Analysis of the tissue-specific distribution of mRNAs encoding the plasma membrane calcium-pumping ATPases and characterization of an alternately spliced form of PMCA4 at the cDNA and genomic levels.

The plasma membrane Ca(2+)-pumping ATPase (Ca(2+)-ATPase) mRNAs are encoded on four different genes designated PMCA1-PMCA4. The primary transcripts from some of these genes are known to be alternately spliced in the region encoding the regulatory domains of the enzymes. The known alternately spliced forms of these Ca(2+)-ATPase mRNAs and a new spliced variant of PMCA4 (PMCA4b), presented here, represent at least nine different mRNAs encoding the Ca(2+)-ATPases. In this report, the examination of the tissue-specific distribution of these alternately spliced mRNAs using polymerase chain reaction amplification of cDNA coupled with Southern blotting revealed that each spliced variant had a unique tissue distribution. PMCA1b and PMCA4a were present in all tissues examined. PMCA1a, PMCA1b, and PMCA4b were expressed in excitable tissues, whereas PMCA1d was expressed only in muscle tissues. PMCA2 was found in liver, adrenal gland, spinal cord, and brain. PMCA3a was present in spinal cord, and PMCA3b in thymus, adrenal gland, spinal cord, and brain. The mRNA for a new spliced variant of PMCA4 (PMCA4b) was detected in this study. Complementary DNAs for this isoform were isolated and characterized from human and bovine brain. This alternately spliced form of the PMCA4 mRNA contained an exon inserted at the splice junction immediately following the sequence encoding the calmodulin-binding domain. As has also been shown for PMCA1a, this insertion produced a shift in the reading frame at the 3'-end of the PMCA4 mRNA that yielded a sequence encoding a Ca(2+)-ATPase lacking a large portion of the C-terminal regulatory domain. When the human PMCA4 gene spanning this region of variable exon splicing was sequenced, it confirmed the intron-exon boundaries where alternate splicing occurs to produce PMCA4a and PMCA4b.

The plasma membrane Ca2+-pumping ATPase (Ca2+-ATPase) mRNAs are encoded on four different genes designated PMCA1-PMCA4. The primary transcripts from some of these genes are known to be alternately spliced in the region encoding the regulatory domains of the enzymes. The known alternately spliced forms of these Ca2+-ATPase mRNAs and a new spliced variant of PMCA4 (PMCA4b), presented here, represent at least nine different mRNAs encoding the Ca2+-ATPases. In this report, the examination of the tissuespecific distribution of these alternately spliced mRNAs using polymerase chain reaction amplification of cDNA coupled with Southern blotting revealed that each spliced variant had a unique tissue distribution. PMCAlb and PMCA4a were present in all tissues examined. PMCAla, PMCAlb, and PMCA4b were expressed in excitable tissues, whereas PMCAld was expressed only in muscle tissues. PMCAS was found in liver, adrenal gland, spinal cord, and brain. PMCA3a was present in spinal cord, and PMCA3b in thymus, adrenal gland, spinal cord, and brain.
The mRNA for a new spliced variant of PMCA4 (PMCA4b) was detected in this study. Complementary DNAs for this isoform were isolated and characterized from human and bovine brain. This alternately spliced form of the PMCA4 mRNA contained an exon inserted at the splice junction immediately following the sequence encoding the calmodulin-binding domain. As has also been shown for PMCAla, this insertion produced a shift in the reading frame at the 3'-end of the PMCA4 mRNA that yielded a sequence encoding a Ca2+-ATPase lacking a large portion of the C-terminal regulatory domain. When the human PMCA4 gene spanning this region of variable exon splicing was sequenced, it confirmed the intron-exon boundaries where alternate splicing occurs to produce PMCA4a and PMCA4b.
Grants NS28406 (to R. L. N.), NS21868 (to T. C. V,), and GM20818 * This work was supported in part by National Institutes of Health (to R. E. R.). 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 GenBankTM/EMBL Data Bank with accession number(s)
$ Supported by Army Predoctoral Fellowship DAAG29-83-G-006. The plasma membrane calcium-pumping ATPase (Ca2+-ATPase) regulates cytosolic calcium concentrations in all cells by extruding calcium ions in an ATP-dependent manner. The Ca2+-ATPase continues to remove calcium from the cell until the free cytosolic concentration drops below M. This final cytosolic calcium concentration is below the Kd(Caz+) for calmodulin, so the many cellular processes regulated by the Ca2+calmodulin complex are affected.
