Presence of Three Distinct Molecular Species of Gi Protein a Subunit STRUCTURE OF RAT cDNAs AND HUMAN GENOMIC DNAs*

We have cloned a new species of rat Gia (Gi3a) cDNA and genomic DNAs for three distinct human Gia proteins (Gila, Gi2a, and Gi3a). Gi3a cDNA codes for a protein of 354 amino acids (M, 40,522) whose sequence is closely related but distinct from that of the previously isolated rat Gia (Gi2a). By screening the human genomic libraries with the two rat Gia cDNAs as probes, clones encoding human Gila, Gi2a, and Gi3a were isolated. The human Gi2a and Gi3a genes are composed of eight coding exons and seven introns and possess a completely identical exon-intron organization. Southern blot analysis indicates that a single copy of each Gia gene is present per haploid human genome.

from bovine adrenal (4) and brain ( 5 ) , and GiaZ cDNA from bovine brain (6). In a previous paper, we reported cloning and sequence determination of cDNAs coding for Gsa, Gia (now referred to as Gi2a), and Goa from rat C6 glioma cells (7). The heterogeneity of Gia species was suggested since the predicted amino acid sequence of rat Gi2a differed with the partial amino acid sequence of rat brain Gia protein  by 2 amino acid residues.
More recently, clones for Gia cDNA have been obtained from several other sources including mouse macrophages (8), bovine pituitary gland (9), human monocytes (lo), human granulocytes (ll), human liver (a), and human basal ganglia (12). Comparison of the nucleotide and predicted amino acid sequences of these Gia cDNAs suggests the occurrence of multiple distinct molecular species of Gia.
The molecular heterogeneity of Gia had already been indicated from the observation that there are multiple pertussis toxin substrates in mammalian brain (13,14). Immunological evidence also suggests that the pertussis toxin substrate in human neutrophils is distinct from brain 41-kDa and 16).
In addition to the structural and immunological distinction of multiple Gia species, functional heterogeneity of Gia proteins has been reported. G proteins which are modified by pertussis toxin are found to be involved not only in inhibition of adenylate cyclase (17,18), but also in activation of phospholipase C (19) and phospholipase A, (20). Recent evidence suggests that they are also involved in gating of K' (21)(22)(23) and Ca2+ channels (24,25). Furthermore, it is shown that a signal leading to cell proliferation is blocked by the pertussis toxin treatment (26). These observations suggest the presence of multiple molecular species of Gia having discrete individual functions.
In this paper, we describe the sequence of a new Gia (Gi3a) cDNA clone which we isolated from the same rat C6 glioma cell cDNA library previously used for isolation of Gi2a cDNA. We then cloned and characterized three distinct human chromosomal genes (designated as the Gila, Gi2a, and Gi3a genes), which code, respectively, for human homologues of bovine Gia (6), rat Giza (7), and rat Gi3a. SSC, 0.1% sodium dodecyl sulfate for high stringency or in 1 X SSC, 0.1% sodium dodecyl sulfate for low stringency. 3ZP-Labeled probes were prepared by random priming (28) or nick translation (29) of DNA fragments purified from agarose gel (Low Gelling Temperature, Bio-Rad).
Rat C6 glioma cell cDNA library was constructed with X g t l O as 0 0.4 0.8

