The CaVβ Subunit Protects the I-II Loop of the Voltage-gated Calcium Channel CaV2.2 from Proteasomal Degradation but Not Oligoubiquitination

CaVβ subunits interact with the voltage-gated calcium channel CaV2.2 on a site in the intracellular loop between domains I and II (the I-II loop). This interaction influences the biophysical properties of the channel and leads to an increase in its trafficking to the plasma membrane. We have shown previously that a mutant CaV2.2 channel that is unable to bind CaVβ subunits (CaV2.2 W391A) was rapidly degraded (Waithe, D., Ferron, L., Page, K. M., Chaggar, K., and Dolphin, A. C. (2011) J. Biol. Chem. 286, 9598–9611). Here we show that, in the absence of CaVβ subunits, a construct consisting of the I-II loop of CaV2.2 was directly ubiquitinated and degraded by the proteasome system. Ubiquitination could be prevented by mutation of all 12 lysine residues in the I-II loop to arginines. Including a palmitoylation motif at the N terminus of CaV2.2 I-II loop was insufficient to target it to the plasma membrane in the absence of CaVβ subunits even when proteasomal degradation was inhibited with MG132 or ubiquitination was prevented by the lysine-to-arginine mutations. In the presence of CaVβ subunit, the palmitoylated CaV2.2 I-II loop was protected from degradation, although oligoubiquitination could still occur, and was efficiently trafficked to the plasma membrane. We propose that targeting to the plasma membrane requires a conformational change in the I-II loop that is induced by binding of the CaVβ subunit.

tion and increasing the probability of opening (5), and second, they increase the trafficking of the channels to the plasma membrane (6). The binding site for Ca V ␤ subunits on Ca V ␣1 was shown to be a highly conserved region of the intracellular loop between domains I and II (I-II loop) of high voltage-activated channels that became known as the ␣-interaction domain (AID) (7). This conserved motif, consisting of 18 amino acids starting 24 residues after the end of the IS6 transmembrane segment of Ca V 2.1, was found to bind to all four Ca V ␤ subunits (8).
The mechanism by which the Ca V ␤ subunit promotes trafficking of the Ca V ␣1 to the plasma membrane is under debate. Apart from Ca V ␤2, which can associate with the plasma membrane either through palmitoylation (9) or by a polybasic region in the case of Ca V ␤2e (10), the Ca V ␤ subunits are cytoplasmic proteins. It was first suggested that the Ca V ␣1 I-II loop contained an ER retention motif that was masked in the presence of a Ca V ␤ (11), but more recently it has been postulated that the I-II loops contain an acidic ER export signal (12). Fusion proteins containing the Ca V ␣1 I-II loops fused to a transmembrane CD4 sequence were not retained in the ER (13), and evidence suggests that binding of the Ca V ␤ subunits to the Ca V 1.2 and Ca V 2.2 I-II loop prevented ubiquitination and degradation of these channels (1,13). Our previous work has shown that a mutant Ca V 2.2 channel that is unable to bind Ca V ␤ subunits (Ca V 2.2 W391A) was subjected to increased proteasomal degradation relative to the wild-type channel (1).
Efficient degradation by the proteasome requires the substrate to be ubiquitinated. Ubiquitin is a protein of 76 amino acids that is attached by a covalent bond to a lysine residue in the substrate; this then serves as a sorting signal (14). A single ubiquitin molecule can bind, resulting in monoubiquitination. More commonly, as ubiquitin contains seven lysine residues and can itself be ubiquitinated, polyubiquitination occurs. Different types of ubiquitination lead to different fates of the proteins; for example, Lys-48-linked ubiquitination is a signal to target the substrate for degradation via the proteasome (14).
In this study, we show that, in the absence of Ca V ␤ subunit, the isolated I-II loop of Ca V 2.2 is directly ubiquitinated on a lysine residue within the loop and is rapidly degraded. Binding of a Ca V ␤ subunit prevents degradation of the ubiquitinated I-II loop but does not prevent oligoubiquitination. Surprisingly, we found that attaching a palmitoylation motif to the N terminus of the I-II loop is insufficient to target it to the plasma membrane in the absence of the Ca V ␤ even when protein degradation is inhibited. This suggests that membrane targetting despite the presence of the palmitoylation sequence. Either the palm Ca V 2.2 I-II-GFP was being expressed at much lower levels in the absence of Ca V ␤1b, or more likely it was being degraded.
When expressed alone, Ca V ␤1b-mCherry was found uniformly distributed throughout the cytoplasm (Fig. 1B, bottom  left panel), but when co-expressed with palm Ca V 2.2 I-II-GFP, The panel on the right shows the merged image; DAPI was used to visualize the nucleus (blue). The scale bar is 20 m, and the grayscale calibration bar is shown below each set of images. B, box and whisker plots showing fluorescence intensity in the absence of MG132 (DMSO control; cyan bar; n ϭ 221 cells) or the presence of MG132 (yellow bar; n ϭ 314) of the cells represented in A. The box shows the interquartile range, whiskers range from 10 to 90%, the median is shown as a solid line, and the mean is shown as a dotted line. GFP fluorescence was measured in all cells that were expressing mCherry. Palm mCherry fluorescence (left) was slightly increased in the presence of MG132 (*, p ϭ 0.04, Student's t test; mean fluorescence ϮS.E. of 8977 Ϯ 463 arbitrary units (a.u.) in DMSO and 10,379 Ϯ 460 arbitrary units in MG132), whereas palm Ca V 2.2 I-II-GFP (right) fluorescence was significantly increased by MG132 (***, p Ͻ 0.0001, Student's t test; mean Ϯ S.E. of 291 Ϯ 40 (DMSO) and 962 Ϯ 63 arbitrary units (MG132)). C, as for A but for cells co-expressing palm Ca V 2.2 I-II-GFP with Ca V ␤1b-mCherry. D, box and whisker plots (as in B) showing fluorescence intensity of the cells in C in the absence (cyan; n ϭ 243 cells) or the presence (yellow; n ϭ 244) of MG132. Ca V ␤1b-mCherry (left) fluorescence was not significantly altered by MG132 (mean Ϯ S.E. of 5067 Ϯ 264 (DMSO) and 5595 Ϯ 329 arbitrary units (MG132); palm Ca V 2.2 I-II-GFP (right) fluorescence was slightly (*, p ϭ 0.02, Student's t test) decreased in the presence of MG132 (mean Ϯ S.E. of 2830 Ϯ 145 (DMSO) and 2317 Ϯ 156 arbitrary units (MG132)). E, control experiments showing the co-expression of Ca V ␤1b-mCherry and palm GFP as for A. F, box and whisker plots (as in B) showing fluorescence intensity of the cells in E; Ca V ␤1b-mCherry (left) or palm GFP (right) in the absence (cyan; n ϭ 86 cells) or presence (yellow; n ϭ 69) of MG132. Ca V ␤1b-mCherry fluorescence was not significantly different (mean Ϯ S.E. of 11,200 Ϯ 940 (DMSO) and 13,190 Ϯ 1160 arbitrary units (MG132)); palm GFP fluorescence was slightly (*, p ϭ 0.03, Student's t test) decreased in the presence of MG132 (mean Ϯ S.E. of 13,410 Ϯ 1019 (DMSO) and 10,340 Ϯ 871 arbitrary units (MG132)). ns, non-significant.
