Identification of the GTP-binding Protein Encoded by Gi3 Complementary DNA*

Three closely related, but distinct, GTP-binding pro- teins (G-proteins) are encoded by cDNAs arbitrarily designated Gil, Gi2, and Gi,. The in vitro translated products of mRNAs prepared from Gilt Gia, and Gi3 cDNAs migrate as 41-, 40-, and 41-kDa proteins, respectively, on sodium dodecyl sulfate-polyacrylamide gels. Antisera were raised against synthetic decapep- tides corresponding to a divergent sequence (residues 169-168 for Gil and Gis; 160-169 for Giz) of the three cDNAs and tested on immunoblots for reactivity with three purified G-proteins, G41 and GqO from brain and G41 from HL-60 cells. LD antisera (Gil peptide) react only with brain G41. LE antisera (Giz peptide) react only with brain G40, and SQ antisera (Gi3 peptide) react exclusively with HL-60 Gql. The results indicate that the 41-kDa G-protein purified from HL-60 cells differs from the purified brain 41-kDa protein and suggest that the HL-60 cell protein corresponds to that encoded by Gi3 cDNA. Heterotrimeric GTP-binding proteins (G-proteins)’ couple cell-surface receptors to effectors (1, 2). Distinct a-subunits bind guanine nucleotides and are thought to confer specificity in G-protein-receptor and -effector interactions. cDNA

major pertussis toxin substrate purified from neutrophils, as the protein encoded by Giz cDNA (6). In order to identify the putative product of Gi3 cDNA, and in an effort to develop reagents capable of distinguishing closely related G-proteins, we raised antisera against unique, synthetic peptides corresponding to Gil and Gi3, respectively. Using these antisera, we report here the tentative identification of a 41-kDa G-protein purified from HL-60 cells as the protein encoded by Gi3 cDNA.

EXPERIMENTAL PROCEDURES
Peptide Synthesis-a-subunit peptide8 (Table I) were prepared exactly as described previously (6). The @-subunit peptide ( Table I) was obtained from Peptide Technologies (Washington, D.C.) and was synthesized as the carboxyl-terminal amide.
Antisera-Peptide conjugation and immunization schedules were as previously described (6). The antisera used in this study are defined in Table I. Protein Purification-41-and 40-kDa G-proteins were purified from bovine brain essentially as described by Katada et al. (7). The 41-kDa G-protein purified from HL-60 cells (8) was kindly provided by R. Uhing, P. Polakis, and R. Snyderman (Duke University and Genentech).
I n Vitro Transcription and Translntwn-Rat cDNAs encoding Go, Gi2, G,, and Go (3) were the kind gift of D. Jones and R. Reed (Johns Hopkins University). The cDNAs were contained in the pGEM-2 vector (Promega Biotec). The sense strands of Gilt Giz, and Gi3 cDNAs were oriented 3' to the SP6 RNA polymerase promoter. The plasmids containing these cDNAs were linearized with HindIII restriction endonuclease (New England Biolabs). The Go cDNA sense strand was oriented 3' to the T7 RNA polymerase promoter; this plasmid was linearized with XmnI restriction endonuclease (New England linearized plasmid DNA and 10 units of either SP6 or T7 RNA Biolabs). I n uitro RNA transcription was performed using 2 pg of polymerase (Promega Biotec). The reaction included (final concentrations): Tris/HCl, (pH 7.4), 40 mM; MgCL, 6 mM; spermidine, 2 mM; NaCI, 10 mM; dithiothreitol, 10 mM; and ATP, CTP, UTP, 0.5 mM each. Methylguanylate cap structure (Pharmacia LKB Biotechnology Inc.) 0.25 mM, was added with 0.1 mM GTP at the beginning of the 2-h incubation at 37 "C. After 20 min, GTP, 0.15 mM, was added to give a final concentration of GTP of 0.25 mM. After 2 h reaction products were extracted with phenol/chloroform, precipitated with '/loth original volume 3 M sodium acetate (pH 6.0) and 2.5 times original volume ethanol, vacuum dried, resuspended in water treated with diethyl pyrocarbonate (Sigma), and stored at -20 "C.
In vitro translation of the RNA (9) was performed with the rabbit reticulocyte lysate system (Bethesda Research Laboratories). K+ and Mg2+ concentrations in the reaction were optimized for translation of 2 pg of each RNA. ~-[~~S]methionine (Amersham Corp.) was used at a final concentration of 1 mCi/ml. Incubation for 1 h at 30 "C was terminated by heating the reaction mixture at 100 "C in SDS-PAGE sample buffer for 5 min.
Other Methods-Cholate extracts of bovine brain membranes were prepared as described (6). SDS-PAGE and immunoblotting were performed as previously described (6), except that 10% polyacrylamide gels contained half of the usual concentration of Bis-acrylamide (final 0.13 g/lOO ml rather than 0.27). This was found to give better resolution of a-subunits (10).

