Pias1 Interaction and Sumoylation of Metabotropic Glutamate Receptor 8*

Group III presynaptic metabotropic glutamate receptors (mGluRs) play a central role in regulating presynaptic activity through G-protein effects on ion channels and signal transducing enzymes. Like all Class C G-protein-coupled receptors, mGluR8 has an extended intracellular C-terminal domain (CTD) presumed to allow for modulation of downstream signaling. In a yeast two-hybrid screen of an adult rat brain cDNA library with the CTDs of mGluR8a and 8b (mGluR8-C) as baits, we identified sumo1 and four different components of the sumoylation cascade (ube2a, Pias1, Piasγ, Piasxβ) as interacting proteins. Binding assays using recombinant GST fusion proteins confirmed that Pias1 interacts not only with mGluR8-C but also with all group III mGluR CTDs. Pias1 binding to mGluR8-C required a region N-terminal to a consensus sumoylation motif and was not affected by arginine substitution of the conserved lysine 882 within this motif. Co-transfection of fluorescently tagged mGluR8a-C, sumo1, and enzymes of the sumoylation cascade into HEK293 cells showed that mGluR8a-C can be sumoylated in vivo. Arginine substitution of lysine 882 within the consensus sumoylation motif, but not other conserved lysines within the CTD, abolished in vivo sumoylation. Our results are consistent with post-translational sumoylation providing a novel mechanism of group III mGluR regulation.

within the same protein family: overexpression of the E2 conjugase ubc9 results in a 4-fold increase of GLUT4 at the membrane, whereas levels of GLUT1 are decreased (24,25). Recently, sumoylation has been found to silence the plasma membrane leak potassium channel K2P1 (26). To our knowledge, no plasma membrane receptors have yet been identified as targets of the sumoylation machinery. Here we report an interaction of family C GPCRs, namely group III mGluRs, with the E3 ligase Pias1. In addition, we provide evidence that the CTD of mGluR8a can undergo in vivo sumoylation on a consensus site in HEK cells.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Screen and Yeast Mating-Yeast two-hybrid screening was carried out using the DupLEX-A system (OriGene Tech. Inc., Rockville, MD) as described previously (27). Briefly, the cDNA fragments encoding the CTDs of mouse mGluR8a (mGluR8a-C) and mGluR8b (mGluR8b-C) were generated by PCR, cloned into pGilda and used as baits to screen an adult rat brain cDNA library cloned into pJG4 -5, according to the user's manual (Version 1.2). Yeast strain EGY48 was transformed using lithium acetate, and protein expression was induced by 1.5% (w/v) galactose. For screening with the mGluR8a and mGluR8b baits, 3 ϫ 10 6 and 1.25 ϫ 10 6 independent recombinants were examined, and 1,385 and 934 clones, respectively, were identified to be positive by Leu Ϫ auxotrophy and ␤-galactosidase expression. Yeast mating was performed with yeast strains EGY48 and RFY206, using the same selection conditions as in the two-hybrid screen. After examining insert sizes and HaeIII digestion patterns, ϳ100 candidate clones identified with each bait were selected for sequencing.