it too was found mainly at the plasma membrane (Fig. 1B, top  left panel). Surprisingly, the palmitoylation motif alone was also insufficient to target mCherry (Fig. 1B, middle row) or GFP (Fig.  1B, bottom row) to the plasma membrane; trafficking to the plasma membrane required both palm Ca V 2.2 I-II and the Ca V ␤ subunit (Fig. 1B).
Whether protein degradation was responsible for reduced palm Ca V 2.2 I-II-GFP expression in the absence of Ca V ␤ was  5) of Ca V ␤1b. B, quantification of band intensities expressed as a scatter plot for palm Ca V 2.2 I-II-HA from Western blots of WCLs, including the example shown in A. Analysis was performed using ImageJ on gels from three separate transfections. The intensity of each band of palm Ca V 2.2 I-II-HA was measured and expressed as a percentage of total palm Ca V 2.2 I-II-HA band intensities on each gel. Mean band intensities (shown by solid lines) show that, in the absence of MG132 (A, lanes 2 and 3), Ca V ␤1b causes a significant increase in the amount of palm Ca V 2.2 I-II-HA (10.22 Ϯ 2.20, black symbols in the absence of Ca V ␤; 28.28 Ϯ 1.27, red symbols in the presence of Ca V ␤1b; **, p ϭ 0.002, Student's t test). In the presence of MG132 (A, lanes 4 and 5), Ca V ␤1b also causes a significant increase in the amount of palm Ca V 2.2 I-II-HA (25.23 Ϯ 0.565, blue symbols in the absence of Ca V ␤; 36.27 Ϯ 1.38, pink symbols in the presence of Ca V ␤1b; **, p ϭ 0.002, Student's t test). C, Western blot showing tsA-201 cells immunoprecipitated (IP) with rabbit anti-HA antibody and immunoblotted (IB) with rat anti-HA antibody. Lanes are the same as those in A. A ladder of ubiquitinated I-II loop proteins (indicated by arrows) is detected. D, quantification of HA-stained Western blots showing the I-II loop bound to one to four ubiquitin molecules as a percentage of total HA protein in each lane. Analysis was performed using ImageJ on gels from three to five separate transfections and includes the example shown in C. Band intensities for palm Ca V 2.2 I-II-HA bound to one to four ubiquitins were combined and expressed as a percentage of total HA proteins per lane. Mean values are represented by solid lines. In the presence of Ca V ␤1b, ubiquitinated products make up a larger percentage of HA-detected proteins (mean Ϯ S.E. of 41.38 Ϯ 1.71%, red symbols, corresponding to lane 3 of C compared with 18.12 Ϯ 1.34% in the absence of Ca V ␤1b, black symbols, corresponding to lane 2; ***, p ϭ 0.0002, Student's t test). In the presence of MG132 (lanes 4 and 5), Ca V ␤1b has no significant effect on the percentage of ubiquitinated products detected with the HA antibody (29.28 Ϯ 3.23% in the absence of Ca V ␤1b, blue symbols compared with 32.88 Ϯ 2.73% in the presence of Ca V ␤1b, pink symbols). E, Western blot showing tsA-201 cells immunoprecipitated with rabbit anti-HA antibody and immunoblotted with anti-Ub antibody for the same samples as those shown in A and C. The ladder of ubiquitinated products detected with the anti-HA antibody (shown in C) is also detected with anti-Ub (indicated with arrows) along with a smear of polyubiquitinated proteins of high molecular mass. Nonspecific IgG bands detected in all lanes are marked with ‡.  3) and presence (lanes 2 and 4) of Ca V ␤1b. GFP-Ub competes with endogenous ubiquitin to bind to palm Ca V 2.2 I-II-HA to give larger products (arrows for ϩ1ϫ or ϩ2ϫ GFP-Ub; also marked with asterisks in bottom blot). ‡ shows nonspecific IgG bands. ns, non-significant. investigated by measuring the levels of fluorescence in the absence and presence of the proteasomal inhibitor MG132 (Fig.  2, A and B). In the presence of MG132, the levels of palm Ca V 2.2 I-II-GFP were significantly increased (Fig. 2, A and B), suggesting that it is usually rapidly degraded by the proteasome. However, it was still not associated with the plasma membrane.
In the presence of Ca V ␤1b tagged with mCherry, palm Ca V 2.2 I-II-GFP was expressed at increased levels and was trafficked to the plasma membrane (Fig. 2C, middle panel). There was no increase in fluorescence in the presence of MG132 (Fig.  2D), suggesting that Ca V ␤ protects palm Ca V 2.2 I-II-GFP from being degraded. The same result was found using untagged Ca V ␤1b (data not shown). Control experiments showed that expression of either Ca V ␤1b-mCherry or palm GFP alone was not increased by MG132 (Fig. 2, E and F).
