Isolation and characterization of cDNAs corresponding to two human calcium, calmodulin-regulated, 3',5'-cyclic nucleotide phosphodiesterases.

cDNAs corresponding to two human calcium, calmodulin (CaM)-regulated 3′,5′-cyclic nucleotide phosphodiesterases (PDEs) were isolated. One, Hcam1 (PDE1A3), corresponds to the bovine 61-kDa CaM PDE (PDE1A2). The second, Hcam3 (PDE1C), represents a novel phosphodiesterase gene. Hcam1 encodes a 535-amino acid protein that differs most notably from the bovine 61-kDa CaM PDE by the presence of a 14-amino acid insertion and a divergent carboxyl terminus. RNase protection studies indicated that Hcam1 is represented in human RNA from several tissues, including brain, kidney, testes, and heart. Two carboxyl-terminal splice variants for Hcam3 were isolated. One, Hcam3b (PDE1C1), encodes a protein 634 amino acids (72 kDa) in length. The other, Hcam3a (PDE1C3), diverges from Hcam3b 4 amino acids from the carboxyl terminus of Hcam3b, and extends an additional 79 amino acids. All the cDNAs isolated for Hcam3a are incomplete; they do not include the 5′-end of the open reading frame. Northern analysis revealed that both splice variants were expressed in several tissues, including brain and heart, and that there may be additional splice variants. Amino-truncated recombinant proteins were expressed in yeast and characterized biochemically. Hcam3a has a high affinity for both cAMP and cGMP and thus has distinctly different kinetic parameters from Hcam1, which has a higher affinity for cGMP than for cAMP. Both PDE1C enzymes were inhibited by isobutylmethylxanthine, 8-methoxymethyl isobutylmethylxanthine, zaprinast, and vinpocetine.

Cyclic nucleotides are involved in a large number of mammalian signal transduction pathways. The intracellular concentrations of cAMP and cGMP reflect their rate of synthesis, by adenylyl and guanylyl cyclases, and their rate of degradation to 5Ј-monophosphate nucleosides, by cyclic nucleotide phosphodiesterases (PDEs). 1 A number of biochemically distinct PDEs have been identified. They fall into seven families distinguished by their allosteric regulation, substrate kinetics, amino acid sequence homology, and interaction with specific inhibitors (1,2). All known mammalian PDEs share a conserved region of approximately 250 amino acids that contains the phosphodiesterase catalytic site (3). In certain PDEs the regions amino-terminal to the catalytic domain are known to be involved in the allosteric regulation caused by the binding of calmodulin or cGMP (3).
Type I PDEs (PDE1, CaM PDEs) are activated by the binding of calmodulin in the presence of Ca 2ϩ (CaM). This binding increases the hydrolysis of both cAMP and cGMP (4). The regulation of cyclic nucleotide hydrolysis by changes in calcium concentration may allow the type I enzymes to integrate the Ca 2ϩ and cyclic nucleotide second messenger pathways within a cell.
Biochemical characterizations have distinguished at least five type I PDEs differing from each other in apparent molecular weight, K m values for cyclic nucleotide substrates, affinities for calmodulin, and regulation by phosphorylation (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16). The cDNAs for several of these type I PDEs have been isolated and characterized (17)(18)(19)(20). The existence of type I PDEs with distinct sequences and properties suggests a diversity of cellular functions for these enzymes. However, this diversity complicates the analysis of type I PDEs in tissues where more than one biochemical form is present. Isolating each of the type I PDE genes helps aid in understanding the role played by each member of this complex family of enzymes.
We report here the isolation and characterization of cDNAs corresponding to two human type I PDE genes. One, Hcam1, corresponds to the bovine 61-kDa CaM PDE. The other, Hcam3, is a novel type I PDE, whose mouse and rat counterparts have also been recently isolated (52). We examine the tissue distribution of the mRNAs for Hcam1 and Hcam3 and the biochemical properties of amino-truncated proteins corresponding to each gene. Hcam3 shows alternative splicing that would generate phosphodiesterase proteins with different carboxyl termini; this is the first example of carboxyl-terminal splice variants for mammalian type I PDEs.

MATERIALS AND METHODS
Library Screening-A bovine 61-kDa CaM PDE cDNA fragment was used as a probe to screen human cDNA libraries. Degenerate oligonucleotides (corresponding to the bovine 61-kDa CaM PDE * This work was supported in part by Grants HL44948 and DK21723 from the National Institutes of Health. 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 GenBank TM  amino acid sequences KMGMMKKK and NMKGTTND) and bovine heart cDNA were used in a PCR reaction to generate an 1108-bp bovine cDNA fragment as previously described (17). Hybridization probes were prepared by isolating DNA fragments from agarose gels and labeling them with [ 32 P]dCTP and [ 32 P]dTTP (800 Ci/mmol, DuPont) using a Boehringer Mannheim random priming kit. Library screening and hybridization conditions were as described elsewhere (21). Three positively hybridizing phage (H2a, H3a, H6a) were isolated from a human hippocampus cDNA library (Clontech). The cDNA inserts from these phage were subsequently used as probes to isolate additional phage. An H6a probe (1.2-kb HindIII/EcoRV fragment) was used to isolate phage A2d from an aorta cDNA library (Clontech). An H3a probe (2.4-kb HindIII/EcoRI fragment) was used to isolate two additional phage, He11a and He19a, from the heart library (Stratagene).