( k b )
FIG. 1. Restriction map and sequencing strategy for the rat G,3a eDNAs. The hatched box depicts the amino acid coding region. The thick lines depict the cDNA inserts of XGX14 and XRGi34. The dashed line indicates the probe for rescreening to isolate XRGi34. The arrows indicate the direction and extent of DNA sequence that was determined by the dideoxynucleotide chain-termination method. described previously (7). To isolate the full length cDNA of rat Gi3a, we constructed the cDNA library using random hexamers as primers.
A clone, XRGi34, was isolated from this cDNA library with rat Gi3a cDNA EcoRI-NcoI 161-bp fragment of XGXl4 as a probe.
To isolate genomic clones of Gia subfamily, we screened the human genomic DNA library (30) under low stringency conditions with the EcoRI fragment from hGX13 containing the rat Gi2a coding sequence (7). Eighteen genomic clones were picked up for restriction analysis and Southern hybridization at high and low stringency. Two clones, hHGill and XHGi24, were hybridized under high stringency and identified as the Gi2a clones. One clone (XHGi67) which hybridized with the Gi2a probe under low stringency conditions was analyzed by restriction mapping and by sequence analysis, and it turned out to be the Gila clone containing exons 2 and 3. Three clones (XHGi2N-6, XHGi2N-1, and hHGi2N-16) contained the genomic sequences of the 5' flanking and exon 1 for human GiBa, G&, and Gila, respectively, were isolated from another human genomic DNA library (31) with the 323-bp EcoRI-PstI fragment of rat Gi2a cDNA containing most of exon 3 and all of exons 1 and 2. Human Gi3a genomic clone hHGi3-13 and XhGi3-81 were obtained from this library with rat Gi3a cDNA (XGX14) as a probe.
Nucleotide Sequencing-For dideoxynucleotide sequencing (32) cDNA and genomic DNA fragments were subcloned into M13mp Ser Glu Leu Ala G1 Val I l e L s A r ' Leu Trp A r As G1 G1 Val Gln A l a C Phe Ser A r Ser A r Glu T r Gln Leu As; 150 As Ser Ala Ser T r T r Leu Asn As Le; As A r I l e Ser Gln Thr Asn Tyr I l e P r i Thr Gln Gln As Val Leu A r Thr A r Vai  The deduced amino acid sequence of rat Gi3a is shown above the nucleotide sequence of XRGi34 cDNA insert. Rat Gi2a nucleotide sequence was obtained from two cDNA clones, XGX13 and XGX12. The 5' flanking was 65 nucleotides longer in XGX12 than in XGX13. In the coding region of rat Gi2a cDNA, h y p h m represent a nucleotide residue that is identical with the residue in rat Gi3a cDNA. On the fourth line of each set of lines are shown the amino acid residues of rat Gi2a that differ from those of rat Gi3a. vector (33,34). Single-stranded templates were prepared, and sequencing reactions were performed with either [cY-~'P]~ATP or [a-32P]dCTP. For sequencing of the G + C-rich region, 2'-deoxy-7deazaguanosine 5"triphosphate was used instead of dGTP as described (35).
Genomic Southern Blotting-High molecular weight DNA was prepared from human HL60 cells according to the method of Gross-Bellard et ~l .
(36). The DNA was digested completely with HindIII and BglII and subjected to electrophoresis in 0.8% agarose gels. DNA fragments were then transferred to a nitrocellulose filter as described by Southern (37) and hybridized with 3ZP-labeled genomic DNA fragments.

Isolation of Second Cia (Gi3a) cDNAs from the Rat C6
Glioma Cell Libraries-In the course of the previous studies (7), we have isolated from a cDNA library of rat C6 glioma cells, a clone, XGX14, whose sequence is highly homologous to but distinct from the previously identified Gia clone (XGX13). We designate the new Gia as Gi3a and the previous one as Gi2a. The Gi3a sequence contained in XGX14, however, lacked the NHn-terminal sequence corresponding to amino acid residues 1-34 of Gi2a. To obtain a cDNA containing the complete NH,-terminal sequence of Gi3a, we constructed a new cDNA library from rat C6 glioma cells using random primers and screened this cDNA library with the DNA fragment containing the 5' region of hGX14. On screening 1 X lo6 recombinants, one clone (XRGi34) was isolated. This clone was identical with XGX14 in restriction map, but longer in length a t both 5' and 3' ends ( Fig. 1).
The nucleotide and deduced amino acid sequences of rat Gi3a cDNA are shown in Fig. 2. The sequence contained an open reading frame of 1062 nucleotides coding for 354 amino acids including an initiator methionine with a calculated molecular weight of 40,522. The nucleotide sequence of Gi3a was about 75% homologous with Giza in the coding region, but their sequences in the 5' and 3' flanking regions are entirely different. Although the deduced amino acid sequence of Gi3a was closer to bovine Gia (6) (94% homology) than rat Giza (85% homology), rat Gi3a is not the counterpart of bovine Gia. The nucleotide homology between rat Gi3a cDNA and bovine Gia cDNA was only 80% in the coding region, and there was no homology in their 5' and 3' flanking regions (data not shown). As will be described, bovine Gia was found to be a counterpart of human Gila.
Isolation and Characterization of Human Chromosomal G,2a Gene-Restriction enzyme analysis of the three Gi2a genomic clones (XHG,11, XHGi24, and XHGi2N-6) isolated as described under "Experimental Procedures," and Southern blot analysis using the fragments of rat Giza cDNA revealed the organization of the human Giza gene as shown in Fig. 3. The coding region of the human Gi2a splits into 8 exons and 7 introns, and spans 16 kilobases of genomic DNA. In the 3' flanking region, a short sequence of about 200 bp was found in upstream of the XbaI site (data not shown), which was 80% homologous with the 3' flanking sequence of rat Gi2a cDNA (7). This may suggest the presence of an additional exon(s) in the 3' flanking region of the human Gi2a gene. Fig. 4 shows the nucleotide sequence of the 5' flanking and the coding regions of the human Giza gene. In the coding region, the homology between human and rat Gi2as was 92 and 99%, in the nucleotide and deduced amino acid sequences, respectively. There were only five amino acid replacements in the total sequence of 355 amino acids; two of them were nonconservative and the three remaining were conservative.