Evidence for Multiple Stages of Ubiquitination of the Ca V 2.2 I-II Loop-To investigate the ubiquitination of the Ca V 2.2 I-II loop in more detail, palmitoylated I-II loop constructs were tagged with hemagglutinin (HA) at the C terminus (palm Ca V 2.2 I-II-HA) and expressed in tsA-201 cells in the presence or absence of Ca V ␤1b as well as in the presence or absence of the proteasomal inhibitor MG132. The anti-HA antibody identified palm Ca V 2.2 I-II-HA from whole cell lysates (Fig. 3A). Quantification showed that there was significantly less palm Ca V 2.2 I-II-HA in the absence of Ca V ␤ than in the presence of Ca V ␤1b (Fig. 3B) in both the absence of MG132 (Fig. 3B, compare columns 1 and 2, which correspond to Fig. 3A, lanes 2 and 3) and the presence of MG132 (Fig. 3B, columns 3 and 4). Ca V ␤, therefore, is likely to protect palm Ca V 2.2 I-II-HA from being degraded, although an effect on expression cannot be ruled out.
Palmitoylated Ca V 2.2 I-II-HA proteins were immunoprecipitated from lysates with anti-HA antibody, and Western blotting analysis with both anti-HA and anti-ubiquitin antibodies was carried out. In this experiment, the protein A-Sepharose beads used for the pulldown were limiting; an excess of lysates was added in an attempt to immunoprecipitate equal amounts of palm Ca V 2.2 I-II loop in the different conditions. Even so, there was less HA-tagged protein present in the absence of both Ca V ␤ and MG132 (Fig. 3C, lane 2 compared with lanes [3][4][5]. In the presence of Ca V ␤1b, palm Ca V 2.2 I-II-HA was found at increased levels (Fig. 3C, lane 3). The anti-HA antibody identified the palm Ca V 2.2 I-II-HA at the correct size (18 kDa) but also revealed a ladder of bands at higher molecular mass (Fig.  3C). Ubiquitin has a molecular mass of ϳ8 kDa. The ladder of bands corresponds to the size expected if one or more endogenous ubiquitins were bound to the Ca V 2.2 I-II loop. Quantification of band intensity for ubiquitinated I-II loops expressed as a percentage of total protein per lane for HA-stained Western blots showed that the ubiquitinated palm Ca V 2.2 I-II-HA represented a significantly lower proportion of total I-II loop protein in the absence of Ca V ␤ than in the presence of Ca V ␤1b (Fig. 3D, columns 1 and 2). This difference was lost in the presence of MG132 (Fig. 3D, columns 3 and 4), suggesting that, in the absence of both Ca V ␤ and MG132, ubiquitinated products are rapidly degraded. However, Ca V ␤1b does not appear to prevent the addition of one to four ubiquitins to the I-II loop (Fig.  3C, lanes 3 and 5).
The ladder of ubiquitinated I-II loop products was also observed using the anti-ubiquitin antibody (Fig. 3E), further evidence that these bands represent ubiquitinated I-II loop. The addition of one to four ubiquitins to palm Ca V 2.2 I-II-HA was more evident in the presence of Ca V ␤1b (Fig. 3, C and E, lanes 3 and 5) than in its absence (lanes 2 and 4). It appears that Ca V ␤1b protects against degradation of these oligoubiquitinated products as well as or better than MG132.
There was no evidence of high molecular mass polyubiquitination in the absence of MG132 and Ca V ␤ (Fig. 3E, lane 2), suggesting that, in the absence of the proteasomal inhibitor, the polyubiquitinated Ca V 2.2 I-II loop was rapidly degraded by the proteasome. In contrast, in the presence of MG132 but the absence of Ca V ␤, palmitoylated Ca V 2.2 I-II loop degradation was prevented, and there was an accumulation of high molecular mass polyubiquitinated proteins as seen by the smear above 76 kDa observed with the anti-ubiquitin antibody (Fig.  3E, lane 4). This smear was reduced in the presence of Ca V ␤1b (Fig. 3E, lane 5), suggesting that Ca V ␤1b protects the Ca V 2.2 I-II loop from polyubiquitination as well as degradation. The Ca V ␤1b, therefore, appears to allow the addition of one to four ubiquitins to the Ca V 2.2 I-II but prevents the polyubiquitination that usually leads to degradation.
Further evidence that the ladder of bands shown in Fig. 3, C and E, represents mono-or oligoubiquitinated products was obtained by including GFP-tagged ubiquitin in the transfection mixture. This competed with endogenous ubiquitin, reducing the density of bands at lower molecular mass (26 and 34 kDa) but producing bands at higher molecular mass (55 and 92 kDa) corresponding to the binding of one or two larger GFP-tagged ubiquitins (Fig. 3F). Addition of GFP-tagged ubiquitin was observed both in the presence and absence of Ca V ␤1b (Fig. 3F). The greater intensity of bands representing GFP-ubiquitin bound to I-II loop in the presence of Ca V ␤1b (Fig. 3F, lane 4 compared with lane 3) suggests that there is less degradation when Ca V ␤1b is present. The smear of polyubiquitin-conjugated proteins visualized by the anti-ubiquitin antibody was much stronger in the presence of the highly expressed GFPubiquitin (Fig. 3F, bottom panel, compared with Fig. 3E). The anti-ubiquitin antibody binds to polyubiquitinated proteins much more efficiently than to monoubiquitinated conjugates (16).

Ubiquitination Occurs on Lysine Residues within the Ca V 2.2 I-II Loop-Ubiquitination involves the formation of a covalent
bond between the C terminus of ubiquitin and the ⑀-amino of a lysine residue on the substrate (17). The palmitoylated Ca V 2.2 I-II-HA construct contains 12 lysines (Fig. 4A, *). To determine whether ubiquitination occurs directly on the I-II loop, all 12 of the lysine residues were mutated to arginines (palm Ca V 2.2 I-II K-R), and confocal and co-immunoprecipitation experiments were performed on the mutated constructs expressed in tsA-201 cells in the presence or absence of Ca V ␤1b subunits.