Protein Expression and Binding Studies-Expression of GST and MBP fusion proteins was performed in Escherichia coli BL21 (Stratagene) as described (28). Bacteria were lysed in phosphate-buffered saline containing the protease inhibitor mixture CompleteTM (Roche Diagnostics) via passage through a French press. The supernatant was collected after centrifugation at 100,000 ϫ g for 45 min at 4°C. GFP, CFP, or YFP fusion proteins were expressed in HEK293 cells and processed as described (27). Protein expression in the bacterial lysates and cell homogenates was confirmed by Western blotting with anti-GST, anti-MBP, and anti-GFP antibodies. For pull-down assays, the lysates were incubated with 25 l of glutathione-agarose beads (Amersham Biosciences) preloaded with GST-mGluR-C in incubation buffer (PBS containing 0.1% (v/v) Triton X-100, 2 mM EDTA, 2 mM EGTA, 2 mM dithiothreitol, and protease inhibitor mixture). After 2 h of rotary agitation, beads were collected by centrifugation and washed three times with incubation buffer. After elution with SDS sample buffer, eluted proteins were resolved by SDS-PAGE, followed by Western blotting with monoclonal anti-MBP (New England Biolabs) or polyclonal rabbit anti-GFP (Clontech). About one-fifth of the glutathione-agarose beads preloaded with GST-mGluR-C were loaded onto another gel and used to evaluate the amount of GST fusion protein bound. After transfer to a nitrocellulose membrane (Schleicher & Schuell, Dassel, Germany), proteins were stained for 4 min with 2% (w/v) Ponceau in 3% (w/v) trichloroacetic acid.

RESULTS
Yeast Two-hybrid Screen and Yeast Mating-To identify mGluR8-Cinteracting proteins, we performed a yeast two-hybrid screen of an adult rat brain cDNA library, using the entire CTD coding regions of the mGluR8a and mGluR8b cDNAs as baits. Twelve and thirteen clones encoding the C-terminal region of protein inhibitor of activated STAT1 (Pias1) were identified with mGluR8a-C and 8b-C, respectively (TABLE  ONE). The shortest cDNA fragment isolated for Pias1 comprised only the coding region for amino acids 514 -721, which corresponds to the extreme C terminus not including the zinc finger domain. Additionally, 12 (mGluR8a) and 9 (mGluR8b) clones were found for the Msx-interacting zinc finger (Miz1), which encodes the C terminus of mammalian Piasx␤. Nine (mGluR8a) and three (mGluR8b) clones contained the entire coding sequence for the ubiquitin-conjugating enzyme E2a (ube2a), the rat homologue of the yeast sumo-specific E2-conjugating enzyme (ubc9). All clones were in-frame with the B42 activation domain, and interactions were confirmed in yeast mating assays using the respective CTDs of mGluR8 (TABLE ONE). For mGluR8b, we isolated three additional clones, two encoding ubiquitin-like 1 (sumo1), and one encoding a C-terminal fragment of Pias␥. These clones were tested in yeast mating experiments and gave positive signals also with mGluR8a-C (TABLE ONE).
GST-mGluR8-C Interacts with MBP-Pias1-To biochemically confirm the results of the yeast two-hybrid screen, we expressed candidateinteracting proteins, except Piasx␤, in bacteria as MBP fusion proteins and used them with GST-mGluR8-C fusion proteins in pull-down assays (Fig. 1). Batch adsorption onto glutathione-agarose, followed by SDS-PAGE and Western blotting with an antibody against MBP revealed that the MBP-Pias1 could be affinity-purified on GST-mGluR8b-C and, to a lesser extent, on GST-mGluR8a-C, but not on GST alone. Normalization of the GST fusion protein levels retained on the beads by Ponceau staining of the nitrocellulose membrane (data not shown) revealed that the amount of GST-mGluR8a-C adsorbed onto the agarose beads was only about 25% of that of mGluR8b-C. A comparatively weak specific interaction was also detected for the MBP fusion of the C-terminal region of Pias␥, which similarly bound to both mGluR8-C isoforms. GST alone failed to bind MBP and all MBP fusion proteins tested. MBP-ube2a did not exhibit detectable binding to any of the GST-mGluR8-C termini. It is, however, important to note that rather stringent washing conditions had to be used in these pull-down assays because of high unspecific binding of MBP (data not shown).
