A Point Mutation in the Ubiquitin Ligase RNF170 That Causes Autosomal Dominant Sensory Ataxia Destabilizes the Protein and Impairs Inositol 1,4,5-Trisphosphate Receptor-mediated Ca2+ Signaling*

Background: RNF170 is an endoplasmic reticulum membrane ubiquitin ligase that when mutated at arginine 199 causes a neurodegenerative disease. Results: The mutation disrupts salt bridges between transmembrane domains, destabilizes the protein, and inhibits Ca2+ signaling via IP3 receptors. Conclusion: These manifestations of the mutation are likely causative to neurodegeneration. Significance: Understanding the mechanism of action of mutant ubiquitin ligases will lead to better therapies. RNF170 is an endoplasmic reticulum membrane ubiquitin ligase that contributes to the ubiquitination of activated inositol 1,4,5-trisphosphate (IP3) receptors, and also, when point mutated (arginine to cysteine at position 199), causes autosomal dominant sensory ataxia (ADSA), a disease characterized by neurodegeneration in the posterior columns of the spinal cord. Here we demonstrate that this point mutation inhibits RNF170 expression and signaling via IP3 receptors. Inhibited expression of mutant RNF170 was seen in cells expressing exogenous RNF170 constructs and in ADSA lymphoblasts, and appears to result from enhanced RNF170 autoubiquitination and proteasomal degradation. The basis for these effects was probed via additional point mutations, revealing that ionic interactions between charged residues in the transmembrane domains of RNF170 are required for protein stability. In ADSA lymphoblasts, platelet-activating factor-induced Ca2+ mobilization was significantly impaired, whereas neither Ca2+ store content, IP3 receptor levels, nor IP3 production were altered, indicative of a functional defect at the IP3 receptor locus, which may be the cause of neurodegeneration. CRISPR/Cas9-mediated genetic deletion of RNF170 showed that RNF170 mediates the addition of all of the ubiquitin conjugates known to become attached to activated IP3 receptors (monoubiquitin and Lys48- and Lys63-linked ubiquitin chains), and that wild-type and mutant RNF170 have apparently identical ubiquitin ligase activities toward IP3 receptors. Thus, the Ca2+ mobilization defect seen in ADSA lymphoblasts is apparently not due to aberrant IP3 receptor ubiquitination. Rather, the defect likely reflects abnormal ubiquitination of other substrates, or adaptation to the chronic reduction in RNF170 levels.

RNF170 is an endoplasmic reticulum membrane ubiquitin ligase that contributes to the ubiquitination of activated inositol 1,4,5-trisphosphate (IP 3 ) receptors, and also, when point mutated (arginine to cysteine at position 199), causes autosomal dominant sensory ataxia (ADSA), a disease characterized by neurodegeneration in the posterior columns of the spinal cord. Here we demonstrate that this point mutation inhibits RNF170 expression and signaling via IP 3 receptors. Inhibited expression of mutant RNF170 was seen in cells expressing exogenous RNF170 constructs and in ADSA lymphoblasts, and appears to result from enhanced RNF170 autoubiquitination and proteasomal degradation. The basis for these effects was probed via additional point mutations, revealing that ionic interactions between charged residues in the transmembrane domains of RNF170 are required for protein stability. In ADSA lymphoblasts, platelet-activating factor-induced Ca 2؉ mobilization was significantly impaired, whereas neither Ca 2؉ store content, IP 3 receptor levels, nor IP 3 production were altered, indicative of a functional defect at the IP 3 receptor locus, which may be the cause of neurodegeneration. CRISPR/Cas9-mediated genetic deletion of RNF170 showed that RNF170 mediates the addition of all of the ubiquitin conjugates known to become attached to activated IP 3 receptors (monoubiquitin and Lys 48 -and Lys 63linked ubiquitin chains), and that wild-type and mutant RNF170 have apparently identical ubiquitin ligase activities toward IP 3 receptors. Thus, the Ca 2؉ mobilization defect seen in ADSA lymphoblasts is apparently not due to aberrant IP 3 receptor ubiquitination. Rather, the defect likely reflects abnormal ubiquitination of other substrates, or adaptation to the chronic reduction in RNF170 levels.
In eukaryotic cells, ubiquitin ligases (E3) 2 work together with ubiquitin-conjugating enzymes (E2) to promote ubiquitination of a range of substrates, often leading to their degradation by the proteasome (1)(2)(3)(4)(5). Mammals express hundreds of E3s, the vast majority of which contain a RING domain, a motif that appears to provide a docking site for E2s and be necessary for ubiquitin transfer to the substrate (3)(4)(5). Endoplasmic reticulum (ER)-associated degradation (ERAD) is the term used to describe the pathway by which aberrant ER lumen or membrane proteins are removed from the cell via ubiquitination and proteasomal degradation (6,7). A recent bioinformatic survey of RING domain-containing proteins that localize to the ER membrane and that could play a role in ERAD identified a group of 24 E3s (8), some of which (e.g. HRD1) appear be capable of mediating the ubiquitination of a broad array of substrates, whereas others (e.g. TRC8) have a much narrower substrate range (3,6,7). Furthermore, several of the E3s identified (e.g. RNF170) have yet to be fully characterized (8).
