Mass Spectrometric Analysis of Type 1 Inositol 1,4,5-Trisphosphate Receptor Ubiquitination*

Inositol 1,4,5-trisphosphate (IP3) receptors form tetrameric channels in endoplasmic reticulum membranes of mammalian cells and mediate IP3-induced calcium mobilization. In response to various extracellular stimuli that persistently elevate IP3 levels, IP3 receptors are also ubiquitinated and then degraded by the proteasome. Here, for endogenous type 1 IP3 receptor (IP3R1) activated by endogenous signaling pathways and processed by endogenous enzymes, we sought to determine the sites of ubiquitination and the composition of attached ubiquitin conjugates. Our findings are (i) that at least 11 of the 167 lysines in IP3R1 can be ubiquitinated and that these are clustered in the regulatory domain and are found in surface regions, (ii) that at least ∼40% of the IP3R1-associated ubiquitin is monoubiquitin, (iii) that both Lys48 and Lys63 linkages are abundant in attached ubiquitin chains, and (iv) that Lys63 linkages accumulate most rapidly. Additionally, we find that not all IP3R1 subunits in a tetramer are ubiquitinated and that nontetrameric IP3R1 complexes form as degradation proceeds, suggesting that ubiquitinated subunits may be selectively extracted and degraded. Overall, these data show that endogenous IP3R1 is tagged with an array of ubiquitin conjugates at multiple sites and that both IP3R1 ubiquitination and degradation are highly complex processes.

Inositol 1,4,5-trisphosphate (IP 3 ) 4 receptors (IP 3 Rs) are ϳ300-kDa endoplasmic reticulum (ER) membrane proteins that tetramerize to form Ca 2ϩ channels that are gated by the co-agonists IP 3 and Ca 2ϩ and that govern Ca 2ϩ release from the ER (1)(2)(3). There are three homologous mammalian IP 3 R types (termed IP 3 R1, IP 3 R2, and IP 3 R3) that can form either homo-or heterotetrameric channels. Each can be divided into an N-terminal ligand-binding domain, a large central regulatory domain that contains several modulatory sites, and a C-terminal channel domain that contains six membrane-spanning helices and the channel pore (see Fig. 1D). IP 3 R1, which is 2749 amino acids in length, is the most widely expressed and best studied of the three types (see Fig. 1D).
In response to activation of certain G protein-coupled receptors that persistently elevate IP 3 , IP 3 Rs are degraded via the ubiquitinproteasome pathway (4 -7). This phenomenon, termed "IP 3 R down-regulation," which has been described for all three IP 3 R types in various cell types and tissues (8,9), appears to protect cells from the deleterious effect of prolonged Ca 2ϩ mobilization (6). The ubiquitin-proteasome pathway is used by cells to maintain homeostasis by degrading both key proteins involved in important cellular processes (e.g. those that govern transcription and cell cycle transitions) and by allowing for ER-associated degradation (ERAD), a mechanism that accounts for the disposal of aberrant or unrequired ER proteins (e.g. misfolded proteins or unassembled subunits of multimeric protein complexes) (10). ERAD is currently being intensely studied and appears to rely on several key proteins, including Ubc7 and Hrd1 (10 -12), which ubiquitinate substrates, and p97, an AAA-ATPase that aids in the removal of substrates from the ER (10,13). It appears that IP 3 Rs are processed by ERAD, because both p97 and Ubc7 are involved in their degradation (14,15).
Ubiquitination refers to the process of conjugating ubiquitin, a 76-amino acid protein, to a target protein through the concerted action of E1, E2, and E3 enzymes (16). The E1 ubiquitinactivating enzyme first activates ubiquitin by forming a thiolester bond with the C-terminal glycine of ubiquitin. The activated ubiquitin moiety is then transferred to a ubiquitinconjugating enzyme (E2 or Ubc), again through a thiolester bond. A ubiquitin-protein ligase (E3) then recognizes the target protein and recruits a charged cognate E2 to its vicinity. Ubiquitin is then conjugated to the target protein via a covalent linkage between the C-terminal glycine of ubiquitin and an ⑀-amino group of a lysine residue. Remarkably, ubiquitin contains seven lysines (Lys 6 , Lys 11 , Lys 27 , Lys 29 , Lys 33 , Lys 48 , and Lys 63 ), which can also be selected for ubiquitination, allowing for the formation of ubiquitin chains. Thus, targeted proteins can be tagged with either monoubiquitin or ubiquitin chains. Although our knowledge of the abundance and role of the various possible chain linkages is far from complete, it is clear that ubiquitin chains formed through Lys 48 linkages signal for recognition and degradation by the proteasome, whereas Lys 63 -linked chains appear to be involved in diverse cellular processes, such as DNA damage response pathways, signal transduction through the NF-B pathway, and protein trafficking (17,18). Ubiquitin is removed from targeted proteins by deubiquitinating enzymes (DUBs), which serve to reverse the consequences of ubiquitination and/or to recycle ubiquitin (19,20).
Little is known about how IP 3 Rs are processed by ERAD, including which of the 167 lysines in IP 3 R1 are targeted for ubiquitin conjugation. Unlike most other post-translational modifications, there is no consensus sequence for ubiquitination, and it remains unclear how lysines are selected (21,22). Further, although it is clear that IP 3 Rs shift to much higher M r values upon stimulation and thus are modified by many ubiquitin moieties (6,7), it is not known whether IP 3 Rs are multiply monoubiquitinated or modified by ubiquitin chains, and if so, through what linkages. Recent advances in mass spectrometry now allow for the determination of both of the lysines occupied by ubiquitin (23) and the composition and quantity of ubiquitin chains associated with substrates (24). The latter relies on a novel technique termed absolute quantification (AQUA) (25) that uses isotopically labeled internal standard peptides to quantitate peptides generated by tryptic digestion. To date, several ubiquitin-AQUA studies have been performed on ubiquitinated substrates and have yielded some interesting results; e.g. that ϳ40% of the ubiquitin associated with overexpressed epidermal growth factor receptor contains Lys 63 linkages (26) and that purified cyclin B1 is first multiply monoubiquitinated and then chains are added at those ubiquitin moieties via Lys 48 , Lys 63 , and Lys 11 linkages (27). However, a caveat with these studies is that the substrates analyzed were either overexpressed or ubiquitinated in vitro and may not accurately reflect the situation in vivo. Nevertheless, these mass spectrometric techniques provide a powerful tool for identifying sites of ubiquitin conjugation and ubiquitin chain linkages.