GFP-Pias1 Interacts with All Group III mGluRs-Based on its frequent detection in the yeast two-hybrid screen, we next focused on Pias1. To examine whether binding of Pias1 is shared by other group III or even group II mGluR members, we performed binary yeast twohybrid assays with the C-terminal coding fragment of the Pias1 cDNA and group II/III mGluR-CTDs (except mGluR7b-C and mGluR8b-C; Fig. 2A). In these assays, mGluR8a-C showed the strongest interaction, followed by mGluR7a-C and then mGluR6-C/mGluR4-C. In contrast, neither mGluR2-C nor mGluR3-C interacted with Pias1. Presuming that Pias1 binds in its function as E3 ligase in sumoylation, we compared group II (29) and group III (28) CTD sequences for the presence of candidate acceptor lysine residues ( Fig. 2A). Notably, the group II mGluR C termini contain only a single lysine (mGluR2, Lys 823 ; mGluR3, Lys 832 ), just C-terminal of the end of the transmembrane domain 7.
Group III mGluRs do not carry a lysine at this position but display multiple lysine side chains throughout their CTDs.
Next, we investigated whether GFP-Pias1 generated in mammalian cells binds other group III mGluR C termini and tested its interaction with the respective GST fusion proteins (Fig. 2B). Whereas GST failed to bind GFP-Pias1 and conversely GFP did not interact with GST-mGluR7a-C (data not shown), all group III mGluR C termini showed some interaction. Strongest binding was seen with mGluR7a-C, mGluR4-C, and mGluR6-C. The weak band recovered with GST-mGluR8a could be attributed to substantially less GST fusion protein being retained on the agarose beads; a sequential Western blot with anti-GST antibody produced only a signal corresponding to ϳ25% of the other GST fusion proteins used in this experiment (data not shown).
In conclusion, all group III mGluR-CTDs were able to bind GFP-Pias1.
The same assay was also used to examine whether the different components of the sumoylation pathway found in the yeast two-hybrid screen can directly interact with GST-mGluR8a-C. For these experiments, we now used cDNA constructs encoding full-length Pias1, ube2a, and sumo1. CFP-sumo1, YFP-ube2a, and GFP-Pias1 all were expressed in HEK293 cells, and Triton X-100 extracts of the transfected cells were used in binding assays (Fig. 3). Interaction could be confirmed for mGluR8a-C and GFP-Pias1, whereas only very little or no YFP-ube2a was recovered in the bound protein fraction. GFP alone did not bind to GST-mGluR8a-C. In contrast to the results obtained in the original two-hybrid screen, GST-mGluR8a failed to bind CFP-sumo1 (molecular mass ϳ40 kDa) under our assay conditions but enriched two high molecular mass (Ն90 kDa) sumo-conjugated proteins from the FIGURE 1. GST-mGluR8-C fusion proteins interact with MBP-Pias1. GST, GST-mGluR8a-C, and GST-mGluR8b-C were immobilized on glutathione-agarose beads and incubated with 50 l of extracts from bacteria expressing MBP fusions of the in-frame cDNA fragments of ube2a, and Pias␥ isolated in the yeast two-hybrid screen and full-length Pias1. Input lanes show protein expression in 20-l bacterial extracts. After washing with incubation buffer, bound proteins were eluted with SDS-sample buffer, and aliquots analyzed by SDS-PAGE followed by Western blotting with an anti-MBP antibody. Note that MBP-Pias1 was retained on GST-mGluR8b-C and, to a lesser extent, on GST-mGluR8a-C, whereas GST failed to bind MBP fusion proteins. Weak binding was also seen with MBP-Pias␥ on both mGluR-Cs. cDNAs identified by yeast two-hybrid screening using mGluR8a-C and mGluR8b-C as baits 3 ϫ 10 6 (mGluR8a-C) or 1.25 ϫ 10 6 (mGluR8b-C) independent recombinants were plated on 15-cm dishes; only in-frame positive ones are listed. Amino acid [aa] numbers indicate the predicted positions of the amino acids encoded by the cDNA fragments. HEK cell lysates. We did not try to disclose the identity of these proteins but an unbiased mass spectrometry-based analysis of sumo-conjugated HEK cell proteins has identified several candidates in the respective molecular mass range (30). Together these results are consistent with Pias1 representing the primary group III mGluR binding partner of the sumoylation machinery.