Inositol 1,4,5-trisphosphate receptors (IP 3 Rs) are ER membrane proteins that form tetrameric Ca 2ϩ channels that govern ER Ca 2ϩ store release (9,10). When persistently activated, a portion of cellular IP 3 Rs are ubiquitinated and degraded by the proteasome apparently via the ERAD pathway, and this IP 3 R "down-regulation" suppresses Ca 2ϩ mobilization (11,12).
Recent studies on the mechanism of IP 3 R processing by the ERAD pathway have shown that a complex composed of the integral ER membrane proteins erlin1 and erlin2 associates rapidly with activated IP 3 Rs (11)(12)(13)(14), as does RNF170 (15). RNF170 is ϳ257 amino acids in length, is highly conserved in mammals, is predicted to have 3 transmembrane (TM) domains, with the RING domain facing the cytosol (15,16), and is constitutively associated with the erlin1/2 complex. RNF170 contributes to IP 3 R ubiquitination (15), although whether it is responsible for the addition of all of the conjugates that become attached to activated IP 3 Rs (monoubiquitin and Lys 48 -and Lys 63 -linked ubiquitin chains) (17,18) is currently unknown. Remarkably, a recent molecular genetic study demonstrated that autosomal dominant sensory ataxia (ADSA), a rare neurodegenerative disease characterized by ataxic gait, reduced sensory perception, and neurodegeneration in the posterior columns of the spinal cord, segregates with an arginine (Arg 199 ) to cysteine mutation in human RNF170 (16,19). Here we examine the effects of this mutation on the properties of RNF170 and find that it destabilizes the protein because of disruption of a salt bridge between TM domains 2 and 3, and that mutant RNF170 disrupts Ca 2ϩ signaling at the IP 3 R locus in ADSA lymphoblasts. Genetic deletion of RNF170 revealed that RNF170 mediates the addition of all ubiquitin conjugates to activated IP 3 Rs and that the ubiquitin ligase activities of wild-type and mutant RNF170 toward activated IP 3 Rs are apparently identical. Thus, aberrant ubiquitination of other substrates, or cellular adaptation to chronically reduced RNF170 levels likely accounts for the ADSA-associated Ca 2ϩ signaling deficit.
Plasmids-Mouse and human RNF170 cDNAs were cloned from mouse ␣T3 cells and human lymphoblasts as described (15). RNF170 FLAG was constructed by ligating the sequence encoding the 257-amino acid mouse RNF170 into the KpnI and BamHI sites of the pCMV14 -3xFLAG expression vector such that a triple FLAG tag (DYKDHDGDYKDHDIDYKDDDDKG) is spliced to the C terminus (15). RNF170 HA was constructed by PCR amplification of mouse and human RNF170 cDNAs, such that an HA tag (GYPYDVPDYAG) is spliced to the C terminus. Additional constructs were created via PCR that encode proteins with an optimal N-glycosylation consensus sequence (NSTMMS) (22) immediately after the HA tag, and with a variety of amino acid substitutions. Primer sequences are available upon request.
Lymphoblasts-To assess protein expression in cell lysates, lymphoblasts were collected by centrifugation (1,000 ϫ g for 5 min), washed once by resuspension in PBS followed by centrifugation (16,000 ϫ g for 1 min), disrupted for 30 min at 4°C with 1% CHAPS lysis buffer, centrifuged (16,000 ϫ g for 10 min at 4°C), and supernatant samples were subjected to SDS-PAGE and immunoblotting as described (13)(14)(15). For measurement of the cytosolic free Ca 2ϩ concentration ([Ca 2ϩ ] c ), lymphoblasts were collected by centrifugation (1,000 ϫ g for 1 min), washed once, and then incubated in culture medium at 2 mg of protein/ml with 5 M fura2-AM for 1 h at 37°C, washed 4 times with Krebs HEPES buffer (23), and finally resuspended at 0.5 mg of protein/ml in Krebs HEPES buffer for measurement of [Ca 2ϩ ] c at 37°C, as described (23). For measurement of IP 3 levels, lymphoblasts were collected by centrifugation (1,000 ϫ g for 5 min), washed once by resuspension in RPMI 1640 (Cellgro) followed by centrifugation (1,000 ϫ g for 5 min), resuspended in RPMI 1640 and preincubated at 37°C for 1 h, then exposed to vehicle (DMSO) or PAF, and IP 3 mass was measured as described (23).
