GTPase Activating Specificity of RGS12 and Binding Specificity of an Alternatively Spliced PDZ (PSD-95/Dlg/ZO-1) Domain*

Regulator of G-protein signaling (RGS) proteins increase the intrinsic guanosine triphosphatase (GTPase) activity of G-protein α subunits in vitro, but how specific G-protein-coupled receptor systems are targeted for down-regulation by RGS proteins remains uncharacterized. Here, we describe the GTPase specificity of RGS12 and identify four alternatively spliced forms of human RGS12 mRNA. Two RGS12 isoforms of 6.3 and 5.7 kilobases (kb), encoding both an N-terminal PDZ (PSD-95/Dlg/ZO-1) domain and the RGS domain, are expressed in most tissues, with highest levels observed in testis, ovary, spleen, cerebellum, and caudate nucleus. The 5.7-kb isoform has an alternative 3′ end encoding a putative C-terminal PDZ domain docking site. Two smaller isoforms, of 3.1 and 3.7 kb, which lack the PDZ domain and encode the RGS domain with and without the alternative 3′ end, respectively, are most abundantly expressed in brain, kidney, thymus, and prostate. In vitro biochemical assays indicate that RGS12 is a GTPase-activating protein for Gi class α subunits. Biochemical and interaction trap experiments suggest that the RGS12 N terminus acts as a classical PDZ domain, binding selectively to C-terminal (A/S)-T-X-(L/V) motifs as found within both the interleukin-8 receptor B (CXCR2) and the alternative 3′ exon form of RGS12. The presence of an alternatively spliced PDZ domain within RGS12 suggests a mechanism by which RGS proteins may target specific G-protein-coupled receptor systems for desensitization.

The mammalian "regulators of G-protein signaling" (RGS) 1 gene family was first identified by sequence and functional similarity to fungal and nematode genes captured in genetic screens for negative regulators of specific G-protein-coupled receptor (GPCR) signals (1)(2)(3). In vitro biochemical analyses soon established that this gene family encodes potent accelerators ("GAPs") of the intrinsic GTP hydrolysis activity of Gprotein ␣ subunits, revealing a molecular mechanism by which RGS proteins drive G-proteins into their inactive GDP-bound form and hence down-regulate GPCR signal transduction in vivo (reviewed in Refs. 4 and 5). However, the mechanisms by which individual RGS proteins desensitize pathways activated by particular GPCRs remain to be elucidated. Tightly regulated transcription has been described for RGS1 (3), RGS2 (6), and RGS3-RGS11 (7), and palmitoylation of the cysteine-rich N terminus of G␣-interacting protein (GAIP) has also been observed (8); however, transcriptional regulation and post-translational modifications of particular RGS family members can each only be expected to afford a gross level of intracellular control over the temporal and spatial expression of G␣-directed GAP activity.
We and others have hypothesized that regions outside the RGS fold contribute to regulation of G␣ GAP activity and/or targeting of individual RGS proteins to particular receptor signaling pathways (4,5,9,10). Here, we report the GAP activity of RGS12 and identify a PDZ-like N-terminal sequence within two splice forms. PDZ domains are protein-protein interaction modules that bind to three or four amino acid motifs at the extreme carboxyl termini of target proteins (reviewed in Ref. 11). PDZ domains of many proteins localize enzymatic activities and other protein-protein interaction domains to specific submembranous regions. Members of the PSD-93/95 protein family cluster NMDA receptors, nitric oxide synthase activity, and Shaker K ϩ channels at neuronal synapses (12). The protein tyrosine phosphatase FAP-1 interacts via a PDZ domain with the C terminus of CD95, receptor of the apoptosisinducing Fas ligand (13). Most recently, the five-PDZ domaincontaining protein InaD was shown to organize the G-proteincoupled phototransduction machinery in Drosophila, with individual PDZ domains binding to phospholipase C, transient receptor potential ion channel, and protein kinase C subunits (14). Using yeast two-hybrid, surface plasmon resonance, and protein overlay analyses, we demonstrate specific binding of the RGS12 PDZ domain to C-terminal (A/S)-T-X-(L/V) motifs. Our findings suggest that, by virtue of an alternatively spliced N terminus, specific isoforms of RGS12 may localize in vivo to * This work was supported in part by Amgen, Inc. (to D. P. S.), National Institutes of Health Grants HL16037 (to R. J. L.) and GM34497 (to A. G. G.), and by the Raymond and Ellen Willie Distinguished Chair in Molecular Neuropharmacology (to A. G. G.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF035151 and AF035152.