Subcloning and DNA Sequencing-cDNA inserts from H6a, H2a, H3a, and A2d were subcloned into Bluescript vectors (Stratagene). Bluescript plasmids containing the cDNA inserts from He11a and He19a were excised in vivo from the ZAP vector (Stratagene) following the manufacturer's instructions. Plasmid DNA was extracted using kits from Promega Biotech Inc. and QIAGEN Inc. Restriction and modification enzymes were purchased from Boehringer Mannheim. Oligonucleotides were synthesized using an ABI 394 DNA synthesizer. The cDNA inserts from H6a, H3a, He11a, and A2d were completely sequenced on both strands using a Sequenase kit (U. S. Biochemical Corp). Those from He19a and H2a were partially sequenced. Additional molecular biological techniques are as previously described (22).
RNA Isolation-Human tissue samples were obtained from the National Disease Research Interchange and the Cooperative Human Tissue Network. Tissue was frozen within 8 h of death or surgery except for three of the four heart samples which were frozen within 30 h of death. The tissues were pulverized under liquid nitrogen, and RNA was extracted by the method of Chomczynski and Sacchi (23).
RT-PCR-First strand cDNA was reverse-transcribed from 2 g of human heart poly(A) ϩ mRNA using a Boehringer Mannheim cDNA synthesis kit. 0.5 l of the 20-l final reaction volume was used per subsequent PCR reaction, which also contained 10 g/ml of each primer, 10 mM Tris, pH 8.3, 1.5 mM MgCl 2 , 50 mM KCl, 0.2 mM each dNTP (Boehringer Mannheim), 0.6 unit Taq polymerase (Boehringer Mannheim) in 25 l. Reaction conditions were 94°C for 4 min, followed by 30 cycles of 94°C for 1 min, and 60°C for 2 min and 72°C for 4 min. The polymerase was added after the reaction temperature reached 94°C. Primary PCR reactions were diluted 1/100, and 1 l was added to a 25-l secondary PCR reaction. Hcam1 primers are shown in Fig. 1B, primers 1 and 4 were used in the primary reaction, and primers 2 and 3 in the secondary reaction. Hcam3 primers are shown in Fig. 2, B and C; primers 1 and 4 were used in the first reaction and primers 2 and 3 in the secondary reaction. The PCR-derived DNA fragments were gel purified and partially sequenced.
RNase Protection Assays-An antisense Hcam1 RNA probe (made from a plasmid containing nucleotides 1252-1939; Fig. 1B The reaction was extracted sequentially with phenol/chloroform and chloroform and then precipitated (50 l of 7.5 M ammonium acetate, 300 l of ethanol).
Northern Analysis-A multiple tissue blot was purchased from Clontech. The probes were prepared as described for the RNase protections except that no unlabeled UTP was present in the reaction. Prehybridizations were performed in 50% formamide, 50 mM sodium phosphate, pH 6.8, 5 ϫ SSC, 1 mM EDTA, 2.5 ϫ Denhardt's solution, 200 g/ml denatured salmon testes DNA, 0.5 mg/ml tRNA, and 20 units/ml heparin at 65°C; fresh solution containing the probe was added for the hybridization overnight. 5 ϫ SSC and 2.5 ϫ Denhardt's solutions are as defined in Sambrook et al. (22). The filters were washed first at room temperature and then at 65°C in 0.1 ϫ SSC containing 0.1% SDS prior to autoradiography. The filter was stripped between uses by immersion in boiling water. The Hcam1 riboprobe was the same as that used in the RNase protection analysis. The Hcam3a-specific probe extended from nucleotides 2068 -2686 ( Fig. 2C) and the Hcam3b-specific probe extended from nucleotides 2062-2652 (Fig. 2B). Fragments corresponding to these regions were generated by PCR and subcloned prior to their use in preparing an antisense riboprobe (as described above).
Expression Constructs-The cDNAs were cloned into the yeast expression vector pBNY6n and expressed in Saccharomyces cerevisiae. This vector, which provided EcoRI and XhoI cloning sites, was constructed from pADNS (25) and Yeplac181 (26) as follows. The polylinker of pADNS was replaced with an EcoRI/NotI/XhoI linker (5Ј-AGCTC-GAATTCTGCGGCCGCTCGAGA-3Ј and 5Ј-AATTTCTCGAGCGGCCG-CAGAATTCG-3Ј). This linker was ligated into pADNS that had been cut to completion with HindIII and partially digested with EcoRI. The resultant plasmid was cleaved with BamHI and the fragment that contained the ADHI promoter and terminator flanking the new polylinker was inserted into the BamHI site of a YEplac181 derivative. In this derivative the EcoRI site of YEplac181 had been removed by digestion with HpaI and SmaI followed by ligation.