Confocal imaging experiments showed that, when expressed alone in tsA-201 cells, the palmitoylated Ca V 2.2 I-II K-R mutant tagged with GFP (palm Ca V 2.2 I-II K-R-GFP) was either expressed at higher levels or was less degraded than the wildtype I-II loop ( Fig. 4B compared with wild type in Fig. 2A). However, it was not trafficked to the plasma membrane in the absence of Ca V ␤ subunit, again suggesting that the palmitoylation motif alone is not sufficient for trafficking the I-II loop to the plasma membrane. In the presence of MG132, the K-R mutant I-II loop still accumulated in aggregates (Fig. 4B) like the wild-type I-II loop, indicating that the mutant is also misfolded in the absence of Ca V ␤. Unlike the wild-type, however, palm Ca V 2.2 I-II K-R-GFP fluorescence was not increased by MG132 (Fig. 4C), suggesting that the K-R mutant undergoes less proteasomal degradation than wild type. In the presence of Ca V ␤1b, both the palmitoylated Ca V 2.2 I-II K-R-GFP and the mCherry-tagged Ca V ␤1b were found at the plasma membrane ( Fig. 4D), indicating that Ca V ␤ was still able to interact with the I-II loop when all lysine residues were mutated to arginines. Neither palm Ca V 2.2 I-II K-R-GFP nor Ca V ␤1b-mCherry fluorescence was increased in the presence of MG132 (Fig. 4E). The fluorescence intensity of palm Ca V 2.2 I-II K-R-GFP when expressed without Ca V ␤1b was significantly higher than that of wild-type Ca V 2.2 I-II-GFP (Fig. 4F), whereas in the presence of The panel on the right shows the merged image; DAPI was used to visualize the nucleus (blue). The scale bar is 20 m, and the grayscale calibration bar is shown below. These images were taken from the same experiment using the same settings as those shown in Fig In the absence of Ca V ␤1b (F), mean free mCherry fluorescence (left) was not significantly different between WT (1.74 Ϯ 0.07) and the K-R mutant (1.68 Ϯ 0.06), whereas GFP fluorescence (right) was significantly lower for WT (0.091 Ϯ 0.011; light blue; n ϭ 441 cells) than for K-R mutant (0.581 Ϯ 0.040; orange; n ϭ 552; ***, p Ͻ 0.0001, Student's t test). In the presence of Ca V ␤1b (G), mean Ca V ␤1b-mCherry fluorescence was slightly lower for the K-R mutant (0.912 Ϯ 0.027; n ϭ 534) than for WT (1.00 Ϯ 0.034; n ϭ 518; *, p ϭ 0.04, Student's t test), whereas GFP fluorescence was slightly higher (1.189 Ϯ 0.045 (K-R) and 1.00 Ϯ 0.040 (WT); **, p ϭ 0.002, Student's t test). ns, non-significant. SEPTEMBER 23, 2016 • VOLUME 291 • NUMBER 39

JOURNAL OF BIOLOGICAL CHEMISTRY 20407
Ca V ␤1b-mCherry, the difference in GFP fluorescence intensities was much reduced (Fig. 4G).
Western blotting of the palmitoylated Ca V 2.2 I-II K-R-HAtagged construct immunoprecipitated with anti-HA antibody showed that the direct binding of ubiquitin to the I-II loop was abolished by mutation of all lysine residues (Fig. 5, A and B). The ladder of bands corresponding to the size expected for the addition of one or more ubiquitins to the palmitoylated I-II loop (Fig. 5A, lane 3) was absent for both the anti-HA (Fig. 5A,  lanes 4 -7) and anti-ubiquitin antibodies (Fig. 5B).
The Ca V ␤ subunit is known to interact with the AID within the first half of the I-II loop (7). To identify which part of the I-II loop is involved in ubiquitination, two further K-R mutant HAtagged constructs were made: one in which the first four lysine residues were mutated to arginines, referred to as 5Ј K-R, and the second in which the last eight lysines were mutated to arginines, 3Ј K-R. Western blots of the palm Ca V 2.2 I-II K-R-HAtagged constructs immunoprecipitated with the anti-HA and blotted with anti-HA and anti-ubiquitin antibodies show that both constructs were able to bind ubiquitin (Fig. 5, C and D). This indicates either that ubiquitin binds to more than one lysine residue or that it is able to switch to another lysine if the preferred lysine is no longer available.
To test whether the ladder of bands in Figs. 3 and 5 represents monoubiquitination of multiple lysine residues or oligoubiquitination of a single residue, we used a GFP-tagged mutant ubiquitin construct in which all seven lysine residues were mutated to arginines, GFP-Ub KO (18). This mutant ubiquitin can no longer form polyubiquitinated chains but can still compete with endogenous ubiquitin to monoubiquitinate the substrate. When included in the transfections, a single GFP-Ub KO was found to bind to the palmitoylated Ca V 2.2 I-II loop (Fig. 6A, lanes 3 and 4; data quantified in Fig. 6B), whereas the ladder of multiple GFP-ubiquitin (Ub) moieties bound was absent (Fig. 6,  A and B, compare lanes 4 and lanes 6). Taken together, these data suggest that only a single lysine residue on the I-II loop is oligoubiquitinated sequentially, but if the preferred lysine has been mutated to arginine, this oligoubiquitination can occur on another available lysine.
Ca V ␤1b Does Not Protect Ca V 2.2 I-II Loop from Oligoubiquitination-Ca V ␤1b protected palm Ca V 2.2 I-II-GFP from being degraded (Fig. 2, A-D). Western blotting, however, showed that palm Ca V 2.2 I-II-HA was ubiquitinated even when co-transfected with Ca V ␤1b (Fig. 3, C and E). Although ubiquitination is usually the initial step on the proteasomal degradation pathway, mono-or oligoubiquitination can also lead to different outcomes for the protein (14). We therefore asked whether the I-II loop that was interacting with the Ca V ␤1b was protected from ubiquitination or was also ubiquitinated. To do this, we co-expressed palm Ca V 2.2 I-II-HA with either GFPtagged Ca V ␤1b (Ca V ␤1b-GFP) or with GFP (without Ca V ␤) as a control in tsA-201 cells, immunoprecipitated the Ca V ␤ with anti-GFP antibody, and immunoblotted with anti-HA and anti-Ub antibodies.