The activity of the plasma membrane Ca2+-ATPase is modified by a large number of agents that interact either with the enzyme directly or through some integrated regulatory pathway. The enzyme found in human erythrocytes is stimulated by Caz+-calmodulin (Bond and Clough, 1973), calbindin-9K (Walter, 1989), phosphorylation by CAMP-dependent kinase (Neyses et al., 1985)) protein kinase C (Smallwood et ul., 1988), thyroid hormone (Galo et al., 1981), acidic phospholipids (Carafoli and Zurini, 1982), mild proteolysis Sarkadi et al., 1986;James et al., 1989), and selfassociation (Kosk-Kosicka and Bzdega, 1988). Oxytocin depresses -43% of the Ca2+-ATPase activity in myometrium and -15% in fat pad adipocytes, but has no effect on the enzyme activity found in duodenum (Soloff and Sweet, 1982). In adipocytes, insulin causes the Ca2+-ATPase to be less active; but in liver plasma membrane preparations, it has no effect (Lotersztajn et al., 1985). In isolated primary hepatocytes and membranes isolated from whole livers perfused with vasopressin, phenylephrine, epinephrine, or angiotensin 11, the Ca2+-ATPase activity is observed to be depressed (Lin et al., 1982;Prpic et al., 1984). Glucagon appears to inhibit the Ca2+-ATPase in isolated liver plasma membranes and does not act through an intermediate such as CAMP (Lotersztajn et al., 1985).
For the cases where it is known, these agents appear to exert their regulatory effects by interaction with a region near the carboxyl terminus of the enzyme. The calmodulin-binding site is in the carboxyl-terminal 9 kDa of the Ca2+-ATPase (Zurini et aZ., 1984). Cleavage in this 9-kDa region by trypsin, chymotrypsin, and calpain activates the enzyme Sarkadi et al., 1986;James et al., 1989). Calbindin-9K protein has been shown to interact with the calmodulinbinding site in this 9-kDa region of the enzyme (James et aZ., 1991). The CAMP-dependent protein kinase and protein kinase C phosphorylation sites also have been found in this 9-kDa region (James et dl., 1989;Wang et al., 1991).
Recently, several cDNA clones representing the coding region of the Ca2+-ATPases have been isolated (Brandt et al., 1988;Shull and Greeb, 1988;Verma et al., 1988, Greeb andShull, 1989;Strehler et al., 1990). It appears that there are four families of Ca2+-ATPase mRNAs, each derived from a that occur in the region near the end of the coding sequence are shown. These exon combinations are based on isolated cDNA sequences, except for PMCASb, which was hypothesized by Greeb and Shull (1989). The expected sizes of the PCR amplification products from each of these different mRNAs are shown at the right. a Shull and Greeb (1988); Verma et al. (1988); Strehler et al. (1989); Greeb and Shull(1989); e Strehler et al. (1990); f this study. unique gene. The primary transcript from each gene can be alternately spliced in the region encoding the regulatory domain at the carboxyl terminus of the enzyme to generate these mRNA families. The mRNAs all appear to have the common exon structure A-B-C, where exons A and C are invariant and exon B varies in an isoform-and tissue-specific manner. The known possible exon combinations for mRNAs from the four gene families are summarized in Fig. 1. Since the splicing of these mRNAs results in alterations in the primary structure of the regulatory domain, this may lead to changes in the response of the enzyme to different agonists and antagonists. Thus, the variations in the response of the Ca2+-ATPase to the effectors discussed above may be a reflection of the great diversity of the enzyme at the genetic level.
Since these effectors modify the Ca2+-ATPase activity differently in different tissues and cell types, it is possible that the mRNAs for the Ca2+-ATPases may be differently expressed in response to the metabolic or regulatory use of Ca2+ by particular cells and tissues. For this reason, we elected to examine the tissue distribution of Ca2+-ATPase mRNAs.
