In Vivo Generation of an Adenylylcyclase Isoform with a Half-molecule MotiP

A truncated form of adenylylcyclase (type V-a) has been cloned from a cardiac cDNA library. It constitutes a half-molecule of type V adenylylcyclase diverging at the end of the first cytoplasmic loop. Northern blotting study has revealed the presence of such a mRNA spe- cies (approximately 3.5 kilobases in size) in the heart. Genomic sequence analysis has revealed that type V-a is generated via usage of a polyadenylation signal located within an intronic sequence of type V adenylyl- cyclase gene. When type V-n is co-expressed with an artificially generated half-molecule constituting the latter half of type V adenylylcyclase, the catalytic activity in transfected cell membranes is significantly higher than that of controls. However, when either alone is overexpressed, no significant increase in cat- alytic activity results. These results indicate that a half-molecule of adenylylcyclase, i.e. a protein con- taining six-transmembrane spans followed by a single cytoplasmic domain, can be generated in vivo, but cat- alytic activity is lacking unless heterodimerization can occur. This finding identifies another potential mech- anism for generating diversity within this enzyme family. of mammalian

I to VI) (1-8). Although different in amino acid sequence, biochemical characteristics, and tissue distribution, they share a common structure of sixtransmembrane spans followed by a large cytoplasmic domain repeated in tandem. Although the amino acid sequence differs significantly in the transmembrane regions among the isoforms, that of the cytoplasmic and putative catalytic domains is relatively well conserved (9). Indeed the first and the second cytoplasmic regions are similar to each other (6). These re-* This work was supported in part by United States Public Health Service Grant HL-38070 and by a grant from American Cyanamid Company-Lederle Laboratories. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the e e n B u n k T~~~M B L Datu Bank with accession numberis) M97886.
)I To whom correspondence should be addressed Dept. of Phar-630 W. 168th St., New York, NY 10032. macology, College of Physicians and Surgeons, Columbia University, gions also show homology to the C-terminal, catalytic domain of mammal~an guanylyl cyclases (1). We have recently described a dendrogram based upon the homology score among the various mammalian adenylylcyclase family members (6). Following divergence from guanylyl cyclase, the adenylylcyclase types I and I11 diverged from the others; thereafter types II/IV and V/VI diverged forming subgroups within the family.
The guanylyl cyclase family contains members that exist in both the particulate and soluble fractions of cell homogenates. Soluble forms of the enzyme recognize nitric oxide or its related products, while diversity within the extracellular domain of the plasma membrane forms has resulted in a series of guanylyl cyclases that are activated by different peptide ligands, including atrial natriuretic factor (10,11). However, both forms possess a C-terminal, catalytic domain, whose sequence is homologous among all the family members. Soluble forms of guanylyl cyclase have two distinct subunits, cy and , B (12-14). Although each subunit contains an apparent catalytic domain, it does not function independently; previous studies have suggested that heterodimer formation is necessary (13). This finding is consistent with recent observations on the calmodulin-sensitive form of adenylylcyclase (type I) found in brain. It has been shown that artificially created molecules, constituting either the initial or latter half of the original molecule, must be expressed together to generate catalytic activity (15).
We have now cloned a novel type V adenylylcyclase cDNA; it contains the first half of the molecule while totally lacking the latter half and instead terminates in a short stretch of novel amino acid sequence. In addition to its cloning, we also describe the putative mechanis~ for generating such a molecule.