Ca V ␤1b-GFP was able to co-immunoprecipitate non-ubiquitinated palm Ca V 2.2 I-II-HA (Fig. 6C, lanes 4 and 5, shown with an arrow). In addition, the anti-HA antibody also detected a ladder of bands likely to represent ubiquitinated I-II loop products (Fig. 6C, top, lanes 4 and 5) that were also detected using the anti-Ub antibody (Fig. 6C, bottom, marked with asterisks). Both non-ubiquitinated and ubiquitinated palm Ca V 2.2 I-II-HA products immunoprecipitated directly with the anti-HA antibody are shown in lanes 6 -9 for comparison. These results, quantified in Fig. 5D, could indicate that the I-II loop does not need to be ubiquitinated to interact with Ca V ␤ subunits but that Ca V ␤ is able to interact with both ubiquitinated and non-ubiquitinated I-II loop. Alternatively, ubiquitin may interact with the Ca V 2.2 I-II after it has bound to Ca V ␤.
Mutation of all Lysine Residues in the I-II Loop Reduces Degradation of Full-length Ca V 2.2 Channels-To determine whether ubiquitination of the I-II loop plays a role in degradation of the full-length channel, the 12 lysine residues in the I-II loop of Ca V 2.2 (Fig. 4A) were mutated to arginines. The wildtype and mutated channels were expressed in tsA-201 cells and examined by confocal imaging. In this experiment, the Ca V 2.2 ␣1 subunits contained an extracellular HA tag (6), and GFP was fused to the N terminus to give GFP-Ca V 2.2-HA (wild type (WT)) or GFP-Ca V 2.2 K-R-HA (containing 12 lysine-to-arginine mutations). Full-length channels were expressed with auxiliary Ca V ␣ 2 ␦-1 subunits in the presence or absence of MG132 and the presence or absence of Ca V ␤1b-mCherry (Fig. 7, A and C). As with the palm Ca V 2.2 I-II loop, MG132 caused a significant increase in the total GFP-Ca V 2.2-HA fluorescence measured in the absence of Ca V ␤1b (Fig. 7, A and B), whereas MG132 had no effect on the GFP fluorescence of GFP-Ca V 2.2 K-R-HA (Fig. 7, C and D). In the absence of Ca V ␤1b, total GFP fluorescence intensity was significantly lower for GFP-Ca V 2.2-HA WT than for GFP-Ca V 2.2 K-R-HA (Fig. 7E). In contrast, in the presence of Ca V ␤1b, GFP fluorescence intensities were not significantly different between the two conditions (Fig. 7F). This suggests that the K-R mutations within the I-II loop protect the full-length channel from degradation in the absence of Ca V ␤.
We next wanted to determine whether mutation of the lysines in the I-II loop had any effect on trafficking of the fulllength channel. Full-length GFP-Ca V 2.2-HA (either WT or K-R mutant) was expressed in tsA-201 cells together with Ca V ␤1b and Ca V ␣ 2 ␦-1. In non-permeabilized conditions, an anti-HA antibody was used to label the channel in the proximity of the plasma membrane (Fig. 8A). HA labeling was measured on a line (width of 10 pixels) drawn around the cell, and this was compared with the intracellular GFP fluorescence of the same construct measured within the cell and excluding that at the plasma membrane. Although intracellular GFP fluorescence was slightly higher for GFP-Ca V 2.2 K-R-HA than for the WT channel in this experiment, the HA fluorescence intensities at the plasma membrane were unchanged (Fig. 8B), indicating that the 12 mutations in the I-II loop have little effect on Ca V 2.2 trafficking.
Electrophysiological examination of Ca V 2.2-HA channels expressed in tsA-201 cells together with Ca V ␤1b-GFP and Ca V ␣ 2 ␦-1 showed that the K-R mutant produced a decrease in current density compared with the WT channel (Fig. 8, C and  D). The reduction in current density attributed to the K-R mutation was paralleled by a reduction in whole cell conductance (G max ) through the mutated channel (Fig. 8D, left). The K-R mutant also had a significantly depolarized V 50, act (Fig. 8D, right) compared with WT Ca V 2.2. As a hyperpolarization in V 50, act is observed in the presence of the Ca V ␤1b (19), the effect of the K-R mutation on Ca V 2.2 activation kinetics plausibly reflects a perturbed interaction between the Ca V ␣1 and Ca V ␤ subunits.

Discussion
It is well established that the Ca V ␤ subunit binds to the Ca V ␣1 channel via the AID in the intracellular I-II loop (7) and that this interaction is required for trafficking the Ca V ␣1 subunit to the plasma membrane (4). Previously we have shown that a mutant Ca V 2.2 channel that is unable to bind Ca V ␤ subunits (Ca V 2.2 W391A (19)) was rapidly degraded (1). In the present study, we have shown that expression of a construct containing the isolated I-II loop of Ca V 2.2 with a palmitoylation motif was also rapidly degraded when expressed alone but was efficiently trafficked to the plasma membrane in the presence of the Ca V ␤1b subunit.
The I-II loops of Ca V 1.3 (20) and other L-type Ca V ␣1 subunits (21) have been shown to be targeted to the plasma membrane even in the absence of Ca V ␤ subunits. Trafficking required the presence of a polybasic plasma membrane binding motif consisting of four arginine residues in the distal portion of the Ca V 1.2 I-II loop (21). In contrast, our experiments show that the palmitoylated I-II loop of Ca V 2.2 is not targeted to the plasma membrane unless Ca V ␤ is present despite the fact that it too has similar clusters of basic amino acids in its distal portion (Fig. 4A).
To target the I-II loop to the plasma membrane, we therefore included a palmitoylation motif derived from the G protein G␣ q . G␣ q subunits contain two cysteines close to the N termi-  nus, and palmitoylation of both residues leads to greater stability at the plasma membrane (22). G protein ␣ subunits follow a two-signal model for targeting to the plasma membrane (23). For non-myristoylated G proteins, such as G␣ q , an ␣-helical polybasic motif at the N terminus provides the initial signal (24,25), and this signal, in conjunction with an association with the G protein ␤␥ subunit (G␤␥) (26), allows targeting to the plasma membrane and subsequent palmitoylation (25). Tethering of the Ca V ␤2e subunit to the plasma membrane also requires a polybasic motif at the N terminus (10). However, although the I-II loop of Ca V 2.2 contains multiple basic amino acids at the N terminus (Fig. 4A), this region alone is insufficient to target it to the plasma membrane, even when it is palmitoylated, in the absence of Ca V ␤ subunits.  F). In the absence of Ca V ␤1b (E), mCherry fluorescence (left) was not significantly different between WT (light blue; n ϭ 1678 cells) and the K-R mutant (orange; n ϭ 1046) with mean normalized fluorescence values ϮS.E. of 1.138 Ϯ 0.033 and 1.064 Ϯ 0.037, whereas normalized GFP fluorescence (right) was significantly lower for WT than the K-R mutant (***, p Ͻ 0.0001, Student's t test) with mean values of 0.402 Ϯ 0.032 and 1.132 Ϯ 0.069. In the presence of Ca V ␤1b (F), although Ca V ␤1b-mCherry expression was slightly lower for K-R mutant (orange; n ϭ 734; mean, 0.841 Ϯ 0.033) than WT (light blue; n ϭ 738; mean, 1.00 Ϯ 0.039; **, p ϭ 0.0018, Student's t test), GFP fluorescence was not significantly different (mean values, 1.00 Ϯ 0.065 for WT and 0.973 Ϯ 0.065 for K-R mutant). ns, non-significant.