The plasmid Hcam3a(met150) was constructed by ligating the cDNA insert of He11a (nucleotides 610-2067 of Fig. 2B followed by 2068 -2686 of Fig. 2C) into the EcoRI site of pBNY6n. New junctions in the construct were verified by DNA sequencing. The first ATG in the cDNA insert is nucleotide 624 of Hcam3a (Fig. 2B) which corresponds to amino acid 150.
The plasmid Hcam1(met141) contains an EcoRI/XhoI fragment (nucleotides 505-1692, Fig. 1B) inserted into pBNY6n. The first ATG in the cDNA insert is nucleotide 505 of Hcam1 (Fig. 1B) which corresponds to amino acid 141. DNA sequences generated by PCR and new junctions were verified by DNA sequencing.
Yeast Strains-The genotypes of the yeast strains used in this work are shown in Table I. YKS42 was generated from SX50 -1C by deleting The presence of the pep4-3 allele in YKS76 was confirmed by sequencing the region of the gene that contains the nonsense mutation present in this allele (34). Tests (35,36) for the presence of the prb1-1122 and prc1-126 alleles gave ambiguous results because pep4-3 is epistatic to their expression.
Biochemistry-YKS76 was transformed (37) with the plasmids Hcam1(met141) and Hcam3a(met150) and the transformants were grown at 30°C in synthetic medium lacking leucine to 1-2 ϫ 10 7 cells/ml, collected by centrifugation, washed once with water, and frozen in a dry ice-ethanol bath.
The yeast cell pellets were lysed and phosphodiesterase activity was assayed as described elsewhere (38) with the following modifications. [ 32 P]cAMP and [ 32 P]cGMP were used (25 Ci/mmol, ICN Biochemical). The 32 P i product was separated from the substrate using charcoal (39,40). Two volumes of 25 mg/ml activated charcoal in 0.1 M KH 2 PO 4 were added to terminate the reaction. Following centrifugation, the supernatant was removed and quantitated by Cerenkov counting. Extracts prepared from YKS76 containing only the pBNY6n vector yielded ac-tivities indistinguishable from assay background.
Kinetic and inhibitor analyses were performed as described previously (38). Cyclic nucleotide hydrolysis was measured at substrate concentrations ranging from 0.03 to 100 M. The specific activity of the substrate was held constant (0.2 Ci/mmol). The kinetic data were fitted to the Michaelis-Menton model using TableCurve (Jandel Scientific). Inhibitor analyses were performed at 0.1 M cGMP. At this substrate concentration the observed IC 50 should approximate the apparent inhibitor constant (K i ) for both competitive and noncompetitive inhibitors. The reaction time and amount of enzyme in the assay were adjusted such that less than 25% of the substrate was consumed in the reaction. Dose-response curves were fitted (TableCurve ) using a four-parameter or a two-parameter logistic model (38). The four-parameter model derives values for the minimum PDE activity, the maximum PDE activity, and the IC 50 , as well as a parameter that determines the slope of the fitted curve at the IC 50 . The two-parameter model derives values for the maximum PDE activity and the IC 50 . The minimum PDE activity is set to 0, and the parameter affecting the slope is set to 1. The two-parameter model was used when the highest concentration of inhibitor was unable to inhibit fully the enzymatic activity. All other data sets utilized the four-parameter model. Stocks of inhibitors were prepared in dimethyl sulfoxide (Aldrich), and the final solvent concentrations in the PDE assay never exceeded 2% (v/v). Vinpocetine was obtained from Andard Mount Co.; IBMX from Sigma; rolipram and zaprinast were gifts from Paul Feldman, Glaxo Inc. Research Institute (Research Triangle Park, NC); and 8-methoxymethyl IBMX was a gift from Jack Wells, Vanderbilt University (Nashville, TN). Cilostamide was synthe- The open reading frame encoded by a composite of these cDNAs is diagrammed, and the putative CaM binding region and the phosphodiesterase catalytic region are indicated. B, nucleotide sequence and predicted protein sequence of Hcam1 cDNAs (PDE1A3). Nucleotide and amino acid residues are numbered on the right. The methionine at amino acid position 141 is underlined to indicate the initiation point of the truncated protein Hcam1(met141). Also underlined is the insertion in the human Hcam1 relative to the bovine 61-kDa CaM PDE (see Panel C). The exact boundaries of the insertion are uncertain as it is flanked by AG nucleotides only one pair of which is part of the insertion. The positions of oligonucleotides used for RT-PCR are indicated above the sequence by numbered arrows. C, an alignment of the predicted human Hcam1 amino acid sequence (top) and the bovine 61-kDa CaM PDE amino acid sequence (bottom). A dot in the bovine sequence indicates that the amino acid in that position is identical to that found in the human sequence. The dashes in the bovine sequence indicate a region present in the human sequence that is not present in the bovine sequence. The two different