Splice variants of type VIII adenylyl cyclase. Differences in glycosylation and regulation by Ca2+/calmodulin.

Three alternatively spliced type VIII adenylyl cyclase messages have been identified by cDNA cloning and amplification from rat brain cDNA. Type VIII-A was previously referred to simply as type VIII (Cali, J. J., Zwaagstra, J. C., Mons, N., Cooper, D. M. F., and Krupinski, J.(1994) J. Biol. Chem. 269, 12190-12195). The types VIII-B and -C cDNAs differ from that of type VIII-A by deletion of 90 and 198 base pair exons, respectively, which encode a 30-amino acid extracellular domain with two consensus sites for N-linked glycosylation and a 66-amino acid cytoplasmic domain. Stable expression of types VIII-A, -B, and -C cDNAs in human embryonal kidney 293 (HEK-293) cells leads to the appearance of novel proteins, which are recognized by type VIII-specific antibodies and which co-migrate with immunoreactive species detected on immunoblots of rat brain membranes. Types VIII-A and -C are modified by N-linked glycosylation, while type VIII-B is insensitive to treatment with N-glycosidase F. An influx of extracellular Ca2+ stimulates cAMP accumulation in HEK-293 cells stably expressing type VIII-A, -B, or -C, but not in control cells. Adenylyl cyclase activity of each of the variants is stimulated by Ca2+/calmodulin and the EC50 for activation of type VIII-C is one fourth of that for either type VIII-A or -B. Type VIII-C also has a distinct Km for substrate, which is approximately 4-12-fold higher than that for types VIII-A or -B depending on whether Mn2+ or Mg2+ is the counterion for ATP. The differences in the structural and enzymatic properties of these three variants are discussed.

A variety of hormones, neurotransmitters, and olfactants regulate the synthesis of cAMP by adenylyl cyclases (ACs). 1 Many of these agents act through three component, G proteincoupled systems that are capable of modulating the activity of an AC (for review, see Ref. 1). Alternatively, seemingly independent signaling pathways may generate other second messengers or activate kinases that subsequently regulate AC activity by signal cross-talk. The importance of the latter class of regulatory mechanisms has become apparent with the realization that the different AC isoforms are distinguished by their ability to provide a unique integrated response to coincident stimuli.
Eight full-length mammalian ACs have been characterized (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14), and other partial cDNA sequences have been reported (15,16). The existence of additional forms which are derived by alternative splicing of the messages is consistent with the sequence differences in the amino-terminal domains encoded by cDNA clones for both type V (6,17) and type VI (7)(8)(9)(10). A half-molecule variant of the canine type V has been described (18), and the expression of other AC variants is suggested by the detection of multiple type V and type VIII transcripts on RNA blots under high stringency conditions (6,9,11,13). Thus far, unique functional properties have not been ascribed to any specific splice variant.
The mammalian ACs share a common topography. The amino-terminal domains are all predicted to be cytoplasmic, but they vary dramatically in both sequence and length. The only well conserved sequences among the ACs are found in the so-called C 1a and C 2a regions, which are two large cytoplasmic domains of over 200 amino acids each (5,19,20). The conserved domains are homologous to each other and to the catalytic domain of the guanylyl cyclases (2,21). Based on this similarity, they are considered to be nucleotide binding domains and have recently been shown to be sufficient to confer enzymatic activity (22). Each of these domains is preceded by a large hydrophobic region of variable sequence, which includes six transmembrane spans based on hydropathy analysis. Consensus N-linked glycosylation site(s) are always present in at least one putative extracellular domain associated with the second set of six transmembrane spans (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14). On the carboxyl-terminal side of the first putative nucleotide binding domain is a highly variable region called C 1b (5,19,20), which is a site for type-specific regulation (23)(24)(25). The corresponding region following the second nucleotide binding domain, C 2b , is only present in types I, III, and VIII (2,4,13). While similarities among these isoforms may underlie their common enzymatic function, their differences at the amino acid level presumably account for variable responsiveness to a variety of regulatory influences.
Type VIII AC is a Ca 2ϩ /calmodulin-stimulated isoform abundantly expressed in discrete regions of rat brain (13). The distribution of type VIII message and its enzymatic properties are most closely related to those of type I (13,19,26). Here we describe the isolation of two additional type VIII cDNAs from rat brain, which are derived by alternative mRNA splicing. The differences in the structure and enzymatic properties of the variants are described.

EXPERIMENTAL PROCEDURES
cDNA Cloning-The strategy used to isolate type VIII AC cDNA clones was described previously (13). Clones were separated into three classes called type VIII-A, -B, and -C based on internal differences in their restriction maps that contrasted with nucleotide sequence identity at their ends. One incomplete cDNA encoding type VIII-B and another encoding type VIII-C AC were selected from a rat brain cDNA library in the Zap II vector (Stratagene), and a partial type VIII-B cDNA was selected from a bovine brain cDNA library in the GT11 vector (Clontech). Each rat clone was sequenced on both strands using dye-labeled chain terminator chemistry and an automated sequencer (Applied Biosystems).