Binding of Ca V ␤ to the AID is thought to induce a dramatic change in secondary structure of the I-II loop (27). In the absence of Ca V ␤, circular dichroism (CD) measurements showed that the AID exists as a random coil (28), whereas in the presence of Ca V ␤, the AID was shown to have an ␣-helical structure (28,29). Secondary structure predictions suggested that the entire linker between the last transmembrane domain of domain I (IS6) and the AID could form a continuous ␣-helix (28), and CD measurements showed that this was the case (30). The helical content was found to be significantly higher for this region of Ca V 2.2 than that of Ca V 1.2 (31). It appears that Ca V ␤ is able to interact with an unfolded I-II loop via its AID, inducing a conformational change and extending the ␣-helical structure along the I-II linker toward the IS6 (28). The N terminus of the I-II loop contains a number of basic amino acids (see Fig.  4A). Fig. 9 shows two helical wheel diagrams of the region of the I-II linker upstream of the AID (left) and including the AID (right); formation of an ␣-helix would allow two clusters of basic amino acids to align along one side and form a polybasic region. This may act as a trafficking signal to direct the construct to the plasma membrane and promote stable dipalmitoylation. Interestingly, the Trp-391 residue, which is critical for the interaction between the I-II loop and the Ca V ␤ subunit, is located along the same face of the helix although beyond the polybasic regions (Fig. 9).
Alternatively, formation of an ␣-helix may allow another binding partner to interact with the I-II loop and direct it to the plasma membrane. The I-II linkers of Ca V 2.1 (32) and Ca V 2.2 (33) were found to interact directly with G␤␥. We have previously shown that Ca V ␤ and G␤␥ subunits bind at the same time and that Ca V ␤ must be present for the G␤␥ subunit to induce voltage-dependent modulation of Ca V 2.2 (5). Binding of G␤␥ subunits was found to be dependent on the formation of a rigid IS6-AID linker induced by the binding of Ca V ␤ (34). The G protein G␣ q has been shown to require an association with G␤␥ before it can be palmitoylated and stabilized at the plasma membrane (26). To examine whether G␤␥ subunits were involved in the trafficking of the I-II loop constructs, we included a minigene derived from the C terminus of ␤-adrenergic receptor kinase (5) to bind and remove free G␤␥ subunits. Confocal experiments showed that transfection of ␤-adrenergic receptor kinase had no effect on the trafficking of palm Ca V 2.2 I-II in the presence of Ca V ␤1b, 4 providing no evidence that G␤␥ binding is involved.
In the absence of Ca V ␤, the I-II loop remains in a misfolded state. The Ca V ␤ subunit is able to act as a molecular chaperone to induce correct folding of the I-II loop of the Ca V ␣1 subunit. Misfolded proteins are more likely to aggregate as they have exposed hydrophobic domains, and a key role of the chaperone is to interact co-translationally as the nascent protein emerges from the ribosome and prevent misfolding and aggregation (35). In the absence of the chaperone, misfolded proteins are rapidly targeted to the ubiquitin-proteasome system for degradation.
Substrates for the ubiquitin-proteasome system are tagged with ubiquitin; addition of ubiquitin moieties is thought to occur sequentially, one at a time, with the rate-limiting step being the addition of the first ubiquitin (36). It has previously been suggested that a chain of four ubiquitins is the minimum signal for targeting the substrate to the proteasome (37), but more recently it has been shown that only a single ubiquitin is needed for small proteins (20 -150 amino acids) (38). Ubiquitination also has other functions, however, such as protein trafficking (14), endocytosis, and signaling (39), which are independent of the proteasome. Our data show that the palmitoylated I-II loop carries up to four ubiquitins (Fig. 3) even in the presence of the Ca V ␤ subunit and the proteasomal inhibitor MG132; conditions in which the protein is not degraded. Relevant to this, a mass spectrometry study has shown that in almost 50% of cases, ubiquitination of proteins has a non-proteasomal function (40).
In the absence of Ca V ␤, the I-II loop was rapidly degraded as shown in Fig. 2. It has been shown that nascent proteins can be ubiquitinated co-translationally and targeted for degradation before translation has finished (41). Ubiquitination does not necessarily lead to degradation of newly synthesized ubiquitinated proteins; if correct folding does occur with the help of a chaperone, the ubiquitins may be removed by the action of deubiquitinating enzymes (42). In this regard, binding of the Ca V ␤ subunit to the ubiquitinated I-II loop clearly rescues it from degradation, although our evidence does not indicate it promotes deubiquitination. By contrast, it is possible that oligoubiquitination is functionally important for the effect of Ca V ␤ because the full-length Ca V 2.2 K-R mutant showed reduced functional expression, possibly reflecting reduced interaction with the Ca V ␤ subunit.
Mutation of the 12 lysine residues in the I-II loop of the full-length Ca V 2.2 channel protected the channel from degradation. The full-length K-R mutant channel still contains multiple internal lysine residues that were not mutated and were therefore available as ubiquitination sites. However, the fact that the K-R channel was less degraded than WT shows that the 12 lysines within the I-II loop have a critical role in ubiquitination and degradation.