MATERIALS AND METHODS
isotution and Sequencing of Clones-A cDNA library was prepared according to standard procedures in a X g t l O vector using poly(A)+ R N A prepared from canine ventricular tissue (16). In the primary library screening, 2.0 X IO6 independent clones were screened with a EcoRI-SphI fragment from type V adenylylcyclase cDNA as a probe.
The hybridization was carried out in a solution containing 50% formamide, 5 X SSC, 5 X Denhardt's solution, 25 mM NaPO, (pH 6.5), 0.25 mg/m~ calf thymus DNA, and 0.1% SDS at 42 "C. Hybridization was carried out in the same buffer at 42 "C for 14-20 h with cDNA fragments labeled with 3zP, followed by washing under increasingly stringent conditions. All DNA sequencing (using either universal or synthetic oligonucleo~ide primers) was carried out bidirectionally at least twice using either Sequenase (17) or Taq polymerase (18). For certain GC-rich sequences, such as the 5'-untranslated region, the reaction was carried out both with Sequenase and Taq polymerase with or without 7-deaza-dGTP (19). Electrophoresis was carried out in a polyacrylamide gel containing 8 M urea and 20% formamide, when necessary, to eradicate problems with hand compression. For genomic DNA cloning, 2 X 106 recombinant clones from an EMBL3 canine genomic DNA library were screened with a 0.4-kb' AuaII fragment as a probe that contained the unique sequence from the type V-a cDNA. Screening and sequencing were subsequently performed in a manner analogous to that described above. 'The abbreviations used are: kb, kilobase(s); bp, base pair(s); GTPrS, guanosine 5'-3-O-(thio)triphosphate. fragment designated as 113-0 (for type V-0) was constructed by digesting clone 113 with EcoRI and Ssp1 (nucleotide 1639-4303 from type V adenylylcyclase) (5). There is a Met at amino acid residue 581 with a reasonable Kozak consensus sequence (. . . C C A A G m A . . .) in 113-p; thus it would translate a protein of 604 amino acids constituting the latter half of type V adenylylcyclase starting in the Clb domain. Each fragment was subcloned into unique polylinker sites of the plasmid pcDNA I-amp and designated pcDNA113-cr and -0, respectively. Twenty micrograms of the purified plasmid were transfected into CMT cells (20) by the method previously described (21). Control cells were mock-transfected and were otherwise treated in the same way.
synthesized based on the novel sequence of 6L (AGCCGTCCCGGA Northern Blotting-A 60-mer antisense oligonucleotide probe was ATCAG~GGGCCCTTCCTTACAAAGAACCGTAGGCGAAGGA AGAGCAG) (Fig. l b ) and was labeled with 32P. Poly(A)+ mRNA was isolated from canine heart tissue by standard techniques using oligo(dT)-cellulose. Hybridization was performed as described above, except that a 30% formamide solution was used.
Adenylylcyclase Assay-The transfected CMT cells were washed twice with phosphate-buffered saline and then collected into 1 ml of cold buffer containing 50 mM Tris-HC1 (pH 8.0), 2 mM EGTA, 10 p~ phenylmethylsulfonyl fluoride, 100 units of leupeptin, and 50 units of egg white trypsin inhibitor. The cells were homogenized with a Polytron (setting 6 for 10 s) and centrifuged at 800 X g for 10 min at 4 "C. The supernatant was further centrifuged at 100,000 X g for 40 min at 4 "C. The resultant pellet was resuspended in 50 mi% Tris (pH 8.0), 1 mM EDTA, 1 pM phenylmethylsulfonyl fluoride, 50 units of leupeptin, and 50 units of egg white trypsin inhibitor to a concentration of 5 pg/pl. This crude membrane preparation was used in the adenylylcyclase assay.
Adenylylcyclase activity was measured by the method of Salomon AMP was used. Cyclic AMP was separated from ATP by passing through Dowex and alumina columns, and the radioactivity was measured by liquid scintillation counting. Protein concentration was measured by the method of Bradford (23) using bovine serum albumin as the standard.

RESULTS AND DISCUSSION
Isolation of Type V-a Clones-We obtained 12 primary cDNA clones using an EcoRI-SphI probe from type V adenyiylcyclase. Each was subcloned into pUCl8 and sequenced.
Eleven of the 12 cDNAs were identical to type V adenylylcyclase. Clone 6L was identical to type V adenylylcyclase in the 5'-region but totally diverged at its 3'-end; it contained 570 bp of unique sequence followed by a poly(A) tail. The 5'-end of clone 6L extended to a position 150 bp downstream of the translation initiation site found in the conventional type V adenylylcyclase cDNA. We later identified clones which were 5' extended and contained the putative translation initiation site. This novel sequence contains a stop codon (TAG) in the same open reading frame as the conventional type V cDNA and therefore encodes a putative protein product of 596 amino acids, 25 of which (residues 572-596) are unique to this isoform (type V -a ) (Fig. l a ) . Hydropathy analysis by the method of Kyte and Doolittle (24) reveals a structure of sixtransmembrane spans followed by a large cytoplasmic domain (data not shown).
In order to delineate the mechanism for generating such a molecule, we screened a canine genomic library using a 0.4 kb of novel sequence from 6L as the probe. Two million independent colonies were screened. Five positives were obtained; one clone, 123, was further analyzed. A 10-kb insert from clone 123 was restriction-mapped, and the fragments were subcloned into pUC18. A 4-kb EcoRI-EcoRI fragment from the 10-kb insert hybridized to the 0.4-kb probe. Its sequence  is shown in Fig. 16. The unique sequence found in clone 6L was located adjacent to an exon-intron border identified within the genomic fragment and in fact spanned the consensus sequence of the 5'-donor splice site (AAGC/aCCGG) (25).
These data suggest that during the processing of the precursor RNA a cryptic polyadenylation site within the intron, 260 bp downstream from the conventional exon-intron border, can be utilized to generate a novel truncated mRNA (26). Thus the mRNA of this novel form of adenylylcyclase, designated type V-a, incorporates a part of the intron into the 3"terminal sequence (Fig.  2). The consensus sequence for polyadenylation utilized in this isoform is not the most preferred one, likely contributing to the relatively small amount of mRNA which is generated (27).
mRNA Expression-In order to confirm that the mRNA encoding this unique isoform is indeed expressed, Northern blotting was performed using a 60-mer antisense oligonucle-otide whose sequence is unique to clone 6L as a probe (Fig.  la). The results indicate that such a mRNA species (approximately 3.5 kb in size) does exist in heart (Fig. 3). For comparison, the conventional isoform has two apparent splicing variants, 5 and 7 kb, as previously reported (5). Cells-The adenylylcyclase activities of   three different constructs were compared pcDNA113-72, which encodes the conventional, full-length molecule of type V adenylylcyclase; pcDNA113-a, which encodes the initial half of the protein (type V-a); and 113-@, which encodes the latter half of the protein (type V-p). When expressed alone, neither pcDNA113-a nor pcDNA113-fi had increased activity over the mock-transfected control. However, when co-transfected together, the activity was significantly enhanced. In the presence of forskolin, there was a 1.6-fold increase over the mock-transfected control, although this activity was less than half of that observed with the conventional type V adenylylcyclase (Fig. 4).

