RGSZ1, a Gz-selective RGS Protein in Brain

We cloned the cDNA for human RGSZ1, the major Gz-selective GTPase-activating protein (GAP) in brain (Wang, J., Tu, Y., Woodson, J., Song, X., and Ross, E. M. (1997)J. Biol. Chem. 272, 5732–5740) and a member of the RGS family of G protein GAPs. Its sequence is 83% identical to RET-RGS1 (except its N-terminal extension) and 56% identical to GAIP. Purified, recombinant RGSZ1, RET-RGS1, and GAIP each accelerated the hydrolysis of Gαz-GTP over 400-fold withK m values of ∼2 nm. RGSZ1 was 100-fold selective for Gαz over Gαi, unusually specific among RGS proteins. Other enzymological properties of RGSZ1, brain Gz GAP, and RET-RGS1 were identical; GAIP differed only in Mg2+ dependence and in its slightly lower selectivity for Gαz. RGSZ1, RET-RGS1, and GAIP thus define a subfamily of Gz GAPs within the RGS proteins. RGSZ1 has no obvious membrane-spanning region but is tightly membrane-bound in brain. Its regulatory activity in membranes depends on stable bilayer association. When co-reconstituted into phospholipid vesicles with Gz and m2 muscarinic receptors, RGSZ1 increased agonist-stimulated GTPase >15-fold with EC50<12 nm, but RGSZ1 added to the vesicle suspension was <0.1% as active. RGSZ1, RET-RGS1, and GAIP share a cysteine string sequence, perhaps targeting them to secretory vesicles and allowing them to participate in the proposed control of secretion by Gz. Phosphorylation of Gαz by protein kinase C inhibited the GAP activity of RGSZ1 and other RGS proteins, providing a mechanism for potentiation of Gz signaling by protein kinase C.

both G z and its GAPs may be involved in the control of transmitter release. Although G z can respond to G i -coupled receptors and inhibit adenylyl cyclase (17)(18)(19), its scarcity in tissues that are already well endowed with G i suggests that it is not just an ersatz G i and that the physiological signaling target for G z remains to be discovered.
As the next step toward understanding the function of G z GAP and its G z target, we cloned the cDNA that encodes the principal brain G z GAP isoform and compared the properties of the recombinant protein with those of the brain G z GAP and of related RGS proteins. This work establishes the major G z GAP as RGSZ1, an RGS protein whose specificity for G␣ z is unusually high among the RGS family. The data also show that RGSZ1 is closely related to RET-RGS1 and GAIP, suggests that all three are primarily GAPs for G z , and clarifies the interactions of RGSZ1 with membranes and the role of that interaction in its function. We further show that phosphorylation of G␣ z by protein kinase C (20,21) blocks the GAP activity of RGSZ1 and other RGS proteins, thereby providing a mechanism for the amplification of G z signals.

EXPERIMENTAL PROCEDURES
Isolation of Bovine Brain G z GAP-G z -selective GAP was purified from bovine brain through the final phenyl-Sepharose step exactly as described (12). Protein from several preparations, ϳ30,000 -50,000 units in GAP activity, was pooled and precipitated with acetone at Ϫ20°C to remove most of the Triton X-100. The precipitate was resuspended in Laemmli sample buffer (22) containing 20 mM DTT, and protein was denatured at room temperature for 30 min. The sample was then electrophoresed on a 12% acrylamide gel that contained 0.1% SDS. The gel lane was cut in about 30 slices (1 mm each), and a small piece of each slice was eluted and assayed for G z GAP activity as described (12). Slices with the highest GAP activity were eluted into SDS sample buffer and individually electrophoresed again to remove any residual detergent (other than SDS) and any other low molecular weight contaminants. Protein was then transferred electrophoretically in Tris glycine buffer to nitrocellulose (Schleicher & Schuell BA85) and visualized by Amido Black staining.
Peptide Analysis and Sequencing-Protein samples were digested on the nitrocellulose membranes essentially as described previously (23,24). Briefly, nitrocellulose bands were excised, cut in 1-mm squares, and incubated for 18 h at 37°C in 15-20 l of 15 mM N-ethylmorpholine, 5 mM acetic acid, 1% Zwittergent 3-16 (Calbiochem) that contained 1 g of sequencing grade modified trypsin (Promega). The supernatant was collected, and the membrane pieces were washed with another 15 l of the digestion buffer. After brief centrifugation, the supernatant was again collected, and both extracts were pooled for peptide analysis.
Peptide analysis was performed by reverse-phase high performance liquid chromatography using an Ultrafast Microprotein Analyzer micro-high performance liquid chromatography (Michrom BioResources Inc.) equipped with a 0.5 ϫ 50 mm Monitor C18 column (Column Engineering). Chromatography solvents were H 2 O:CH 3 CN:CH 3 COOH: CF 3 CF 2 CF 2 COOH, 97.5:2.5:0.05:0.005, v/v (solvent A), and H 2 O: CH 3 CN:CH 3 COOH:CF 3 CF 2 CF 2 COOH, 20:80:0.045:0.005, v/v (solvent B), and the flow rate was 15 l/min throughout the gradient. The column was washed isocratically for 2 min at 0% solvent B and then developed with a gradient of 0 -60% solvent B in 48 min. Peptide elution was monitored at 214 nm, and selected fractions were collected according to their UV absorbance. On-line mass spectrometric and tandem mass spectrometric analysis were performed on a Finnigan-MAT TSQ 7000 triple quadrupole mass spectrometer operated as described previously (25) using a 2% flow split between the UV cell and the fraction collection (26). The microsprayer was operated at 1.2 kV (no sheath gas), and the capillary temperature was set at 200°C.