In the present study, we sought to define the molecular details of IP 3 R1 ubiquitination by first determining the sites at which IP 3 R1 is ubiquitinated and then the composition of attached conjugates. For this work, we focused on ␣T3-1 mouse pituitary gonadotropes, in which gonadotropin-releasing hormone (GnRH) induces a robust stimulation of endogenous IP 3 R1 ubiquitination and degradation (7), and on Rat-1 fibroblasts, in which endothelin-1 has a similar effect (14). Mass spectrometric analysis of ubiquitinated IP 3 R1 identified 11 lysines that can be ubiquitinated and revealed that IP 3 R1 is conjugated with both monoubiquitin and Lys 48 -and Lys 63linked ubiquitin chains. We also show that not all IP 3 R1 subunits in a tetramer are ubiquitinated and provide evidence that individual subunits in a tetramer can be selectively degraded. Taken together, this indicates that IP 3 R1 ubiquitination and degradation are highly complex events.
Immunoprecipitation for Mass Spectrometric Analysis-␣T3-1 and Rat-1 cells were grown to confluence in 15-cm-diameter dishes, and 24 h prior to harvest, Rat-1 cells were cultured in serum-free medium. After stimulation, the cells were harvested by adding 3-6 ml/dish of ice-cold lysis buffer (50 mM Tris, 150 mM NaCl, 1% Triton X-100, 1 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride, 10 M pepstatin, 0.2 M soybean trypsin inhibitor, pH 8.0). The lysates were then incubated with 2.5 mM N-ethylmaleimide for 1 min to inhibit DUBs, quenched by adding 5 mM dithiothreitol, incubated at 4°C for 30 min with occasional mixing, and cleared by centrifugation (16,000 ϫ g for 10 min at 4°C). Supernatants were then incubated overnight at 4°C with protein A-Sepharose CL-4B beads and anti-IP 3 R1. Immune complexes were then isolated by centrifugation (1000 ϫ g for 1 min at 4°C), washed three to five times with 1 ml of ice-cold lysis buffer, resuspended in gel loading buffer, boiled for 5 min, subjected to SDS-PAGE, and either Coomassie Blue-stained (10% acetic acid, 45% ethanol, 2.5 g/liter Brilliant Blue R) followed by destaining (10% acetic acid, 45% ethanol) or transferred to nitrocellulose and immunoblotted and analyzed as previously described (7).
Ubiquitin Site Identification and AQUA Analysis-Regions corresponding to unmodified and/or ubiquitinated IP 3 R1 were excised from gels, destained (50 mM ammonium bicarbonate, 50% acetonitrile), and then incubated overnight at 37°C with 20 g/ml trypsin (Promega) diluted in 50 mM ammonium bicarbonate, 0.005% n-dodecyl ␤-D-maltoside. Following digestion, ubiquitin and/or IP 3 R1 isotope-labeled internal standard AQUA peptides were added to samples, and then peptides were extracted from the gel regions twice using 50% acetonitrile, 5% formic acid and dried in a Speedvac. Ubiquitin-AQUA peptides include diglycine signature peptides corresponding to the seven ubiquitin chain linkages, as well as a battery of unbranched tryptic peptides from ubiquitin as previously described (27,28). The IP 3 R1-AQUA peptides LEELGDQR (amino acids 538 -545) and LQDIVSALEDR (amino acids 1603-1613) were generated by Cell Signaling Technologies (Danvers, MA). Extracted peptides were desalted with Stage-Tip as described (29), dried, resuspended with 5% formic acid, 0.01% H 2 O 2 , and injected onto a 100-m-internal diameter fused silica column pulled to a tip and packed with Magic C18 material (Michrom Bioresources, Auburn, CA). Peptides injected onto the column were separated using a linear gradient of buffer B (acetonitrile, 0.125% formic acid) in buffer A (3% acetonitrile, 0.125% formic acid), and eluting peptides were analyzed using either an LTQ-FT or LTQ-Orbitrap mass spectrometer (Thermo, San Jose, CA) operating in data-dependent MS/MS (tandem mass spectrometry) mode. During each duty cycle, the mass spectrometer collected a high resolution pre-cursor ion scan, followed by MS/MS on the 10 most abundant precursor ions. For IP 3 R1 ubiquitination site identification, MS/MS spectra were searched using Sequest with a variable mass addition of 114.0429 Da on lysines to match the diglycine signature, and spectral matches corresponding to IP 3 R1 ubiquitination sites were validated based on the mass accuracy of the precursor ion and manual inspection of the MS/MS spectra. For AQUA analysis, the ion-extracted chromatogram was drawn with 10 ppm mass accuracy, and cell-derived IP 3 R1 and ubiquitin peptides were quantified from the ratio of the area under the curve for these peptides versus that for ubiquitin-or IP 3 R1-derived AQUA peptides. Throughout the AQUA analysis, conditions were optimized to avoid differences in the reference peptide recovery as described (25,27). This allowed for a Ͻ50% coefficient of variation for quantitation of total ubiquitin amount using the four ubiquitin-AQUA peptides, and a Ͻ20% coefficient of variation for quantitation of IP 3 R1 amount using the two IP 3 R1-AQUA peptides.