G"J!%L---------A-
1 Met G1 C s Thr Val Ser Ala Glu As L i Ala Ala Ala Glu Ar Ser L s Met I l e As' L s Asn Leu Ar Glu As G1 Glu L s Al; Leu Gln Ile As Phe Ala As Pro Ser Ar' Ala As As Ala Ar Gln Leu Phe Ala Le; Ser C s Thr Ala Glu Glu Gln G1 Val Le; 151 As Ser Ala Ala T r T r Leu Asn As Le; Glu A r Ile Ala Gln Ser As T r Ile P r l Thr Gln Gln    search, a remarkable homology was found between the nucleotide sequence of the human Gi2a gene at positions -162 to -189 and the 5' flanking region of human c-Ha-ras protooncogene at positions 510-537 (5' CCCGCCCCCGCCCCG-CCCCGGCCTCGG 3'). The degree of identity between these two sequences is 86%, 24 out of 28 nucleotides being identical. This sequence of c-Ha-ras proto-oncogene occurs immediately upstream of the transcriptional initiation site (41). Isolation and Characterization of Human Chromosomal GJa Gene-Three genomic clones XHGi2N-1, XHGi3-13, and XHGi3-81 were isolated as described under "Experimental Procedures." The comparison of the rat Gi3a cDNA sequence with the sequence of the genomic subclones revealed that XHGi3-13 and XHGi3-81 overlapped and contained exons 2-4 and exons 4-8 of human Gi3a, respectively. On the other hand, XHGi2N-1 contained exon 1 and the 5' flanking region (see Fig. 3). The nucleotide sequence homology of the latter clone with rat Gi3a cDNA were 80 and 93% in the 5' flanking (94 bp) and the coding (118 bp) sequences, respectively (see Fig. 5). The nucleotide sequence of the coding and 5' flanking regions of the human Gi3a gene is presented in Fig. 5. In the coding region, 92% nucleotide sequence homology was found between the human and rat Gi3as. Human Gi3a gene differs from rat Gi3a cDNA at 5 of 354 amino acid residues, and all of the differences are conservative amino acid substitutions.