Tandem mass spectra were processed after each run with SEQUEST software (27,28) using the non-redundant protein data base OWL (29) for the first pass. In case of inconclusive results, selected tandem mass spectra were processed against the genpept and the dbEST data bases (National Center for Biomolecular Information). Abundant peptides (Ͼ5 pmol) were also analyzed by automated Edman degradation using an Applied Biosystems protein sequencer model 477A following standard protocols.
cDNA Cloning-EST AA242973, prepared from RNA pooled from human fetal heart, pregnant uterus, and melanocytes, was purchased from Genome Systems. Pooled human placental cDNA tagged with priming sequences for rapid amplification of cDNA ends was obtained from CLONTECH. A human fetal brain cDNA library in ZAP II was purchased from Stratagene, and a HeLa cell cDNA library (30) in EXlox was a gift from Xiaodong Wang. Cloning and preparation of cDNAs for RET-RGS1 (31), GAIP (32), and RGS4 (33, 34) have been described by others. cDNA was amplified by PCR using the primers shown in Table I and the reaction conditions described previously (35). Products were evaluated in some cases by restriction mapping and then sequenced either directly or after cloning into appropriate vectors. mRNA from rat tissues, human retina, and cell lines was purified as described (35). A blot of mRNA from several human tissues and two blots of mRNA from human brain were purchased form Origene and Stratagene, respectively. The blots were hybridized with 32 P-labeled human RGSZ1 cDNA (entire coding sequence, 654 bp; 5 ϫ 10 6 cpm/ml) at 42°C for 16 h (35) and exposed either to a PhosphorImager plate for 16 h at room temperature or to film for 15 days at Ϫ85°C. The blots were then stripped (36) and rehybridized sequentially with 32 P-labeled cDNA probes for human G␣ z (NcoI fragment, 2476 bp; 5 ϫ 10 6 cpm/ml) and ␤-actin (Origene, 5 ϫ 10 6 cpm/ml) and exposed to film at Ϫ85°C for 5 days or 14 h, respectively.
Expression and Purification of Recombinant Proteins-To express RGSZ1 as a GST fusion protein, restriction sites for BamHI and HindIII were introduced immediately adjacent to the initiation and termination codons ( Table I) by PCR of the human fetal brain library. The product was ligated into pGEX-KG (Amersham Pharmacia Biotech). For His 6tagged versions of RGSZ1 and RET-RGS1, a restriction site for NcoI and a His 6 tag just before the original start codons and a restriction site for HindIII immediately after the termination codon (Table I) were introduced by PCR, and the products were ligated into pQE60 (Qiagen). DNA sequences were confirmed by sequencing. Expression of GST-RGSZ1, His 6 -RGSZ1, His 6 -RET-RGS1, His 6 -GAIP, and His 6 -RGS4 were induced by adding 1 mM isopropylthiogalactoside to mid-log cultures (A 600 ϭ 0.7-0.8). Cultures were harvested after 3 h at 37°C. A baculovirus encoding His 6 -RGSZ1 was prepared as described (18). Sf9 cells expressed this protein 80% in the soluble fraction when lysed as described (18). To purify His 6 -RGSZ1, Escherichia coli cells were allowed to lyse for 30 min at 0°C in buffer A (20 mM NaP i (pH 7.8), 300 mM NaCl, 10 g/ml leupeptin, 1 g/ml aprotinin, 0.1 mM PMSF) plus 1 mg/ml lysozyme and then sonicated with four 30-s bursts using a microprobe. After centrifugation at 35,000 ϫ g for 30 min, 2-mercaptoethanol was added to 10 mM and the supernatant was applied to NTA-Ni 2ϩ agarose (Qiagen) that had been equilibrated with buffer A plus 10 mM 2-mercaptoethanol. The column was washed sequentially with 1 M NaCl, 1 M NaCl and 10 mM imidazole, and 10 mM of imidazole alone in buffer A. RGSZ1 was then eluted with a 10 -150 mM imidazole gradient in buffer A. Pooled peak fractions were diluted 5-fold with buffer B (20 mM NaHepes (pH 7.5), 1 mM EDTA, 1 mM DTT, 0.1 mM PMSF, 10% glycerol) and adsorbed on a Mono Q column equilibrated in buffer B. Both RGSZ1 and substantial protein were eluted as a broad peak with a 50 -600 mM NaCl gradient in buffer B. Pooled active Mono Q fractions were concentrated on an Amicon PM30 membrane, incubated at 0°C for 60 min with 20 mM DTT and 5 mM EDTA, and chromatographed on an AcA34 column that had been equilibrated with buffer C (50 mM NaHepes (pH 7.5), 100 mM NaCl, 5 mM DTT, 1 mM EDTA, 0.1 mM PMSF). RGSZ1 was aggregated at this step and eluted before impurities. His 6 -RGSZ1 expressed in Sf9 cells using a baculovirus vector (18) was purified from the lysis supernatant (18) by the identical protocol. His 6 -RET-RGS1 and His 6 -GAIP were purified by the same procedure. His 6 -GAIP was over 80% pure after these steps, but His 6 -RET-RGS1 was only about 20% pure. His 6 -RGS4 was purified as described by Berman et al. (34). Protein concentrations were estimated according to Amido Black binding (37).
G␣ z , G␣ i1 , G␣ q , and G␣ s , with or without N-terminal myristoylation, were expressed in either Sf9 cells or E. coli as described previously (12,14). Unless stated in the text, all G␣ z and G␣ i subunits are myristoylated. M2 muscarinic cholinergic receptor (18) and G␤ 1 ␥ 2 (19) were expressed in Sf9 cells and purified as described. Other proteins were purified as described (12,14). G␣ z was phosphorylated to about 1.8 mol/mol by PKC␣ exactly as described previously (19). Phosphorylation was mapped by tryptic cleavage of GTP␥S-protected G␣ z after residue 29 (14).
Activity Assays-GAP assays, using either G␣ z -GTP or G␣ i -GTP as substrate, were performed and analyzed as described (12,14). For the standard single-turnover GAP assay, data are presented either in GAP units, an increase in the rate constant for GTP hydrolysis of 1 min Ϫ1 (12), or as moles of G␣-GTP hydrolyzed per min. GAP activities are also analyzed in analogy with enzyme kinetics to yield a V max and K m , where the GAP is considered as an enzyme that binds G␣-GTP substrate and produces G␣-GDP and P i products (12). Reconstitution of purified muscarinic cholinergic receptors and heterotrimeric G proteins into phospholipid vesicles, with or without GAPs, was performed as described (12,18).
Electrophoresis and Immunoblotting-Antiserum against intact His 6 -RGSZ1 (U114) was raised in rabbits and used at 1:40,000 dilution. Antiserum against GAIP (R381; used at 1:2500) (39) was a gift from Susanne Mumby and antiserum against RGS4 (N-16; used at 1:2500) was purchased from Santa Cruz Biotechnology. Immunoblots from 12% acrylamide gels were prepared as described by Mumby et al. (40). For RET-RGS1 and GAIP, which were only partially purified, amounts of protein applied to electrophoresis gels was normalized according to Coomassie Blue staining of the specific band. Samples were denatured, reduced by DTT, and alkylated by N-ethylmaleimide (18). Silver staining was performed according to Wray et al. (41). Blots were probed with antiserum in 10% blocking solution (40) and developed according to instructions in the ECL kit (Amersham Pharmacia Biotech).