Expression in CMT
The above data suggest that type V-a adenylylcyclase alone cannot catalyze the conversion of ATP to cyclic AMP. However, when co-expressed with the latter half of the molecule, this function is restored, implying that, at the very least, heterodimerization of the two molecules is required. A similar mechanism of catalytic activation requiring heterodimerization of distinct subunits has previously been described for the soluble form of guanylyl cyclase which shares sequence homology to adenylylcyclase in its catalytic domain.
There is a superfamily of transporter molecules including P-glycoprotein (28), the cystic fibrosis gene product (29), and major histocompatibility complex-encoded peptide transporters (30), which utilize ATP for energy and share the common motif of tandem repetition of six-transmembrane spans followed by a large cytoplasmic domain. Although adenylylcyclase neither shows homology in its putative catalytic domains to this family of molecules nor does it possess an ATP-binding consensus domain, it does share the same molecular topology as these other transmembrane proteins. Several studies have suggested that the interaction of distinct catalytic domains labeled with 32P and was used as a probe (Fig. Ib). Hybridization was performed in a solution containing 30% formamide, 5 X SSC, 5 X Denhardt's solution, 25 mM NaPOl (pH 6.5), 0.25 mg/ml calf thymus DNA, and 0.1% SDS at 42 "C for 14-20 h, followed by washing under increasingly stringent conditions. MWS, molecular weight standard (kb). Values (pmol/min.mg.protein) f S.E. are shown. *, control < transfected, p < 0.05. For each experiment, the results represent the average obtained from 6-9 independently transfected plates of CMT cells. either contained in the same molecule or in discrete subunits may underlie the catalytic activation of this class of molecules. An example of the latter is provided by RING4 and RING11, two major histocompatibility complex-encoded peptide transporters, which are the products of distinct genes. Apparently these two proteins assemble to form a complex since a defect in either protein results in the formation of an unstable molecule and loss of function (31).
Our data suggest that a single gene encoding an adenylylcyclase isoform can generate either a full or half-molecule motif via usage of alternative polyadenylation signals. The findings that such a half-molecule does exist for type V adenylylcyclase has certain implications for the generation of functional diversity in the signal transduction pathway culminating in adenylylcyclase activation. In addition to the existence of multiple isoforms (some of which appear to be expressed in the same cell type) of adenylylcyclase, which facilitates diversity in signal transduction, it is possible that half-molecules of different adenylylcyclase isoforms might also be encoded, functionally couple, and thereby create heterodimers with novel properties. Gilman and co-workers (15) have published data indicating that co-expression of artificial half-molecules of type I and type I1 adenylylcyclase can produce, putatively via heterodimerization, a novel adenylylcyclase activity that possesses the fir sensitivity of type I yet maintains the other characteristics of type I1 adenylylcyclase. We have, as yet, no evidence that other adenylylcyclase genes (Type I-IV and VI) either via the mechanism described in this study or by other routes (e.g. alternative promoters) can generate such half-motif adenylylcyclase molecules.