturbs Ca 2ϩ signaling (11) and causes neurodegeneration and autosomal-dominant sensory ataxia (45). Second, as the T65I mutation reduces the PI(3)P-binding capacity of the erlin1/2complex,PI(3)PmetabolismandPI(3)P-dependentprocesses may be impaired in T65I-expressing cells with a deleterious effect on cell health.

Plasmids
Mouse erlin1 and erlin2 and human erlin1 cDNAs were subcloned from pCMV-Sport6 plasmids originally purchased from ATCC (12,13). Human erlin2 cDNA was cloned from HeLa cells by RT-PCR as described (14). Human erlin2-HA (E2HA WT ) was constructed by PCR amplification of human erlin2 cDNA such that an HA tag (GYPYDVPDYAG) was spliced to the C terminus. FLAGHA-tagged mouse erlin1 and erlin2 (E1FLAGHA and E2FLAGHA) and erlin1/2 chimeras (13) were created by PCR amplification such that a triple FLAGHA tag (DYKDHDGDYKDHDIDYKDDDDKGYPYD-VPDYA) was spliced to the C termini. T65I human erlin2 (E2HA T65I ) was created using mutagenic PCR. Primer sequences are available upon request.

Erlin1/2 complex-mediated ubiquitination and PI(3)P binding
0.2 M soybean trypsin inhibitor, 1 mM DTT, pH 8.0), and centrifuged (16,000 ϫ g for 10 min at 4°C), and supernatant samples were subjected to SDS-PAGE and immunoblotting. Membrane fractions were prepared by harvesting cells in hypotonic homogenization buffer (10 mM Tris base, 1 mM EGTA, 10 M pepstatin, 0.2 mM phenylmethylsulfonyl fluoride, 0.2 M soybean trypsin inhibitor, 1 mM DTT, pH 7.4), sonication for 1 min, and centrifugation at 40,000 ϫ g for 10 min at 4°C. To analyze the interactions among the erlin1/2 complex, RNF170, and IP 3 R1, cells were disrupted using lysis buffer containing 1% CHAPS rather than Triton X-100. For analysis of IP 3 R1 ubiquitination, cells were incubated with CHAPS lysis buffer lacking DTT but supplemented with 5 mM N-ethylmaleimide for 30 min at 4°C followed by addition of 5 mM DTT and centrifugation (16,000 ϫ g for 10 min at 4°C). Lysates were then incubated with anti-IP 3 R1 to immunoprecipitate IP 3 R1 as described (31), and complexes were heated to 37°C for 30 min before being subjected to SDS-PAGE in 5% gels for ubiquitin conjugate analysis or in 10% gels for analysis of additional proteins. [Ca 2ϩ ] c was measured as described (11) using cells loaded with 5 M Fura2-AM for 1 h at 37°C.

Data presentation
All experiments were repeated at least once, and representative images of gels or lipid arrays are shown. Immunoreactivity was detected and quantitated using enhanced chemiluminescence reagents and a GeneGnome (SynGene Bio Imaging) or a Chemidoc (Bio-Rad). The migration positions of molecular mass markers (size in kDa) are shown to the right of immuno-blot images. Quantitated data are expressed as mean Ϯ S.E. of n independent experiments.