RESULTS
Isolation of cDNA That Encodes the Principal Brain G z GAP-When G z GAP that had been purified approximately 10,000-fold from bovine brain was analyzed by SDS-gel electrophoresis, both GAP activity and stained protein were distributed in two broad peaks in the region that corresponded to 22-28 kDa. When the gel was sliced and the contents of each slice eluted and electrophoresed again, both activity and protein migrated as distinct single bands at their original positions, indicating that the purified GAP is heterogeneous in size (12). Because the heterogeneity might result from partial proteolysis, other covalent modifications, or the existence of multiple distinct GAP proteins, we separated the purified GAP into separate size fractions and analyzed each fraction separately (ϳ20 1-mm slices; Ͻ1 kDa each). To obtain peptide sequence from each electrophoretic fraction, eluted protein was electrophoresed a second time and transferred to nitrocellulose, and individual Amido Black-stained bands were digested with trypsin and analyzed as described under "Experimental Procedures." The recovery of peptides was unusually low based on the intensity of Amido Black staining, suggesting that the GAP protein(s) was refractory to proteolysis or that its peptides stuck irreversibly to the nitrocellulose (or both). Inefficient extraction is common for peptides from hydrophobic proteins. Nevertheless, multiple peptides were sequenced from nine separate protein size fractions. Peptide sequence from three separate protein bands indicated the presence of an RGS protein closely related to both human GAIP (32) and bovine RET-RGS1 (31). One peptide that was found in three gel slices, with sequence VREVINREVSLDSR, corresponded exactly to a human EST (AA242973). Upon complete sequencing, the EST was found to encode the C-terminal region of an RGS protein, defined according to the sequence of the RGS box, approximately 120 amino acid residues that are conserved throughout the RGS family (3,4). The EST included the last 310 base pairs of the ORF followed by a termination codon, 746 base pairs of 3Ј-non-coding sequence, a consensus polyadenylation signal, and a poly(A) track (Fig. 1B). We then used primers based on this sequence to PCR-amplify the complete cDNA. We refer to the product of this cDNA as RGSZ1 based on its selective GAP activity toward G␣ z and the likelihood that more than one RGSZ homolog exists in mammals (see below).
We amplified the remaining 5Ј region of RGSZ1 cDNA (residues Ϫ68 through 423; Fig. 1B) from a HeLa cell cDNA library using an antisense primer in the EST ORF (AS2) and a sense primer in the cloning vector (ST7). The resulting cDNA contained one long ORF that encodes a new RGS protein. Its length, 1497 base pairs plus poly(A), agrees with that of the only detectable mRNA with which it hybridized ( Fig. 2A). The first methionine codon in the major ORF falls after an in-frame termination codon and at the position of the proposed N terminus of GAIP (32). To confirm the validity of the cDNA, we amplified it independently from a second library derived from human fetal brain. The sequence of the entire ORF was confirmed using two independent PCR products from each library generated with the primer pairs S1/AS3 and S1/AS1 or S2/AS1. 3Ј-Non-coding sequence from the EST was confirmed by PCR of both libraries using primers S3/AS7 to produce products of the correct size that contained predicted restriction sites. Noncoding sequence 3Ј to the primer AS7 was not independently isolated.
Multiple PCR products from the HeLa cell library prepared using several antisense primers in the ORF and a vector sense primer (SEXLOX) produced only 68 bp of 5Ј-non-coding region (Fig. 1B). The overall length of the RGSZ1 cDNA is 1665 bp (Fig. 1B), the same size as the mRNA detected in multiple brain tissues assuming the presence of about 100 bases of 3Ј poly(A). The length of the 5Ј-untranslated region was confirmed by 5Ј-rapid amplification of cDNA ends of placental cDNA using two antisense primers in the ORF, AS2, and AS5 (not shown). Although one PCR product from the fetal brain library contained a divergent 5Ј-non-coding sequence, it appears to be an artifact of either library construction or of incomplete splicing. It diverged from all other cDNA isolates at a consensus splice acceptor site, was rich in Alu repeats, and extended for about 1100 bp. It would thus give a total mRNA size of 2.6 kilobase pairs, much longer than detected by Northern blot ( Fig. 2A). We therefore believe that the 5Ј-non-coding region of the RGSZ1 cDNA is that shown in Fig. 1B.
RGSZ1 mRNA Is Expressed at Low Levels in Select Brain Regions-The major RGSZ1 mRNA, 1700 nucleotides in length, was detected by Northern blotting only in brain, where it is  Table I. B, cDNA and protein sequences (Gen-Bank TM accession number AF060877). The underlined protein sequence was determined in peptide fragments of G z GAP purified from bovine brain. Asterisks denote sites where three introns were found (partial sequences of introns have Gen-Bank TM accession numbers AF071507, AF071508, and AF071509). The same cDNA has been cloned by Glick et al. (65). C, alignment of the protein sequences of human RGSZ1, bovine RET-RGS1 (residues 157-374) (31), and human GAIP (32). present at low levels ( Fig. 2A). Spleen and retina both contain G␣ z and substantial G z GAP activity (12), but we did not detect RGSZ1 mRNA in either tissue ( Fig. 2A, retina blot not shown). However, the lower limit for detection of this or other mRNAs was only about 20% the amount found in brain, and low levels of RGSZ1 or greater amounts of RGSZ1 homologs may be present elsewhere.
RGSZ1 mRNA was found in multiple areas of human brain (Fig. 2B), with the highest concentrations in the caudate nucleus and the temporal lobe. We also probed the same RNA blots for G␣ z mRNA, whose localization in various brain regions had not been described. There was a rough correlation of expression of G␣ z and RGSZ1, but several regions that express significant amounts of G␣ z mRNA contained negligible amounts of RGSZ1 mRNA. In rat brain, G z GAP activity was also broadly distributed in cerebellum, olfactory bulb, hippocampus, midbrain, caudate nucleus, cortex, and basal ganglia (11-47 units/mg protein in crude membrane fractions, listed in order of increasing activity). It is therefore plausible that additional G z GAPs are expressed in regions where RGSZ1 is not abundant. Consistent with this possibility, we have detected a peptide in the brain G z GAP preparation that is related, but not identical, to RGSZ1. We have not yet isolated its cDNA. Alternatively, RGSZ1 may serve a regulatory role in some brain region and not in others.