Northern Blot Analyses-Human MTN blots (CLONTECH), containing 2 g of poly(A) ϩ RNA from multiple adult human tissues and brain regions, were hybridized consecutively with a 686-bp hRGS12 3Ј-UTR cDNA probe (nucleotides 4185-4870 of GenBank AF035152), a 330-bp Reactions were performed at 4°C for A-C and 20°C for D. Aliquots were taken at the indicated times for measurement of free 32 P.
cDNA probe spanning the entire PDZ domain (nucleotides 55-384 of GenBank AF035152), and a 350-bp alternative 3Јend probe (nucleotides 2441-2790 of GenBank AF030109). Blots were then hybridized with a 1.3-kb fragment of human GAPDH cDNA to correct for RNA loading. Probe labelings and blot hybridizations were performed as described previously (15).
Overlay Assay-GST-h␤ 2 AR tail (80 amino acids) and GST-hCXCR2 tail (40 amino acids) fusion proteins were produced via PCR amplification of the tails and insertion of the PCR products into a pGEX-2T vector (Amersham Pharmacia Biotech). Mutation to the hCXCR2 tail was induced by the use of a mutant sequence oligonucleotide during PCR and confirmed by sequencing. GST fusion proteins were expressed FIG. 2. Northern blot analyses of four hRGS12 isoforms. Blots of poly(A) ϩ RNA (2 g) from various human tissues (A) and brain anatomical features (B) were probed with randomly primed 32 P-labeled human RGS12 cDNA fragments corresponding to the 3Ј-UTR described previously (9), the N-terminal PDZ domain (PDZ), and the alternative 3Ј exon (Alt 3Ј-end) described under "Results and Discussion." Blots were then probed with 32 P-labeled human GAPDH cDNA (hGAPDH) to assess loading. C, schematic representation of the four hRGS12 transcripts detected by Northern blot analyses. CC, location of predicted coiledcoil heptad repeats previously described (9). ATFV, C-terminal four amino acids encoded by the alternative 3Ј end. and purified as described above. GST and GST fusion proteins were resolved by SDS-polyacrylamide gel electrophoresis, electroblotted onto nitrocellulose, and incubated with recombinant thioredoxin-rRGS12-(1-440) fusion protein. Bound protein was detected using an antithioredoxin mouse monoclonal antibody, 3 Horseradish peroxidase-conjugated goat anti-mouse secondary antibody, and enhanced chemiluminescence (Amersham Pharmacia Biotech).
Biosensor Measurements-Surface plasmon resonance measurements were performed on the BIAcore 2000 (Biacore Inc., Piscataway, NJ). N-terminally biotinylated, synthetic polypeptides were bound to a streptavidin-coated sensor surface (Sensor Chip SA) as per manufacturer's instructions to a density of 300 response units (RU). GST and thioredoxin fusion proteins were 2-fold serially diluted in running buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.005% (v/v) surfactant P20) and injected at a 10 l/min flow rate; surface regeneration was performed by 10-l injections of 1 M NaCl in 50 mM NaOH at a 20 l/min flow rate. Relative binding to test peptides (test) was calculated based on the increase in response units relative to simultaneous measurement of binding to wild-type CXCR2 peptide (wt) and negative control peptide (ctrl) surfaces at 286 s after the start of a 300 s total protein injection: relative binding ϭ (RU 286s test Ϫ RU 286s ctrl )/(RU 286s wt Ϫ RU 286s ctrl ).