Generation and Analysis of RNF170 Knock-out and Reconstituted Cell Lines-The CRISPR/Cas9 system (24,25) was used to target an exon within the mouse RNF170 gene that was common to all predicted splice variants (exon 6). Oligonucleotides that contained the RNF170 target CRISPR sequence (GATACTGGCGATACGGGTCCTGG) were annealed and then ligated into AflII-linearized gRNA vector (Addgene). ␣T3 mouse pituitary cells were transfected (18) with a mixture of the gRNA construct and vectors encoding Cas9 (Addgene) and EGFP (Clontech). Two days post-transfection, EGFP-expressing cells were selected by fluorescence-activated cell sorting and plated at ϳ1 cell/well in 96-well plates. Colonies were expanded and screened for ablation of RNF170 expression by immunoblotting with anti-RNF170 (raised against amino acids 50 -65 of RNF170) (15). Of the cell lines screened, ϳ30% lacked RNF170 and 2 of those were expanded for characterization of IP 3 R1 processing with essentially identical results; clone IC8 was used for the experiments shown in Fig. 7. Reconstituted cell lines were obtained by transfecting clone IC8 with cDNAs encoding mouse RNF170 and R198C RNF170 (18), followed by selection in 1.3 mg/ml of G418 for 72 h, plating at ϳ1 cell/well in 96-well plates, colony expansion, screening with anti-RNF170, and maintenance in 0.3 mg/ml of G418. 2 clones expressing each construct were characterized with essentially identical results.
For analysis of IP 3 R1 ubiquitination and protein levels, cells were incubated with 1% CHAPS lysis buffer lacking dithiothreitol, but supplemented with 5 mM N-ethylmaleimide, for 30 min at 4°C, followed by addition of 5 mM dithiothreitol 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 (15) and complexes were heated at 37°C for 30 min prior to SDS-PAGE in 5% gels for ubiquitin conjugate analysis, or 100°C for 5 min prior to SDS-PAGE in 10% gels for analysis of other proteins. [Ca 2ϩ ] c was measured as described (23).
Data Presentation-All experiments were repeated at least once, and representative images of gels or traces are shown. Immunoreactivity was detected and quantitated using Pierce ECL reagents and a GeneGnome Imager (Syngene Bio Imaging). Quantitated data are expressed as mean Ϯ S.E. or range of n independent experiments.

Results
Analysis of the Membrane Topology of RNF170 -The predicted topology of mouse RNF170, derived from bioinformatic analysis, together with the amino acid sequences of TM domains 2 and 3 are depicted in Fig. 1A. The sequence of human RNF170 is very similar to that of mouse RNF170 (91% identical) and human RNF170 has the same predicted topology (15,16). Note that Arg 199 in the human sequence corresponds to Arg 198 in the mouse sequence, because of the absence of Asp 16 from the latter (15,16). The existence of TM domains 2 and 3 was confirmed experimentally, using a series of truncation mutants containing a C-terminal HA/glycosylation tag that together with an assessment of sensitivity to the deglycosylating activity of endo H, can provide information on orientation across the ER membrane (22,26) (Fig. 1B). That fulllength RNF170 ( N267 RNF170 HA ) is insensitive to endo H (and is therefore not glycosylated) localizes the C terminus of this construct to the cytosol (Fig. 1B). In contrast, the sensitivity of N230 RNF170 HA to endo H shows that the C terminus of this construct is glycosylated and is within the ER lumen, demonstrating that TM2 traverses the ER membrane (Fig. 1B). Following the same logic, that N200 RNF170 HA is insensitive to endo H and is not glycosylated localizes the C terminus of this construct to the cytosol and confirms the orientation of TM2 (Fig.  1B). Finally, to confirm the location of the N terminus, G8N RNF170 FLAG was created, which due to replacement of Gly 8 with asparagine, contains an N-glycosylation consensus sequence (NQS) (22,26) very near the N terminus. This construct was partially glycosylated, indicating that the N terminus is located in the ER lumen (Fig. 1B); that the glycosylation was relatively weak may be because the consensus sequence is suboptimal in comparison to the optimal sequence (NSTMMS) present in the other constructs (22). Overall, these data are consistent with the topology of TM domains 1, 2, and 3 shown in Fig. 1A.