Antibodies to Recombinant RGSZ1 Recognize Brain G z GAP Proteins-Although it is likely that multiple G z GAPs are expressed in peripheral tissues and in distinct brain regions, RGSZ1 appears to account for the majority of the activity purified from cerebral cortex. We raised antibodies to recombinant RGSZ1 and used them to analyze the protein composition of partially purified G z GAP activity from bovine brain. This antiserum is quite sensitive and is more than 10-fold selective for RGSZ1 over both GAIP and RET-RGS1 (Fig. 3B, upper panel; note that blots of GAIP, RGS4, and RET-RGS1 were exposed 10 times longer than blots of RGSZ1 or of bovine brain G z GAP).
When partially purified bovine brain G z GAP was electrophoresed in SDS, activity was distributed in two peaks with apparent sizes of 22 and 28 kDa (12) (protein stain in Fig. 3B, upper panel, lane g). Anti-RGSZ1 antibodies reacted with two components of the same sizes, a somewhat fuzzy band at about 22 kDa and a broad band at about 28 kDa. The upper band is sometimes smeared and sometimes is an obvious doublet (Fig.  3B, upper panel). In this and multiple other experiments, integration of signal from the multiple cross-reacting bands from bovine brain ranged from 30 to 80% that obtained from samples of pure RGSZ1 that had the same G z GAP activity. It thus appears that RGSZ1 or a closely related protein contributes all or most of the G z GAP activity purified found in brain. The After incubation with secondary antibody and chemiluminescence reagents, lanes with RGSZ1 or brain G z GAP (lanes a, b, and f) were exposed to film for 2 min and lanes with GAIP, RGS4 or RET-RGS1 (lane c, d and e) were exposed for 20 min. The blots were then stripped and reprobed sequentially with an antibody against GAIP (middle panel) and RGS4 (lower panel).
slightly lower signal obtained probably reflects limited crossreactivity of anti-human antibodies with bovine RGSZ1, but it may result from the presence of other G z GAPs in brain.
Structure of RGSZ1-The RGSZ1 cDNA encodes an RGS protein of 218 amino acid residues (calculated M r ϭ 24,822) that is closely related to both RET-RGS1 and GAIP (Fig. 1C). The cDNA contains four in-frame Met codons near the 5Ј end, the first of which appears to have the best consensus eukaryotic translational initiation site (42). If the first Met is the initiation site, then the N terminus of RGSZ1 is likely to be myristoylated Gly 2 in animal cells (43). The first Met codon corresponds to the only appropriate in-frame Met codon in GAIP, although it and two others appear in RET-RGS1 (which has a 156-amino acid N-terminal extension such that none of the three is the initiation site (31)) (Fig. 1C). Multiple initiation sites may help explain the size heterogeneity of G z GAP found in brain (Fig. 3B). N-terminal heterogeneity would not have been detected in recombinant RGSZ1 described here because of the N-terminal His 6 and GST fusions.
The region of RGSZ1 N-terminal to the RGS box (residues 1 to ϳ87) is distinguished by a cysteine string (9 of 13 residues) that is also found in GAIP and RET-RGS1. Cysteine strings were originally noticed on a DnaJ-related protein that is tightly bound to secretory vesicles (44,45), but they can be found in the sequences of multiple membrane proteins. Cysteine strings are potential sites for multiple palmitoylation (46 -48) and might contribute to the size heterogeneity and hydrophobicity of RGSZ1, although the size dispersion of brain G z GAP was not altered by treatment with NH 2 OH (data not shown). The Nterminal region of G z GAP is itself not notably hydrophobic and is unlikely to span the membrane bilayer.
A G z GAP Subfamily of RGS Proteins-Comparison of predicted protein sequences indicates that RGSZ1, GAIP, and RET-RGS1 form a distinct subfamily within the RGS proteins. Human RGSZ1 is 83% identical to the cognate region of bovine RET-RGS1 and 58% identical to human GAIP, including Nterminal regions outside the RGS box. They also share multiple sequence similarities within the RGS box that diverge from other RGS proteins, notably a serine instead of a conserved asparagine at the contact site with the switch II region on the G␣ subunit. This residue is important for GAP function (49 -51). Such conservation of sequence predicts that both GAIP and RET-RGS1 are also G z -selective GAPs.
Sequence comparison in the G z GAP subfamily further defines four regions of differing similarity that potentially constitute subdomains in each protein as follows: the RGS box, the C-terminal 20 residues, and two distinguishable regions in the N-terminal 100 residues. Over the generally conserved RGS box and the C-terminal extension, RGSZ1 is 92% identical to RET-RGS1 and 75% identical to GAIP. Between residues 33 and 87, similarity is somewhat less, 75% identity with RET-RGS1 and 45% with GAIP. This putative domain includes the cysteine string sequence. Over the first 32 residues, RGSZ1 is 59% identical to RET-RGS1 but not obviously related to GAIP. RGSZ1 and GAIP both lack the 156-residue N-terminal extension found in RET-RGS1. The great similarity of these three proteins suggested that the N-terminal extension of RET-RGS1 may be encoded by a distinct and potentially variably spliced exon. However, we have not found an RGSZ1 cDNA that encodes sequence similar to the N-terminal domain of RET-RGS1 nor did Faurobert and Hurley (31) report a short form of RET-RGS1 mRNA. We also did not detect a RGSZ1 transcript large enough to encode a protein the size of RET-RGS1 ( Fig. 2A). We did detect a small amount of G z GAP activity in bovine brain with an electrophoretic mobility corresponding to about 43,000 Da (12). This may correspond to RET-RGS1 itself, a novel member of the RGSZ1 family, or a splicing isoform of RGSZ1 itself whose mRNA was undetectable by Northern blot.
G z GAP activity of RGSZ1, RET-RGS1, and GAIP-To study the biochemical properties of RGSZ1, it was expressed in three forms as follows: an N-terminally His 6 -tagged protein in both E. coli and Sf9 insect cells and an untagged protein (after cleavage of GST) in E. coli. Each form was highly purified (Fig.  3). For comparison, we also expressed and purified His 6 -tagged forms of GAIP and RGS4 (80 and 95% pure) and partially purified RET-RGS1 (ϳ20% pure). RGSZ1 was expressed at high levels (ϳ10% of soluble protein) in both E. coli and Sf9 cells and was found mostly in the supernatant fraction (ϳ80% by activity assays and Western blots). Soluble RGSZ1 was readily purified from either source in the absence of detergent despite a tendency to aggregate (see below). We have not observed significant differences in the behavior of recombinant RGSZ1 from E. coli or Sf9 cells, with or without the His 6 tag. Residual RGSZ1 in the lysis pellets has not been characterized and may include RGSZ1 that is either legitimately bound to membranes or cytoskeleton, aggregated, or artifactually trapped.