open reading frames represented by these cDNAs are diagrammed. The portion of cDNA He19a represented as a dashed line has not been completely sequenced. B, nucleotide sequence and predicted protein sequence of Hcam3b (PDE1C1). The Hcam3a/3b divergence follows nucleotide 2067. The D marked with a diamond is the first amino acid that is specific to the Hcam3b splice variant. cDNA He11a begins at nucleotide 611 and the protein expressed by the truncated Hcam3a(met150) construct begins at the methionine underlined at amino acid position 150. The positions of oligonucleotides used for RT-PCR are indicated above the sequence by numbered arrows. Nucleotide and amino acid residues are numbered on the right. C, nucleotide sequence and predicted protein sequence specific to Hcam3a (PDE1C3). The first nucleotide of this sequence corresponds to nucleotide 2067 of Panel B. Following this nucleotide (G) the two sequences diverge. The Hcam3a protein contains a glycine (G) rather than the aspartic acid (D) marked with a diamond in Panel B and extends another 79 amino acids before terminating. The cDNA insert from He19a contains a G at position 2246, whereas that from He11a contains an A. This change does not affect the amino acid sequence of the protein. Nucleotide and amino acid residues are numbered on the right. sized using published procedures (41). Protein concentrations were assayed using a protein assay kit (Bio-Rad) and published methods (42).

RESULTS
Three phage were isolated from a human hippocampus cDNA library screened with a cDNA fragment derived from the bovine 61-kDa CaM PDE (17). DNA fragments derived from these phage were used to screen heart and aorta human cDNA libraries, and three additional phage were isolated. Subcloning and sequencing revealed that these six cDNAs correspond to two different genes, Hcam1 and Hcam3.
Hcam1-Three Hcam1 cDNAs (H2a and H6a from hippocampus and A2d from aorta) provide a composite sequence of 2008 nucleotides (Fig. 1, A and B). This composite cDNA encodes a 535-amino acid protein with a predicted molecular mass of 61,251 Da that is similar to the bovine 61-kDa CaM PDE protein (5, 17) (Fig. 1C).
The three cDNAs used in building the composite (Fig. 1B) differ from it as follows (Fig. 1A). The composite sequence is identical to the sequence of cDNA H6a except that H6a is lacking nucleotides 626 and 627. These nucleotides are present in H2a and in the bovine 61-kDa CaM PDE. Deletion of them alters the reading frame. We believe their absence in H6a is a cloning artifact.
The cDNA insert in H2a contains a 47-bp insertion following nucleotide 74 and diverges from the Hcam1 composite sequence following nucleotide 807 (Fig. 1A). The 47-bp insertion is in the 5Ј-untranslated region and does not alter the open reading frame of the cDNA. It contains sequences that show a good match to consensus splice donor and acceptor sequences (44), which suggests it may be an intron that was not spliced out of the mRNA from which the cDNA was made. It was not included in the Hcam1 composite sequence (Fig. 1B). The sequences following nucleotide 807, which show a good match to consensus splice donor sequences (44), may also represent an intron. It is known from an analysis of genomic DNA sequences that an intron is present at this position. 2 These sequences were also not included in the composite sequence (Fig. 1B).
RT-PCR was used to isolate a portion of Hcam1 from human heart cDNA. A PCR-derived band extending from primer 2 to primer 3 (Fig. 1B) was sequenced. The 2 bp (626 -627) were present, and the presumed intron following nucleotide 807 was absent, thus confirming our interpretation of the cDNA structures.
The sequence of the cDNA insert in A2d contains 15 nucleotides at its 5Ј-end that represent an inverted repeat of sequences found elsewhere in the cDNA and which are presumably a cloning artifact. It also contains 4 additional nucleotides located just 5Ј to the poly(A) tail (AGCT following nucleotide 1999 of Fig. 1B). These may reflect the use of a second polyadenylation site. The composite Hcam1 sequence (Fig. 1B) has been submitted to GenBank TM as HSPDE1A3.
The human Hcam1 protein sequence contains residues that are identical to a calmodulin-binding region identified in the bovine 61-kDa CaM PDE (amino acids 24 -42, Fig. 1, A and B) (5,17). Hcam1 also contains a region conserved in all mammalian phosphodiesterases that corresponds to the phosphodiesterase catalytic domain (amino acids 194 -447, Fig. 1, A and B) (3). Although Hcam1 is most similar to the bovine 61-kDa CaM PDE (PDE1A2), the human protein contains two regions that are not found in the bovine protein. One region is a 14-amino acid insertion in the human protein following amino acid 458. The second region is the carboxyl terminus; the carboxyl-terminal 14 amino acids of the human protein have no homology to the carboxyl-terminal 23 amino acids of the bovine protein.