Identification of Type VIII AC Splice Variants by PCR-An RNA PCR kit (manufacturer's protocol, Perkin Elmer) was used to generate randomly primed first strand cDNA template from rat brain poly(A) ϩ RNA (Clontech). Sense and antisense primers, AGCCACACTGGATTGC-CTCA and TTATAGTCTTCTGCCGTGGT, corresponding to regions of the type VIII-A cDNA encoding amino acids 551-556 and 756 -762 (13), respectively, and flanking the region deleted from the type VIII-C cDNA, were used at 375 nM in a reaction at pH 8.5 containing 60 mM Tris-HCl, 15 mM (NH 4 ) 2 SO 4 , 3.5 mM MgCl 2 , 250 M dNTPs, and cDNA reverse transcribed from 0.25 g of rat brain poly(A) ϩ RNA. The thermocycle consisted of an initial 2 min incubation at 94°C, 34 cycles of 1 min at 94°C, 30 s at 50°C, 1 min at 70°C, and a final cycle in which the incubation at 70°C was extended to 7 min. Reaction components and the thermocycle for a second amplification were the same, except that MgCl 2 was 2.5 mM and the primers were TCTCCATCCTGATCATGCT-GCACTC (sense strand) and TCAGTAGCAGTGATGCCTCCTTGGT (antisense strand), corresponding to regions of the type VIII-A cDNA encoding amino acids 742-749 and 894 -901, respectively, and flanking the regions deleted from the type VIII-B cDNA. Amplified products were subcloned into the pCR II vector (Invitrogen) and sequenced as described above.
Confirmation of Type VIII-B and -C AC 5Ј cDNA Ends by PCR Amplification and Southern Blot Analysis of Rat Genomic DNA-A sense strand primer corresponding to nucleotides 284 -303 in the 5Јuntranslated region of the type VIII-A cDNA (CCATCCTAGCCATCG-GTCAC) and a primer antisense to the sequence encoding amino acids 237-243 (CCGCTGTATTGCAGGTAAGT) were used to amplify rat genomic DNA. Reaction components were as described above with 0.1 g of rat genomic DNA (2 separate preparations from Clontech) and 2.0 mM MgCl 2 . Amplification and sample processing were as described above except that the annealing temperature was 60°C, and the extension temperature was 72°C. Rat genomic DNA (8.5 g, Clontech) was digested with PstI alone or PstI and SfaNI, and subjected to Southern hybridization analysis. The probes were two restriction fragments including nucleotides 324-1387 of the type VIII cDNA that were labeled to a specific activity of 0.88 Ci/ng with [␣-32 P]dCTP using the Prime-A-Gene kit (Promega). Electrophoresis, transfer, and hybridization were essentially as described (27). The final wash was for 1 h at 65°C in 0.1 ϫ SSC (0.15 M NaCl, 0.015 M sodium citrate) and 0.1% SDS. The membrane was exposed to x-ray film with an intensifying screen for 50 h at Ϫ70°C.
Expression of Type VIII-A, -B, and -C AC-The type VIII-A expression construct in the pCMV5-neo vector is the same as described previously (13), but the -A designation has been added to distinguish this construct from those encoding types VIII-B and -C. Full-length type VIII-B and -C expression constructs in the pCMV5-neo vector were prepared by ligation of a restriction fragment from the start of the type VIII-A reading frame, to a restriction fragment from an overlapping region of the type VIII-B or -C cDNAs. Transfection into HEK-293 cells, tissue culture, and selection of polyclonal populations of Geneticinresistant cells were as described previously (13). HEK-293 cells transfected with the pCMV5-neo vector without a cDNA insert served as the negative control in the expression experiments.
Preparation of Whole Cell Lysates-Stably transfected HEK-293 cell populations were grown to confluence on 150-mm culture dishes, washed with phosphate-buffered saline, scraped from the plates, and collected by centrifugation. The cell pellet was resuspended in 20 mM NaHepes, pH 7.5, 2 mM MgCl 2 , 1 mM EDTA, 2 mM dithiothreitol, 22 g/ml each of L-1-tosylamido-2-phenylethyl chloromethyl ketone, 1-chloro-3-tosylamido-7-amino-2-heptanone and phenylmethylsulfonyl fluoride, 3.2 g/ml each of leupeptin and lima bean trypsin inhibitor, and 0.5 g/ml aprotinin. The resuspended cells were lysed by vigorous vortexing after the addition of sodium dodecyl sulfate to 1% final concentration. Genomic DNA was fragmented by passage through syringe needles (10 passages each through an 18-, 23-, and 26-gauge needle), and sonication using a 50% pulsed duty cycle at 20% power output with an Ultrasonics model W-225 sonicator. The protein concentration of the lysates was determined with the DC Protein Assay kit (Bio-Rad).