Effects of the Arg 198 to Cys mutation on RNF170 -Remarkably, the Arg 198 to Cys mutation significantly reduced RNF170 expression. This was seen when the mutation was introduced into either untagged RNF170 or RNF170 FLAG ( Fig. 2A). Furthermore, the mutants migrated ϳ1 kDa more rapidly than their wild-type counterparts ( Fig. 2A). Additional mutants were made to explore the basis for the reduction in expression and the migration shift, using RNF170 FLAG as a template (Fig. 2B). To examine the possibility that these changes resulted from disulfide bond formation with the newly introduced cysteine,  (15) with the TM2/3 region expanded to show the mouse amino acid sequence. Note that the amino acid that corresponds to Arg 199 of human RNF170 is Arg 198 in mouse RNF170, and that the sequences of human and mouse TM2/3 regions are identical, with the exception that Met 202 of mouse RNF170 is Ile 203 in human RNF170 (15,16). The RING domain, the three TM domain regions, and the N and C termini are indicated, and Arg 198 is identified with an asterisk. The precise limits of the predicted TM domains have not been defined experimentally, but the scheme shown is predicted by multiple programs (e.g. TMHMM, TOPCONS, etc). B, N-glycosylation of RNF170 mutants. An HA/glycosylation tag (black box) was introduced at the C terminus of fulllength RNF170 ( N267 RNF170 HA ), or truncated RNF170 lacking putative TM domain 3 ( N230 RNF170 HA ), or putative TM domains 2 and 3 ( N200 RNF170 HA ). These, and G8N RNF170 FLAG (FLAG tag indicated by a gray box and the G8N mutation with a circle) were expressed in HeLa cells, lysates were incubated without or with 1 unit/l of endo H for 3 h at 37°C, and samples were subjected to SDS-PAGE. Blots were then probed with anti-HA or anti-FLAG to identify the exogenous RNF170 constructs, or anti-erlin2 to identify endogenous erlin2, which is known to be N-glycosylated (13,14), and which serves as a positive control for endo H. The migration positions of unmodified and N-glycosylated species are indicated with arrows and arrowheads, respectively; degylcosylation of erlin2, N230 RNF170 HA , and G8N RNF170 FLAG by endo H reduces their apparent molecular masses by ϳ2 kDa.
Arg 198 was replaced with serine, which is similar in size to cysteine, but which cannot form disulfide bonds. However, R198S RNF170 FLAG was also expressed poorly and migrated rapidly (Fig. 2B, lane 3), ruling out a role for disulfide bonds. Likewise, to examine whether the changes were due to loss of the bulky side chain of arginine, or the positive charge, Arg 198 was replaced with glutamine, which is relatively bulky, but is uncharged, albeit polar. R198Q RNF170 FLAG was also expressed quite poorly and migrated rapidly (lane 4), suggesting that it is not the bulk of arginine, but rather its positive charge, that is necessary for normal expression and migration. This was confirmed by replacing Arg 198 with lysine, which has a slightly shorter side chain, but is still positively charged; R198K RNF170 FLAG was expressed and migrated similarly to WT RNF170 FLAG (lane 5).
Examination of the amino acid sequences of TM domains 2 and 3 (Fig. 1A) revealed the presence of an additional arginine in TM2 (Arg 200 ) and two aspartic acids in TM3 (Asp 231 and Asp 232 ) (Fig. 1A), suggesting that ionic interactions between TM2 and TM3 could be required for wild-type behavior. Mutation of these amino acids confirmed this notion, as R200C RNF170 FLAG , D231A RNF170 FLAG , D232A RNF170 FLAG , and D231A/D232A RNF170 FLAG all expressed poorly and migrated more rapidly than RNF170 FLAG (lanes 6 -9). Thus, ionic interactions between arginines in TM2 and aspartic acids in TM3 appear to be required for normal expression of RNF170. Furthermore, it appears that 2 charged amino acids in each TM are required for normal expression, because mutation of one arginine and one aspartic acid in R198C/D231A RNF170 FLAG did not rescue expression (lane 10). Finally, it is noteworthy that accelerated migration correlated well with reduced expression (Fig.  2B) and was caused by both arginine and aspartic acid loss and is therefore not due to a change in net protein charge. Rather, these data, together with the fact that RNF170 migrates at 21.5 kDa rather than the predicted 30 kDa (15), suggest that RNF170 is not fully denatured during SDS-PAGE and that the mutations affect the structural organization that is retained. The mutations may also alter the structure of RNF170 in vivo, which in turn, could account for the reductions in expression, perhaps because of protein destabilization.