Mapping of the Pias1 Interaction Domain of mGluR7a-C and mGluR8a-C-To determine which domains of mGluR7a-C and mGluR8a-C interact with Pias1, we tested binding of mammalian expressed CFP-Pias1 to respective truncated GST fusion proteins (Fig.  4B). A schematic drawing of the truncation mutants used is shown in Fig. 4A. The mGluR7a-C truncation constructs did not overlap whereas those for mGluR8a-C overlapped by three amino acids. Also, the positions of the truncations were different in the respective CTDs: for mGluR7a-C, GST-mGluR7a-N38 ends, and GST-mGluR7a-C27 starts, just before the conserved lysine. For mGluR8a-C, GST-mGluR8a-N24 only included the proximal signal transduction domain with the G-protein ␤␥ and Ca 2ϩ /calmodulin binding sites (28), whereas GST-mGluR8a-C44 contained all conserved lysines outside of this signaling domain. GST fusion protein levels were normalized and tested semiquantitatively for the amount retained on beads (Ponceau stain on nitrocellulose membrane, Fig. 4B, lower panel). Lower protein levels were seen particularly for GST-mGluR7a-N38, GST-mGluR7a-K889R, GST-mGluR8a-C, and mGluR8a-K882R. Binding of CFP-Pias1 was found with GST-mGluR7a-N38 and mGluR8a-C44, whereas the complementary truncations GST-mGluR7a-C27 and GST-mGluR8a-N24 failed to interact. In sequence alignments ( Figs. 2A and 4A), mGluR7a-N38 and mGluR8a-C44 overlap by 17 amino acids but are only identical in the last six residues preceding the consensus sumoylation motif (sequence DRPNGE; see amino acids 875-880 of mGluR8a). We therefore deduce that these residues are important for Pias1 recruitment to group III mGluRs.
In Vivo Sumoylation of mGluR8a-C Requires Lys 882 -To demonstrate that mGluR8a-C is sumoylated in vivo, GFP-mGluR8a-C and the FIGURE 2. Pias1 interacts with all group III mGluR. A, alignment of rat group II and group III mGluR-C termini starting from the predicted end of transmembrane domain VII. The first three lysines of all group III mGluR (mGluR4, -6, -7, and -8) CTDs are located in the highly conserved N-terminal G-protein ␤␥/calmodulin binding region. In the more variable C-terminal region, additional lysines are found, some of which are conserved among isoforms. Bold letters, consensus sumoylation motif ⌽KXE. Note that group II mGluRs (mGluR2 and -3, bottom) only contain one conserved lysine in position ϩ4 after the predicted end of the last transmembrane domain. B, mammalian expressed GFP-Pias1 interacts with GST fusion proteins of all group III mGluR C termini. GFP-Pias1 was expressed in HEK293 cells, and an aliquot (10 l) of the cell lysate separated in the left lane. Pull-down assays with 100 l of HEK lysate on GST, or GST-mGluR-Cs as indicated, were performed as described under "Experimental Procedures." Bound protein was detected after SDS-PAGE by Western blotting with an anti-GFP antibody. Note that GFP-Pias1 binds to all GST-mGluR-C termini but not to GST. The GFP reactive band labeled with an asterisk represents a degradation product of GFP-Pias. Amounts of immobilized GST fusion protein were similar for all fusion proteins, as indicated by Western blotting with anti-GST (data not shown), except for GST-mGluR8a for which only 25% of the average protein level was bound. . GFP-Pias1 interacts with GST-mGluR8a-C but not ube2a or sumo1. Ube2a, Pias1, and sumo1 tagged with different GFP variants were expressed in HEK293 cells (input: 10 l of HEK cell extracts, protein concentration, 3 mg/ml; left panel). 50 l of Triton X-100 extracts of the transfected cells were incubated with GST or GST-mGluR8a-C, respectively (right panel). After the pull-down procedure described in the legend to Fig. 1, bound proteins were separated by SDS-PAGE, and immunoreactive bands were detected by Western blotting with an anti-GFP antibody. Note that GFP-Pias1 binds to GST-mGluR8a-C but not to GST. Only a very weak interaction is detected with YFP-ube2a. Asterisks mark high molecular mass bands of unknown identity that were bound from extracts expressing CFP-sumo1. Note that free sumo1 was not recovered.