Yeast Two-hybrid Analysis-Yeast strain PJ69 -4A and the plasmids pAS1 and pACTII have been described (19,20). Rat RGS12 cDNA (amino acids 1-94) was amplified from rat brain cDNA using PCR, cloned into pCR2.1 (Invitrogen), sequenced, and subcloned in-frame downstream of the Gal4p DNA binding domain (DBD) (amino acids 1-147) in pAS1 using the EcoRI and SalI sites. cDNA encoding amino acids 322-359 of the rat high affinity interleukin-8 receptor B (rCXCR2; Swiss-Prot no. P35407) was amplified from rat brain cDNA using PCR, cloned and sequenced in pCR2.1, and subcloned in-frame downstream of the Gal4p activation domain (AD) (amino acids 768 -881) in pACTII using the EcoRI and XhoI sites. Nonsense and missense mutations to the rCXCR2 C-terminal tail were engineered within the pCR2.1 plasmid using QuikChange TM site-directed mutagenesis (Stratagene, La Jolla, CA), sequenced, and subcloned into pACTII as above.
Yeast strain PJ69 -4A was co-transformed with pAS1-and pACTIIbased plasmids using the Gietz lab transformation kit (Bio/Can Scientific, Mississauga, Ontario, Canada) and plated on synthetic complete drop-out medium (SC) lacking tryptophan and leucine (BIO 101, Vista, CA). Six independent yeast colonies from at least two independent transformations were then plated onto SC lacking tryptophan, leucine, histidine, and adenine, supplemented with 2 mM 3-aminotriazole.

RESULTS AND DISCUSSION
We recently described the cloning of Rgs12 from rat brain using a degenerate PCR strategy directed toward conserved regions of the RGS domain (9). To ascertain whether RGS12 is capable of stimulating the GTPase activity of G-protein ␣ subunits, the conserved RGS domain of rRGS12 (amino acids 664 -885) was expressed as a GST fusion protein in E. coli and purified by glutathione-Sepharose chromatography. The ability of GST-rRGS12-(664 -885) protein to enhance G␣ GTP hydrolysis was measured in single turnover assays using [␥-32 P]GTPloaded, recombinant G i ␣ 1 , G o ␣, G q ␣, and G s ␣ proteins. The slow rate of GDP dissociation and GTP binding observed for wild-type G q ␣ protein relative to its intrinsic GTPase activity makes it difficult to prepare GTP-G q ␣ for use in single turnover studies (21). Thus, a GTPase-deficient mutant of G q ␣ (R183C) was employed; this protein serves as an adequate substrate for detection of the GAP activities of RGS2 and RGS4. 2 The GTPase activities of myr-G i ␣ 1 and myr-G o ␣ proteins were increased by the addition of GST-rRGS12-(664 -885) protein, 3 G. M. Brothers and D. P. Siderovski, unpublished data.

FIG. 3. Specific binding of the rat RGS12 PDZ domain to the C terminus of interleukin-8 receptor B (CXCR2) as measured by SPR and protein blot overlay assays.