Recombinant RGSZ1 displays the characteristic features of the G z GAP purified from brain and shares extensive functional similarity with RET-RGS1. RGSZ1 and RET-RGS1 both are GAPs for G␣ z with K m values of about 2.6 nM (Fig. 4A), similar to that reported for G z GAP from brain. Such a low K m , approximately equal to K s , indicates both that RGSZ1 binds to G␣ z -GTP with high affinity and that the rate of GTP hydrolysis in the RGSZ1⅐G␣ z ⅐GTP complex is not faster than the rate at which the GAPs bind G␣ z -GTP (12,52). The maximum GAPstimulated hydrolytic rate, calculated from V max , is approximately 4.5 mol of G␣ z -GTP hydrolyzed per min/mol of RGSZ1 and 4.0 mol/min/mol of RET-RGS1 (Fig. 4A). GAIP has K m (2.6 nM) and V max (4.0 min Ϫ1 ) values similar to those of RGSZ1 and RET-RGS1 (not shown). When a fixed concentration of G␣ z -[␥-32 P]GTP was titrated with increasing concentrations of RGSZ1, the maximum hydrolytic rate constant was 6 min Ϫ1 (not shown). This value is presumably more accurate because its calculation is based both on the concentration of G␣ z -[␥-32 P]GTP, which is determined directly, and on the shape of the titration curve (12). The slight discrepancy between the two values, 4.5 versus 6.0 min Ϫ1 , presumably indicates that only about 75% of the purified GAP is active, perhaps because of aggregation (see below). Both RGSZ1 and RET-RGS1 thus stimulate the rate of hydrolysis of G␣ z -bound GTP over 400fold, to a maximum of 6 min Ϫ1 at 15°C. This is somewhat higher than the V max reported for G z GAP purified from brain (12), probably reflecting an overestimate of the purity of that preparation. We have not measured the GAP-stimulated rate of hydrolysis of G␣ z -GTP at physiological temperatures. However, if basal and GAP-stimulated hydrolytic rates for G␣ z are similarly temperature-dependent, then the GAP-stimulated turnoff rate for G z will be above 40 min Ϫ1 at 30°C, giving a deactivation t1 ⁄2 of about 1 s.
The GAP activities of RGSZ1 and RET-RGS1 are stimulated about 4-fold by Mg 2ϩ , with a sharp optimum at 1 mM (Fig. 4B). This response is identical to that of G z GAP purified from brain (12) or of RGSZ1 expressed in Sf9 cells (not shown). Individual size fractions of brain G z GAP separated by SDS-gel electrophoresis (12) also displayed this pattern of Mg 2ϩ sensitivity, as did total G z GAP activity in bovine brain membranes and in partially purified fractions prior to G␣ z affinity chromatography (not shown). GAIP was much less sensitive to Mg 2ϩ than RGSZ1; it was reproducibly stimulated less than 50% at 0.1 mM Mg 2ϩ . It therefore appears that GAIP does not contribute substantially to total G z GAP activity in brain membranes or extracts because its behavior would decrease the relative Mg 2ϩ response in these fractions.
Fine control of GAP activity by Mg 2ϩ is apparently unique to the interaction of G z -selective GAPs with G␣ z -GTP. The GAP activity of RGSZ1 with G␣ i -GTP increased monotonically only 50 -70% as the Mg 2ϩ concentration over the range 0 -5 mM (not shown). RGS4 and RGS10 were not stimulated by Mg 2ϩ under any conditions tested (RGS10 data not shown). The G z GAP activities of all RGS proteins tested were inhibited by Mg 2ϩ concentrations above 1 mM (Fig. 4B).
Selectivity of RGSZ1 for G␣ z -Recombinant RGSZ1 is highly selective for G␣ z over other G␣ subunits and actually displays somewhat greater selectivity than does the G z GAP purified from brain. RGSZ1 displays a K m for G␣ z -GTP of 2.6 nM (Fig.  4A). In contrast, we were unable to demonstrate saturation with G␣ i1 -GTP (not shown). Maximal stimulation of the G␣ i -GTP hydrolytic rate was also low, 30 -50% acceleration by 15-20 nM RGSZ1 in multiple experiments. This concentration of RGSZ1 accelerates hydrolysis of G␣ z -GTP several hundredfold ( Fig. 4A) (Ref. 12). For comparison, 1.6 nM RGS4 accelerated activity by 50 -75%; it is thus at least a 10-fold better G i GAP than is RGSZ1, and it is about 10% as good a G z GAP. The activity of RGSZ1 on either G␣ z or G␣ i was not altered by the presence of the His 6 tag (not shown).
Much of the specificity of RGSZ1 for G␣ z stems from its high selectivity of binding. Binding can be measured according to competitive inhibition in a G z GAP assay, where the inhibition constant K i is equal to the equilibrium binding constant K d (12,14). In such assays, both RGSZ1 and G z GAP from brain display a K i for G␣ z -GTP␥S of about 2 nM, equal to or somewhat lower than the K m for G␣ z -GTP (Fig. 5). RGSZ1 binds G␣ i1 -GTP␥S about 100-fold less tightly than G␣ z (K i Ͼ 150 nM, Fig.  5B). RGSZ1 was essentially insensitive to the GTP␥S-bound forms of G␣ o , G␣ q , and G␣ s (Table II), again indicating greater selectivity than the brain G z GAP (compare with Ref. 12). The selectivity of recombinant RGSZ1 for G␣ z thus exceeds that of G z GAP purified from brain. RGSZ1 produced in E. coli and Sf9 cells, with and without the His 6 tag, displayed similar high selectivity among G␣ subunits. G␣ z -GDP-AlF 4 bound RGSZ1 with the same affinity as did G␣ z -GTP␥S (data not shown; see Ref. 12). The K i for G␣ i1 -GDP-AlF 4 was 20 -40 nM, indicating tighter binding than displayed by the GTP␥S-bound form but still 10-20-fold lower affinity than that of G␣ z . The difference between the two active forms of G␣ i may reflect the similarity of G␣ i1 -GDP-AlF 4 to the transition state (49,53).