CCCAGCAAT~GACGGTGTC~CAGCCTGCC~AGCCGCAGT~TCCGTGGTG~G~~AGTGAGTC~GGGCCCGTG~CCCCTCTCC~GCCGCCGCC -----------GA-----C+--_------G-T*-T-T"__G_-C__-C----TG-____C_A______*__-______G___-___T_
1 Met GI C s Thr Leu Ser Ala Glu As L Ala Ala Val Glu Ar Ser L s Met I l e AS; A r Asn Leu A r Glu Asp GI Glu L s AI; Human amino acid sequence of human Gi3a is shown above the aligned human genomic DNA and rat cDNA sequences. Numbers on the left indicate residue numbers for the nucleotide and amino acid sequences. Hyphens represent a nucleotide residue in rat Gi3a cDNA that is identical to the residue shown in human Gi3a genomic DNA. Asterisks indicate the gaps introduced for optimal alignment. 5' flanking GC boxes are underlined, and TATA box homologue is underlined and overlined. The position at which an intron occurs in the human gene is indicated by a triangle above the amino acid sequence.
Gi3a cDNA was 80%. In the 5' flanking region of the human Gi3a gene, TATA-like sequence, TTTAAA, occurs at position -182, and GC boxes are found three times at positions -87, -300, and -410. There is no typical CCAAT box. Isolation and Sequencing of Partial Human Gila Gene-Two genomic clones, XHGi2N-16 and XHGi67, were isolated as the clones weakly hybridizing with rat Gi2a cDNA, as described under "Experimental Procedures." The nucleotide sequence analysis of genomic subclones revealed that these two genomic clones coded for the amino acid sequence more homologous to bovine Gia (6) than Giza or Gi3a. As can be seen in Fig. 6, homologies of this human gene and bovine Gia cDNA in the coding and the 5' flanking regions are remarkable. In the 5' flanking region, from positions -1 to -115, the homology in nucleotide sequence was 71%. On the other hand, homologies in the partial nucleotide and amino acid sequences of the coding region (exons 1, 2, and 3) were 96 and loo%, respectively. We therefore concluded that this gene is a human homologue of bovine Gia (Gila). Within about 1050 bp upstream of the ATG initiator condon, several transcription signals were noted. Five GC boxes occur at positions -118, -182, -354, -364, and -533, and a consensus hexamer sequence for the steroid receptor binding, TGTTCT (42), at position -784. There was no typical TATA-like sequence nor CCAAT box.
Conservation of the Exon Junctions in All G;a Genes- Table   I shows the sequences at the exon-intron boundaries of the human genes for Gila, Gi2a, and Gi3a. All introns begin with the sequence GT at the 5' end and end at the 3' end with the sequence AG. The deduced amino acid sequence of human G i l a is shown above the nucleotide sequence up to exon 3. Hyphens represent a nucleotide residue in bovine Gia cDNA (6) that is identical to the residue shown in human Gila genomic DNA. Asterisk indicates the gap introduced to obtain maximal homology. 5' flanking GC boxes are underlined. The steroid receptor binding consensus sequence is underlined and ouerlined. The position at which an intervening sequence occurs is indicated by a triangle above the amino acid sequence.

TABLE I Exon-intron organization of the human Gia genes
The nucleotide sequence of exon-intron junctions were determined from genomic clones according to the strategy shown in Fig. 3. Gila intron 2, Gi2a introns 3 and 6, and Gi3a intron 2 were sequenced in their entirety. Exon sequences are in capital letters, intron sequences are in lowercase letters. The consensus sequence was obtained from Padgett et al. (50).

Intron
No. Species 5' Boundary Intron size 3' Boundary Exon size 1 2 3 TGCTCG g t a a g g . . TGTTGG g t g a g g . . TACTCG g t g a g g . . GATGAA g t a a g t . . GATGAA g t a a g t . . GATGAA g t a a g t . . CGGGCG g t a a g t . . AGAGCG g t a t g t . . AGGGCA g t a a g t . .
TGCCTA g t g a g t . . TTCATA g t a a g t . .
CTTCAA g t g a g c . . CTTCAA g t a a g t . . GAGATG g t g a g a . . GAGATG g t a t g t . .
ACACAG g t g t g g . . ACACAG g t a a g g . . GGGATG g t g a t c . . TGAAAG g t a a a g . . . . t g t t t a t a t t t t a g GTGCTG . . g c c c t c t g t t c c a g GTGCTG . . t t c t t t t c t t a c a g GTGCTG . . t t t t g t c t c a t t a g AATTAT . . g t c c t g g c t a t c a g GATCAT . . t c t t t g t t t t a a a g AATCAT ..gccactgtgcccag GACGAC . . a c t g c t t t c t t c a g GATGAT . . c c c c c c a t c c c c a g CTACCT . . t g t c t t t t a t t t a g TTATCT .......... ctgcag GATGTT . . t t c c c c t t g c g c a g GATGTT . . t a t t c t a c c c c c a g AACCGC . . t a t t t c t t t t t c a g AACCGA ..tttCtCtCcCCcag GGGCCA . . g g t g t c C g t t t t a g GTTCCA