A, sequence similarity between the N-terminal PDZ domain of human RGS12 and PDZ domain 1 of human NHERF. Asterisk denotes conserved histidine residue starting the second ␣-helix (␣B) of the PDZ fold, previously shown within the PSD-95 PDZ-3/peptide co-crystal structure to interact with threonine at position Ϫ2 of the QTSV-COOH peptide binding site (26). B, simultaneous SPR measurement of binding to C-terminal peptides from rat interleukin-8 receptor B (rCXCR2), ␤ 1 -and ␤ 2 -adrenergic receptors (r␤1AR, r␤2AR), and mouse Notch1 protein (mNotch1) after injection (time 0 s, flow rate 10 l/min) of 50 l of 2.5 M GST-rRGS12-(1-94) fusion protein at 25°C (see Table II for peptide sequences). C, equal amounts (25 g) of GST and GST fused to the C termini of human ␤ 2adrenergic receptor (h␤2AR), interleukin-8 receptor B wild-type (hCXCR2 WT), or a mutant of the interleukin-8 receptor B where the final residue was changed from leucine to alanine (hCXCR2 L360A) were resolved on an SDS-polyacrylamide gel, blotted, and overlaid with thioredoxin-rRGS12-(1-440) fusion protein (50 nM). Bound RGS12 was detected by Western analysis using an anti-thioredoxin monoclonal antibody. The left panel shows a Coomassie stain of the fusion proteins; the right panel shows the binding of RGS12 following a blot overlay experiment. The positions of molecular weight markers are shown on the left. albeit modestly in comparison to recombinant, full-length RGS4 (Fig. 1, A and B). In contrast, G s ␣ and G q ␣(R183C) GTPase activities were not enhanced by RGS12 (Fig. 1, C and  D). Partially purified preparations of thioredoxin-rRGS12-(1-1387) and thioredoxin-rRGS12-(440 -1108) fusion proteins also demonstrated G i ␣ 1 and G o ␣ GAP activity (data not shown). Based on these in vitro results, we conclude that RGS12 is a bona fide member of the RGS family, in that it acts as a GAP for at least G i class ␣ subunits.
In our initial report of the cloning of rat Rgs12, we also identified the human ortholog (9) within cosmid sequences mapping to human chromosome 4p16.3. PCR was used to amplify both the predicted open-reading frame and 3Ј-UTR of hRGS12 from human brain cDNA. These two cDNA clones overlapped to form a contiguous sequence encompassing 4870 nucleotides and encoding 1376 amino acids (GenBank AF035152). Comparison of rat and human RGS12 protein sequences previously revealed nearly identical RGS domains and a highly conserved N terminus with similarity to a C-terminal region of mouse rhophilin (9); subsequent sequence and biochemical analyses (see below) indicate that the latter region encodes a PDZ domain. During the cloning of rRgs12 cDNA, we observed an alternatively spliced variant lacking the PDZ domain (GenBank AF035151); an alternative exon 1 with a unique 5Ј-UTR was found to replace exons 1 and 2 of the longer rRgs12 cDNA (GenBank U92280). Sequence of several human expressed sequence tags suggests that alternative exon usage also occurs in the same position in human RGS12 (e.g. Gen-Bank N31659 and AA455449).
To explore the tissue distribution of human RGS12 mRNA isoforms containing or lacking the PDZ domain region, Northern blot analyses were performed with probes specific to either VGSSSANTSTTL-COOH    the 3Ј-UTR or the PDZ domain. Hybridization of the 3Ј-UTR cDNA probe to poly(A) ϩ RNA prepared from various human tissues ( Fig. 2A) and brain regions (Fig. 2B) revealed two major hRGS12 transcripts of 6.3 and 3.7 kb. The 6.3-kb transcript was expressed abundantly in spleen, testis, ovary, cerebellum, and caudate nucleus, whereas the 3.7-kb transcript was most abundant in whole brain, kidney, thymus, and prostate, and detected at lower levels in all other tissues examined. Within specific brain regions, the 3.7-kb mRNA was most abundant in the cerebral cortex, occipital pole, frontal lobe, temporal lobe, putamen, amyglada, caudate nucleus, and hippocampus. Hybridizing the same Northern blots with a PDZ domain-specific probe revealed two hRGS12 transcripts of 6.3 and 5.7 kb (Fig.  2, A and B); the absence of hybridization to lower molecular weight mRNA indicates that the 3.7-kb hRGS12 transcript corresponds to an isoform lacking the PDZ domain. The 6.3-kb mRNA detected by the PDZ domain-specific probe corresponds to the size and tissue distribution of the 6.3-kb transcript detected by the 3Ј-UTR probe. The additional, 5.7-kb transcript was observed in both the same tissue distribution and abundance as seen for the 6.3-kb transcript.