Mechanism of Reduced Expression-To explore possible destabilization mechanisms, we examined the effects of the proteasome inhibitor MG-132. When incubated with cells expressing exogenous RNF170 FLAG constructs (Fig. 3A), MG-132 caused a marked accumulation of R198C RNF170 FLAG (lanes 4 -6), although leaving WT RNF170 FLAG levels essentially unaltered (lanes 1-3). Thus, the proteasome appears to mediate the reduced expression of R198C RNF170 FLAG . Furthermore, as E3 autoubiquitination is a commonly observed phenomenon (27), we examined the effects of inactivating ligase activity on R198C RNF170 FLAG expression. Previously, we have shown that mutation of Cys 101 and His 103 in the RING domain of mouse RNF170 blocks ubiquitin ligase activity (15), and introduction of these mutations into R198C RNF170 FLAG (creating R198C/⌬RING RNF170 FLAG ) normalized expression (Fig. 3B, lane  4), suggesting that ligase activity, and most likely autoubiquitination, mediates the reduced expression of R198C RNF170 FLAG . Interestingly, the R198C/⌬RING RNF170 FLAG mutant still migrated faster than WT RNF170 FLAG , suggesting that the putative structural change has not been reversed by mutation of the RING domain (Fig. 3B, lane 4). A role for autoubiquitination was supported by the observation that WT RNF170 HA expression was not differentially affected by co-expression of WT RNF170 FLAG and R198C RNF170 FLAG (Fig. 3C, lanes 2 and 3); this result shows that the apparent destabilizing effect of the R198C mutation is intramolecular and is not transmitted to co-expressed WT RNF170 HA molecules. Importantly, R198C RNF170 FLAG exhibited in vitro ubiquitin ligase activity similar to that of WT RNF170 FLAG (Fig. 3D, lanes 2 and 4), indicating that it is fully capable of autoubiquitinating. Finally, to determine protein half-lives, transfected cells were incubated with the protein synthesis inhibitor cycloheximide (Fig. 3E). This revealed that the half-life of R198C RNF170 FLAG is much shorter than that of WT RNF170 FLAG , and that MG-132 blocks R198C RNF170 FLAG degradation. Overall, these data indicate that the R198C mutation decreases RNF170 expression by triggering the molecule to autoubiquitinate and then be degraded by the proteasome.  6 -10). B, lysates were probed with anti-FLAG and the histogram shows combined quantitated immunoreactivity (mean Ϯ S.E., n Ն 3).

The R198C Mutation Does Not Alter Membrane Topology, Subcellular Localization, or Interaction with the Erlin1/2
Complex-To explore why mutation-induced salt bridge disruption triggers RNF170 autoubiquitination, we first examined whether subcellular localization was altered. However, the extent of N-glycosylation seen for the relevant constructs shown in Fig. 1B ( N267 RNF170 HA and N230 RNF170 HA ) was not substantially altered by introduction of the R198C mutation (Fig. 4A), indicating that localization to the ER and insertion of TM domains is normal. Likewise, interaction with the endogenous erlin1/2 complex was normal, because immunoprecipitation with anti-erlin2 (15) recovered WT RNF170 HA and R198C RNF170 HA equally well (Fig. 4B). Furthermore, membrane association was normal, as both WT RNF170 HA and FIGURE 3. The R198C mutation reduces RNF170 expression via autoubiquitination and the proteasome. A-C, cDNAs encoding tagged WT and mutant RNF170 constructs transfected into HeLa cells were treated as indicated, and cell lysates were probed with anti-FLAG or anti-HA to recognize exogenous RNF170 constructs, or anti-erlin2, which served a loading control. The histograms show combined quantitated immunoreactivity (mean Ϯ S.E., n Ն 4). D, RNF170 FLAG constructs were immunopurified from transfected HeLa cells and incubated with E1 (UBE1), E2 (UbcH5b), and HA-ubiquitin as indicated for 30 min at 30°C, with the exception of lane 1, which lacked E2. Samples were then probed with anti-HA to assess ubiquitination (upper panel), or anti-FLAG to assess the levels of RNF170 FLAG constructs (lower panel). The asterisk marks a background band. E, transfected HeLa cells were treated as indicated with 20 g/ml of cycloheximide (CHX), without or with 10 M MG-132. Cell lysates were then probed with anti-p53 as a positive control for cycloheximide action, anti-p97 as a loading control, and anti-FLAG to recognize exogenous RNF170 constructs (long and short exposures are shown to facilitate visualization of immunoreactivity changes). The graph shows combined quantitated FLAG immunoreactivity, using the long exposure for R198C RNF170 FLAG and the short exposure for WT RNF170 FLAG (mean Ϯ range, n ϭ 2). R198C RNF170 HA were fractionated with membranes (Fig. 4C). Thus, the R198C mutation does not dramatically alter the localization of RNF170, indicating that a subtle change in its properties (e.g. in the way that it folds) accounts for its apparent autoubiquitination and proteasomal degradation.