Pias1 Interaction with Group III mGluRs
following tagged components of the sumoylation pathway were co-expressed in HEK293 cells: CFP-sumo1, E1 components aos1/uba2, YFP-ube2a, and GFP-Pias1. After detergent extraction of the transfected cells in the presence of protease inhibitors and 20 mM N-ethylmaleimide, which blocks sumo1-deconjugating enzymes (31), the extracts were separated by SDS-PAGE and Western-blotted with an antibody directed against the mGluR8 CTD. In the transfected, but not in untransfected cells, a significant fraction of mGluR8a-C immunoreactivity displayed a size shift to ϳ70 kDa, consistent with the addition of a single CFP-sumo1 molecule (Fig. 5A). Parallel Western blotting with an anti-sumo antibody confirmed that the 70-kDa band was indeed sumoylated (Fig. 5B). We also examined whether co-transfection of all six cDNAs was necessary for this sumoylation and found that only ube2a was not endogenously expressed at levels high enough to yield visible sumoylation of the overexpressed mGluR8a-C (data not shown). Inclusion of the E1 enzyme (aos1/uba2) proved to be second most important, whereas overexpression of Pias1 appeared not to be required and hence was omitted from subsequent transfections.
Substitutions of target lysines by equally charged arginines are commonly used to identify motifs for sumoylation, whereas corresponding alanine substitutions have been shown to result in reduced binding of E2 to substrate proteins like Ran GTPase-activating protein (RanGAP1) (32). Here, sumoylation of mGluR8a-C was also abolished upon replacing specific lysines by arginines within the CTD (for positions of lysine substitutions, see Fig. 4A). Substitutions were selected based on two criteria: location within C44 of mGluR8a and conservation in both, mGluR8a and 8b. We found that single or triple arginine substitutions, including Lys 882 , abolished sumo-conjugation, whereas single or combined substitution of Lys 868 and Lys 872 did not interfere with this modification (Fig. 5A). Notably, sumo-conjugation did not occur on the neighboring lysines Lys 868 or Lys 872 when the consensus sumoylation lysine 882 had been substituted. Also, K882R substitution or the triple mutation K868R/K872R/K882R did not lead to sumoylation of one of the remaining four lysines in the CTD of mGluR8a (Fig. 5A). Thus, in transfected cells, sumoylation of mGluR8a-C occurs specifically at lysine 882 located within the conserved consensus sumoylation motif. Notably, arginine substitution of Lys 882 in mGluR8a-C and of the homologous lysine Lys 889 in mGluR7a-C did not affect binding of CFP-Pias1 in the GST pull-down assay (Fig. 4B). This further confirms that the interaction of Pias1 with mGluRs does not depend on an intact sumoylation consensus motif in the CTD.

DISCUSSION
In this study, we identified components of the sumoylation pathway as interactors of the two splice variants of the C terminus of mGluR8. Sumoylation is a highly conserved protein modification that has been shown to be essential for cell cycle function in yeast (22). The following sumo-associated proteins were found to bind to the CTD of mGluR8 in the yeast two-hybrid assays: Pias1, Pias␥, Piasx␤, ube2a, and sumo1. Based on the number of clones that were found in the two-hybrid screen, we focused on the involvement of Pias1 in group III mGluR sumoylation. The interaction of Pias1 and mGluR8-C was confirmed in vitro and upon transfection into mammalian cells, whereas recombinant ube2a and sumo1 failed to bind in GST pull-down assays. This may indicate that Pias1 is the only protein of the sumoylation machinery that binds with high affinity to mGluR8-C, consistent with a more general role of the E3 ligase in mGluR regulation. Binding of Pias1 could be demonstrated for all group III mGluR C termini in vitro. Thus, sumoylation may be a common post-translational modification of these presynaptic receptors.