We chose to utilize a very sensitive and selective PCR'based mRNA amplification method coupled with Southern blotting. This method detected several possible novel spliced forms of the PMCA mRNAs. We have isolated cDNA for one of these new spliced forms of the PMCA4 mRNA. This new isoform mRNA (designated PMCA4b) contained an additional segment in the region encoding the regulatory domain. We present here the isolation of cDNAs from human and bovine brain and human genomic DNA sequences for PMCA4b and show that the carboxyl terminus encoded by The abbreviations used are: PCR, polymerase chain reaction; bp, base pair(s); SDS, sodium dodecyl sulfate. the PMCA4b mRNA is truncated. PMCA4b thus may represent another example of regulation of Ca2+-ATPase activity by the use of alternate mRNA splicing to change the encoded regulatory domain.
[T-~'P]ATP (6000 Ci/mmol) and L U -~~S -~A T P (800 Ci/ mmol) were purchased from Du Pont-New England Nuclear. Sephadex G-100, Sepharose CL-4B, oligo(dT)-cellulose, Escherichia coli DNA polymerase I, RNase  Amplification of mRNAs by Polymerase Chain Reaction-RNA was isolated from frozen tissues obtained from a 20-22-week-old human male fetus by the guanidine isothiocyanate method . The integrity of these RNAs has been verified by Northern blot analysis using an amyloid precursor protein cDNA probe (Neve et al., 1988). For synthesis of DNA complementary to the mRNAs of Ca2+-ATPase isoforms PMCA1, PMCA2, and PMCA4,2 pg of total RNA from each tissue were resuspended in 12 pl of RNase-free water and added to 6 pl of 100 mM methylmercury hydroxide. This was allowed to stand at room temperature for 7 min, after which 3.1 pl of 0.7 M 2-mercaptoethanol were added, and the mixture was incubated at room temperature for 5 min. The following components were then added to give a final reaction volume of 50 pl: 50 units of RNasin, deoxynucleotide triphosphates to a 500 p~ final concentration, 50 units of avian myeloblastosis virus reverse transcriptase, and 1 pg of the composite 3'-end primer: This primer was complementary to positions 3421-3452 in hPMCAlb (human) (Verma et al., 1988), 3907-3938 in rPMCA2 (rat) (Shull and Greeb, 1988), and 3379-3410 in hPMCA4 (Strehler et al., 1990). The reaction was incubated at 42 "C for 2 h, at which point the reverse transcriptase was inactivated by incubation at 65 "C for 10 min. The same reaction was carried out for synthesis of DNA complementary to the PMCA3 isoform mRNA, except that the oligonucleotide 5'-TCTGTGTCATCGATGAGCGGAAT-3' (positions 3578-3600) (Greeb and Shull, 1989) was used as the 3'-primer. The PCR was performed by adding 1 pl of cDNA to 49 pl of a reaction mixture containing the following: 10 mM Tris-HC1, pH 8.3 (at 25 "c), 50 mM KC1, 1.5 mM M&12, 0.001% gelatin, 200' pM deoxynucleotide triphosphates, 1.25 units of native Taq DNA polymerase, 0.5 pg of the 3'-end primer (the composite primer for PMCA1, PMCA2, and PMCA$ the PMCA3-specific primer for PMCA3), and 0.5 pg of the composite 5'-end primer: The composite 5'-end primer was homologous to positions 3294-3325 in hPMCAlb (Verma et al., 1988), 3780-3810 in rPMCA2 (Shull and Greeb, 1988), 3242-3272 in rPMCA3 (Greeb and Shull, 1989), and 3252-3582 in hPMCA4 (Strehler et al., 1990). The DNA was amplified in a Perkin-Elmer Cetus thermal cycler under the following conditions: 94 "C for 1 min, 51 "C for 2 min, and 72 "C for 3 min for 40 cycles. It was empirically determined that under these conditions, 35-40 cycles were needed to detect all the  Strehler et al. (1989).
Ca2+-ATPase isoforms unambiguously because of the low abundance of their mRNAs.
Southern Blotting of PCR Products-Fifteen microliters of the PCR products were resolved on a 3% composite agarose gel (2% NuSieve low melting agarose (FMC, Rockland, ME) and 1% medium EEO agarose (Type 11, Sigma)) in the presence of 0.5 pg/ml ethidium bromide. The resolved DNA fragments were denatured by soaking the gel in 1 M NaOH for 5 min, followed by neutralization in 1 M Tris-HC1,3 M NaCl, pH 7.5, for 5 min. The DNA was then transferred to a nylon membrane (Hybond-N) in 20 X SSC (1 X SSC = 150 mM sodium chloride, 15 mM sodium citrate, pH 7.0) overnight. The DNA was fixed to the membranes by cross-linking with 300 nm light for 3 min.