Immunochemical Analysis-Two peptides with sequence CDFDKSI-PLK (amino acids 804 -813 of type VIII-A AC) and KRIEHTIDLRSG-DKLRR (amino acids 666 -682) were synthesized (Applied Biosystems model 413A synthesizer). These peptides were coupled to the tuberculin purified protein derivative and used to raise polyclonal antisera Ab VIII-A 804 -813 and Ab VIII-A 666 -682 in New Zealand White rabbits (13). Ab VIII-A 804 -813, Ab VIII-A 666 -682, and Ab VIII-A 1229 -1248 (antiserum D88 in Ref. 13) were affinity-purified as described (28), except that the initial incubation of diluted antisera with the appropriate Sepharose 6MB-peptide was for 1 h at room temperature with gentle rocking. Where indicated, membrane proteins were treated with N-glycosidase F (Boehringer Mannheim) prior to electrophoresis (13). The contents of the polyacrylamide gels were electrophoretically transferred to Immobilon-P membranes (Millipore), which were incubated with the indicated affinity-purified primary antibody for either 1 h (Ab VIII-A 1229 -1248 and Ab VIII-A 666 -682 diluted 1:25,000) or overnight (Ab VIII-A 804 -813 diluted 1:250) at room temperature in 20 mM Tris-HCl, pH 7.6, 137 mM NaCl, 0.1% Tween 20. Prior to incubation with Ab VIII-A 804 -813, the Immobilon-P membrane was boiled for 30 min in distilled H 2 O (29). When rat brain membrane samples were included on the blot, 5% nonfat dry milk was included in the blotting buffer, affinity-purified Ab VIII-A 1229 -1248 was diluted 1:250 and the incubation was overnight. Detection of immune complexes was by enhanced chemiluminescence (manufacturer's protocol, Amersham Corp.). Band intensities were quantified with a Molecular Dynamics personal laser densitometer and average values from three separate experiments were used to calculate relative expression levels. Signal intensity varied linearly with the protein load for the relative signals reported.
Cyclic AMP Accumulation-HEK-293 cells were prepared for measurement of cAMP accumulation and analyzed essentially as described (10). Assay medium was DMEM (Life Technologies, Inc.) buffered with 25 mM Hepes, pH 7.4, instead of sodium bicarbonate. Where indicated, CaCl 2 was omitted from the medium. When used, the phosphodiesterase inhibitor Ro 20-1724 (final concentration 100 M) was added to the cells in assay medium 15 min prior to the addition of either ATP or 4-bromo-A23187. Reported values are mean Ϯ S.E. (n ϭ 4).
Fura-2 Fluorescence Measurements-Time-dependent changes in intracellular free Ca 2ϩ were measured with the fluorescent indicator, fura-2 (30), using a Spex Fluorolog DM3000 dual excitation spectrofluorometer. The autofluorescence of the stably transformed HEK-293 cells was recorded prior to fura-2/AM loading, and was subtracted from all subsequent readings. Confluent monolayers of cells on glass coverslips were loaded with the Ca 2ϩ indicator by incubation for 1 h at 37°C in Hepes-buffered DMEM containing 4.2 M fura-2/AM, but lacking sodium bicarbonate and phenol red. Traces labeled ϪCaCl 2 were obtained from cells that were transferred to fresh loading buffer that lacked CaCl 2 during the last 20 min of the 1 h incubation. The timing of washes and agonist addition was adjusted to correspond to the sequence used during measurement of intracellular cAMP accumulation. The linear time-dependent changes in 340 F 505 / 380 F 505 measured in the absence of added agonist, which may reflect metabolism of intracellular fura-2 (31), were subtracted from the time course of agonist-induced changes in the 340 F 505 / 380 F 505 ratios.
Assay of Adenylyl Cyclase-Preparation of HEK-293 cell membranes and the assay of AC activity was as described (13). The program EQCAL (Biosoft) was used to estimate values of 17 M free Ca 2ϩ , 5.74 mM free Mg 2ϩ and 0.23 mM free Mn 2ϩ under the assay conditions. For the ATP concentration dependence, the concentration of exogenous calmodulin was 1 M, while the concentrations of free Ca 2ϩ and Mg 2ϩ or Mn 2ϩ were kept constant by varying total concentrations of CaCl 2 and MgSO 4 or MnCl 2 . Activity measured in membranes prepared from cells expressing types VIII-A, -B, and -C was corrected for endogenous activity measured in membranes from vector-transformed HEK-293 cells. The program, Prism (GraphPad), was used to fit the data to the Michaelis-Menten equation.
The calmodulin concentration dependence was also determined in the presence of 17 M free Ca 2ϩ and 5.74 mM free Mg 2ϩ . Calmodulinindependent activity was measured in the presence of excess EGTA and was subtracted from all values. The concentration dependence for calmodulin activation was adequately represented by a hyperbolic equation if the concentration of calmodulin was considered to be the sum of the known concentration of added calmodulin and a parameter representing the unknown concentration of endogenous calmodulin in the membrane preparation. The activity value determined at each concentration of calmodulin was divided by the maximal activity computed for a given membrane preparation under these conditions to visualize the differences in EC 50 values.