Our pull-down assays with truncation or point mutants of the mGluR7a-and mGluR8a-CTDs show that binding of Pias1 to the receptor C termini can occur independently of the presence of the actual sumoylation site (mGluR7a-N38) or the target lysine residue (mGluR8a-K882R, mGluR7a-K889R). We therefore suggest that a min- The position of the consensus sumoylation motif is indicated in bold. Note that mGluR7a, which binds GFP-Pias1, does not contain the consensus sumoylation site but overlaps with mGluR8a-C44 in the region proximal to the consensus motif. Single or multiple point mutations were introduced into the mGluR8a-C cDNA at all lysine codons that are conserved between mGluR8a and mGluR8b (K868R, K872R, K882R). B, mapping of the Pias1 interaction site with mGluR8a-C. Upper panel, GST fusion proteins of mGluR7a-C and mGluR8a-C immobilized on glutathione-agarose beads were used in pull-down assays with 60 l of CFP-Pias1 expressed in HEK293 cells. Western blots were stained with an anti-GFP antibody. Pias1 failed to bind to GST-mGluR7a-C27 and GST-mGluR8a-N24, whereas the K882R (mGluR8a) and K889R (mGluR7a) substitutions within the consensus sumoylation motif, had no effect. The GFP reactive bands labeled with asterisks represent degradation products of GFP-Pias. Lower panel, relative amounts of GST or GST fusion protein bound onto the beads were revealed by Ponceau protein staining. Note that low levels of immobilized GST-mGluR8a-C strongly bound large amounts of CFP-Pias1, whereas high levels of immobilized GST-mGluR8a-N24 failed to bind under identical conditions. FIGURE 5. mGluR8a-C is sumoylated on lysine 882. A, GFP-mGluR8a-C and tagged enzymes of the sumoylation pathway were co-expressed in HEK293 cells. 48 h after transfection, the cells were solubilized in the presence of protease inhibitors and 20 mM N-ethylmaleimide. After centrifugation, the extracts were subjected to SDS-PAGE followed by Western blotting with an antibody directed against the mGluR8-CTD. In the presence but not absence of enzymes of the sumoylation cascade, a size shift of mGluR8a-C to ϳ70 kDa, i.e. the approximate size of the mGluR8a-C-CFP-sumo1 conjugate, was observed. Single or combined arginine substitutions of Lys 882 abolished sumo-conjugation, whereas substitution of Lys 868 and Lys 872 had no effect. B, 70-kDa mGluR8 immunoreactive band is detected by an antibody against mGluR8a-C. Parallel sections of the same nitrocellulose strip were blotted with anti-sumo1. NOVEMBER 18, 2005 • VOLUME 280 • NUMBER 46 imal binding sequence must exist outside of the consensus conjugation site. Using partial constructs of mGluR8a-C and mGluR7a-C, this minimal binding sequence could be shown to reside within six amino acids preceding the consensus conjugation site (mGluR8a 875-880, DRPNGE), a motif that is conserved among mGluR7 and -8 isoforms and, to a lesser extent, in mGluR4. Notably, Pias1 binding was also found with mGluR6 in both yeast two-hybrid and pull-down assays. Among group III mGluRs, mGluR6 is unusual for several reasons: it is localized postsynaptically, is exclusively expressed in retina, and lacks the ability to interact with Ca 2ϩ /calmodulin, which recognizes all other group III mGluRs (27,33,34). As mGluR6 also lacks the consensus sumo-conjugation motif and harbors only two (Pro, Asn) of the six amino acids within the DRPNGE motif common to all other group III mGluRs, we speculate that these two amino acids, which are separated by a glutamine, may be sufficient to mediate Pias1 binding. Alternatively, other more homologous regions of the receptor CTDs that are also present in mGluR6 may contribute to the binding of Pias1, although our studies using truncations of mGluR7a and mGluR8a do not provide evidence for the existence of additional interacting sequences. Group II mGluRs, which lack both, the consensus sumoylation site and the Pias1 interaction domain, did not show an interaction with Pias1 in the yeast twohybrid system. To investigate whether the sequence DRPNGE could be a common binding motif for Pias1, we performed a data base search for short, nearly exact matches. This revealed that only one other mammalian gene family, the ING tumor suppressor protein family, contains a Ͼ80% identical