RNF170 Expression in Control and ADSA-affected Individuals-Lymphoblast lines from control and ADSA-affected individuals were examined for RNF170 expression (Fig. 5A). In controls, an anti-RNF170 immunoreactive band at ϳ21.5 kDa was observed ( lanes  1-3), the same size as endogenous HeLa cell RNF170 and untagged mouse RNF170 (Fig. 2A). Remarkably, in affected lymphoblasts the strength of the 21.5-kDa band was reduced and an additional weaker band at ϳ20.5 kDa was also observed (lanes 4-6). These results are consistent with the expression level and migration differences seen between exogenous mouse WT RNF170 and R198C RNF170 ( Fig. 2A) and between exogenous human WT RNF170 and R199C RNF170 (Fig.  5B), and with the knowledge that affected individuals are heterozygote (16). Thus, it appears that the R199C RNF170 encoded by the mutant allele in affected individuals migrates at ϳ20.5 kDa and is relatively poorly expressed. This leads to an ϳ27% reduction in total RNF170 immunoreactivity in affected lymphoblasts (Fig. 5A).
Ca 2ϩ Mobilization via IP 3 Receptors Is Impaired in Affected Lymphoblasts-Because IP 3 Rs are the only known substrates for RNF170 (15), we examined whether IP 3 R function was different in control and affected lymphoblasts, using PAF to trigger IP 3 R-mediated Ca 2ϩ mobilization (28) (Fig. 6). Remarkably, PAF-induced increases in [Ca 2ϩ ] c were significantly suppressed in affected lymphoblasts (Fig. 6A), suggesting that IP 3 R-mediated Ca 2ϩ mobilization might be impaired, although the suppression could also be due to an effect on Ca 2ϩ entry (29). To rule out any role for Ca 2ϩ entry, EGTA was added immediately prior to PAF to chelate extracellular Ca 2ϩ to ϳ100 nM (23), and under these conditions the suppression of PAFinduced increases in [Ca 2ϩ ] c was still clearly evident (Fig. 6B). To examine if the suppression results from ER Ca 2ϩ stores being smaller, cells were exposed to the sarcoplasmic reticulum Ca 2ϩ -APTase pump inhibitor thapsigargin, which allows Ca 2ϩ leak from the ER in an IP 3 R-independent manner (30). However, thapsigargin caused the same [Ca 2ϩ ] c increase in control and affected lymphoblasts (Fig. 6C), indicating that Ca 2ϩ stores are of equal size. Likewise, reduced IP 3 formation was not the reason for the suppression, as PAF-induced increases in IP 3 mass were the same in control and affected lymphoblasts (increases over basal resulting from 0.5 min exposure to 100 nM PAF were 24 Ϯ 8 and 35 Ϯ 15%, respectively; n ϭ 3). Finally, measurement of the levels of IP 3 R1-3 did not reveal any consistent differences between control and affected lymphoblasts (Fig. 6D). Overall, these data indicate that Ca 2ϩ mobilization via IP 3 Rs is impaired in affected lymphoblasts, but not because of a change in Ca 2ϩ store size, or the abundance of IP 3 Rs, and point toward a mechanism that involves a regulation of IP 3 R activity.
Ubiquitin Ligase Activity of RNF170 and R198C RNF170 in Cells-R198C RNF170 exhibits apparently normal ubiquitin ligase activity in vitro when mixed with purified enzymes (Fig.  3D), but that tells us almost nothing about how its activity might differ from that of WT RNF170 in vivo, where the intracellular milieu contains a full complement of E2s, substrates, and other factors. To date, endogenous activated IP 3 Rs are the only known substrates for RNF170 (15), but it has yet to be resolved whether RNF170 is responsible for the addition of all of the ubiquitin conjugates that become attached to activated IP 3 Rs, or just some (17,18). Initially, we sought to compare the ligase activities of WT RNF170 and R199C RNF170 toward IP 3 Rs in control and ADSA lymphoblasts, but pilot experiments showed that to be unfeasible, primarily because of the paucity of IP 3 Rs therein (data not shown). Thus, we employed ␣T3 cells, the cells in which we first identified RNF170 (15) and that exhibit very robust IP 3 R1 ubiquitination in response to the IP 3 generating cell surface receptor agonist GnRH (17,18). First, to determine which ubiquitin conjugates RNF170 adds, we deleted RNF170 using CRISPR/Cas9-mediated gene editing (Fig. 7). RNF170 "knock-out" was specific (Fig. 7A), as other pertinent proteins (e.g. the ERAD pathway proteins p97, Hrd1, and gp78 (6, 7) and erlins 1 and 2) were expressed at the same level in control and "␣T3 RNF170KO" cells. Interestingly, IP 3 R1 expression was enhanced ϳ65% by RNF170 deletion (Fig.   