Cloning of Type VIII Adenylyl Cyclase Splice Variants-
Screening of cDNA libraries with a type VIII adenylyl cyclase cDNA probe (10) resulted in the selection of three classes of clones. The first class, referred to here as type VIII-A cDNA clones, and the screening strategy were described previously (13), where the cDNAs were simply said to encode type VIII adenylyl cyclase. The two additional classes of clones are said to encode type VIII-B and type VIII-C adenylyl cyclase, and differ from overlapping type VIII-A clones by the deletion, in-frame, of 90 and 198 bp, respectively, from within their coding sequences. The deleted regions correspond to nucleotides 3180 -3269 and 2679 -2876 within the type VIII-A cDNA (GenBank accession number L26986). Otherwise these cDNAs are identical to that of type VIII-A at the nucleotide level, consistent with the proposal that they are derived from a single gene by an exon skipping mechanism.
The mRNAs encoding types VIII-A, -B, and -C adenylyl cyclase can be amplified from rat brain poly(A) ϩ RNA by reverse transcription PCR using oligonucleotide primers designed to distinguish the alternatively spliced messages (Fig. 1A). Products of 480 and 390 bp are amplified with primers flanking the region deleted from the type VIII-B cDNA, while products of 636 and 438 bp are amplified with primers flanking the region deleted from the type VIII-C cDNA. DNA sequencing verified that the 480-and 636-bp PCR products are derived from the type VIII-A message and the 390-and 438-bp PCR products arise from the type VIII-B and -C messages, respectively, as predicted based on the sequence of the cDNA clones. This confirms that the messages encoding the three splice variants are expressed in rat brain.
Both the rat type VIII-B and -C cDNA clones extend into the 3Ј-untranslated region but are incomplete at their 5Ј ends, terminating 13 and 585 nucleotides, respectively, 3Ј of the ATG encoding the initiator methionine in the type VIII-A cDNA. Therefore, PCR and Southern hybridization analyses were combined to demonstrate that the three type VIII messages share the same 5Ј end (Fig. 1, B-D). Rat genomic DNA was used as a template for amplification with a sense strand primer corresponding to 5Ј-untranslated sequence that is 5Ј of stop codons in all reading frames in the type VIII-A cDNA, and an antisense strand primer complementary to a region 3Ј of the start of both the partial type VIII-B and -C cDNAs. A single 1220-bp product is detected corresponding both in size (Fig. 1B) and exact sequence to that reported for the type VIII-A cDNA. Southern hybridization was performed with a probe internal to the genomic PCR product (Fig. 1, C and D). When the rat genomic DNA is digested with PstI and SfaNI, a single fragment of the size predicted based on the cDNA sequence is observed, consistent with the PCR results (Fig. 1C, lane 1). This indicates that all three splice variants share the same initiator methionine. Digestion of the genomic DNA with PstI alone indicates that there are introns in the type VIII gene since the probe hybridized to a single band of approximately 2600 bp rather than a 1640-bp fragment that is predicted based on the cDNA sequence (Fig. 1C, lane 2).
The open reading frames within the type VIII-B and -C messages encode proteins of 1218 and 1182 amino acids, respectively, while the type VIII-A cDNA encodes the 1248-amino acid protein previously described (13). Each of the type VIII splice variants conforms well to the topography typical of the membrane-bound, mammalian adenylyl cyclases (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14), including the two clusters of 6 potential transmembrane spans and putative nucleotide binding domains (3,21). The 90-base pair region spliced out of the type VIII-B message encodes amino acids 802-831 in type VIII-A adenylyl cyclase, which includes two consensus N-linked glycosylation sites (Asn-814 and Asn-818). This 30-amino acid domain is flanked on the NH 2 -terminal and COOH-terminal sides by putative transmembrane helices 9 and 10, respectively, and is predicted to be on the extracellular side of the membrane (Fig. 2). The 198-bp region spliced from the type VIII-C message encodes amino acids 635-700 in type VIII-A, which includes two consensus casein kinase II phosphorylation sites (Ser-659 and Ser-695) and a consensus bipartite nuclear targeting sequence (amino acids 666 -682) as determined by sequence analysis with the program, Prosite. However, there is no evidence to indicate the sites have these functional roles in vivo. This 66-amino acid stretch is in the C 1b region of the first large cytoplasmic domain (Fig. 2), which can vary considerably among the adenylyl cyclases (5,19,20). Other considerations of topography and sequence homology discussed previously for type VIII-A are the same for types VIII-B and C (13).