RESULTS
To define the molecular weights of the proteins encoded by G-protein a-subunit cDNAs, we prepared mRNAs from the corresponding cDNAs and translated these in uitro with the reticulocyte lysate system. [35S]Methionine-labeled protein products were subjected to SDS-PAGE and visualized by fluorography. A labeled protein of approximately 47 kDa (migrating just above the ovalbumin marker protein) was present in the reticulocyte lysate without addition of exoge-

Encoded by
Gi3 cDNA 6477  nous mRNA (control lune, Fig. 1). This protein serves as a convenient internal marker for comparing the mobility of translation products. The translated products of all four mRNAs migrate as approximately 40-kDa proteins, but subtle differences in migration are apparent (Fig. 1). The Go cDNA product migrates most rapidly, consistent with the estimated molecular mass, 39 kDa, of purified Go (5). Gi, and Gi3 products migrate most slowly, at approximately 41 kDa. The product of Gi2 cDNA shows intermediate mobility, about 40 kDa, consistent with our earlier identification (6) of a 40-kDa protein purified from neutrophils as the protein encoded by To identify the proteins encoded by Gi cDNAs, we prepared antisera against synthetic peptides corresponding to sequences predicted by Gi cDNAs (Table I). Previously, we raised antisera, AS/6 and AS/7, against the carboxyl-terminal Giz cDNA. decapeptide of transducin-a. These antisera recognize not only transducin but also the 41-kDa G-protein (G41) purified from brain as well as the 40-kDa G-protein (G40) purified from neutrophils (6). The latter two proteins correspond to Gil and Gi2, respectively; these cDNAs have an identical carboxyl-terminal sequence, differing by a single residue from transducin-a (Table I). To distinguish Gi2 from Gil, we generated antisera, LE/2 and LE/3, against a peptide corresponding to residues 160-169 predicted by Gi2 cDNA. This decapeptide differs in sequence for all three forms of Gi cDNA (Table  I). LE antisera reacted specifically with the neutrophil G40 and not the brain G41. This allowed tentative identification of the former as the product of Gi2 cDNA (6).
In an effort to develop Gil-and Gi3-specific reagents, we immunized rabbits with the synthetic peptides corresponding to residues 159-168 predicted by Gil and Gi3 cDNAs ( Table  I). The resultant antisera (LD and SQ for Gil and Gi3, respectively), as well as several others (Table I), were tested for specific reactivity with three purified G-holoproteins: G-proteins with a-subunits of 41 and 40 kDa (G41 and G40, respectively) purified from brain and a G-protein with 41-kDa asubunit (G41) purified from HL-60 cells (8).