When these two regions are excluded from comparison, the human and bovine proteins share 94% amino acid identity (Fig. 1C).
Hcam3-The sequences of three Hcam3 cDNAs (He11a and He19a from heart and H3a from hippocampus) yield a composite that predicts two splice variants with different carboxyl termini ( Fig. 2A). One predicted splice variant, Hcam3b, is represented by cDNA H3a. This splice variant is designated HSPDE1C1 and was submitted to GenBank TM . The 2694-nucleotide cDNA contains an open reading frame that encodes a 634-amino acid protein with a predicted molecular mass of 72,207 Da (Fig. 2B). The second splice variant, Hcam3a, is represented by cDNAs He11a and He19a. This splice variant is designated HSPDE1C3. Both cDNAs (He11a and He19a) encode a protein with a carboxyl terminus that diverges from that found in Hcam3b, and both cDNAs are lacking the 5Ј-end of the open reading frame. He11a (nucleotides 611-2067 of Fig. 2B followed by nucleotides 2068 -2686 of Fig. 2C) was submitted to GenBank TM as HSPDE1C3.
RT-PCR using human heart cDNA was performed to determine if the Hcam3a 3Ј-end is associated with the Hcam3b 5Ј-end. A DNA fragment was amplified in two rounds of PCR using first primers 1 and 4 (Fig. 2, B and C) and then primers 2 and 3 (Fig. 2, B and C). Sequence analysis revealed that the the DNA fragment had a Hcam3b 5Ј-end and a Hcam3a 3Ј-end. This full-length Hcam3a would encode a 709-amino acid protein with a predicted molecular mass of 80,759. Although the sequence of Hcam3 shares certain characteristics with the human (this work) and bovine (5, 17) 61-kDa CaM PDEs and to the human (19), bovine (6,18), mouse (19,20), and rat (19) 63-kDa CaM PDEs, it appears that Hcam3 represents a novel CaM PDE gene.
The divergence of the human Hcam1 sequence from the bovine 61-kDa CaM PDE sequence and the presence of two different carboxyl-terminal sequences for Hcam3 raised a number of questions about the structures of these cDNAs. Northern and RNase protection analyses were performed both to confirm the cDNA structures and to examine the tissue distribution of the CaM PDE transcripts.
Northern Analysis and RNase Protection Assay of Hcam1-In Northern blots a probe for Hcam1 hybridized to two different sized mRNAs in brain, heart, liver, skeletal muscle, and kidney samples (Fig. 3). The larger mRNA is 4.8 kb. Little of it is present in the liver and skeletal muscle samples. The smaller mRNA is 2.4 kb in the brain sample and 2.6 kb in 2 P. Snyder, unpublished observation. the other samples. (The 2.4-kb/2.6-kb size difference was seen on two different blots (data not shown).) Lung and pancreas mRNA samples contain a small amount of the 2.6-kb mRNA. No hybridization signal was seen in the placenta sample. The transcripts were most abundant in mRNA from the brain, heart, kidney, and skeletal muscle.
We used RNase protection to determine if the specific mRNA structure predicted from the Hcam1 cDNAs was present in human RNA samples. The RNase protection probe extended across the two regions where the bovine and human sequences differed from each other (42-nucleotide insertion in human sequence and divergent 3Ј-ends, Fig. 4B). Using this probe a 688-nucleotide band was seen (Fig. 4A). This band was the size expected for protection of the entire Hcam1 probe. This confirmed that both the 42-nucleotide insertion and the human type of 3Ј-end were found in RNA from these tissues. A human mRNA lacking the 42-nucleotide insertion would have given rise to a 207-209-nucleotide band. No such band was detected. The smaller bands included a prominent 397-nucleotide band (Fig. 4A). This was the size expected for protection of the probe 5Ј to the human/bovine divergence point and suggested the possibility that an additional form(s) of human Hcam1 mRNA might exist. Isolation of additional cDNAs will be required to test this suggestion.
Hcam1 mRNA expression was detected by RNase protection in RNA extracted from temporal cortex, hippocampus, brain stem, cerebellum, occipital cortex, heart ventricle, aorta, femoral and renal arteries, psoas skeletal muscle, and uterus (Fig.  4A). No expression was detected in the heart atrium, liver, or spleen.
Some of the RNA samples used in this RNase protection (indicated by stars in Fig. 4A) were partially degraded, possibly due to the length of time between donor death and tissue acquisition. The presence of protected bands using these samples indicated that the RNA included molecules that were long enough to span the RNase protection probe, although the signals obtained may be underrepresented or absent.