The membrane was probed with the isoform-and exon-specific probes listed in Table I. The membranes were prehybridized in 6 X SSC, 5 X Denhardt's solution, 0.3% SDS, and 100 pg/ml denatured salmon sperm DNA for 6-18 h at the same temperature at which the probe would be hybridized. After prehybridization, the probe was added to the bag containing the membrane to a final activity of -10' cpm/ml, and this was hybridized for 6-18 h. The temperature of hybridization for each probe was determined using the formula: T = ((A + T) X 2 "C) + ((G + C) X 4 "C) -5 "C, where A, T, G, and C refer to the number of each of these nucleotides in the probe. After hybridization, the membrane was washed in 6 X SSC, 0.1% SDS at room temperature for 30 min, followed by another wash in 6 X SSC, 0.1% SDS at the hybridization temperature for 30 min. The membrane was then placed on x-ray film for 2-20 h at -80 "C with one intensifying screen. Before a membrane was probed with a different oligonucleotide, the previous probe was stripped from the filter by incubation in 0.2 M NaOH at 65 "C for 1-2 h. The removal of the probe from the filter was monitored by autoradiography.
Construction of cDNA Libraries-RNA was extracted from frozen bovine brain by the guanidine isothiocyanate method and centrifuged through a CsCl cushion . Poly(A)-containing RNA was purified on oligo(dT)-cellulose as described by Aviv and Leder (1972). Double-stranded cDNA was made using the protocol of Gubler and Hoffman (1983). The first library constructed had firststrand cDNA synthesis primed with the oligonucleotide 5'-TGA-CACGTCCGAGGG-3' (Oligo A) (Fig. 2), which is complementary to positions 451-465 in the cDNA sequence of CAATPl (Brandt et al., 1988).* Blunt ends were generated on the double-stranded cDNA with Since the isolation of the cDNA sequence designated CAATPl ATPases has been adopted (Greeb and Shull, 1989 FIG. 2. Summary of PMCA4b DNA inserts and their positions relative to each other. The relative positions of bovine and human brain cDNAs of the plasma membrane Ca2+-ATPase isoform PMCA4b are shown. CAATPl, BB-4, and BB-6 were isolated from bovine brain cDNA libraries. HB-4 was isolated from a human fetal brain cDNA library. Oligo A and Oligo B refer to oligonucleotides that were used to prime first-strand cDNA syntheses for libraries constructed by primer extension. Their sequences and positions are described in the text. The arrows underneath the clones represent the strategy that was employed to sequence the cDNAs. The positions of exons A-C are shown at the top. The 5'-end of exon A and the 3'end of exon C have not been determined and are so indicated. E, EcoRI; H, HindIII; Hn, HincII; N, NsiI; S, SacI. T4 DNA polymerase. The cDNA was then treated with EcoRI methylase, and EcoRI linkers were ligated on with T4 DNA ligase. Cohesive EcoRI ends were generated by digestion with EcoRI, and the digestion products were chromatographed on Sephadex G-100. The cDNA was ligated into EcoRI-digested dephosphorylated vector (XgtlO), and the ligation products were packaged with an in uitro packaging kit.