To determine how the 5.7-and 6.3-kb isoforms differed, the sequence of human RGS12 was compared with the GenBank data base. Alignments to the predicted amino acid sequences of human expressed sequence tags (e.g. GenBank AA523013) and several other hRGS12 cDNA clones (e.g. GenBank AF030109 and AF030111) revealed several isoforms of RGS12 possessing an alternative 3Ј end. Alternative splicing removes the last four amino acids of RGS12 (GenBank AF035152) and inserts an additional 75 amino acids of open-reading frame and an alternative 3Ј-UTR. Hybridization of Northern blots with a probe specific for the alternative 3Ј end detected hRGS12 mRNA of 5.7 and 3.1 kb ( Fig. 2A and B), with the size and tissue distribution of the 5.7-kb transcript identical to that detected with the PDZ domain-specific probe. The 3.1-kb isoform was observed to be expressed with the same tissue distribution and abundance as the 3.7-kb isoform. These data indicate the presence of at least four alternatively spliced isoforms of hRGS12, containing or lacking an N-terminal PDZ domain, and encoding one of two alternative C termini (Fig. 2C).
Alternative splicing of the N terminus of human and rat RGS12 suggests that this region may be functionally important. Sequence analysis revealed highest homology of this region to the PDZ domains of the Na ϩ /H ϩ exchanger regulatory factor (NHERF; Fig. 3A), a protein that has recently been shown to bind the C-terminal tail of the ␤ 2 -adrenergic receptor (22). Thus, we examined whether RGS12 might also bind the tail of ␤ 2 AR or other G-protein-coupled receptors (GPCRs). Full-length and truncated forms of human and rat RGS12 were expressed as GST or thioredoxin fusion proteins and their binding to various peptides was investigated by surface plasmon resonance (SPR) analysis. Streptavidin-coated biosensor surfaces were pre-adsorbed with N-terminally biotinylated, synthetic peptides encompassing the last 12 amino acids of various GPCRs. Neither thioredoxin-rRGS12-(1-1387), thioredoxin-rRGS12-(1-440), nor GST-rRGS12-(1-94) fusion proteins bound appreciably to a biosensor surface coated with rat ␤ 1 -or ␤ 2 -adrenergic receptor tail peptides. However, the RGS12 fusion proteins did bind the C terminus of the rat interleukin-8 receptor B (rCXCR2; Fig. 3B and data not shown). GST and thioredoxin proteins alone did not bind any peptide surface tested, whereas full-length NHERF and NHERF domain 1 proteins bound both the rat ␤ 2 AR and CXCR2 tail peptides (data not shown).
We tested the C-terminal-specific nature of this interaction in a protein blot overlay using a point mutant of the CXCR2 C-terminal motif. Thioredoxin-rRGS12-(1-440) protein bound a GST-hCXCR2 tail fusion protein in blot overlay experiments but failed to interact with either GST alone or GST-␤ 2 AR tail (Fig. 3C); mutation of Leu to Ala at the terminal position within the hCXCR2 tail (L360A) abolished all binding. Specificity for the extreme C terminus of CXCR2 was also shown by yeast two-hybrid analysis. Rat RGS12 PDZ domain fused to the Gal4p transcription factor DBD and wild-type or mutated rCXCR2 tails fused to the Gal4p AD were expressed together in a yeast strain containing HIS3 and ADE2 reporter genes under the control of the GAL1 and GAL2 promoters, respectively (19). Growth on selective media lacking both histidine and adenine was compared with a positive control strain co-expressing the strongly interacting yeast proteins Snf1p and Snf4p (23) and a negative control strain co-expressing solely the Gal4p DBD and AD proteins. Co-expression of rRGS12 PDZ domain with wild- STTL type rCXCR2 receptor tail allowed prototrophic growth (Table  I); however, truncation of the last five C-terminal amino acids of the rCXCR2 tail (⌬355-359), or mutation of the Ϫ2 or terminal residues, abolished the interaction. Several missense mutations at the Ϫ1 position within the rCXCR2 tail did not disrupt the interaction, consistent with previous reports of PDZ domain binding specificity (24). We were intrigued by the selectivity shown by RGS12 for threonine