Immunochemical Detection of Type VIII AC in HEK-293 Cells and Rat Brain Membranes-HEK-293 cells were initially transfected with the type VIII-A, -B, or -C cDNA constructs in the pCMV5-neo vector and polyclonal populations of stably transformed cells were selected with the antibiotic, Geneticin. The expression of the splice variants was assessed with specific antipeptide antisera. Affinity-purified Ab VIII-A 1229 -1248 was raised against a peptide with the sequence of type VIII-A  1 and 2) or that deleted from the type VIII-C cDNA (lanes 3 and 4). The reactions were carried out in the presence (lanes 1 and 3), or the absence (lanes 2 and 4) of cDNA reverse transcribed from 0.25 g of rat brain poly(A) ϩ RNA. 4% of the total reaction volume was loaded into lanes 1 and 2 and 20% of reaction volume into lanes 3 and 4. Products were fractionated on a 1% agarose, 0.5 ϫ TBE gel and stained with ethidium bromide. B, amplification reactions were performed using a sense strand primer, which hybridizes to a site in the 5Ј-untranslated region of the type VIII cDNA constructs and an antisense primer to a region on the 3Ј side of the 5Ј ends of the type VIII-B and -C cDNA clones as described under "Experimental Procedures." Reactions contained 0.1 g of rat genomic DNA template from two separate preparations (lanes 1 and 2) or no genomic DNA (lane 3). The products from 10% of the final reaction volume were fractionated as in A. C, 8.5 g of rat genomic DNA was cut with PstI and SfaNI (lane 1) or PstI (lane 2) and subjected to Southern hybridization analysis as described under "Experimental Procedures." D, schematic representation of the structure of the 5Ј end of the rat type VIII AC gene. The open box represents the exon that includes the 5Ј coding sequence, and the broken line represents an intron. A, B, and C indicate positions corresponding to the 5Ј ends of the type VIII-A, -B, and -C cDNAs, respectively. Relative positions are indicated for the PstI and SfaNI cleavage sites, and the forward (F) and reverse (R) PCR primers used for amplification of genomic DNA (see panels B and C, and "Experimental Procedures"). amino acids 1229 -1248, a COOH-terminal peptide common to all three splice variants. HEK-293 cells do not express type VIII AC (13,16), and no immunoreactive species can be detected when whole cell lysates or membranes prepared from vector-transformed HEK-293 cells are immunoblotted with this antibody (Fig. 3, A and B). Immunoreactive species varying in size from 125 to 165 kDa are specifically detected in rat brain membranes and whole cell lysates or membranes prepared from the stably transformed HEK-293 cells (Fig. 3). All immunoreactive species co-migrate with the 125-kDa species when the preparations are first treated with N-glycosidase F. Note that type VIII-B/HEK-293 cells only express the 125-kDa species, and its mobility is not altered by treatment with N-glycosidase F (Fig. 3B). This indicates that at least one of the two N-linked glycosylation sites missing from type VIII-B is utilized in types VIII-A and -C. The fact that a single 125-kDa band is observed after deglycosylation of type VIII-C/HEK-293 cell membranes implies that the additional species detected in these cells arise from differences in glycosylation (Fig. 3B). Densitometric analysis of the immunoblots, such as those shown, indicates that in whole cell lysates the relative amounts of the immunoreactive species are approximately 8:1:2 for type VIII-A:-B:-C while in membranes the relative amounts are approximately 3:1:1 (Fig. 3, A and B, and data not shown). This suggests that there is differential recovery of the splice variants in the membrane preparations.
Ab VIII-A 804 -813 was raised against a peptide that should not be present in type VIII-B because the sequence encoding it has been removed from the message by alternative splicing. A nonspecific species of approximately 120 kDa is detected by this antiserum in all of the membranes including those prepared from the vector/HEK-293 cells (Fig. 3B). As expected, no 125-kDa immunoreactive species can be detected with this primary antibody in membranes prepared from type VIII-B/ HEK-293 cells. In membranes prepared from type VIII-A/and -C/HEK-293 cells, this antiserum recognizes the deglycosylated 125-kDa species more readily than the glycosylated 165-kDa species. This result can be rationalized if the interaction of Ab VIII-A 804 -813 with the 165-kDa species is sterically hindered by the presence of N-linked glycans adjacent to the epitope in types VIII-A and -C.
Ab VIII-A 666 -682 was raised against a peptide that is not present in type VIII-C because the sequence encoding it has been spliced out of the message. The patterns of immunoreactivity observed with this antibody in HEK-293 cell membranes are identical to those observed with Ab VIII-A 1229 -1248 except in type VIII-C/HEK-293 cells where no bands can be detected (Fig. 3B). The immunoblots confirm the expression of the type VIII-A, -B, and -C splice variants in the stably transformed HEK-293 cell populations. Neither Ab VIII-A 804 -813 nor Ab VIII-A 666 -682 was sufficiently sensitive or specific to unambiguously detect endogenous type VIII in a rat tissue, but the three immunoreactive species observed with Ab VIII-A 1229 -1248 (Fig. 3C) are consistent with the expression of the type VIII AC variants in rat brain.