Immunoprecipitation from Microsomal Preparations-For immunoprecipitations involving anti-ubiquitin, we first prepared microsomes to remove free ubiquitin. Control or stimulated cells in 15-cm-diameter dishes were rinsed with ice-cold 155 mM NaCl, 10 mM HEPES, 1 mM EDTA, pH 7.4, were detached with 10 ml of 10 mM Tris, 1 mM EGTA, 0.2 mM phenylmethylsulfonyl fluoride, 10 m pepstatin, 0.2 M soybean trypsin inhibitor, pH 7.4), and were processed in a Dounce homogenizer (20 strokes in the presence of 2.5 mM N-ethylmaleimide and two additional strokes after the addition of 5 mM dithiothreitol), and the microsomes were isolated by centrifugation (38,000 ϫ g for 10 min at 4°C). They were then solubilized by incubation in 5 ml of lysis buffer for 30 min at 4°C and then centrifuged (16,000 ϫ g for 10 min at 4°C), and supernatants were incubated overnight at 4°C with either anti-IP 3 R1 or anti-ubiquitin and protein A-Sepharose CL-4B beads. Immune complexes were then processed for immunoblotting as previously described (7).
Blue Native PAGE-All of the reagents for blue native PAGE were obtained from Invitrogen, and the methods used are described in the Invitrogen NativePAGE TM Novex Bis-Tris gel system user manual. Briefly, ␣T3-1 cells in six-well Falcon plates were harvested with 200 l of ice-cold lysis buffer and after 30 min at 4°C were centrifuged (16,000 ϫ g for 10 min). 75 l of supernatants were mixed with 25 l of 4ϫ sample buffer and 5 l of 5% G-250 sample additive, and 20 g of protein for each sample was loaded into wells of a 3-12% NativePAGE TM Novex Bis-Tris gel and electrophoresed for 1 h at 150 V in dark cathode buffer and then for 45 min at 250 V in light cathode buffer. The gel was then washed for 10 min in blotting buffer (7) and transferred to polyvinylidene difluoride membrane, which was activated by washing in methanol. After transfer, the membrane was washed in 40% methanol, 10% acetic acid for 15 min, rinsed in water, and processed for immunoblotting as previously described (7). Duplicate aliquots of supernatants were also subjected to regular SDS-PAGE and processed for immunoblotting as described (7).

At Least 11 of the 167 Lysines in IP 3 R1 Can Be Ubiquitinated-
We have previously demonstrated that exposure of ␣T3-1 cells to GnRH results in the ubiquitination and degradation of IP 3 R1 (7), which quantitative immunoblotting (8) reveals, represents ϳ99% of the IP 3 R complement in these cells (data not shown). To further our understanding of the mechanisms of IP 3 R ubiquitination, we initially sought to identify the sites of ubiquitin conjugation by subjecting ubiquitinated IP 3 R1 to mass spectrometry. Samples were prepared from ␣T3-1 cells treated with 100 nM GnRH for 7 min, which causes maximal IP 3 R1 ubiquitination (7). In the absence of GnRH, immunopurified IP 3 R1 migrates as a single band at ϳ260 kDa (Fig. 1A, lane 1) and does not exhibit ubiquitin immunoreactivity (lane 2). In contrast, after exposure to GnRH, immunopurified IP 3 R1 is now strongly recognized by anti-ubiquitin, with immunoreactivity beginning at ϳ270 kDa and extending upward (lane 4), consistent with the addition of multiple ubiquitin moieties; the reason that IP 3 R1 immunoreactivity is only partially shifted upward (lane 3) is that only ϳ10% of IP 3 R1 is ubiquitinated under these conditions (7). For mass spectrometry, immunopurified IP 3 R1s were excised from analogous Coomassie Blue-stained gels (lane 5) and subjected to in-gel trypsin digestion, which cleaves ubiquitin between Arg 74 and Gly 75 and leaves diglycine motifs (Gly 75 -Gly 76 ) covalently attached to modified lysines ( Fig. 1B) (23). This and missed cleavages that result from the modification of lysines generates signature peptides that allows for identification of lysines as ubiquitin conjugation sites. Using this method, 11 lysines were identified as sites of ubiquitination: Lys 916 , Lys 962 , Lys 1571 , Lys 1771 , Lys 1884 , Lys 1885 , Lys 1886 , Lys 1901 , Lys 1924 , Lys 2118 , and Lys 2257 (Fig. 1C). In the eight independent analyses performed, peptides were detected that covered ϳ64% of the IP 3 R1 sequence and included 128 of the 167 lysines in IP 3 R1 (supplemental Fig. S1). These detectable lysines were distributed randomly throughout IP 3 R1, indicating that there was no bias for or against observing ubiquitination of lysines in a particular region. Clearly, however, given that 39 lysines were not covered, it remains a possibility that there are ubiquitination sites additional to the 11 described in Fig. 1C. As a control, we also analyzed IP 3 R1 immunoprecipitated from unstimulated cells, and as expected, almost no ubiquitination sites were identified (Lys 962 only was identified in two of four experiments; data not shown).
How the 11 ubiquitination sites are located within the IP 3 R1 primary sequence is depicted in Fig. 1D; as yet, high resolution structural data for full-length IP 3 R1 have not been defined (3), so these sites cannot yet be mapped in three dimensions. However, all of the ubiquitination sites are located within the regulatory domain and are within or adjacent to binding sites for modifiers (1) or regions predicted to be surface-exposed loops on the basis of their susceptibility to mild trypsinization ( Fig. 1D) (30,31). Specifically, four sites, Lys 916 , Lys 962 , Lys 1571 , and Lys 1924 are located near these trypsin cleavage sites, Lys 1884 -1886 and Lys 1901 flank a caspase-3 cleavage site (Asp 1891 ), and Lys 1771 is located between a cAMPdependent protein kinase phosphorylation site (position 1755) and an ATP-binding site (positions 1773-1775) and is in a glycine-rich region thought to form a flexible connection between the N-terminal cytosolic region of IP 3 R1 and the region encompassing the channel domain (1, 3). Additionally, Lys 1571 is adja-cent to a phosphorylation site (position 1588) and within a calmodulin-binding site (positions 1565-1586), and Lys 2118 and Lys 2257 are near a critical Ca 2ϩ -binding site, Glu 2100 (1). Over-all, these data show that the ubiquitination sites in IP 3 R1 are clustered in the regulatory domain and are found at sites likely to be surface-exposed.