Because of difficulty in obtaining complete cDNA methylation in the protol described above, it was necessary to eliminate the need for EcoRI treatment of the cDNA. A second bovine brain library was constructed using the oligonucleotide 5"GGTGGGGGGAGATTTG-3' (Oligo B) ( Fig. 2) to prime first-strand cDNA synthesis. The sequence for this oligonucleotide is complementary to positions 1341-1356 in the insert of a recombinant bacteriophage isolated from the bovine brain cDNA library constructed as described above (BB-4) (Fig. 2). The DNA was made double-stranded and blunt-ended as described above. At this point, a blunt-to-EcoRI adaptor was ligated onto the cDNA. The adaptor consisted of the two oligonucleotides 5'-CTCGTGCCG-3' (oligonucleotide 1) and 3"GAGCACGGCT-TAA-5' (oligonucleotide 2). Oligonucleotide 1 was first phosphorylated with polynucleotide kinase and ATP (Tabor, 1987). After the kinase reaction was completed, the polynucleotide kinase was inactivated by incubation at 95 "C for 5 min. Equimolar amounts of phosphorylated oligonucleotide 1 and nonphosphorylated oligonucleotide 2 were added to the blunt-ended cDNA at a 100-fold molar excess and ligated with T4 DNA ligase. The ligation products were chromatographed on Sepharose CL-4B, precipitated with sodium acetate and ethanol, and dissolved in 20 pl of water. The 5'-end of the EcoRI site was phosphorylated with polynucleotide kinase and ATP. The cDNA was extracted with 1 volume of saturated phenol and 1 volume of chloroform and precipitated with ethanol in the presence of ammonium acetate. The cDNA was then ligated into the vector X g t l O and processed as described above.
Screening of cDNA Libraries-The first bovine brain cDNA library was screened with an EcoRI/SocI fragment from positions 1 to 597 in the cDNA sequence of CAATPl (Brandt et al., 1988). The second bovine brain cDNA library and a human fetal brain cDNA library (Neve et al., 1986) were screened with the upstream EcoRI fragment isolated from the insert of XBB-4 (Fig. 2). The screening techniques were essentially as described by Maniatis et al. (1982).
Zsolatwn of PMCAl DNA-containing Phage from Human Genomic Library-The homologous recombination system of Seed (1983) was used to isolate human genomic clones, with the exception that rAN7 was used instead of sVX, and E. coli MC1061 [p3] was used instead of E. coli W311O[p3] (Neve and Kurnit, 1983). First, an EcoRI/ HindIII fragment (positions 1-1075 in CAATP1) (Brandt et al., 1988) was ligated into the EcoRI and HindIII sites of the homologous recombination vector rAN7. The resulting ligation reaction was treated with PstI to reduce background transformants and then was used to transform E. coli MClO6l[p3]. The ampicillin-, tetracycline-, and kanamycin-resistant transformants containing the inserted sequence were selected by in situ colony hybridization using the 32P-labeled EcoRI/HindIII fragment as a probe (Maniatis et al., 1982).
One million bacteriophage from a human genomic library (Lawn et al., 1978) were used to infect E. coli MC1061[p3] cells containing the rAN7-CAATP1 recombinant plasmid. After overnight growth, bacteriophage were eluted from the plate by incubation in suspension medium (50 mM Tris, 10 mM MgS04, 0.2% gelatin, pH 7.5) at room temperature for 4 h. The eluted bacteriophage were titered on E. coli LE392, and 10' bacteriophage were used to infect E. coli MC1061 [p3]. The bacteriophage that were viable on MClO6l[p3] were tested for recombination by in situ colony hybridization with the 32P-labeled CAATPl EcoRI/HindIII fragment.
A restriction map of the insert of one of the recombinant bacteriophage isolated by homologous recombination (XgH6) was generated (data not shown). Based on this map, a 593-bp PstIISphI fragment that was in the intron between exons B and C was used to probe a different human genomic library. One million bacteriophage from this library were screened by in situ plaque hybridization, and the positive plaques were carried to clonal purity by successive rescreening as described above for the screening of bacteriophage containing cDNA inserts.
DNA from bacteriophage containing genomic inserts was examined by Southern blot analysis. The DNA was digested with EcoRI, electrophoresed on a 1% agarose gel, and transferred to a nylon membrane (Hybond-N). The blot was prehybridized in 6 X SSC, 5 X Denhardt's solution, 0.1% SDS, 100 pg/ml denatured salmon sperm DNA at 42 "C overnight. The blot was then hybridized with one of the following hPMCA4-specific probes under the same conditions as those for prehybridization: 5'-GAACCGTATCCAGACTCAG-3' (exon A-specific), 5'-CATCTCCTCCTATGGGCA-3' (exon B-specific), and 5'-ATCAAAGTGGTCAAAGCG-3' (exon C-specific). After hybridization, the filter was washed in 6 X SSC, 0.1% SDS at room temperature for 30 min, followed by a wash in fresh solution at 42 "C for 30 min. After exposure, the probes were stripped from the filter by incubation in 0.2 M NaOH at 65 "C for 1 h. The removal of probe was determined by autoradiography.