Effects of Ca 2ϩ Mobilizing Agents on cAMP Accumulation in HEK-293 Cells Stably Expressing Type VIII Splice Variants-
Addition of 1 M 4-bromo-A23187, a Ca 2ϩ ionophore, stimulates cAMP accumulation in HEK-293 cells expressing any of the type VIII splice variants (Fig. 4A). Ionophore-dependent increases in cAMP correlate with the sustained, receptor-independent mechanism of Ca 2ϩ influx mediated by 4-bromo-A23187 as reflected in the increases in fura-2 fluorescence (Fig.   FIG. 3. Immunoblot analysis of the expression of types VIII-A,  -B, and -C AC in HEK-293 cells and rat brain membranes. A, whole cell lysates prepared from HEK-293 cells stably transformed with the type VIII expression constructs were subjected to SDS-PAGE and immunoblotted as described under "Experimental Procedures." The amount of total cell protein loaded was 10 g. The amount of this derived from a particular transformed cell population is indicated, and the difference was made up by the addition of vector/HEK-293 cell lysate where needed. B, membranes prepared from the stably transformed HEK-293 cells were subjected to SDS-PAGE and immunoblot analysis directly or after treatment with N-glycosidase F as described under "Experimental Procedures." When the primary antibody was Ab VIII-A 1229 -1248 or Ab VIII-A 666 -682, 2.5 g of vector/HEK-293 cell membranes, 0.5 g of type VIII-A/HEK-293 cell membranes, 1.5 g of type VIII-B/HEK-293 cell membranes, and 2.5 g of type VIII-C/HEK-293 cell membranes were loaded. Vector/HEK-293 cell membranes were added to equalize the loads where needed. When the primary antibody was Ab VIII-A 804 -813, 6 times more membrane protein was loaded. C, 25 g of rat brain membranes were loaded in each lane (see "Experimental Procedures"). The primary antibody was incubated with the antigenic peptide (0.1 g/ml) for 1 h prior to incubation with the Immobilon membrane in the lane designated "ϩpeptide."

FIG. 2. Schematic representation of predicted structures of types VIII-A, -B and -C AC.
The differences in the topography of the three type VIII splice variants are depicted based on analysis of their sequences and comparison to a model initially proposed for type I AC (2). A glycosylated 30-amino acid extracellular domain is predicted between transmembrane spans 9 and 10 in types VIII-A and -C, but is not present in type VIII-B. The hatched lines indicate where alternative splicing of the message has resulted in the excision of sequence encoding 66 amino acids in the C 1b region of type VIII-C. 4F). This is consistent with the Ca 2ϩ -dependent activation of type VIII-A reported previously (13) and extends the observation to types VIII-B and -C. Phorbol esters did not mimic the response induced by 4-bromo-A23187, implying that secondary activation of protein kinase C was not involved in the response to the ionophore, and the Ca 2ϩ /calmodulin-dependent kinase II inhibitor, KN93, also failed to alter the response (data not shown).
The P 2 purinergic agonist, ATP, stimulates cAMP synthesis in type VIII-A/, -B/, or -C/HEK-293 cells, although there are quantitative differences in the responses (Fig. 4B). In the presence of 100 M Ro 20-1724, a cAMP phosphodiesterase inhibitor, maximal cAMP accumulation is achieved 2-5 min after the addition of ATP in type VIII-A/or -C/HEK-293 cells and at 1-2 min in type VIII-B/HEK-293 cells (Fig. 4B), with no response in control cells. The time course of the ATP-induced cAMP changes lags behind the increase in intracellular Ca 2ϩ monitored by fura-2 fluorescence (Fig. 4F). cAMP content decreases at times greater than 5 min, regardless of which type VIII splice variant is expressed. Addition of a mixture of five phosphodiesterase inhibitors attenuates, but does not eliminate the decrease in cAMP content that occurs at longer times (data not shown), indicating that incomplete inhibition of phosphodiesterases by Ro 20-1724 contributes to this decay. ATP-dependent cAMP synthesis can be measured in the absence of any phosphodiesterase inhibitor, but the decay is more rapid and the maximal values are reduced to about half those measured in the presence of Ro 20-1724 (Fig. 4D). Intracellular Ca 2ϩ concentration changes are also reduced in the absence of Ro 20-1724 (Fig. 4F). This suggests that a component of the Ca 2ϩ influx observed at longer times may be stimulated in a cAMPdependent manner, and that the decreases in cAMP accumulation in the absence of Ro 20-1724 may reflect both enhanced degradation and decreased stimulation because of reduced Ca 2ϩ .