FIGURE 4. Lack of effect of the R198C mutation on RNF170 membrane association and topology, and interaction with the erlin1/2 complex. cDNAs encoding HA-tagged WT and mutant RNF170 constructs were transfected into HeLa cells. A, N-glycosylation of N267/R198C RNF170 HA and N230/ R198CRNF170 HA , R198C-containing versions of the constructs shown in Fig. 1B, was assessed as described in the legend to Fig. 1B. B, interaction with the erlin1/2 complex. Erlin1/2 complex was immunoprecipitated with anti-erlin2 was probed for RNF170 constructs (lower panels). Note that the amounts of WT RNF170 HA and R198C RNF170 HA that co-immunoprecipitate are proportional to the amounts in input lysates, indicating that they interact with the erlin1/2 complex equally well. C, cells were lysed and centrifuged into cytosol (C) and membrane (M) factions as described (13), and probed as indicated. A and D), indicating that in addition to its role in mediating the degradation of activated IP 3 Rs (15), RNF170 also plays a role in basal IP 3 R1 turnover. Ca 2ϩ mobilization in ␣T3 RNF170KO cells in response to GnRH was normal (Fig. 7B), indicating that the IP 3 -dependent signaling pathway and IP 3 R activation are not perturbed by the absence of RNF170. Remarkably, deletion of RNF170 completely blocks GnRH-induced IP 3 R1 ubiquitination (Fig. 7C), observed using either FK2 antibody, which detects all ubiquitin conjugates (monoubiquitin, and Lys 48 -and Lys 63 -linked chains), or antibodies specific for Lys 48 -and Lys 63 -linked chains (18). In contrast, erlin2 association with IP 3 R1 was not blocked, indicating that association of the erlin1/2 complex with activated IP 3 R1 is unimpaired, consistent with the notion that it is the erlin1/2 complex that recruits RNF170 to activated IP 3 R1, rather than vice versa (12,15). Thus, RNF170 does indeed catalyze the formation of all ubiquitin conjugates on activated IP 3 R1, and as would be expected, GnRH did not cause IP 3 R down-regulation in ␣T3 RNF170KO cells (Fig. 7D).

7,
To directly assess the ligase activity of R198C RNF170, we sought to reconstitute IP 3 R ubiquitination in ␣T3 RNF170KO cells by stably expressing exogenous RNF170 constructs (Fig.  7E). Cell lines were obtained, although exogenous expression was less than that seen for endogenous RNF170 in control ␣T3 cells (data not shown). Both WT RNF170 and R198C RNF170 were capable of ubiquitinating activated IP 3 Rs, as indicated by increases in Lys 48 -and Lys 63 -linked ubiquitin chains and total ubiquitin after exposure to GnRH (Fig. 7E). That less ubiquitination was seen in R198C RNF170-expressing cells is most likely a consequence of lower expression level and IP 3 R binding of the mutant (Fig. 7E). Thus, the ligase activity of R198C RNF170 toward IP 3 Rs receptors in vivo appears to be qualitatively normal, and aberrant IP 3 R receptor ubiquitination is unlikely to account for the Ca 2ϩ signaling deficit seen in ADSA lymphoblasts.

Discussion
Our data show that the Arg 198 to Cys mutation in mouse RNF170 reduces the stability and expression level of the protein. The mutation appears to disrupt a salt bridge between TM domains 2 and 3 and leads to autoubiquitination and enhanced turnover via the ubiquitin-proteasome pathway. This mechanism also likely applies to human R199C RNF170 in ADSA-affected individuals, and accounts for the ϳ27% reduction in the total cellular complement of RNF170. An equivalent decrease in cellular ligase activity attributable to RNF170 is likely, which could be critical to development of the disease. Alternatively, the ligase activity of mutant RNF170 in vivo could be abnormal.  6) were probed with anti-RNF170 and anti-erlin2. For the panels shown, immunoreactivity was quantitated and is plotted as total RNF170/erlin2 immunoreactivity (arbitrary units). Multiple quantitations of total RNF170 immunoreactivity from these and other lymphoblast lines showed that affected individuals contained 73 Ϯ 5% of the immunoreactivity seen in control lymphoblasts (n Ն 5). B, cDNAs encoding human WT RNF170 HA and R199C RNF170 HA were transfected into HeLa cells and cell lysates were probed as indicated. Erlin2 served as a loading control. We did not obtain support for this notion from studies with purified components in vitro (Fig. 3D), or from analysis of IP 3 R1 ubiquitination in vivo (Fig. 7E), but that does not rule out the possibility that mutant RNF170 acts abnormally toward other substrates. Mutant RNF170 acting in this manner (i.e. dominantly) would be consistent with zebrafish studies, in which expression of exogenous mutant RNF170 causes aberrant development (16).