Trafficking of the palmitoylated Ca V 2.2 I-II loop to the plasma membrane required the presence of the palmitoylation signal, the I-II loop, and the Ca V ␤ subunit. Although the palmitoylated K-R mutant I-II loop was no longer ubiquitinated and degraded, it also was not trafficked to the plasma membrane in the absence of the Ca V ␤ subunit; trafficking was restored when Ca V ␤ was also co-expressed. It is possible that correct folding, which is only achieved in the presence of the Ca V ␤ subunit, exposes a trafficking signal. Typical trafficking signals have been identified as short linear amino acid sequences, but it has been shown that export of the Kir2.1 channel from the Golgi to the plasma membrane requires the correctly folded N and C termini of the channel (43). The trafficking signal was only exposed when two separate domains were folded correctly in the tertiary structure of the channel.
In conclusion, our experiments indicate that, in the absence of Ca V ␤ subunit, the I-II loop of Ca V 2.2 is directly ubiquitinated and rapidly degraded by the proteasome system. When Ca V ␤ is present, it is able to bind to the I-II loop even if the loop is already ubiquitinated. Binding appears to induce a conforma- . GFP fluorescence at the plasma membrane (overlapping with membrane HA labeling) was excluded from the analysis to give intracellular GFP rather than total GFP. Data have been taken from three separate experiments involving three transfections. Although intracellular GFP fluorescence may be slightly increased for the K-R mutant compared with WT (mean fluorescent values ϮS.E. of 1.000 Ϯ 0.069 for WT and 1.263 Ϯ 0.092 for K-R; *, p ϭ 0.022, Student's t test), extracellular HA staining is unchanged (mean values, 1.000 Ϯ 0.025 for WT and 1.003 Ϯ 0.029 for K-R). C, current-voltage relationships of full-length WT and K-R mutant Ca V 2.2 channels co-expressed with Ca V ␤1b-GFP and Ca V ␣ 2 ␦-1 in tsA-201 cells. Top, example traces of WT Ca V 2.2/␤1b-GFP/␣ 2 ␦-1 (Ϫ40 to ϩ10 mV in 5-mV increments; black traces) and K-R Ca V 2.2/␤1b-GFP/␣ 2 ␦-1 currents (Ϫ40 to ϩ15 mV in 5-mV increments; red traces) recorded from a holding potential of Ϫ80 mV. The charge carrier was 1 mM Ba 2ϩ . Cells expressing wild-type or the K-R mutant channel had the same holding currents of Ϫ3 Ϯ 4 pA (n ϭ 12) and Ϫ3 Ϯ 4 pA (n ϭ 13), respectively. The scale bars refer to both traces. Bottom, mean current density-voltage relationships (pA/pF) were calculated from current amplitude (pA) measurements taken 15 ms into a 50-s depolarizing voltage pulse. Error bars represent S.E. I Ba at ϩ5 mV was 277 Ϯ 36 pA/pF (n ϭ 12) and 148 Ϯ 27 pA/pF (n ϭ 13) for WT and K-R Ca V 2.2, respectively (p ϭ 0.0081, Student's t test). D, characteristic properties of the current density-voltage relationships of WT Ca V 2.2 and the K-R mutant channel. Values were derived from modified Boltzmann fits of individual current density-voltage plots describing WT and K-R Ca V 2.2 currents. Left, G max was 7.7 Ϯ 0.9 nS/pF (mean (solid line) Ϯ S.E.; n ϭ 12) and 4.5 Ϯ 0.7 nS/pF (n ϭ 13) for WT and K-R Ca V 2.2, respectively (**, p ϭ 0.0079, Student's t test). Right, V 50 values were Ϫ4.7 Ϯ 1.6 (n ϭ 12) and 1.3 Ϯ 1.3 mV (n ϭ 13) for WT and K-R Ca V 2.2, respectively (**, p ϭ 0.0071, Student's t test). ns, non-significant. tional change and the formation of an ␣-helix in the I-II loop, thereby allowing the construct to associate with the plasma membrane.

Experimental Procedures
Molecular Biology-The calcium channels used were rabbit Ca V 2.2 (GenBank Accession number D14157), rat Ca V ␣ 2 ␦-1 (M86621), and rat Ca V ␤1b (X61394), all expressed in the pMT2 vector. The green fluorescent protein mut3bGFP (GFP) (44) or mCherry was fused to the C terminus of Ca V ␤1b. GFP-Ub and GFP-Ub KO (18) were obtained from Addgene. The first intracellular loop of Ca V 2.2 (amino acids 356 -483, beginning ESGEF . . . and ending . . . KAQ) had a palmitoylation sequence, MTLESIMACCL, added to the N terminus and an HA tag (TSYPYDVPDYA) added to the C terminus to give a construct termed palm Ca V 2.2 I-II-HA in this study. Further constructs were made where the HA tag was substituted with GFP or mCherry. The mutant palm Ca V 2.2 I-II K-R-HA was made by mutating all 12 lysine residues in the I-II loop to arginines, and two further constructs were made by mutating either the first four (palm Ca V 2.2 I-II 5Ј K-R-HA) or the last eight (palm Ca V 2.2 I-II 3Ј K-R-HA) lysines to arginines. The I-II loop containing the 12 lysine-to-arginine substitutions was inserted into the full-length Ca V 2.2, which also contained an extracellular HA tag (6) and GFP fused to the N terminus to give GFP-Ca V 2.2 K-R-HA. The sequences of all constructs were confirmed by DNA sequencing.
Cell Culture and Transfection-tsA-201 cells (European Collection of Authenticated Cell Cultures (ECACC)) were cultured in Dulbecco's modified Eagle's medium in the presence of 10% fetal bovine serum, penicillin, streptomycin, and 2% GlutaMAX (Invitrogen). For immunocytochemistry and co-immunoprecipitation experiments, transfections were carried out using PolyJet (SignaGen) according to the manufacturer's instructions using equal ratios of constructs unless otherwise stated. When included, the protease inhibitor MG132 (Calbiochem) was added at a concentration of 4 M 24 h after transfection, and the cells were harvested or fixed a further 16 -18 h afterward. At the concentration of MG132 used (4 M) and the time of incubation (16 -18 h), MG132 was found to be effective at inhibiting proteasomal degradation without being detrimental to cell survival. Total protein yields from transfections in the absence and presence of MG132 were measured (for Western blotting and co-immunoprecipitation analysis) and found to be similar. For electrophysiological experiments, transfections were performed using FuGENE 6 (Promega) with GFP-Ca V 2.2-HA, Ca V ␣ 2 ␦-1, and Ca V ␤1b subunits in a ratio of 3:2:2.