Monoubiquitin and Lys 48 -and Lys 63 -linked Ubiquitin Chains Are the Predominant Modifications on IP 3 R1-To
quantitate and characterize the attached ubiquitin conjugates, we used ubiquitin-AQUA (23,25). This method relies on the fact that digestion of unmodified ubiquitin, either a single ubiquitin residue attached to the substrate (monoubiquitin) or the last ubiquitin of a chain (end cap ubiquitin), produces a series of peptides with unmodified lysine residues, whereas digestion of a ubiquitinated ubiquitin (i.e. that found in a chain) generates peptides containing the characteristic diglycine motif associated with the modified lysine (Fig. 1B). This revealed that in the absence of stimulus, the gel region encompassing IP 3 R1 contained 148 Ϯ 34 fmol of total ubiquitin (i.e. unmodified plus modified ubiquitin), which increased dramatically to 1812 Ϯ 473 fmol after exposure to GnRH, and remarkably, that 76 Ϯ 3% of IP 3 R1-associated ubiquitin under stimulated conditions is unmodified, with the bulk of the remainder containing Lys 48 and Lys 63 linkages (10.4 and 11.1%, respectively) (Fig. 1E). To determine whether IP 3 R1 is similarly affected in another cell type, we also examined IP 3 R1 from endothelin-1-stimulated Rat-1 cells. Again, most of the ubiquitin associated with IP 3 R1 was unmodified (82 Ϯ 2%), with Lys 63 and Lys 48 linkages making up the bulk of the remainder (Fig. 1E). In both cell types, other possible linkages were either undetectable or present only in very small amounts (Fig. 1E). Overall, these data show that in stimulated ␣T3-1 cells, only ϳ25% of ubiquitin is ubiquitinated and, thus, that a maximum of ϳ50% of total ubiquitin is involved in chains (Fig. 1F). Conversely, and remarkably, it can be concluded that the predominant modification on IP 3 R1 is monoubiquitin, which, at minimum (if diubiquitin is assumed to be the longest chain), comprises ϳ50% of total IP 3 R1-associated ubiquitin (Fig. 1F). This is the theoretical minimum, however, and the real value could be higher, because if longer chains are attached, the proportion of monoubiquitin increases (Fig. 1F).
Kinetics and Stoichiometry of IP 3 R1 Ubiquitination-It was remarkable to find that ubiquitinated IP 3 R1 was modified so strongly by monoubiquitin and that only ϳ10% of attached ubiquitin was Lys 48 -linked (Fig. 1E). Our expectation was that ubiquitin would be attached predominantly via Lys 48 -linked chains, because the current consensus is that it is Lys 48 -linked chains that direct substrates to the proteasome (17,18,32), and the only known consequence of IP 3 R ubiquitination is to evoke IP 3 R degradation (6).
To extend this work we first examined the kinetics of IP 3 R1 ubiquitination and also, by quantitating IP 3 R1 abundance, its stoichiometry. For these studies, immunopurified IP 3 R1 was separated on 4% gels to allow for clear separation between unmodified IP 3 R1 and the smear of ubiquitinated IP 3 R1, which was most readily visualized when nitrocellulose was probed simultaneously with anti-IP 3 R1 and anti-ubiquitin ( Fig. 2A). Ubiquitin immunoreactivity was evident 1.5 min after GnRH addition, peaked at 7 min, and then declined, at least in part because of IP 3 R1 degradation, as evidenced by a marked decrease in the unmodified IP 3 R1 band at 20 min (lanes 2-5). Consistent with the role of the proteasome in IP 3 R degradation, bortezomib both inhibited IP 3 R1 degradation and caused the accumulation of ubiquitinated species (lane 6). For mass spectrometry, regions corresponding to unmodified IP 3 R1 (low molecular weight (LMW)) and ubiquitinated IP 3 R1 (high molecular weight (HMW)) were excised from Coomassie Bluestained gels (lane 7) and analyzed for ubiquitin and IP 3 R1 content using AQUA (Fig. 2, B and C). Total IP 3 R1 levels (the sum of that present in the HMW and LMW regions) was reduced by GnRH treatment, to ϳ47% of control levels after 20 min (Fig.  2B). This decline was inhibited by pretreatment with bortezomib and was largely accounted for by a decline in the IP 3 R1 content of the LMW region, in which the vast majority of IP 3 R1 was found (Fig. 2B). Conversely, total ubiquitin increased dramatically after GnRH treatment, peaking at ϳ1500% of control levels after 7 min and at ϳ2000% of control levels in the presence of bortezomib plus GnRH, and this was largely accounted for by an increase in the ubiquitin content of the HMW region (Fig. 2C). Importantly, these increases in ubiquitin content paralleled an increase in the IP 3 R1 content of the HMW region, confirming that ubiquitinated IP 3 R1 was present in the HMW  1-4). The bracket indicates ubiquitinated IP 3 R1 migrating at ϳ270 -400 kDa, the arrow indicates unmodified IP 3 R1 migrating at ϳ260 kDa, and the dashed box in lane 5 indicates the region excised for mass spectrometry. B, a schematic representation of the trypsinization of a ubiquitinated substrate. Cleavage of ubiquitin and the target protein at the lysine (K) or arginine (R) residues indicated (black arrowheads) generates a target protein-derived signature peptide with an added mass of 114.04 Da (from the covalently linked Gly 75 -Gly 76 motif of ubiquitin) and a missed cleavage at the modified lysine. Comparison of mass spectra with data bases allows for the identification of ubiquitination sites in substrate proteins. Application of the same logic to peptides derived from ubiquitin allows for the determination of ubiquitin chain linkages; in the example shown, a linkage via Lys 48 yields a signature peptide. C, listed are the 11 ubiquitin-conjugated lysines found in IP 3 R1 and the peptide sequences in which they were identified. The periods indicate the trypsin cleavage sites, and the asterisks indicate the ubiquitinated lysines. The data shown are from eight independent analyses, and lysines were defined as ubiquitination sites only if they were identified in two or more of the independent analyses. Seven additional lysines, Lys 720 , Lys 1544 , Lys 1951 , Lys 2079 , Lys 2650 , Lys 2700 , and Lys 2719 , were each identified only once and thus have not been listed. D, a schematic representation of IP 3 R1, with ubiquitination sites indicated by K in the main diagram or by arrows in the expanded regions. The three domains of IP 3 R1 are the ligand-binding domain, the regulatory domain, and the channel domain, which contains six membrane-spanning helices (indicated by vertical lines) and