EcoRI fragments that hybridized with the exon-specific probes were subcloned in pGEM7Zf(+), and the intron-exon boundaries were sequenced using primers that were complementary to sequences in the exons and that were close to expected junctions.
Lubeling of DNA Probes-Large double-stranded DNA probes were labeled with [(u-~'P]~CTP by the method of Feinberg and Vogelstein (1983). Free nucleotides were separated from labeled DNA by chromatography over Sephadex G-100. Oligonucleotide probes were labeled with [y3'P]ATP and T4 polynucleotide kinase (Greene and Struhl, 1987), and free label was removed by two precipitations with ammonium acetate and ethanol in the presence of 2 pg of carrier tRNA.

RESULTS
PMCAl mRNA Distribution-The mRNAs derived from the PMCAl gene have four known spliced forms designated PMCAla-PMCAld (Shull and Greeb, 1988;Strehler et al., 1989). The four mRNAs arise by the insertion of various exons at a common point between exons A and C. The inserted exons are designated B, B', and B"; and their combinations that contribute to the different PMCAl mRNAs are shown in Fig. 1. A Southern transfer of the PCR products was probed with oligonucleotides specific to the known exons of PMCAl that should occur within the amplified region. Fig. 3A shows the autoradiograph of a transfer that was probed with a PMCAl exon C-specific probe. This probe hybridizes to all four of the P M C A l spliced forms. PMCAla was found in spinal cord, brain, skeletal muscle, and heart. PMCAlb was in all tissues examined. PMCAlc was found in skeletal muscle, heart, spinal cord, and brain. PMCAld was found only in heart and skeletal muscle. The results presented in To validate the data obtained with the PMCAl exon Cspecific probe, the transfer was hybridized with probes that were specific for each of the PMCAl variable exons. In Fig.   3B, the transfer was reprobed with an oligonucleotide specific to PMCAl exon B. This probe should have detected PMCA la-, PMCAlc-, and PMCAld-derived products, and it is seen that bands corresponding to the expected sizes of these products hybridized with the probe. This is also seen in Fig. 3C, where an exon B'-specific probe hybridized with PMCAlaand PMCAld-derived products, and i n Fig. 30, where an exon B"-specific probe detected PMCAla-derived products. To further verify the product thought to be derived from PMCAla, we reprobed the membrane with an oligonucleotide that would only hybridize with the exon B"-exon C junction. The results of this hybridization are shown in Fig. 3E. The PMCAladerived material was found in brain and spinal cord as expected. Products corresponding to PMCAla could also be seen in skeletal muscle and heart in a longer exposure of this autoradiograph (data not shown). Several faint bands around 510 bp were detected with the exon B-, B'-, B", and C-specific probes, but not with the exon B"-exon C junction probe (Fig.   3). These may be due to as yet uncharacterized PMCAl isoform mRNAs that have exons inserted between exons B" and C. PMCA2 mRNA Distribution- Fig.   4A shows the same transfer used in Fig. 3 reprobed with a PMCAP exon Cspecific probe. The expected amplification product derived from the known PMCAP mRNA was found in brain, liver, spinal cord, and adrenal gland. In spinal cord, there were bands at -344 and 390 bp that also hybridized with the PMCA2 exon C3-specific probe. These bands may correspond to other spliced forms of PMCAP. PMCAB mRNA Distribution-The amplification products derived from the PMCA3 mRNA were resolved on a separate gel and probed with a PMCA3 exon C-specific probe. Fig. 4B shows the result of this hybridization. Based on the cDNA sequence, the PMCA3 mRNA would have a 154-bp exon between exons A and C, resulting in a 358-bp amplification product. However, by homology to other Ca2+-ATPase cDNAs, Greeb and Shull (1989) predicted another mRNA that would lack this exon and designated it PMCASb, whereas the original isolate was designated PMCA3a. The amplification product generated from the PMCA3b mRNA would be 204 bp in length. PMCA3a-derived material was seen in spinal cord and brain. Products derived from the putative PMCA3b mRNA were present in adrenal gland, spinal cord, and brain. In addition to amplification products derived from PMCA3a and PMCA3b, three other bands were observed primarily in spinal cord, which may be additional alternately spliced versions of PMCA3.