Myristoylation of the G␣ z -GTP substrate enhances its affinity for RGSZ1, as described for the natural G z GAP (12,14). In competition studies, the selectivities of RGSZ1 and brain G z GAP for myristoyl-G␣ z were similar, but RGSZ1 was less selective than brain Gz GAP between the myristoylated and non-myristoylated forms of G␣ i . This behavior is unexplained but has been reproduced in multiple experiments. Palmitoylation of G␣ z markedly inhibits its response to the GAP activity of RGSZ1, as described previously for brain G z GAP (14). We have not studied the G protein selectivity of GAIP or RET-RGS1 on other G␣ subunits in detail. GAIP acts as a GAP for G i and, somewhat less well, for G q (34,54) and seems less selective for G z in our hands than do RGSZ1 and RET-RGS1. The core domain of RET-RGS1 was shown to act as a G t GAP, but it accelerated hydrolysis of G␣ t -GTP only about 5-fold with EC 50 ϳ1 M (31); RGS9 appears to be a better candidate for the physiological G t GAP (55,56).
Association of RGSZ1 with Phospholipid Bilayers-Although soluble RGSZ1 is active in the single turnover GAP assay that uses detergent-solubilized G␣ z -GTP as substrate, it depends on tight association with bilayers for its ability to regulate membrane-bound, heterotrimeric G z . We initially found that addition of brain G z GAP did not accelerate the steady-state GTPase activity of reconstituted phospholipid vesicles that contained G z and m2 muscarinic cholinergic receptor. However, co-reconstitution of the GAP into the vesicle membrane increased agonist-stimulated GTPase activity about 2-fold (30 pM final GAP concentration; Ref. 12). When purified, recombinant RGSZ1 was co-reconstituted with lipid, receptor, and G z , GTPase was accelerated about 17-fold (Fig. 6B). Reconstituted RGSZ1 was half-maximally effective at or below about 12 nM, relatively near its intrinsic K d for G␣ z -GTP. (This concentration is a functional overestimate because we cannot distinguish co-reconstituted RGSZ1 from aggregated RGSZ1 that co-eluted in the vesicle fraction.) When RGSZ1 was simply added to the vesicles, however, it only accelerated GTPase activity up to a maximum of 3.5-fold. There was no effect until its concentration reached 20 nM, and stimulation did not saturate at 3 M RGSZ1 (Fig. 6A). Reincorporation of RGSZ1 into membranes thus increases its potency for trimeric, membrane-bound G z by about 1000-fold.
The functional importance of the association of RGSZ1 with membrane bilayers is consistent with its general behavior as an integral membrane protein. Brain G z GAP was not solubilized by either high ionic strength or by 1% cholate or several other detergents. Its solubilization required incubation with 2% Triton X-100 at 30°C; the Triton-dispersed protein was still badly aggregated, and detergent was required throughout the purification (12). Surprisingly, the RGSZ1 sequence did not indicate obvious membrane-spanning regions, and overexpressed RGSZ1 is found predominantly in the supernatant fraction of E. coli and Sf9 cell lysates. However, recombinant RGSZ1 aggregated markedly at all stages of purification. Apparently soluble RGSZ1 eluted in the void volume of Ultrogel AcA34 (Stokes radius Ͼ80 Å; apparent M r Ͼ10 6 ). Although the RGSZ1-GST fusion protein remained largely soluble, RGSZ1 with or without the His 6 tag aggregated further during storage and eventually became resistant to dissociation with SDS (Fig.  3A, for example).
Aggregation of RGSZ1 was partially reversed by addition of SDS, particularly when cholate or Triton X-100 was added to  . 6. Potentiation of G z GAP activity by co-reconstitution into phospholipid vesicles. M2 muscarinic cholinergic receptors and trimeric G z , either alone (A) or with increasing concentrations of RGSZ1 (B), were co-reconstituted into unilamellar phospholipid vesicles. Steady-state GTPase activity was assayed in the presence of 1 mM carbachol (CCh) or 10 M atropine (Atr). Note that steady-state GAP activity is not apparent unless agonist is added to promote exchange of GDP for GTP. A, detergent-free RGSZ1 was added to the vesicles at the concentration shown on the abscissa immediately before assay. B, the amount of RGSZ1 in each of the seven batches of vesicles was determined by an activity assay on a separate aliquot that was lysed in 0.5% Triton X-100. Concentrations of G z and receptor were assayed as described (2,12). All assays contained 1.1-1.5 nM G z and 0.27-0.40 nM receptor. Data shown are means of duplicates where the range was less than 5% of the mean. The complete titration of reconstituted RGSZ1 with careful maintenance of the concentrations of receptor and trimeric G z (B) has been performed only once as shown (vesicles assayed twice, in duplicate), and the potentiative effects of reconstitution on maximum steady-state GAP activity and on RGSZ1 potency have been confirmed in multiple experiments. The experiments shown were performed with His 6 -RGSZ1 but dependence on co-reconstitution was confirmed using unmodified RGSZ1 (first purified as the GST fusion protein) at a single concentration. Co-reconstituted, untagged 14 nM RGSZ1 stimulated GTPase activity 11.6-fold, compared with 9.7-fold stimulation shown for 13 nM His 6 -RGSZ1. In a separate experiment, added untagged RGSZ (400 nM) stimulated activity 2.0-fold, whereas His 6 -RGSZ1 stimulated 2.3-fold. all subsequent buffers. After such treatment, about 10% of RGSZ1 was monomeric without detergent and about 75% was monomeric in 1% cholate (Fig. 7). We also found that extended storage of RGSZ1 in 0.1% Triton both slowed aggregation and gradually increased GAP activity as much as 3-fold in comparison to RGSZ1 that had been stored without detergent. The structural basis of such apparently hydrophobic behavior is unknown. It is unlikely that the cysteine-rich tridecapeptide cross-linked RGSZ1 either covalently or by multi-protein chelation of contaminating metal ions because it was not disaggregated by incubation with 5 mM EDTA plus 20 mM DTT followed by gel filtration in the same buffer (data not shown). Although palmitoylation of RGSZ1, either in the cysteine-rich sequence or elsewhere, might add to its apparent hydrophobicity (48,57), most of the palmitate would have been removed by treatment with DTT (14). Thus, RGSZ1 behaves as a functionally hydrophobic membrane protein despite its lack of obvious, strongly hydrophobic contiguous sequences.