The results of the immunoblots comparing LD, LE, and SQ reactivity are shown in Fig. 2. A P-subunit-specific antiserum, MS/1, was included in all first antiserum incubations to visualize the P-subunits of the purified G-proteins. MS/1 antiserum alone revealed only the &subunits (Fig. 2, upper  right). A "G-a common" antiserum, GA/1, similar to that used by Mumby et al. (11) was used to visualize all three asubunits (Fig. 2, upper left). Antiserum AS/7 reacts equally well with brain G41 and G40 but reacts significantly less well with HL-60 G41 (compare Fig. 2, upper middle uersw upper  left). These results suggest that brain ($0, like neutrophil G40, corresponds to Gi2, since AS antisera react equivalently with the identical carboxyl-terminal decapeptide of Gi, and Gi2. Gi3 differs from Gil and Giz by two additional conservative substitutions in the carboxyl-terminal decapeptide ( Table I). The lower reactivity of AS/7 with HL-60 G41 is consistent with tentative identification of this protein as Gi3-a. AS/7 does not react at all with Go-a which differs from the peptide antigen by 5 of 10 residues in the carboxyl terminus (6). LD antiserum (Fig. 2, bottom left) reacts only with brain G41, consistent with its identification as Gil. LE antiserum (Fig. 2, bottom middle) reacts exclusively with brain G40, presumptive evidence that this protein is Gi2. Finally, SQ antiserum (Fig. 2, bottom right) reacts only with the HL-60 Gd1. This strongly suggests that this protein represents the product encoded by Gi3 cDNA.
To provide further evidence for the specific reactivity of the peptide antisera, we assessed the ability of the synthetic peptides used as antigens (residues 159-168 for Gil and Gi3, 160-169 for Gi2, Table I) to compete for antibody binding.  Table   I) were added to the antisera as follows: lane 1, 0 peptide; lunes 4, 5, and 6, 1, 10, and 100 pg/ml, respectively, of Gia peptide for both LD and LE antisera; for LD antisera, lanes 2 and 3, 1 and 10 pg/ml, respectively, of Gil peptide; for LE antisera, lanes 2 and 3, 1 and 10 pg/ml, respectively, of Gi2 peptide. The arrows indicate the positions of the 41-and 40-kDa proteins detected with LD and LE antisera, respectively. Additional bands detected (particularly with LD/2 and LE/3) are presumed to represent nonspecific reactivity, as judged by failure of corresponding peptide antigens to block binding (see lanes 2 and 3 in each panel). alent concentration of either of the other two peptides did not affect antibody binding. A similar experiment assessed the specific reactivity of LD and LE antisera with plasma membrane proteins from bovine cerebral cortex. Two different LD antisera detect the same 41-kDa protein, and two different LE antisera reveal a 40-kDa protein (Fig. 4). Reactivity with the specific protein band is blocked by as little as 1 pg/ml of the corresponding peptide antigen (Fig. 4, lanes 2 and 3 ) , but the Gil peptide does not affect LE reactivity nor does Gi2 peptide affect LD reactivity (not shown). This is entirely consistent with the differential reactivity of LD and LE antisera with purified brain G4l and G40, respectively (Fig. 2). Concentrations of Gi3 peptide as high as 100 pg/ml, moreover, fail to block specific reactivity of LD or LE antisera with brain membranes (Fig. 4, lanes 4-6). As yet, we have not detected specific reactivity of SQ antisera with protein(s) in brain membranes.

Brain
H L-60  ( l a n e 4 ) .