PMCA4 mRNA Distribution-The Southern blot used to generate the data in Figs. 3 (A-E) and 4A was reprobed with an oligonucleotide specific for PMCA4 exon C. The autoradiograph resulting from this hybridization is shown in Fig.  4C. Amplification products derived from the PMCA4a mRNA (Strehler et al., 1990) were present in all tissues. Also, an additional band at 342 bp was detected primarily in skeletal muscle, small intestine, heart, spinal cord, and brain. This band was found to be derived from a new spliced form of PMCA4, (PMCA4b), reported here. After the sequence of hPMCA4b was determined (see below), the same transfer used in Fig. 4C was hybridized with a probe specific for PMCA4 exon B (PMCA4b-specific). The results, shown in Fig. 40, verify that the band seen at 342 bp with the exon Cspecific probe was derived from the new isoform (PMCA4b). Characterization of Bovine and Human PMCA4 cDNA Clones-In our initial studies (Brandt et aL, 1988), screening of a Xgtll bovine brain cDNA expression library with rabbit anti-human erythrocyte Ca2+-ATPase antibodies resulted in the isolation of a 1.5-kilobase pair cDNA (designated CAATP1). This cDNA represented the 3"terminus of what is now known as the PMCA4a mRNA.2 Our subsequent cloning strategy was to use this 3'-end cDNA sequence to obtain adjacent sequences by primer extension of the mRNA. This strategy was adopted to avoid the potential problem of generating a chimeric structure based on the isolation of cDNA fragments derived from other Ca2+-ATPase isoform mRNAs with similar sequences. These experiments resulted in the isolation of a partial cDNA for an alternately spliced form of PMCA4 (here designated PMCA4b).
The screening of 200,000 bacteriophage from a bovine brain cDNA library constructed using Oligo A (Fig. 2) as a primer yielded six positive plaques. The recombinant phage containing the largest insert was designated XBB-4. This insert terminated at a naturally occurring EcoRI site at position 942 ( Fig. 2 and contained in the sequence shown in Fig. 5). A second bovine brain cDNA library was constructed using an oligonucleotide complementary to sequences in the insert of XBB-4 as a primer for cDNA synthesis. The screening of 300,000 bacteriophage from this library resulted in the isolation of 28 positive clones, one of which (XBB-6) contained an insert that yielded the rest of the sequence shown in Fig. 5.
The strategy employed in sequencing the inserts of XBB-4 and XBB-6 is presented schematically in Fig. 2 Fig. 3 was probed with a probe specific for exon C of rPMCA2. B, a Southern blot prepared as described for Fig. 3, but with hPMCA3specific PCR products made using the PMCA3-specific primers described under "Experimental Procedures." The blot was probed with a probe specific to exon C of rPMCA3. C, the blot used in

G T S T A R A I A T K C G I V T P G O O F L C L E G K E F N 3 0 R L I R N E K G E V E Q E K L O K l U P -K L R V L A R S S~-6 0 CGC CTC ATC CGA AAC GAG AAG GGC GAG GTA GAG CAG GAA AAG CTG GAC AAG ATC TGG CCT AAG CTC CGA GTG CTG GCG CGG TCT TCC CCC
. .  FIG. 5. Partial cDNA sequence of CaZ+-ATPase isoform (bPMCA4b) from bovine brain and its deduced protein sequence. This is the combined DNA sequence of the cDNA inserts from XBB-6 (positions 1-1356) and XBB-4 (positions 943-1497) isolated from bovine brain cDNA libraries. In keeping with the nomenclature proposed by Greeb and Shull(1989), the combined sequences of these bacteriophage X cDNA inserts are designated bPMCA4b (see Footnote 2). Only the sequence of BB-4 up to the second EcoRI site is presented because the sequence downstream of that point has previously been reported as CAATPl (Brandt et aL,  1988). The nucleotides are numbered below the sequence, and the amino acids are numbered to the right of the sequence.

P T T S V A A A V S S P T L G N Q S G q S V P ' CCC ACC ACT TCT GTT GCT GCT GCT
the insert from XBB-6 that was sequenced only in one direc-the bovine sequence unambiguously. The combined sequences tion (nucleotides 1-942 in Fig. 5) was confirmed by compari-of the inserts from XBB-6 and XBB-4 up to the EcoRI site son with the human homolog of this sequence described below. that marks the beginning of CAATPl are shown in Fig. 5.