Consensus sequence
A G : e a a g t . .

cccccccccc
. . t t t t t t t t t t n c s sequence of cDNA for G,2a and Gi3a were completely identical. Likewise, the junctions of exons 1,2, and 3 of Gila were identical with those in Gi2a and Gi3a. Southern Blot Analysis of the Human Gila, G,2a, and Gi3a Genes-To determine the copy number of the three human Gia genes and also to exclude the possibility of DNA rearrangement in the course of clonings, we carried out the Southern blot analysis of human genomic DNA digested with Hind111 and BglII. Hybridization was performed under high stringency conditions, using as probes the genomic sequences containing the 5' flanking region and exon 1 of each human Gia gene (see the legend to Fig. 3). The discrete bands were observed, and the sizes of the bands were consistent with those of the genomic clones (Fig. 7). We conclude that each of the three Gia genes occurs as a single copy per haploid human genome.  1,3, and 5) and BglII (lanes 2 , 4 , and 6) and analyzed by blot hybridization to human Gi2a (lanes 1 and 2 ) , Gi3a (lanes 3 and 4 ) , and Gila (lanes 5 and 6) genomic DNA fragments containing the exon 1 (see Fig. 3 legend). HindIII-digested DNA size markers (kilobases) are shown.

DISCUSSION
In this paper, we isolated rat Gi3a cDNA and three distinct human Gia genes, i e . the Gila, Giza, and Gi3a genes. Fig. 8 shows the comparison of the deduced amino acid sequences of Gila, Gi2a, and Gi3a obtained from different mammalian sources. In addition to the remarkable homologies of the overall structure, the strong conservation of the amino acid sequence in each group of Gia subfamily was striking. More  (43), human Gi2a (Gi2a), human Gi3a (GJa), human Gila (Gila) (see Fig. 8 legend), rat Goa (Goa),' bovine rod Gta (Gtla) (47), and bovine cone Gta (Gt2a) (48)  FIG. 10. Relationships among the gene structure and their products. Top, the protein structure. Hatched areas indicate region of the amino acid sequence responsible for GTP hydrolysis and GTP binding. The large asterisk shows the site of ADP-ribosylation by pertussis toxin, termed islet-activating protein (IAP). The smull asterisk indicates the site homologous to that of Gta which is ADPribosylated by cholera toxin (CT). Middle, the mRNA structure.
Latticed box denotes the coding region. The AUG initiator and the UGA terminator are shown above the box. Bottom, the structure of human Gi2a gene. Latticed boxes denote the coding exons, and the open boxes denoti the 5' and 3' noncoding region. Exons are numbered with Roman numerals. The functional domain structure of Ga protein has been described by Masters et al. (49). than 98% homologies were maintained among each Gia subfamily. The amino acid sequence of G.a is also strongly conserved between human and rat; only 1 out of 394 amino acids being different (43). The strong conservation of the amino acid sequence of each G protein a subunit among distant mammalian species may reflect the presence of evolutional pressure to maintain the specific physiological function of each G protein gene product. Each Ga protein may be linked to a specific receptor and thereby involved in a specific signal transducing pathway.
The occurrence of multiple forms of Gia was first noticed when we determined the sequence of Gi2a cDNA from rat C6 glioma cells (7). In one tryptic peptide obtained from the purified rat brain Gia , the amino acid sequence P r~l~~-A s n '~~ was different from Ser166-A~p'67 predicted from the nucleotide sequence. This particular amino acid sequence was found neither in Gi2a nor in Gi3a (Fig. 8), but it was present in bovine (6) and human (12) brain Gila. We consider that Gila may correspond to the 41-kDa protein, one of the two major pertussis toxin substrates in mammalian brain. In rat C6 glioma cells, this molecular species (Gila) does not seem to be a predominant one since Gila cDNA was not isolated from the C6 cell library, and Gila mRNA was not detected by Northern blot analysis of C6 cell mRNAs (data not shown).
Recently, Katada et al. (14) purified the 40-kDa Gia species from porcine brain. The partial amino acid sequence analysis of the porcine 40-kDa protein revealed that the protein was different from Gila and Gi3a, but identical with Gi2a (44). These results suggest that, of the pertussis toxin substrates present in mammalian brain, the major 41-kDa species and the minor 40-kDa species may correspond to Gila and Giza, respectively. On the other hand, the product of the Gi3a gene has not yet been isolated, although the gene is expressed in a variety of tissues and cell lines (data not shown). Very recently, Didsbury and Snyderman (11) and Suki et al.