Prior to the current study, our approach to defining the function of RNF170 (and other proteins suspected to play a role in IP 3 R ERAD) has been to deplete them using RNA interference. Although we have had some success with this approach (13)(14)(15)31) it has major limitations; in particular, proteins are only depleted and the effects of residual proteins are hard to assess and complicate interpretation of data from reconstitution experiments, and cells expressing short interfering RNA are often unhealthy and are available only in limited quantities. Use of the CRISPR/Cas9 system (24,25) allowed us to delete RNF170 and demonstrate, for the first time, that RNF170 catalyzes the addition of all ubiquitin conjugates to activated IP 3 R1. Intriguingly, this suggests that RNF170 interacts with multiple E2s, most likely Ubc13 and Ubc7, because Ubc13 is the only E2 known to build Lys 63linked chains (5, 32) and Ubc7, which builds Lys 48 -linked chains (5,32), is already strongly implicated in mediating IP 3 R1 ubiquitination and degradation (33).
Why does the Arg 198 to Cys mutation destabilize RNF170? It appears that a network of salt bridges couple TM domains 2 and 3 together, because mutation of either Arg 198 or Arg 200 in TM2 reduces protein expression, as do mutations to Asp 231 and Asp 232 in TM3. A hint as to why these mutations are apparently destabilizing is provided by the fact that they also cause RNF170 to migrate more rapidly on SDS-PAGE, indicative of a structural change in the protein that either causes more compact folding and/or alters interactions with SDS. The structural change appears to be subtle, as the Arg 198 to Cys mutation did not significantly affect TM2 and TM3 insertion into the ER membrane, or interaction with the erlin1/2 complex, but could still be sufficient to trigger autoubiquitination and targeting for ERAD. Interestingly, many other E3s are known to be regulated by ubiquitination (3,27), including the ER membrane ligase gp78 (27), which is relatively unstable and is controlled both by autoubiquitination and an additional ER membrane ubiquitin ligase Hrd1 (27). It will be interesting to see whether mutations in the TM domains of gp78 and other ER membrane ligases (8) are also destabilizing.
Although the presence of charged residues in TM domains is energetically unfavorable, such residues are often found therein, where they play important functions, often involving salt bridges (34). Intriguingly, TM domain-located charged residues are those most likely to cause disease when mutated, arginine has the highest propensity for disease causation, and the Arg to Cys mutation is relatively common because cytosine to thymine transition occurs with relatively high frequency (35). Situations very similar to that described here for RNF170 are seen in other proteins. For example, the pore architecture of cystic fibrosis transmembrane conductance regulator is maintained by salt bridges between arginines in TM6 (Arg 347 and Arg 352 ) and aspartic acids in adjacent TMs (36), and naturally occurring, charge altering mutations of Arg 352 cause cystic fibrosis (37). Likewise, the naturally occurring R279C mutation in the prostacyclin receptor dramatically reduces protein expression (38).
Remarkably, the expression of R199C RNF170 correlated with a reduction in PAF-induced Ca 2ϩ mobilization in lymphoblasts that was not due to a change in IP 3 R levels, but rather appears to result from reduced signal transduction efficiency at the IP 3 R locus. This was apparently not due to aberrant IP 3 R ubiquitination, as the ubiquitination of activated IP 3 R1 in cells expressing wild-type or mutant RNF170 was qualitatively identical, at least in terms of the addition of total ubiquitin and Lys 48 -and Lys 63linked chains. Rather, the reduction in signal transduction efficiency at the IP 3 R locus could be an indirect effect, if RNF170 turns out to ubiquitinate additional substrates (e.g. proteins that regulate IP 3 Rs), or if long-term adaptation to the ϳ27% reduction in total RNF170 expression seen in ADSA lymphoblasts alters the expression of genes that govern Ca 2ϩ signaling.
Dysregulation of Ca 2ϩ metabolism and IP 3 R function is often mooted to be the cause of the neurodegeneration that underpins certain spinocerebellar ataxias and neurodegenerative diseases (39 -43) and our data suggest that the same could be true for ADSA. If so, therapies aimed at boosting Ca 2ϩ mobilization could be contemplated, similarly to the way that manipulating Ca 2ϩ metabolism is being examined as a therapy for Huntington disease and other spinocerebellar ataxias (41)(42)(43). Interestingly, mutations to erlin2 also cause neurodegenerative diseases (44 -47). Erlin2 is the dominant partner in the erlin1/2 complex, to which RNF170 is constitutively associated (11,12,15), and the erlin1/2 complex mediates the interaction of RNF170 with IP 3 Rs (15). Clearly, defining the mechanisms by which mutations to the erlin1/2 complex-RNF170 axis cause neurodegeneration will be fascinating topics for future study.