the pore loop (between helices 5 and 6). Mild trypsinization of IP 3 R1 generates five fragments by virtue of cleavage at amino acids 346, 924, 1583, and 1932 (indicated by arrowheads). A glycine-rich region, a caspase-3 cleavage site, ATP-and calmodulin-binding sites, the Ca 2ϩ sensor, sites of protein kinase A phosphorylation, and a coiled-coil region are all highlighted. E, ␣T3-1 and Rat-1 cells were incubated with 100 nM GnRH for 7 min or 10 nM endothelin-1 for 10 min, respectively, were prepared as in Fig. 1A (lane 5), and were subjected to ubiquitin-AQUA analysis. The values presented are percentages of total ubiquitin (modified plus unmodified) associated with IP 3 R1, from at least four independent experiments (means Ϯ S.E.). Unmodified ubiquitin refers to either a single ubiquitin moiety (monoubiquitin) or the terminal ubiquitin in a chain (end cap ubiquitin). Modified ubiquitin refers to ubiquitin residues that are ubiquitinated. For ␣T3-1 and Rat-1 cells, total ubiquitin after stimulus was 1812 Ϯ 473 and 290 Ϯ 92 fmol, respectively. F, depicted for an individual IP 3 R1 subunit are models of how ubiquitin could be attached at maximal ubiquitination. In the examples given, IP 3 R1 subunits are tagged with 4, 8, or 12 ubiquitin moieties at a 3:1 ratio of unmodified to modified ubiquitin, and in each case the number and percentage of ubiquitin moieties involved in chains or present as monoubiquitin is tabulated. DECEMBER 19, 2008 • VOLUME 283 • NUMBER 51 JOURNAL OF BIOLOGICAL CHEMISTRY 35323 region. Additionally, after 20 min of exposure to GnRH, both the IP 3 R1 and ubiquitin content of the HMW region declined markedly, consistent with ongoing degradation of ubiquitinated IP 3 R1 (Fig. 2, B and C). Finally, it should be noted that the mass spectrometry and immunoblotting data concur very well (Fig. 2, A-C), indicating that the mass spectrometric methods accurately quantitate the proteins extracted from gels.

Mass Spectrometry of IP 3 Receptor Ubiquitination
The stoichiometry of ubiquitination was calculated from the molar ratio of ubiquitin and IP 3 R1 in each gel region (Fig. 2D). The total ubiquitin/IP 3 R1 ratio was ϳ0.05 in unstimulated cells and ϳ1 under conditions of maximal ubiquitination. This information is of limited value, however, because only a small proportion of IP 3 R1 (ϳ10%) is found in ubiquitinated form in stimulated cells (7), and both ubiquitinated and unmodified IP 3 R1 are included in this calculation. Much more meaningful are data from the HMW region, which should contain only ubiquitinated IP 3 R1. In this region, the ubiquitin/IP 3 R1 ratio was ϳ0.2 in unstimulated cells and ϳ4 at maximal stimulation, indicating that, on average, a maximally ubiquitinated IP 3 R1 is coupled to ϳ4 molecules of ubiquitin. This is likely to be an underestimate, however, because it appears that separation of unmodified receptor from ubiquitinated IP 3 R1 was not as complete as anticipated and that some unmodified IP 3 R1 migrates in the HMW region. This is evident from the fact that, in unstimulated cells, 11 Ϯ 3% of total IP 3 R1 was found in the HMW region (Fig. 2B); why this "tailing" occurs is not clear, but it is certainly not due to IP 3 R1 ubiquitination because the ubiquitin/IP 3 R1 ratio in the HMW region under unstimulated conditions was very low (ϳ0.2). Correcting for this aberration increases the ubiquitin/IP 3 R1 ratio to ϳ8 at peak ubiquitination (Fig. 1D). Overall, these data show that ubiquitin is rapidly attached to activated IP 3 R1 in a manner that precedes IP 3 R1 degradation and that under conditions of maximal stimulation, an average of ϳ8 ubiquitin moieties are attached to those receptors modified. However, it is important to note that this is the average value for species present in the HMW region and that the likelihood is that some IP 3 R1s will be tagged with Ͼ8 ubiquitin moieties, and some Ͻ8.
Next we examined, using AQUA analysis, the time dependence of ubiquitin chain formation on IP 3 R1 and, again, focused on the HMW region, because this region contains the vast majority of ubiquitin (Fig. 2C). Consistent with the data in Fig.  1E, the GnRH-induced increase in ubiquitin content is predominately accounted for by unmodified ubiquitin, which at each time point comprised ϳ70% of total ubiquitin (Fig. 3, A  and B). Thus, monoubiquitin plus end cap ubiquitin are the predominant species, and it can be deduced, as in Fig. 1F, that at FIGURE 2. Kinetics and stoichiometry of IP 3 R1 ubiquitination. A, ␣T3-1 cells were incubated with 100 nM GnRH for 0, 1.5, 3, 7, and 20 min, or preincubated with 1 M bortezomib (Bz) for 60 min and then treated with 100 nM GnRH for 60 min. The cell lysates were then prepared and incubated with anti-IP 3 R1 to immunoprecipitate IP 3 R1, which was then electrophoresed by 4% SDS-PAGE, and either transferred to nitrocellulose and probed simultaneously for anti-ubiquitin and anti-IP 3 R1 (lanes 1-6) or stained with Coomassie Blue (lane 7; only data from the bortezomib/GnRH incubation condition are shown). The arrow and bracket indicate unmodified and ubiquitinated IP 3 R1, respectively, and the dashed boxes in lane 7 delineate the HMW and LMW regions excised for mass spectrometry. B and C, mass spectrometric quantification of IP 3 R1 and ubiquitin levels. The data shown are the IP 3 R1 and ubiquitin levels in the HMW and LMW regions and total ubiquitin (the sum of that present in both HMW and LMW regions). In each independent experiment, the data were normalized to the amount of total IP 3 R1 or ubiquitin at t ϭ 0, and normalized values were then combined and graphed (mean Ϯ S.E., n ϭ 3). Total ubiquitin and total IP 3 R1 levels at t ϭ 0 were 176 Ϯ 36 and 3436 Ϯ 872 fmol, respectively. D, the total, HMW, and LMW ubiquitin/IP 3 R1 ratios were determined from the fmol values of IP 3 R1 and ubiquitin at each time point. Corrected HMW ratios were obtained from corrected fmol values of ubiquitin and IP 3 R1. For ubiquitin, these were determined by subtracting the fmol value present at t ϭ 0 from the fmol values at each time point. For IP 3 R1, the values were determined by subtracting 11% of the total amount of IP 3 R1 at each time point from that present in the HMW regions, which accounts for the incomplete separation of unmodified IP 3 R1 from ubiquitinated IP 3 R1. each time point, a minimum of ϳ40% of total ubiquitin associated with IP 3 R1 is monoubiquitin.