The results obtained in the RNase protections and the Northern blot agree reasonably well, although a few discrepancies were seen. The liver RNA used in the RNase protection showed no Hcam1 signal, while a different liver sample used in the Northern blot was positive. The skeletal muscle signal was also weaker in the RNase protection than might have been expected from the Northern blot results (again, a different sample). In the RNase protections, different samples of the heart and the testes gave qualitatively different results. Whether these differences reflect different expression levels in the different tissue donors, differing locations of the tissue within the organ, or some variation due to technical reasons is unknown.
Northern Analyses of Hcam3-For Hcam3 the cDNAs had predicted two distinct transcripts, one corresponding to Hcam3a and one corresponding to Hcam3b. Hybridization probes specific for each of these were prepared (see "Materials and Methods") and used on Northern blots.
The Hcam3a-specific probe hybridized to a 5.6-kb mRNA present in the heart and brain samples (Fig. 5A). A weaker hybridization signal was detected in the lung, liver, kidney, and skeletal muscle samples. The Hcam3b-specific probe hybridized to an approximately 10-kb mRNA present in the brain and heart, and to a lesser extent in the lung mRNA (Fig. 5B). Additional faint bands were also detected with this probe. A Hcam3 catalytic region probe, which should hybridize to both transcripts, hybridized to both a 10-and a 5.6-kb transcript as well as additional smaller transcripts in the heart sample (data not shown). These results confirmed that both Hcam3a and Hcam3b represent transcripts that are present in human tissue mRNA. Hcam3 expression was also detected in the uterus and testes by RNase protection (data not shown).
Sequence Comparison of CaM PDEs-Human Hcam1, bovine 63-kDa CaM PDE (18), and human Hcam3 show 59% identity of amino acids when the gaps and the divergent carboxyl termini are excluded from the comparison (Fig. 6). Compared pairwise, Hcam1 and bovine 63-kDa CaM PDE are 65% identical as are bovine 63-kDa CaM PDE and Hcam3. Hcam1 and Hcam3 are 77% identical. A partial sequence for the human counterpart of the bovine 63-kDa CaM PDE has been reported (19), and it is 97% identical to the bovine sequence. Thus the greater identity of human Hcam1 and Hcam3 to each other than to the bovine 63-kDa CaM PDE is probably not a species difference. Hcam3 differs from Hcam1 and the 63-kDa CaM PDE by the presence of a 9-amino acid insertion in the putative calmodulin binding domain. Hcam3a and 3b are also larger than Hcam1 and the 63-kDa CaM PDE because of the longer carboxyl termini found in the Hcam3 proteins.
Biochemical Characterization of Hcam1 and Hcam3 Proteins Expressed in S. cerevisiae-The proteins encoded by Hcam1, Hcam3a, and Hcam3b were expressed in a strain of yeast that lacked endogenous PDE activity (YKS76). Expression of fulllength proteins resulted in very low PDE activity that was not stimulated by 10 M calmodulin (data not shown). This cannot be explained by the simple absence of enzyme, because fulllength Hcam1, Hcam3b, and Hcam3a (with the Hcam3b amino terminus) proteins were detected in these yeast extracts by Western analysis (data not shown). This result is also not a simple consequence of the yeast expression system because full-length human Hcam2 (a human counterpart of the bovine 63-kDa CaM PDE) can be expressed in yeast 3 and the activity of the expressed protein is stimulated by calcium and calmodulin. In contrast to the full-length proteins, amino-truncated proteins beginning at methionine 141 of Hcam1 (Fig. 1B) or methionine 150 of Hcam3a (Fig. 2, B and C) gave measurable of PDE activity and hydrolyzed both cAMP and cGMP (truncated Hcam3b was not expressed).  Fig. 2C). B, Hcam3b-specific probe (nucleotides 2062-2652 of Fig. 2B).
The effect of substrate concentration on the initial velocity of cyclic nucleotide hydrolysis was examined. Hcam1(met141) had a K m for cAMP 15-fold greater than its K m for cGMP (Fig.  7A, Table II). The specific activity or maximal rate of hydrolysis for cGMP and cAMP was determined for the yeast extract (nmol⅐min Ϫ1 ⅐mg protein Ϫ1 ). This enzyme had a higher maximal rate of hydrolysis for cAMP than for cGMP, although an accurate estimate for cAMP was not obtained because of the high substrate concentrations required.
The kinetic parameters of Hcam3a(met150) differed significantly from those of Hcam1(met141) (Fig. 7B, Table II). Hcam3a(met150) had a K m of 0.6 M for cGMP and 0.3 M for cAMP, among the lowest ever reported for a calmodulin-dependent PDE (4). The specific activity or maximal rate of hydrolysis for cGMP was approximately 1.2 times that for cAMP. Eadie-Hoffstee plots (not shown) gave very similar results to those reported in Table II.