(2) isolated a cDNA clone for a new G protein a subunit from the differentiated HL60 cell library and human liver, respectively. The deduced amino acid sequence of their cDNAs were identical with that of our Gi3a. Identification of protein products and assignment of functions for each molecular species of Gia have to be carried out.
We have determined the entire organization of the human Gi2a and Gi3a genes consisting of eight exons and seven introns. An additional exon is also found in the 3' flanking region of Gi2a, but this exon does not code for protein. It is remarkable that two genes have exactly the same organization although the lengths of introns are different. The position of the exon junctions in their cDNA sequences are exactly identical. The same splice sites were also conserved in the partial sequence (exons 1, 2, and 3) of the human Gila gene as well as in the human Goa gene? Furthermore, it was found that, in the human Gsa gene, 3 out of 12 splice sites were shared with Gia (43). The results suggest that these genes may be evolved from a common ancestral gene.
An evolutionary tree of G protein a subunits can be drawn based on the homologies of the predicted amino acid sequences obtained from various mammalian sources (Fig. 9). It is remarkable that the homologies among three Gia species are higher than that between rod (G,la) and cone (Gt2a) transducin a subunits (3). Fig. 10 illustrates the relationship between the exon-intron organization of the human Giza gene and the functional domain structure of its protein product. A domain of NH,terminal residues 1-39 encoded by exon 1 is hydrophilic and contains the site for limited tryptic digestions. Although this region may be involved in interaction with subunits, its precise function has not yet been shown. Exon 2 encodes a short length region (14 amino acid residues), which is the most conserved among all Ga proteins and responsible for GTP hydrolysis. The most structurally divergent domain is encoded by exons 3 and 4. The amino acid sequences of residues 80-130 is remarkably diverse. Polypeptide secondary structure analysis (45) of this region in Gi2a is different from other two Gia gene products (data not shown). Probably this region may serve as a site for interaction with an effector. Exon 5 encodes the sequence containing ArgI7' which is ADPribosylated by cholera toxin in bovine G,a. ADP-ribosylation of Gsa by cholera toxin causes a decrease of affinity for By subunits (46). The domain encoded by exons 6 and 7 is strongly conserved among all Ga proteins. Presumably, this domain is involved in formation of a core structure for GTP binding together with that coded by exon 2 (49). The sequence, Asn-Lys-X-Asp, consensus to all guanine nucleotide binding proteins, occurs at positions 270-273 of exon 7. The domain encoded by exon 8 may be involved in interaction with a receptor, since the cysteine residue ADP-ribosylated by per-tussis toxin is present in this region and also the structure of this region is heterogeneous. Further studies including sitedirected mutagenesis and construction of chimeric genes may throw more light on the structure-function relationship of Ga proteins.
The promoter sequence of the human Gia genes has some characteristic features. The regions up to 500 bp upstream of the ATG codon are G + C-rich, and several GC boxes are present in all three Gia genes. Thus, Gia genes may be regarded as "housekeeping genes" which are G + C-rich in their promoter sequences. Both Gi2a and Gi3a mRNAs are expressed in all tissues so far examined (data not shown), and the pertussis toxin substrates are ubiquitously present in all tissues. Recently, Murayama andUi (26) reported the involvement of a pertussis toxin substrate in proliferation of 3T3 cells. It is of interest to note that the sequence homologous with human c-Ha-ras proto-oncogene is present in 5' flanking region of the human Gi2a gene. Furthermore, it was found that the promoter of the human G.a gene shares several blocks of sequences homologous with the human c-Ki-ras2 gene (43). A common regulatory mechanism may exist in the gene expression of G protein a subunits and ras proteins. If so, it is attractive to speculate that both gene products might function cooperatively in the signal transduction directed toward cell proliferation.