The increase in modified ubiquitin is predominately accounted for by increases in Lys 48 -and Lys 63 -linked ubiquitin, with only a minimal contribution from other linkages (Lys 11 , Lys 29 , Lys 33 , Lys 27 , and Lys 6 ) (Fig. 3, A and C). Interestingly, Lys 63 -linked ubiquitin initially accumulates on IP 3 R1 at approximately twice the rate of Lys 48 -linked ubiquitin, such that at early time points (1.5 and 3 min) Lys 63 linkages are predominant (Fig. 3, A and C). Thereafter, Lys 48 -and Lys 63 -linked ubiquitin accumulate to equal levels and then decline in parallel (Fig. 3, A and C). In contrast, in the presence of bortezomib plus GnRH, levels of Lys 48 -linked ubiquitin exceed Lys 63 -linked ubiquitin (Fig. 3, A and D). Overall, these data show that Lys 48and Lys 63 -linked ubiquitin chains are the predominant forms associated with IP 3 R1 and that the accumulation of these chains is subject to differential regulation.
Individual IP 3 R1s in a Tetramer Are Ubiquitinated and Degraded-We next sought to establish whether or not all IP 3 R1s in a tetramer are ubiquitinated. To do this, we immunoprecipitated microsomal proteins from control and GnRH-stimulated ␣T3-1 cells with anti-ubiquitin and probed for IP 3 R1 (Fig. 4A); we reasoned that if both ubiquitinated and unmodified IP 3 R1s were isolated by anti-ubiquitin, it would show that the latter was coimmunoprecipitating with the former and that not all IP 3 R1 subunits in a tetramer are ubiquitinated. This turned out to be the case, because anti-ubiquitin isolated IP 3 R1 immunoreactivity from stimulated ␣T3-1 cells only that migrated as a smear beginning at ϳ260 kDa (lanes 3 and 4); this smear co-migrated with both ubiquitinated IP 3 R1 migrating at ϳ270 -400 kDa (lane 6) and with unmodified IP 3 R1 migrating at ϳ 260 kDa (lanes 1 and 2). Thus, unmodified IP 3 R1 co-immunoprecipitates with ubiquitinated IP 3 R1, indicating that not all IP 3 Rs in a tetramer are ubiquitinated.
We next wondered whether this might lead to the degradation of individual subunits in a tetramer and the appearance of nontetrameric IP 3 R1s in stimulated cells. To assess this possibility, we subjected lysates from control and stimulated cells to nondenaturing blue native PAGE, which allows for the separation of proteins while still in complexes (Fig. 4B). Under these conditions, IP 3 R1 immunoreactivity in unstimulated cells migrated primarily at ϳ1.1 MDa, consistent with IP 3 R1 existing as tetramers (lane 1). GnRH caused a reduction in the ϳ1.1 MDa immunoreactivity, consistent with the induction of IP 3 R1 degradation (lanes 2-4), and remarkably, the appearance of a substantial amount of immunoreactivity at ϳ 600 and ϳ800 kDa, perhaps because of the formation of dimeric and trimeric IP 3 R1 complexes ( lanes  2-4). The appearance of this immunoreactivity at ϳ600 and ϳ800 kDa was strongest after 7-20 min exposure to GnRH (lane 2 and 3) and declined thereafter (lane 4) and was not due to IP 3 R1 cleavage, because GnRH did not induce the formation of Ͻ260-kDa IP 3 R1 fragments when samples subjected to denaturing SDS-PAGE were probed with antibodies against either the C terminus (Fig. 4C, lanes 1-4) or N terminus of IP 3 R1 (Fig.  4C, lanes 5-8). Taken together, these data show that not all IP 3 R1s in a tetramer are ubiquitinated and suggest that as IP 3 R1 degradation proceeds, trimeric and dimeric IP 3 R1 complexes form.

DISCUSSION
In summary, the data presented represent the first comprehensive analysis of the ubiquitination of an endogenous substrate by endogenous enzymes and reveals a far more complex scenario than originally anticipated. Specifically we show that at least 11 of the 167 lysines in IP 3 R1 are sites of ubiquitination, that these sites are occupied primarily by monoubiquitin, that both Lys 48 -and Lys 63 -linked ubiquitin chains are added, that Lys 63 -linked chains accumulate most rapidly, and that not all subunits in an IP 3 R1 tetramer are ubiquitinated.
Typically, post-translational modification sites are selected by adjacent sequence information (e.g. NX(S/T) for N-linked glycosylation), but there is no known consensus sequence for ubiquitination (21). There is, however, growing evidence that selection of lysines for ubiquitination may rely on structural features rather than sequence, because in a systematic analysis  DECEMBER 19, 2008 • VOLUME 283 • NUMBER 51 of 135 ubiquitination sites in proteins from Saccharomyces cerevisiae, it was found that ubiquitin was preferentially added to lysines in surface-exposed loops, as compared with other structural features like ␣-helices and ␤-sheets (21). Additionally, for human epidermal growth factor receptor overexpressed in porcine aortic endothelial cells, six ubiquitination sites were identified in the kinase domain, and all of these were located on exposed surfaces (26). For IP 3 R1, it also appears that the ubiquitination sites are located in exposed regions, because many are found adjacent to sites cleaved by mild trypsinization or are near binding sites for modifiers of IP 3 R1 (Fig. 1D). At present, only the structure of the IP 3 R1 ligand-binding domain, in which we were unable to detect any ubiquitination sites, has been solved (33). As more of the structure is solved, it will be fascinating to see how the ubiquitination sites we identified are arranged in three dimensions. Finally, all of the ubiquitinated lysines we identified in mouse IP 3 R1 are completely conserved in IP 3 R1 from other mammalian species (rat and human), but there is only partial conservation of these lysines in IP 3 R2 and IP 3 R3 (data not shown). Nevertheless, it is likely that IP 3 R1, IP 3 R2, and IP 3 R3 are ubiquitinated in similar regions, because the three receptor types are 60 -70% identical at the amino acid level (34) and are likely to adopt similar conformations, and IP 3 R2 and IP 3 R3 do have lysine residues in regions corresponding to those ubiquitinated in IP 3 R1.