The effects of different PDE inhibitors on the truncated Hcam1(met141) and Hcam3a(met151) proteins were examined. IBMX, a relatively nonselective inhibitor, and 8-methoxymethyl IBMX, an inhibitor reported to be more selective for type I PDEs (45), were equipotent inhibitors of both Hcam1(met141) and Hcam3a(met150) (Fig. 8, Table III). Cilos-tamide, a type III PDE inhibitor, and rolipram, a type IV PDE inhibitor, were relatively poor inhibitors of these recombinant enzymes. Zaprinast (a type V inhibitor with activity against type I PDE) and vinpocetine (a type I PDE inhibitor) (46,47) inhibit both recombinant enzymes. These data are consistent with the behavior expected of a type I PDE. Although vinpocetine inhibits both recombinant enzymes, it shows greater potency for Hcam1(met141) than for Hcam3a(met150). DISCUSSION We report here the isolation and characterization of cDNAs encoding two human type I PDEs: Hcam1 and Hcam3. Hcam1 corresponds to the bovine 61-kDa CaM PDE (5,17). The 535amino acid human protein differs from the bovine protein in two regions (a 14-amino acid insertion and a divergent carboxyl terminus). RNase protection studies show that the nucleotide sequences predicted by the human cDNAs are present in RNA from human tissues, thus these differences are not likely to be cloning artifacts. The 42-nucleotide insertion is also found in a rat cDNA (data not shown). The RNase protection studies also raised the possibility of an additional splice variant near the 3Ј-end of the open reading frame. The bovine 61-kDa CaM PDE is believed to undergo alternative splicing near the 5Ј-end of the open reading frame since the bovine 59-kDa CaM PDE differs from the 61-kDa CaM PDE at the amino terminus but is otherwise identical to it (6). Two different human mRNA transcripts were observed on Northern blots. It is not known whether these correspond to different splice variants. This question can be addressed by the isolation of additional cDNAs.
The second human type I PDE reported here is Hcam3. Two splice variants, which encode proteins that diverge from each other at the carboxyl terminus, were observed. Hcam3a is predicted to be 709 amino acids in length (80,759 Da), assuming it has the same 5Ј-end as Hcam3b (see above). Hcam3b is 634 amino acids in length (72,207 Da). The Hcam3b protein extends only 4 amino acids beyond the Hcam3a/3b divergence point before terminating, whereas the Hcam3a protein extends an additional 79 amino acids beyond the divergence point.
The carboxyl-terminal region specific to Hcam3a is very basic (43). Whether this unique carboxyl terminus of Hcam3a has any consequences for the enzyme's catalytic activity, interaction with calmodulin, stability, or intracellular localization is not known. Some of these possibilities can be tested by comparing the properties of Hcam3a and Hcam3b when full-length, active proteins are expressed. Preliminary results indicated that the K m values for full-length Hcam3a and Hcam3b are similar to the data for the truncated Hcam3a reported here. 4 A 19-amino acid region that bound calmodulin was identified in the bovine 61-kDa CaM PDE (amino acids 24 -42) by peptide/calmodulin binding studies (5). This region is predicted to form an amphipathic ␣-helix, a feature often found in calmodulin binding domains (48). The human Hcam1 protein also contains this region (Fig. 1C). However the human Hcam3 protein has a different amino acid sequence (6/19 amino acids differ) and contains a nine amino acid insertion (Fig. 6). This insertion is found at the same relative position as the alternative splicing that yields the bovine 61-and 59-kDa CaM PDEs. The Hcam3 protein, both with and without the 9-amino acid insertion, is predicted to form an amphipathic ␣-helix, although the insertion alters the amino acids that would be included. Characterizing the binding of CaM to Hcam3 will help determine the effect of the amino acid insertion.
Two sites of phosphorylation by cAMP-dependent protein kinase A have been identified in the bovine 61-kDa CaM PDE When the standard error is small the symbols obscure the error bars. The data presented in Panel A are summarized in experiment 4 (Hcam1(met141)) of Table II. The data presented in Panel B are summarized in experiment 3 (Hcam3a(met150)) of Table II. protein (49). Phosphorylation of the serine at amino acid position 120 reduced the affinity of the 61-kDa CaM PDE for calmodulin. A serine is also found in this position in Hcam1 and Hcam3, although the context is slightly different in each protein. The serine at amino acid 138 is also phosphorylated in the bovine 61-kDa CaM PDE. A threonine is found in this position in Hcam1 and Hcam3. It is not known whether either the serine, threonine, or any additional sites are phosphorylated in the human proteins. Excluding gaps and divergent termini, Hcam1 and Hcam3 proteins are 77% identical, whereas Hcam1 and the bovine 61-kDa CaM PDE proteins are 94% identical. When one compares the DNA sequences over the corresponding regions of the cDNAs, they show 76 and 93% identity, respectively. Thus the analogous genes in two different mammals show higher levels of identity than do two different type I CaM PDEs in the same organism. Presumably, this reflects the conservation of both the regulatory interactions and the specific functions that these PDEs fill in mammalian cells.