sequence (i.e. DRPNG). ING proteins are proteins rich in non-consensus lysines and belong to the zinc finger protein family. Zinc finger proteins (22), but not ING proteins, have been reported to constitute a target of Pias1-mediated sumoylation. In fact, one of the lysines in the vicinity of the DRPNG motif is predicted to be available for sumoylation (www.abgent.com/doc/sumoplot). The domains of Pias1, which mediate the interaction with group III mGluRs, are not yet defined. The fact that our two-hybrid screen isolated a Pias1 fragment encoding only the C-terminal amino acids 514 -721 suggests that binding to the target sequence occurs downstream of the zinc finger domain of Pias1 (residues 401-453) that carries the sumo ligase activity. Thus, the interaction and catalytic domains of Pias1 seem to be distinct.

Pias1 Interaction with Group III mGluRs
Overexpression of components of the sumoylation machinery in HEK293 cells allowed us to demonstrate conjugation of sumo1 to GFP-mGluR8a-C, provided protease inhibitors that prevent de-sumoylation (31) were added during extract preparation. The fact that sumoylation was also seen without Pias1 co-transfection is in agreement with the ubiquitous expression of this protein and of other Pias family members in many cell types, including HEK293 cells where Pias1 is readily detected by Western blotting. 5 Convincing evidence for a crucial role of Pias1 in sumo1 conjugation in vivo comes from experiments in COS-7 cells in which sumoylation of STAT1 was decreased upon co-expression of a dominant-negative Pias1 mutant (35). Also, in vitro bacterially expressed GST-Pias1 has been shown to produce a dose-dependent stimulation of sumo-conjugation of in vitro translated p53 (21). In conclusion, all presently available data are consistent with an essential role of Pias1 in diverse sumoylation reactions, and hence we propose that the sumoylation of GFP-mGluR8a-C demonstrated here depends on its interaction with Pias family members.
Whereas ubiquitination of GPCRs is a well documented phenomenon (36) that appears to play a role in group I mGluR degradation (37), the related but functionally distinct sumoylation cascade has not yet been linked to any plasma membrane receptor. To our knowledge, our results represent the first evidence for sumo-conjugation of a GPCR. All group III mGluRs except mGluR6 contain a consensus sumoylation motif ⌽KX(D/E) (⌽, hydrophobic; K, acceptor lysine; D/E, acidic; X, any residue). Arginine substitution of the lysine 882 residue within the consensus sumoylation motif of the GFP-mGluR8a-C protein revealed that sumoylation occurs at the predicted consensus site. Consensus site sumoylation has been documented for many sumo targets, but sumoylation at lysines outside of consensus motifs has also been described for polypeptides such as huntingtin (38), the promyelocytic leukemia gene product (39), or proliferating cell nuclear antigen (40). In contrast to ubiquitination, sumo1 conjugation never results in formation of sumo chains at the conjugation site. The size shift seen here in SDS-PAGE (molecular mass of sumo1 ϳ15 kDa) upon overexpression of components of the sumoylation cascade is consistent with the attachment of a single CFP-tagged sumo1 molecule to GFP-mGluR8a-C. Whereas in huntingtin all three non-consensus lysines could be conjugated in HeLa cells and identification of the most relevant residue was possible only by multiple lysine substitution (38), single or combined mutations of the different lysine residues in the C-terminal portion of the mGluR8a CTD reliably produced sumoylation of only the consensus lysine 882. Inversely, substitution of Lys 882 did not redirect the sumoylation machinery to neighboring lysines, which underlines the sequence specificity of this covalent modification. Thus, the sumoylation complex, once attached to the mGluR8-CTD, does not modify non-consensus lysine side chains upon substitution of the primary target residue. This cannot be attributed to a masking of alternative acceptor sites by constitutively bound Ca 2ϩ /calmodulin, because our experiments were performed in the presence of divalent cation chelators, which fully prevent calmodulin binding to group III mGluR CTDs (28).