Mass Spectrometry of IP 3 Receptor Ubiquitination
Interestingly, our data differ considerably from a previous immunoprecipitation/immunoblotting-based analysis of the sites of ubiquitination in IP 3 R1. In that study (35), examining rat IP 3 R1 overexpressed in CHO-K1 cells, it was concluded that the region C-terminal to Asp 1891 was monoubiquitinated and that the region N-terminal to Asp 820 was polyubiquitinated. Our data, in contrast, did not consistently detect any ubiquitination sites N-terminal to Asp 820 and found multiple sites C-terminal of Asp 1891 . This discrepancy is unlikely to be due to mouse and rat IP 3 R1 being different (they are 98% identical at the amino acid level (34)) or that we were unable to assess the ubiquitination status of every lysine in IP 3 R1 (the 39 lysines we were unable to assess were spread evenly throughout IP 3 R1). Rather, it is likely due to the fact that we analyzed endogenous IP 3 R1, whereas Bhanumathy et al. (35) analyzed overexpressed exogenous receptors. Our previous work has indicated that endogenous and exogenous IP 3 Rs are ubiquitinated with different characteristics, with highly overexpressed exogenous receptors being ubiquitinated even in the absence of stimulus, and with stimulus-dependent ubiquitination occurring only at very low levels of expression (7). Why highly overexpressed IP 3 R1s are ubiquitinated differently than endogenous IP 3 R1 is currently unknown but could result from misfolding or atypical interactions with the endogenous ubiquitination machinery.
During the course of this work, it was shown that iodoacetamide-based alkylation methods can cause 2-acetaminoacetamidation of lysines, a modification with an atomic composition (C 4 H 6 N 2 O 2 ) and mass addition (114.0429 Da) identical to that seen with diglycine conjugation and that this can result in the false identification of lysines as ubiquitination sites (36). One of the experiments performed in this study (the data from which was later omitted) employed iodoacetamide and contained evidence for iodoacetamide-derived lysine modifications identical to those described (36), apparently because of incomplete removal of iodoacetamide prior to trypsin digestion. In that experiment, 48 lysines in IP 3 R1 were found to have a mass addition of 114.0429 Da, and of those, 10 were found at the C termini of peptides. Because trypsin cannot cleave adjacent to ubiquitin-modified lysines (37), these lysine modifications must result from the effects of iodoacetamide. Furthermore, extracted ion chromatograms generated for the Lys 48 and Lys 63 linkage-derived diglycine signature peptides revealed peak doublets for each, with one member of the doublet corre- A, ␣T3-1 cells were incubated with 100 nM GnRH for 7 min, and microsomes were prepared, lysed, and incubated with anti-IP 3 R1 to immunoprecipitate IP 3 R1 or with anti-ubiquitin to immunoprecipitate ubiquitinated proteins. Immunoprecipitated proteins were then electrophoresed by 4% SDS-PAGE, transferred to nitrocellulose, and probed with either anti-IP 3 R1 or anti-ubiquitin. The bracket indicates ubiquitinated IP 3 R1 at ϳ270 -400 kDa and the arrow indicates unmodified IP 3 R1 at ϳ260kDa. B, ␣T3-1 cells were incubated with 100 nM GnRH for 0,7, 20, or 60 min. The cell lysates were then electrophoresed by blue native PAGE and probed with anti-IP 3 R1. The arrows indicate the migration positions of tetrameric IP 3 R1 at ϳ1.1 MDa and immunoreactivity that corresponds in size to trimeric and dimeric IP 3 R1 complexes. Also shown is the percentage of total IP 3 R1 immunoreactivity at each time point that migrated in the trimer/dimer region (mean, n ϭ 2). C, 7% SDS-PAGE of the same samples, probed with anti-IP 3 R1 (lanes 1-4) and NT1 (lanes 5-8), which recognize the C and N termini of IP 3 R1, respectively.
sponding to the bona fide diglycine-modified signature peptide and the other corresponding to a 2-acetamidoacetamide-modified lysine. Additionally, we noticed that some lysines were modified with a single acetamide group (C 2 H 3 NO) as evidenced by a mass increase of 57.021 Da. Thus, if iodoacetamide is employed as an alkylating agent, we recommend monitoring for these three indicators of possible iodoacetamide-induced artifacts to ensure accurate identification of ubiquitination sites.
Our AQUA analysis of ubiquitinated IP 3 R1 yielded some intriguing results. First, IP 3 R1 is predominately tagged with monoubiquitin (ϳ40 -50% of total ubiquitin, if diubiquitin chains are the longest present, but Ͼ50% if longer chains are attached). This is remarkable, because the consensus view has been that tetraubiquitin chains linked though Lys 48 is the minimal requirement for efficient proteasomal degradation (32). However, it is becoming apparent that the proteasome may recognize different ubiquitin conjugates, including monoubiquitin, through interactions with selective ubiquitin-binding domains found in proteasomal subunits (17). Indeed, recently, a novel ubiquitin-binding domain, called Pru for pleckstrin-like receptor for ubiquitin, that preferentially interacts with monoubiquitin and diubiquitin chains was identified in Rpn13, a proteasomal subunit (38,39). Thus, the monoubiquitin and short ubiquitin chains that are most likely to be attached to IP 3 R1, could be acting as signals for proteasomal degradation.