Immunocytochemistry-Cells were transfected on polylysine-coated coverslips and fixed with 4% paraformaldehyde in Tris-buffered saline (TBS; 20 mM Tris, 150 mM NaCl, pH 7.4) for 5 min. 4Ј,6-Diamidino-2Ј-phenylindole dihydrochloride (DAPI) was used to stain the nuclei. When used, the plasma membrane stain CM-DiI (ThermoFisher Scientific) was used at a dilution of 1:200 for 20 min at room temperature. Coverslips were mounted in VectaShield (Vector Laboratories). When anti-HA staining was used, cells were incubated with blocking buffer (20% goat serum, 4% BSA in TBS) for 1 h at room temperature before being incubated with rat anti-HA (Roche Applied Science) diluted 1:200 in 0.5ϫ blocking buffer at 4°C overnight. After washing, samples were incubated with secondary antibody anti-rat Alexa Fluor 594 at a dilution of 1:500 for 1 h at room temperature before being stained with DAPI and mounted. Samples were viewed on an LSM 780 confocal microscope (Zeiss) using a 63ϫ/1.4 numerical aperture oil immersion objective in 16-bit mode. The tile function (3 ϫ 3 tiles; each tile consisting of 1024 ϫ 1024 pixels) was used, and every transfected cell within the image was analyzed to remove collection bias. Confocal optical sections were 1 m, and acquisition settings were kept constant. Images were analyzed using NIH ImageJ. Images that were analyzed were not saturated.
Western Blotting Analysis-Transfected tsA-201 cells were harvested in phosphate-buffered saline (PBS) containing protease inhibitors (Complete tablet from Roche Applied Science). Cells were lysed by sonication for 10 s with 1% Igepal in PBS in the presence of protease inhibitors followed by incubation on ice for 30 min, and whole cell lysates (WCLs) were collected after centrifugation (14,000 ϫ g for 30 min at 4°C). Samples were incubated for 15 min at 55°C with 100 mM dithiothreitol and 2ϫ Laemmli sample buffer, and proteins were separated by SDS-PAGE on 4 -12% Bis-Tris gels and then transferred to polyvinylidene fluoride membranes. Membranes were incubated in blocking buffer (10 mM Tris, pH 7.4, 500 mM NaCl, 0.5% Igepal, 3% BSA) for 1 h followed by incubation with the primary antibody. The following primary antibodies were used: rat anti-HA (Roche Applied Science) at 1:1000 overnight, mouse anti-ubiquitin (P4D1, Santa Cruz Biotechnology) at 1:500 overnight, and mouse anti-GFP (Roche Applied Science) at 1:1000 for 1 h, all at 4°C. The appropriate secondary antibodies coupled to horseradish peroxidase were incubated at a dilution of 1:2000 for 1 h at room temperature. Proteins were detected using the enhanced ECL Plus reagent (GE Healthcare) on a Typhoon 9410 scanner (GE Healthcare). Quantification of Western blotting was carried out using the gel analysis tool of ImageJ. A rectangular section was placed over the entire lane, and each selected lane was plotted as a line profile. The peak FIGURE 9. ␣-Helical model of the 5 region of the Ca V 2.2 I-II loop. ␣-Helical wheel models of the 5Ј region of the Ca V 2.2 I-II loop after the IS6 transmembrane domain and immediately preceding the AID (left) and the subsequent sequence including the AID (right; in the same orientation) show two clusters of basic amino acids along one face of the helix. Basic residues are shown in blue, acidic residues are shown in red, and the remainder are shown in white. Residues involved in the interaction with Ca V ␤ subunits lie along the dotted line; Trp-391, Ile-392, and Tyr-388, which are critical for this interaction (29), are labeled with asterisks.
band intensities above background were measured. This allows the changes in background along the length of the lane to be taken into account and accurately subtracted.
Immunoprecipitation-The total protein concentration was determined (Bradford assay, Bio-Rad) for each sample, and 1 mg of WCL was precleared on 50 g of protein A-Sepharose (GE Healthcare) for 2 h at 4°C. Sepharose beads were discarded, and supernatants were incubated with rabbit anti-HA (Sigma) or rabbit anti-GFP (Clontech) at a dilution of 1:200 overnight at 4°C. Immunoprecipitated proteins were captured on 60 g of protein A-Sepharose beads for 2 h at 4°C. Beads were washed five times with PBS containing 0.1% Igepal and incubated for 15 min at 55°C with 100 mM dithiothreitol and 2ϫ Laemmli sample buffer. Eluted proteins were then resolved by SDS-PAGE.
Electrophysiology-Whole cell voltage clamp recordings were performed on tsA-201 cells at room temperature (20 -24°C). Single cells were clamped using an Axopatch 200B patch clamp amplifier (Axon instruments). Borosilicate glass patch pipettes were filled with a solution containing 140 mM cesium aspartate, 5 mM EGTA, 2 mM MgCl 2 , 0.1 mM CaCl 2 , 2 mM K 2 ATP, and 10 mM HEPES. CsOH was added to achieve pH 7.2. The external solution contained 150 mM tetraethylammonium bromide, 3 mM KCl, 1 mM NaHCO 3 , 1 mM MgCl 2 , 10 mM HEPES, 4 mM glucose, and 1 mM BaCl 2 . pH was adjusted to 7.4 with Tris base. Current density-voltage relationships were fitted with a modified Boltzmann equation as follows: I ϭ G max ϫ (V Ϫ V rev )/(1 ϩ exp(Ϫ(V Ϫ V 50, act )/k)) where I is the current density (in pA/pF), G max is the maximum conductance (in nS/pF), V rev is the reversal potential, V 50, act is the midpoint voltage for current activation, and k is the slope factor.
Author Contributions-A. C. D. and K. M. P. designed the experiments. K. M. P. made the cDNA constructs and performed all confocal microscopy, immunoprecipitation, and Western blotting experiments. K. M. P. analyzed data. S. W. R. designed, performed, and analyzed electrophysiology experiments. K. M. P. and A. C. D. wrote the manuscript.