Hcam1 expression was seen in brain, testes, and kidney samples. Lower levels of expression were seen in heart and uterus. The pattern of human Hcam1 expression resembled that of the bovine 59-and 61-kDa CaM PDEs (17). Hcam3 also showed a similar pattern of expression and was most readily detected in brain, heart, and testes (data not shown) but was also found in lung, uterus (data not shown), and kidney. Hcam1 and Hcam3 were coexpressed in human tissues. Whether this coexpression extends to the cellular level is unknown.
The biochemical characterization of Hcam1(met141) is consistent with its identification as the human counterpart of the bovine 61-kDa CaM PDE. The K m value for cGMP (3.5 M) that is much lower than the K m value for cAMP (51 M), and the response of Hcam1 to inhibitors, are in general agreement with the literature values for the bovine 61-kDa CaM PDE (4,50,51).
The biochemical properties of Hcam3a(met150) are distinct from those of Hcam1 and the bovine 63-kDa CaM PDE. Of these proteins, Hcam3a(met150) has the highest affinities for both cAMP and cGMP. A variety of CaM PDEs with high affinity for both cAMP and cGMP have been reported. A CaM PDE with a 1 M K m for cAMP and cGMP was detected in immature rat testis (12) and in the germ cells of male mice (15). A 68 -70-kDa CaM PDE was partially purified from mouse testis (13). This enzyme had a K m of 2 M for both cAMP and cGMP. It cross-reacted with polyclonal rabbit antisera raised against the bovine brain CaM PDEs. In view of the size of the human Hcam3a protein (72 kDa), the presence of Hcam3 mRNA in the testis and the low K m for both cAMP and cGMP of Hcam3a(met150), one of the splice variants of the Hcam3 gene may correspond to the rat testis CaM PDE activity described. CaM PDEs with low K m values for cAMP and cGMP have also been reported in heart (8,9,14), pancreas (7), and in olfactory receptor neurons (10) and may also correspond to Hcam3 gene products. The mouse and rat counterparts of Hcam3 have been recently isolated and expression of the rat gene in olfactory neurons has been detected (52).
A 74-kDa bovine brain CaM PDE has also been identified (11). Although Hcam3 and this 74-kDa protein are similar in size and are both expressed in the brain, their kinetic properties differ. Thus the 74-kDa protein is not likely to be Hcam3 and may represent a type I PDE for which cDNAs have not yet been isolated.
Although Hcam1(met141) and Hcam3a(met150) have distinctive kinetic parameters, they show similar sensitivities to inhibitors. The IC 50 values for IBMX, 8-methoxymethyl-IBMX, rolipram, and cilostamide differed less than 2-fold among these CaM PDEs. The apparent affinity of Hcam1(met141) for zaprinast was about 3-fold lower than that observed for Hcam3a(met150). The largest difference seen was with vinpocetine where the IC 50 value for Hcam1(met141) (8.1 M) was 6-fold lower than that observed for Hcam3a(met150) (50 M).  b The IC 50 and its standard error were obtained using a four-parameter model, except as noted in footnote c. The parameter that determines the slope of the fitted curve at the IC 50 ranged from 0.8 to 1.
c These IC 50 values were obtained using a two-parameter model. cetine vary from 20 to 95 M among enzymes isolated from bovine aorta, brain, coronary artery, kidney, and rabbit aorta (46,47). This range of values may reflect the particular mixture of CaM PDEs present in each tissue. The inhibition of the truncated Hcam1 and Hcam3 PDEs, which do not contain calmodulin binding regions, is consistent with the hypothesis that vinpocetine is a catalytic region inhibitor (46). In contrast to the full-length proteins, amino-truncated Hcam1 and Hcam3a were active when expressed in yeast. It has been reported that amino-truncation by proteolysis leads to activation of bovine type I PDEs (5,50,51). The kinetic properties of the proteolyzed bovine enzymes were similar to the calmodulin-activated full-length enzymes (50,51). The amino termini of the proteolyzed bovine 61-kDa CaM PDEs (5) are similar in position to the amino terminus of Hcam1(met141), suggesting that the recombinant truncated Hcam1 is activated as well. Hcam3a(met150) is truncated at the corresponding methionine and thus may also be activated.
The presence in tissues of proteins derived from more than one CaM PDE gene and the observation that different splice variants of the same gene can generate proteins that appear to be regulated differently has complicated the interpretation of studies using a mixture of CaM PDEs isolated from a tissue. An additional layer of complexity is indicated by the results of the inhibitor analysis presented here in which some, but not all, of the PDE inhibitors show slightly different IC 50 values with two human type I PDEs. Continued characterization of each gene and its products should aid in understanding the cellular roles of CaM PDEs.