Second, we found that Lys 63 and Lys 48 linkages were the predominant ubiquitin chain linkages associated with IP 3 R1. The presence of Lys 63 linkages was a surprise, because these typically play a regulatory role in processes like the DNA damage response pathway and NF-B signaling, rather than in mediating proteasomal degradation (17,18). This raises the possibility that some of the ubiquitin conjugates (i.e. the Lys 63 linkages) on IP 3 R1 may have functions other than signaling for proteasomal degradation. However, Lys 63 -linked ubiquitin chains have also been shown to mediate proteasomal degradation in vitro (27,40), so it remains a possibility that the Lys 63 linkages on IP 3 R1 also act as a degradation signal.  (27). On the other hand, when the purified ligases, E6AP and Nedd4, were paired with UbcH5, the human homologue of Ubc4, homogenous Lys 48 -and Lys 63 -linked chains, respectively, were synthesized (41). At present, we cannot distinguish between these possibilities for IP 3 R1 ubiquitination, because the relevant E3 or E3s have yet to be identified. We do know, however, that Ubc7 is an E2 that mediates IP 3 R1 ubiquitination (15). Interestingly, there are several studies that show in vitro that Ubc7 catalyzes the formation Lys 48 linkages, and there is currently no evidence that this E2 can form Lys 63 linkages (42,43). Thus, it is likely that an additional E2 and multiple E2/E3 pairs mediate the ubiquitination of IP 3 R1.
Third, we found that Lys 63 -linked ubiquitin chains accumulate more rapidly on IP 3 R1 than Lys 48 -linked ubiquitin chains, and in the presence of bortezomib, Lys 48 linkages become predominant. These findings likely result from interplay between the addition of ubiquitin by ligases, the removal of ubiquitin by DUBs, and IP 3 R1 degradation by the proteasome. For example, the more rapid accumulation of Lys 63 linkages may be accounted for by more efficient synthesis of Lys 63 linkages, whether catalyzed by one or multiple E2/E3 pairs, and the marked slowing of Lys 63 -linked chain accumulation after 3 min could be due to an increase in Lys 63 -specific DUB activity that occurs as ubiquitin builds up on IP 3 R1. An interesting alternative is that Lys 63 -linked ubiquitin chains on IP 3 R1 may be edited in a manner similar to that described for A20. This protein, a potent inhibitor of the NF-B pathway, contains both ligase and DUB activities, and for both RIP1 and IRAK1, A20 first cleaves Lys 63 -linked ubiquitin chains and then catalyzes the formation of Lys 48 linkages, which results in proteasomal degradation of RIP1 and IRAK1 (28,44). Further, the selective accumulation of Lys 48 linkages on IP 3 R1 after proteasome inhibition with bortezomib may be due to preferential cleavage of Lys 63 linkages by DUBs; several studies show that certain DUBs preferentially cleave Lys 63 -linked ubiquitin (19,20), including a recent in vitro demonstration that ataxin-3 binds both homogenous Lys 48and Lys 63 -linked chains but preferentially cleaves Lys 63 linkages (45). Further exploration of these possibilities will require the identification and characterization of the DUBs that interact with and mediate the deubiquitination of IP 3 R1.
Finally, because the mass spectrometry used herein assesses the ubiquitination of a large population of IP 3 R1 subunits and yields information that is the average of ubiquitination in that population, it is not yet possible to define which sites are ubiquitinated in an individual subunit or whether an individual subunit is modified with only one kind of ubiquitin conjugate (monoubiquitin, Lys 48 -linked chains, or Lys 63 -linked chains) or a mixture. It is clear, however, that ubiquitinated IP 3 R1s are modified with an average of ϳ8 ubiquitin moieties and that at least ϳ40% of these are monoubiquitin. This leads to the prediction that, on average, IP 3 R1s are modified at 6 lysines. That these lysines are often within or near sites involved in IP 3 R1 regulation raises the intriguing possibility that ubiquitination could, in addition to mediating IP 3 R1 degradation, modulate the activity of IP 3 R1.
With regard to the degradation of IP 3 R1, it was intriguing to find that not all subunits in an IP 3 R1 tetramer are ubiquitinated and that nontetrameric associations of IP 3 R1 appear to be formed during the degradation process. This raises the possibility that ubiquitinated IP 3 R1 subunits are selectively removed from tetramers, leaving behind trimers or dimers. How the ubiquitinated subunits are segregated and removed from the ER membrane likely involves p97, which helps remove ubiquitinated substrates from the ER through its ATPase activity (13,14,46) and which is required for IP 3 R1 ERAD (14). Because the degradation of IP 3 Rs by ERAD serves to protect cells from excessive Ca 2ϩ mobilization (6), and it is highly unlikely that the nontetrameric IP 3 R1 complexes will be active as channels, the selective degradation of individual IP 3 R subunits may be the most efficient way of disabling channels without degrading the entire tetramer. It will now be intriguing to determine whether these putative nontetrameric IP 3 R1 complexes retetramerize or are subject to further degradation.
In summary, upon activation, tetrameric IP 3 R1 channels open and allow Ca 2ϩ to be released from the ER, and this activation also triggers their recognition for ERAD (47). IP 3 R1s are then ubiquitinated at, on average, ϳ6 lysines/subunit and have, on average, ϳ8 ubiquitin moieties attached per subunit. Ubiquitination is likely accomplished by multiple E2/E3 pairs that catalyze the formation of monoubiquitin and Lys 48 -and Lys 63linked ubiquitin chains, on one or more, but not all, subunits in a tetramer. Degradation of the ubiquitinated subunits then appears to lead to the formation of trimeric and dimeric IP 3 R1 complexes. Overall, our data show that IP 3 R1 ubiquitination is far more complex than originally anticipated and suggest that the composition of ubiquitin conjugates is regulated by interplay between ligases, DUBs, and the proteasome. To better understand the mechanisms controlling IP 3 R1 processing by the ubiquitin-proteasome pathway, it will now be necessary to define the ubiquitin ligases and DUBs involved, and the manner in which IP 3 R1s are removed from the ER membrane.