DISCUSSION
The sequences of human (12-16), rat (3,5), and bovine (4) cDNAs encoding three distinct, but closely related, forms of Gi have recently been reported. To define the functions and distribution of the putative proteins encoded by these cDNAs, it would be helpful to identify their protein products. In vitro translation of mRNA transcribed from Gil, Gi2, and Gi3 cDNAs indicates that the respective products migrate on SDS-PAGE as 41-, 40-, and 41-kDa proteins. The Go translation product migrates as a 39-kDa protein. The differences in migration between the translation products are subtle but reproducible. The differences are not directly related to size as Go, Gil, and Gi3 cDNAs encode 354 amino acid proteins, whereas Gi2 cDNA encodes a 355-amino acid protein.
Antisera we raised against synthetic peptides proved capable of distinguishing between various purified G-proteins.
LE, and SQ antisera were tested on immunoblots with the purified proteins, brain G4,, brain G40, and HL-60 G41, respectively, to which they specifically react. For each antiserum, reactivity on immunoblot was effectively blocked by 10 pg/ml of the corresponding synthetic decapeptide antigen. An equiv-Antisera AS16 and AS/7, raised against the carboxyl-terminal decapeptide of transducin-a (6), reacted differentially with G41 proteins purified from brain and HL-60 cells, respectively. Since Gi3 shows lower homology in carboxyl-terminal sequence to the antigenic peptide than Gil and Gi2, the weaker reactivity of AS17 with HL-60 G41 suggested the latter could represent Gi3. Stronger evidence comes from immunoblots with three antisera, LD, LE, and SQ, raised against a decapeptide showing differences in sequence between Gil, Giz,and Gi3,respectively. For G,and GiP,no species differences in the sequence of the relevant decapeptide (residues 159-168 and 160-169, respectively) have been reported (3-5, 12, 13). For the Gi3 peptide (residues 159-168), there is a single amino acid difference between the human (14-16) and rat (3) sequences. We chose the human Gi3 sequence for generation of SQ antiserum, even though it shows one less difference from Gi2 than does the rat Gi3 sequence. This makes it more likely that differences in antibody reactivity would reflect differences in G-protein type rather than species differences.
Each of the three antisera reacted exclusively with a different purified G-protein, LD with brain G41, LE with brain G40, and SQ with HL-60 cell G4,, The specificity of antibody reactivity is further emphasized by competition for antibody binding with the synthetic peptides used as antigens. Even though the amino acid sequences of the peptide antigens show relatively minor differences, only the antigenic peptide proved capable of blocking the reactivity of the corresponding antiserum.
The unique pattern of immunoreactivity of these peptide antisera enable us to identify several purified G-proteins as products of defined cDNAs. Although the peptide antisera we have used (Table I) probe three discrete regions of G-CY sequence, our assignments must be considered tentative until the purified proteins are directly sequenced. LD antisera identify brain G41 as Gil, in good agreement with direct protein sequencing (4,5). LE antisera identify brain G40 as G,. In one of the original reports on purification of pertussis toxin substrates from brain, a 40-kDa protein in addition to G,, and G39 was observed (17). Recently, this protein was purified and shown to differ from brain G41 and G39 (7). Our results strongly suggest that brain G40 is the same protein, namely Giz, as G,, purified from neutrophils (18) and HL-60 cells (8, 19). G4, purified from KL-60 cells was shown by peptide mapping to &€fer from GAO (8). The unique reactivity of HL-60 G,, with AS and SQ antisera suggests that this is the protein encoded by Gi3 cDNA. This is consistent with cloning of Gi3 cDNA from an HL-60 cell library (14) and our previous identification of Gis mRNA in HL-60 cells (20). Whereas Gi2 mRNA appears widely distributed in brain and peripheral tissues, Gi, mRNA is relatively brain-specific (21). In contrast Gi3 mRNA is widely distributed in peripheral tissues but undetectable with the methods we have used in brain.' This is consistent with the apparent absence, or very low abundance, of Gi3 immunoreactivity in brain. Our results suggest that G4, in brain is principally, if not exclusively, Gil. G41 in peripheral tissues likely consists, at least in part, of Gi3.

M. Brann and A. Spiegel, manuscript in preparation.
All three forms of Gi are pertussis toxin substrates (6-8,

17-19)
. Several effector functions, including adenylyl cyclase inhibition, stimulation of a K' channel, and stimulation of phospholipase C in cells such as neutrophils are regulated by pertussis toxin-sensitive G- proteins (1, 2). As yet, it is not clear which, if any, of the forms of Gi regulates these effectors.
Results of reconstitution experiments employing "purified" G-protein preparations must be interpreted cautiously given the possible heterogeneity of such preparations. The immunologic tools described in this report offer the possibility of specifically identifying closely related G-proteins and should prove useful in elucidating the functions of this important class of signal transducers.