The physiological consequences of sumo-modification of mGluR8 are presently unknown. If sumoylation would merely antagonize ubiquitination, mGluRs should be subject to ubiquitination reactions. Indeed, a group I mGluR has been shown to bind a protein involved in photoreceptor cell differentiation via the ubiquitin/proteasome pathway (41). This protein, named mammalian homologue of Drosophila seven in absentia (Siah-1A), has also been found to regulate Ca 2ϩ /calmodulin binding to group I mGluRs, and thus to modulate signal transduction. The functional significance of this finding was questioned because of the fact that truncated versions of Siah-1A lacking domains required for the ubiquitin/proteasome pathway displayed similar regulatory effects (42). However, a recent report (37) provides clear evidence that Siah-1A indeed mediates ubiquitination and subsequent degradation of group I mGluRs. Thus, the regulation of mGluR ubiquitination could be one of the functional roles of sumo-conjugation to group III mGluR CTDs.
Recent studies indicate that sumoylation may also have regulatory effects on protein-protein interactions (17). A sumo-binding motif (SBM) in RanBP2 can bind the sumo-conjugated RanGAP1. This SBM consists of a (V/I)X(V/I)(V/I) sequence, a motif found in many proteins that are related to sumo-dependent processes. Among the proteins known to interact with some of the group III mGluR-C termini, only syntenin (43) and filamin A (44) contain one, or six, bona fide SBMs, respectively. The fact that these proteins only interact with the nonconserved, isoform-specific extreme C termini of some group III receptors is inconsistent with Pias-1 binding to regions conserved in all group III mGluR CTDs as described here. Thus, sumo-interacting proteins recognizing these conserved regions of mGluR CTDs remain to be identified.
A yet unsolved cell biological question is where mGluRs and the sumoylation cascade are likely to meet. There is circumstantial evidence that sumoylation can happen at the plasma membrane. Dynamin has been shown to interact with ubc9, sumo1, and Pias1 without being sumoylated itself (45), and to be involved in internalization of the group III mGluR4 (46). Furthermore, overexpression of the sumoylation cascade in mammalian cells can down-regulate the dynamin-mediated endocytosis of other proteins, e.g. transferrin (45). It therefore is tempting to speculate that dynamin serves to localize the primarily nuclear sumoylation machinery in the vicinity of target plasma membrane proteins. A convincing demonstration that components of the sumoylation machinery are indeed present at the cytoplasmic face of the plasma membrane comes from confocal imaging of Xenopus oocytes expressing the sumo1 conjugation enzyme ubc9 (26). In oocytes, ubc9 shows a uniform, nonpolarized distribution beneath the plasma membrane, thereby confirming that this protein actually reaches this cellular compartment. The same study also disclosed that in oocytes sumoylation enzymes modify the leak potassium channel K2P1 at the intracellular CTD and thereby control its ion channel function. Thus, sumoylation may well play a role in regulating the function and density of different integral plasma membrane proteins.
In summary, sumo modification of G-protein-coupled receptors may have different regulatory functions in targeting and modulating receptor signaling. Further studies will be needed to understand the specific roles, regulation, and kinetics of sumo-conjugation to group III mGluRs.