Regulation of G z GAPs by Phosphorylation of G␣ z -G␣ z can be phosphorylated near its N terminus by PKC in cells (20,58), and PKC-catalyzed phosphorylation alters the affinity of G␣ z for the G␤␥ subunits (59). Because the N terminus of G␣ z is important in determining the affinity of GAPs for G␣ z (14) and because G␤␥ inhibits the action of G z GAP from brain (12), we tested the effects of PKC-mediated phosphorylation of G␣ z on RGSZ1 and other G z GAPs. We first demonstrated that purified PKC readily phosphorylates purified G␣ z in vitro to ϳ1.8 mol of total phosphate per mol of G␣ z (not shown) and showed by tryptic mapping that about 70% of the phosphorylation is found in the N-terminal 29 amino acids (serines 16, 25, or 27), with 30% in the remainder of the protein. Assignment of this site agrees with the CNBr mapping data reported by Lounsbury et al. (20). We then tested the GAP responsiveness of phospho-G␣ z .
Phosphorylation of G␣ z by PKC largely blocked the ability of any of several RGS proteins, RGSZ, RET-RGS1, GAIP, or RGS4, to accelerate GTP hydrolysis and consequent deactivation ( Fig. 8 and Table III). Although maximal stimulation by the various GAPs varied both intrinsically and according to their concentrations, the GAP-stimulated rate of hydrolysis was decreased dramatically in each case. RGSZ1, brain G z GAP, and RET-RGS1 were inhibited by about 80%; GAIP was inhibited more strongly, and RGS4 was completely inhibited. Increasing the concentration of GAP did not decrease the fractional inhibition by phosphorylation (not shown). Preliminary experiments indicate that phosphorylation of G␣ z inhibits GAP activity largely by increasing the K m , but we have not been able to prepare enough phospho-G␣ z to show that there was no additional effect on V max . The fact that RGS4 is more strongly inhibited by phosphorylation of G␣ z is consistent with an effect of phosphorylation on affinity since RGS4 binds G␣ z with the lowest affinity of the several GAPs tested. In contrast to GAPpromoted GTP hydrolysis, phosphorylation of G z did not inhibit the basal rate of GTP hydrolysis by G␣ z (Table III), indicating that the effect of phosphorylation blocks responsiveness to GAPs rather than altering the hydrolysis or deactivation events themselves.
G␤␥ also decreases the affinity of GAPs for G␣ z -GTP, probably by forming the G␣ z ␤␥ heterotrimer (12). This effect of G␤␥ on GAP activity was also general, all recombinant GAPs and G z GAP from brain were inhibited to similar extents ( Fig. 8 and Table III). Again, RGS4 action was blocked most potently by G␤␥, consistent with its lower affinity for G␣ z (IC 50 ϳ20 nM for RGS4 and ϳ200 nM for RGSZ1). We have not been able to titrate inhibition over a wide enough range of G␣ z and G␤␥ concentrations to demonstrate whether G␤␥ and GAPs bind competitively to G␣ z or whether the G␣ z ␤␥ heterotrimer binds GAPs with very low affinity. G␤␥ did not further inhibit GAP activity when added to phospho-G␣ z . Such lack of additivity suggests that whatever GAP interaction is blocked by G␣ z phosphorylation is also blocked upon heterotrimer formation. DISCUSSION This study defines RGSZ1 as the major G z GAP activity in brain and links it with two other known RGS proteins, RET-RGS1 and GAIP, as members of a family of G z GAPs that probably contains at least two other members. Because G z hydrolyzes bound GTP so slowly, the activity of RGSZ1 and other G z GAPs is necessary for G z to function as a physiological signaling molecule with reasonable temporal acuity. Understanding the mechanisms of action and regulation of GAP is crucial to understanding further how G z itself is regulated, and the ability to express and control G z GAP activity may help us focus on the signaling functions unique to G z , which still lacks recognized targets.
RGSZ1, or very closely related homologs, appears by multiple criteria to be the major G z GAP activity in brain. The RGSZ1 cDNA was identified according to peptide sequence obtained from brain G z GAP, and antibodies raised against RGSZ1 cross-react with the multiple polypeptides in partially purified brain G z GAP with nearly quantitative intensity (Fig.  3). The specific activity of RGSZ1 is equal to that of the brain preparation (corrected for fractional purity), and selectivity of RGSZ1 for G␣ z over other G␣ subunits is actually somewhat better that of the brain G z GAP. The GAP activity of GAIP with G␣ z -GTP is as high as that of RGSZ1, and GAIP may also be primarily a G z GAP. GAIP probably does not contribute in a major way to the G z GAP activity in cerebral cortical membranes, however, based on its unresponsiveness to millimolar Mg 2ϩ (Fig. 4). Whereas RGSZ1, the G z GAP activity in brain membranes, and all size fractions of purified brain G z GAP were each stimulated about 4-fold by 1 mM Mg 2ϩ , GAIP was not. If GAIP contributed to total brain G z GAP activity substantially (Ն25%), the activity in crude brain membrane fractions would have had a detectably diminished Mg 2ϩ response. For example, G z GAP activity in C6 glioma cells is stimulated only 2-fold at 1 mM Mg 2ϩ , suggesting that about half of the activity is contributed by a Mg 2ϩ -responsive GAP and half by an unresponsive GAP (or that it all derives from yet another GAP with intermediate Mg 2ϩ sensitivity). Thus, GAIP may act primarily on other G proteins (32,48), or its G z GAP activity may require an additional activating factor, perhaps analogous to the way in which the G t GAP activity of RGS9 in photoreceptor cells is activated by cyclic GMP phosphodiesterase (55).
It is likely that other RGSZ isoforms also exist, perhaps in significant amounts, in brain or elsewhere. Peptide sequence of a still uncloned protein that co-purified with RGSZ1 from bovine brain represents at least one more members of the G z GAP subfamily. Because the distribution of G␣ z in the brain is broader than that of RGSZ1 (Fig. 2), it is likely that other homologs are represented in these areas. For example, RET-RGS1 is probably the GAP for G z in retinal interneurons, where both RET-RGS1 and G␣ z are expressed (31, 60) but where RGSZ1 mRNA was undetectable. Distinct G z GAPs may also be found in peripheral tissues that express G z , adrenal medulla, platelets, etc. (13, 15, 16, 61).