at the Ϫ2 position of the CXCR2 tail, as removal of the C␤ methyl group by mutation of threonine to serine was observed to abolish completely the interaction in yeast (Table  I). To confirm this specificity, we tested three different rat metabotropic glutamate GPCR tails with Ser/Thr-rich carboxyl termini ending in S-(S/T)-L, of which MGR1␣ and MGR5 interact with the PDZ domain of the neuron-specific protein Homer (25). The lack of binding of GST-rRGS12-(1-94) to these three MGR tail peptides, as measured by SPR (Table II, A), confirms the preference of RGS12 for the motif T-X-L versus S-X-L. While necessary, the C-terminal T-X-L motif alone is not the sole determinant of RGS12 PDZ domain binding, as interaction with phospholipase C ␤1 and ␤3 carboxyl termini, both possessing the T-X-L motif, was at least 4-fold weaker than with the CXCR2 tail (Table II, A and B). Alanine/serine scanning mutagenesis of the last six residues of human CXCR2 tail peptide revealed the most important residues for hRGS12 PDZ binding specificity as the Ϫ3, Ϫ2, and 0 positions (S-T-X-L), with lesser contributions from the Ϫ4 and Ϫ5 positions (Table II, B). In addition, the failure of GST-hRGS12PDZ-(1-110) to bind amide-blocked C-terminal CXCR2 peptides confirms a critical role for the free carboxylic acid moiety in the binding interaction, as previously observed for other PDZ domains (26).
The predicted C terminus of the alternative 3Ј end splice isoform of hRGS12 is similar to that of human CXCR2, with both tails conforming to a consensus of (T/S)-S-(A/G)-H-X-(A/ S)-T-X-(L/V). We therefore tested binding of the human RGS12 PDZ domain to synthetic peptides encompassing the last 12 amino acids of the two, alternative carboxyl termini of RGS12. Although no binding was detected to the T-S-R-F C terminus conserved between human and rat RGS12 (hRGS12Tail1; Table II, B), GST-hRGS12PDZ-(1-110) protein was shown to bind the A-T-F-V C-terminal RGS12 peptide (hRGS12Tail2; Table  II, B). While a pattern search of the Swiss-Prot data base identified only CXCR2 as terminating with a polypeptide sequence conforming to the (T/S)-S-(A/G)-H-X-(A/S)-T-X-(L/V) motif, several other GPCRs, membrane-spanning proteins, and intracellular proteins with GAP or other enzymatic activities were identified as terminating in the shorter (A/S)-T-X-(L/V) motif. These potential RGS12 PDZ domain-binding proteins and their C-terminal sequences are summarized in Table III. It must be noted, however, that a terminal (A/S)-T-X-(L/V) motif alone is clearly not sufficient to specify RGS12 PDZ domain binding, as no appreciable binding has been observed to the A-T-N-V C-terminal tail of the human neuropeptide Y receptor type 2 (Table II, B). We are currently evaluating other proteins from Table III for their ability to bind RGS12.
Binding of the RGS12 PDZ domain to the RGS12 A-T-F-V tail was somewhat weaker than binding to the CXCR2 tail. However, among those form(s) of RGS12 containing both the N-terminal PDZ domain and the A-T-F-V C terminus, the in-teraction could be kinetically favored given its intramolecular nature. Such a scenario is reminiscent of the mechanism of reversible autoinhibition of Src family kinases, which adopt a "closed" conformation upon intramolecular association of their N-terminal SH2 and SH3 domains with C-terminal motifs (27). Intramolecular association of the N and C termini within RGS12 isoform(s) may serve as a source of regulation, either of its G i ␣-directed GAP activity or of PDZ-mediated binding to other proteins. The ability of the alternative C terminus to serve as a docking site for PDZ domain-containing proteins also presents the possibility of in vivo concatemerization and/or organization of RGS12 within multi-component signaling complexes. Compartmentalization by virtue of the N-and/or Cterminal domains of RGS12 could thereby associate desensitizing GAP activity within specific receptor complexes, analogous to inclusion of the negative regulatory function of protein kinase C within the Drosophila phototransduction complex by InaD (14).