cAMP content was normalized to type VIII immunoreactivity to obtain an estimate for the relative specific activities of the variants in an intact cell assay (Fig. 4C). There is approximately 8-fold more type VIII immunoreactivity detected in cells expressing type VIII-A than in those expressing type VIII-B (Fig. 3A), and the differences in maximal cAMP accumulation at early time points (Fig. 4B) simply reflect the relative expression levels of these two variants (Fig. 4C). However, maximal increases in cAMP content in HEK-293 cells expressing type VIII-C are approximately 3-fold higher than would be expected if the ratio of cAMP accumulation to immunoreactivity were constant for each variant (Fig. 4C). Expression of the splice variants does not differentially affect the magnitude or time course of ATP-dependent changes in fura-2 fluorescence (additional data not shown) indicating that differential changes in intracellular Ca 2ϩ concentrations can not explain the apparent enhanced cAMP accumulation in type VIII-C/HEK-293 cells. Assays were performed in medium prepared without added CaCl 2 to determine the contribution of extracellular Ca 2ϩ to the stimulation of cAMP synthesis (Fig. 4, E versus B). There is an ATP-dependent release of Ca 2ϩ from intracellular stores under these assay conditions, but the maximal value of fura-2 fluorescence is decreased and the decay is accelerated (Fig. 4F). cAMP accumulation in minimal Ca 2ϩ medium is essentially independent of time and is reduced to approximately one tenth of that measured in the standard assay medium with 1.8 mM CaCl 2 (Fig. 4E). This indicates that the stimulation of the type VIII splice variants by ATP results primarily from the influx of extracellular Ca 2ϩ as previously noted for type VIII-A (13).
Enzymatic Activity of Type VIII Splice Variants Is Stimulated by Ca 2ϩ /Calmodulin-The enzymatic activity of the type VIII splice variants was measured in the presence of increasing concentrations of the substrates, MgATP or MnATP, and 1 M calmodulin (Fig. 5, A and B). For each set of conditions, measurements were made on Geneticin-resistant HEK-293 cell populations selected after transfection with the pCMV5-neo vector (q), type VIII-A (E), type VIII-B (ϫ), or type VIII-C (f) expression constructs. The notation ϩ or Ϫ Ca 2ϩ , indicates that the assay medium was Hepes-buffered DMEM prepared either with or without 1.8 mM CaCl 2 . C, cAMP accumulation presented in B has been scaled to the relative expression of the type VIII variants as determined by densitometric analysis of immunoreactivity (Fig. 3A). F, agonist-induced changes in intracellular Ca 2ϩ were recorded over time in type VIII-B/HEK-293 cells that were loaded with the fluorescent Ca 2ϩ indicator, fura-2. Fluorescence emission was measured at 505 nm in response to excitation at 340 nm ( 340 F 505 ) and 380 nm ( 380 F 505 ) as described under "Experimental Procedures." Corrected fluorescence ratios shown are signal averages of triplicate determinations for each set of conditions indicated.
The K m of type VIII-C is always higher than that of type VIII-A or -B, although the precise mechanism by which the 66 amino acid domain affects interactions with substrate remains to be determined. K m for MgATP of type VIII-C increases 18-fold relative to its K m for MnATP. It must be noted that type VIII-C membrane activity is the most labile of the three, particularly in the presence of detergents (data not shown). Differential stabilization of type VIII-C by Mn 2ϩ might contribute to the more pronounced relative effect of this divalent ion on the K m of the VIII-C variant. The maximal velocities of types VIII-A and -B vary in proportion to their relative expression levels (A:B approximately 3:1, Fig. 3B), and their V max values increase modestly with MnATP as substrate. The V max of type VIII-C is insensitive to this change of substrate.
AC activity in membranes prepared from type VIII-A/, -B/, or -C/HEK-293 cells increases in a concentration-dependent manner with the addition of exogenous Ca 2ϩ /calmodulin (Fig. 5C), while the activity in control HEK-293 cells is insensitive to this treatment (data not shown and Ref. 13). No calmodulin-dependent effect on AC activity is observed in any of the membrane preparations when free Ca 2ϩ is chelated with excess EGTA, although this treatment significantly inhibits the type VIII activity measured in the absence of exogenous calmodulin (data not shown and Ref. 13). The curves of Fig. 5C have been fitted assuming that the total concentration of calmodulin stimulating the enzyme is the sum of the known concentration of exogenously added calmodulin and a parameter representing the unknown concentration of endogenous calmodulin in the membrane preparation. The computed estimate for the latter parameter is 8.7 Ϯ 0.5 nM for the three membrane preparations, and the addition of this single parameter significantly improves the fit (p Ͻ 0.0002), although it increases the EC 50 values (13). The calculated EC 50 values for calmodulin-dependent stimulation of AC activity are 140 Ϯ 21 nM, 116 Ϯ 12 nM, and 30 Ϯ 6 nM for types VIII-A, -B, and -C, respectively, indicating that type VIII-C is approximately 4 times more sensitive to stimulation by Ca 2ϩ /calmodulin than type VIII-A or -B. Under three different sets of experimental conditions, the relative ratio of the calculated EC 50 values for calmodulin-dependent activation of type VIII-A or -B to that of type VIII-C was 4.5 Ϯ 0.5 ( Fig. 5C and data not shown). DISCUSSION Evidence for alternative splicing of the rat type VIII message was originally obtained by cDNA cloning, and expression of type VIII-A, -B, and -C messages in rat brain has been verified by reverse transcription-PCR (Fig. 1). The variant messages can arise if distinct exons are alternatively excised from the complete type VIII-A precursor transcript to generate the type VIII-B and -C messages. While the rat type VIII-B variant has been characterized here, a bovine homolog was also selected from a brain library (data not shown), and a partial human type VIII cDNA clone that apparently has the properties of the -B variant has been discussed (32). The conservation of this variant across species indicates that splicing of the type VIII message is physiologically significant.