The size of brain G z GAP is heterogeneous according to both GAP activity and RGSZ1 immunoreactivity (Fig. 3). Heteroge-neity is not an artifact of purification (12). More than one of the four in-frame Met codons may be used for initiation, the initial translation product may be partially proteolyzed in cells, other covalent modifications may alter electrophoretic mobility, or the size isoforms may reflect multiple RGSZs or RGSZ1 splice products. Although the RGSZ1 gene contains at least three introns (Fig. 1), we found no evidence of multiple RNAs, but rare species would have been undetectable.
The G z GAP Subfamily-The primary structures of RGSZ1, RET-RGS1, and GAIP clearly define a subfamily of RGS proteins, all of which are active, high affinity, and relatively selective GAPs for G z . Their sequences are distinctive in the generally conserved RGS box, which includes a major site of interaction with the Switch II region of the G␣ subunit (49). Their uniqueness includes replacement of a generally conserved Asn at the contact site by a Ser (Ser 157 in RGSZ1). This substitution, which diminishes the activities of other RGS proteins (50,51), may be important for selective regulation of G␣ z , which itself contains several unusual amino acid residues near the active site and which hydrolyzes bound GTP very slowly. The regulation of the G z GAP subfamily by Mg 2ϩ may also reflect the distinct features of their RGS box.
The G z GAP subfamily is also distinctive in its sequence both C-terminal and N-terminal to the RGS box. Divergence of sequence in the N-terminal region of RGS proteins suggests that this domain may determine some of the distinct functions of the group: selectivity among G proteins, subcellular localization, covalent or allosteric regulation, or yet unknown regulatory functions not directly related to G proteins. Potential multiple palmitoylation of the cysteine string (residues 37-49) may influence subcellular localization (48,57), although all three proteins are relatively hydrophobic even when expressed in bacteria. Palmitoylated cysteine strings are thought likely to target the neuronal cysteine string dnaJ homolog to the secretory vesicle (44,46). Elucidation of the domains that confer these functions awaits study of suitable chimeras with sequences of other RGS proteins.
The RGSZ family is primarily marked by high selectivity for G z . RGSZ1 is at least 20-fold, and probably over 100-fold, selective for G␣ z over other G i family members in our standard solution assay and has even lower affinity for G q and G s than for G i . Its tissue distribution also argues for its physiological linkage to G z . Nevertheless, the selectivity of RGS proteins for their G protein targets remains inadequately understood. The activity of RGSZ1 is strongly influenced by its hydrophobic environment (Fig. 6). At least for RGS2, apparent selectivity can be dramatically altered by their association with phospholipid bilayers that contain trimeric G protein (62). Such changes have not been observed for RGSZ1, however. Choice of detergent can also alter apparent preferences of a G protein among various GAPs (52), and subcellular localization may TABLE III Modulation of G z GAP activity of RGS proteins by G␤␥ and by PKC-catalyzed phosphorylation of G␣ z GAP-stimulated hydrolysis rates were measured as described in the legend to Fig. 8 using G␣ z that was or was not phosphorylated by PKC␣. Assays were performed with or without 2.5 M G␤ 1 ␥ 2 . Data show the first-order hydrolysis rate constants, k app (12). Data are means from 4 to 20 determinations, ϮS.D. The concentration of G␣ z -GTP was approximately 1.5 nM in all experiments. Concentrations of GAPs were RGSZ1 expressed in E. coli (Ec), 50 pM; RGSZ1 expressed in Sf9 cells (Sf), 50 pM; brain G z GAP, 50 pM; RET-RGS1, 50 pM; GAIP, 75 pM; RGS4, 750 pM. contribute to specificity in vivo. Interaction of the RGSZ Family with Membranes-The structure and behavior of RGSZ1 raise intriguing questions about its mode of association with membranes and its interaction with membrane-bound G z . G z GAP activity in brain membranes was resistant to solubilization by most detergents under standard conditions and was solubilized incompletely only by 2% Triton X-100 at 30°C (12). It was highly aggregated during all steps in purification. RGSZ1 also aggregated even in detergent solution but has no obviously hydrophobic sequence suggestive of a membrane span. RET-RGS1 also aggregated badly, and GAIP aggregated to a lesser extent (not shown). Moreover, RGSZ1 was essentially unable to regulate heterotrimeric G z in phospholipid vesicles unless it was incorporated into the membranes during their formation. Regardless, all members of the G z GAP subfamily were found mostly in the cytosol of E. coli or Sf9 cells when expressed in large amounts. These observations suggest that the intrinsic hydrophobicity of these proteins may act coordinately with affinity for a specific, low abundance attachment site in natural membranes to confer correct subcellular targeting. For example, preliminary data indicate that RGS4 expressed at very low levels in Sf9 cells is mostly particulate but that about half is soluble when it is overexpressed. 2 Regulation of G z GAPs-Although we have no direct information about how or when the GAP activity of RGSZ1 is regulated in cells, several potentially physiological processes control its function in vitro. Of these, palmitoylation and PKCcatalyzed phosphorylation of G␣ z are potentially the most important. G␣ z is phosphorylated in platelets in response to thrombin and thromboxane agonists (14,20) and, in response to phorbol esters, in AtT20 and RBL-2H3 cells, which are models for anterior pituitary and mast cells, respectively (63). Whereas G␣ z phosphorylation does not alter the basal rate of G␣ z -GTP hydrolysis appreciably, it markedly decreases GAP activity ( Fig. 8 and Table III). Potentially, inhibition of GAPs can prolong and amplify a G z signal many fold. This mechanism allows stimulation of PKC to potentiate G z activation by blocking G z GAPs in response to Ca 2ϩ and diacylglycerol, second messengers of PIP 2 phospholipase C signaling.
In addition to regulation by PKC-catalyzed phosphorylation, G z GAP activity is almost completely blocked by palmitoylation of the G␣ z target, allowing palmitoylation to protect G z from GAP-promoted deactivation (14). Such blockade is of particular interest in terms of the proposed linkage of the palmitoylationdepalmitoylation cycle to G protein activation (Refs. 14 and 64 and references therein). Although G␤␥ can also inhibit GAP activity (12) (Fig. 8), we do not know how the various complexes of receptor, G␣ z , G␤␥, and their effectors may modulate access to RGSZ1. Clearly G␤␥ is not a potent GAP inhibitor in a membrane-bound, multi-protein signaling system (Fig. 6). Last, control of the amount of RGSZ1 present in a cell may be a crucial component of its regulation; it displays high activity even though expressed at levels well below that of G␣ z itself. G z GAPs may prove to be as important and versatile regulators of G z -mediated signaling as are receptors.