Three proteins detected in immunoblots of rat brain membranes appear to co-migrate with the variants as expressed in HEK-293 cells, and all of the immunoreactive species are reduced to approximately 125 kDa by treatment with N-glycosidase F (Fig. 3C). Preincubation of the antibody with the antigenic peptide prevents detection of these brain membrane proteins arguing that recognition of all three species is specific. Unfortunately neither Ab VIII-A 666 -682 nor Ab VIII-A 804 -813 is sufficiently specific to unambiguously identify the presence of additional type VIII sequences in these rat brain membrane proteins. The localization and identification of adenylyl cyclase isoforms in mammalian tissues has been hampered by their low abundance, and most information has therefore come from analysis of mRNA expression (1).
The structural properties of type VIII-B are distinct from those of type VIII-A, although we have not been able to distinguish their enzymatic properties. The type VIII-B variant is insensitive to treatment with N-glycosidase F (Fig. 3), consistent with the fact that it lacks the extracellular domain between transmembrane spans 9 and 10, which includes consensus N-linked glycosylation sites in all cloned ACs (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14). Additional consensus N-linked glycosylation sites are present in the sequence of several ACs in a putative extracellular domain between transmembrane spans 11 and 12 (4,6,10,13,14,33). The single site present in this latter domain in type VIII-B is apparently not utilized when the protein is expressed in HEK-293 cells. The expression of non-glycosylated, membrane-associated forms of AC in bovine brain has been proposed based on purification studies using lectin chromatography (34,35). The absence of functional glycosylation sites in type VIII-B is expected to alter the post-translational processing pathway which the protein follows, and could result in its localization to a distinct membrane compartment relative to the glycosylated types VIII-A and -C. Differential processing, or localization might contribute to the fact that expression of type VIII-B is reduced relative to that of type VIII-A and -C (Fig. 3). Mutation of the glycosylation sites of type VIII-A and immunocytochemical studies with variant-specific antisera will address these possibilities.
The type VIII-C splice variant has unique structural properties that affect its enzymatic activity when assayed both in whole cells and membranes. The exon that is skipped to generate the type VIII-C message encodes 66 amino acids that are found in the C 1b region of type VIII-A or -B (Fig. 2). The calmodulin-binding domain of type I AC has been localized to this region (23)(24)(25). There is no significant sequence homology between the calmodulin-binding domain of type I and the corresponding sequence in type VIII-A, nor is this region in type VIII predicted to have the amphipathic structure typical of calmodulin-binding domains (36). If the sequences of types I and VIII-A are aligned, the calmodulin-binding domain of type I terminates just 7 amino acids before the start of the 66-amino acid region that is missing from type VIII-C. Deletion of this region actually enhances the calmodulin-sensitivity of type VIII-C approximately 4-fold relative to that of type VIII-A or -B as reflected in their different EC 50 values for calmodulin-dependent stimulation (Fig. 5B). This could be a conformational effect, or the additional 66 amino acids in types VIII-A and -B may play an inhibitory role. The enhanced calmodulin-sensitivity of type VIII-C is reflected in whole cell assays; cAMP content in HEK-293 cells expressing type VIII-C is approximately 3-fold higher than would be expected if cAMP accumulation were directly proportional to the relative expression levels of the variants (Fig. 4C). Stimulation of the type VIII variants requires the modest increases in intracellular Ca 2ϩ that result from the influx of extracellular Ca 2ϩ (Ref. 13 and Fig. 4E). Under these conditions the concentration of intracellular Ca 2ϩ is limiting for the formation of the stimulatory Ca 2ϩ /calmodulin complex. The relative differences in the EC 50 values for Ca 2ϩ /calmodulin-dependent stimulation are predicted to result in greater relative activation of type VIII-C, as supported by the data.
Total type VIII message is abundantly expressed throughout the hippocampus of the rat brain, as determined with an oligonucleotide probe that does not distinguish between the splice variants (13). A role for type VIII-A in long term potentiation (LTP) in the hippocampus was suggested based on the distribution of its message and its enzymatic properties (13). Nmethyl-D-aspartate receptor activation, which induces LTP, also causes an influx of extracellular Ca 2ϩ that stimulates cAMP accumulation in the dendritic spines of post-synaptic neurons in the CA1 field of the hippocampus (37,38). Calmodulin-stimulated ACs have also been implicated in the regulation of LTP at the mossy fiber synapses (39). In this case it is the depolarization-dependent activation of Ca 2ϩ channels in the presynaptic axon terminal that leads to a Ca 2ϩ influx and activation of calmodulin-sensitive ACs. This illustrates the need to have calmodulin-stimulated ACs capable of responding to Ca 2ϩ influxes of potentially different magnitudes mediated by two types of channels in distinct neuronal compartments. Alternative splicing of the type VIII AC message may provide a mechanism to generate variants with the requisite properties to contribute to the expression of LTP in both of these pathways.