When differences between bovine and human sequences oc-
Because of the availability of many more high quality curred, the autoradiograph from the bovine cDNA sequencing human brain cDNA libraries than bovine brain libraries, we reactions was examined for ambiguous nucleotide assign-elected to screen one of these libraries to isolate full-length ments. In all cases, it was possible to assign the nucleotide in cDNA clones, rather than to continue in the bovine system. The screening of 300,000 recombinant bacteriophage from a human fetal brain cDNA library with the upstream EcoRI fragment from the insert of XBB-4 resulted in the isolation of 18 positive plaques. Examination of the sequence of the insert from one of these isolates (XHB-4) showed that the sequence was identical to hPMCA4 (Strehler et al., 1990) from residues 2176 to 3403, except that a 178-bp stretch was inserted immediately after the sequences encoding the calmodulinbinding domain at position 3309. The sequence of what is presumably a new exon is shown in Fig. 6 with a portion of flanking sequence from the PMCA4 sequence presented by Strehler et al. (1990). To be consistent with the nomenclature used by Strehler et al. (1989) and Greeb and Shull (1989), we designated the new exon, B, and this new isoform, hPMCA4b (human) and bPMCA4b (bovine). The isoform presented by Strehler et al. (1990) is here designated hPMCA4a. The numbering of the sequence in Fig. 6 is the same as that used by Strehler et al. (1990) up to the point transcribed from the new exon. The deduced amino acid sequences of the carboxylterminal region of the enzymes from just upstream of the insertion site to the stop codons are also shown. The inserted sequence from exon B of PMCA4b led to a shift of the reading frame that generated an in-frame stop codon deleting a large portion of the C terminus of the PMCA4a-encoded protein.

I K V V K A F H S S L H E S I Q K P l N h P K U
Comparison of bovine PMCA4b with the homologous region of human PMCA4b showed that the deduced protein sequences were nearly identical (Fig. 7). In general, only conservative substitutions were observed where there were differences between the sequences. The only other difference found was the insertion of a single alanine residue in the bovine sequence at position 458. This high degree of conservation across species is also seen between rat and human PMCAl and rat and human PMCAZ3 Characterization of Exon ABC-containing Region of PMCA4 Gene-To verify the existence of the new exon B suggested by the sequence of the PMCA4b cDNA, we probed a human genomic DNA library with an EcoRIIHindIII fragment from CAATPl by homologous recombination techniques and chose six isolates for examination. Southern blot analysis with probes from two different regions of PMCA4 showed that a single 4.3-kilobase pair PstI-generated fragment in an isolate P. Brandt and R. L. Neve, submitted for publication.

b P K A l b GlSlARB~TKCGIV7PGODFLCLEblEFNRLIRNEKGEVEPLKLDKIWPKLRVLARSSPlDKHlLVKGIIDSTVUIQRQVVAVlGOGlNDGPlV 95
hPWCA4b KKMYffWGIAGlOVUEASDIILlDONFlSlVKIVmGRNVlOSISKFLQ~lVNV~AVIVClG*CIlPOSPLKAV~YVNLLlDlFML  designated XgH6 contained the probe sequence as well as upstream sequences (data not shown). A smaller PstIISphI fragment of the 4.3-kilobase pair PstI-generated fragment that contained only DNA from the intron between exons B and C was used to probe a human genomic DNA library by filter hybridization. Eight positive plaques were selected and taken to clonal purity, and their DNA was subjected to Southern blot analysis. Probing of the Southern blot with oligonucleotide probes specific to each of the exons in the variable region of PMCA4b (exons A-C) showed that one isolate (XgH5A) contained all three exons and that each exon was on a separate EcoRI fragment (data not shown). The EcoRI fragments that hybridized with the exon-specific probes were subcloned into the plasmid vector pGEM7Zf(+) for sequence analysis.

BBB bPMCAlb K K M V f f I V I G I A G T O V A I E A S D I I L l O O N F l S l V K A V m G R N V Y O S I S K F L Q~l V N V V A V I V I F T G~l l~S P L K A V a C Y V N L l l D l F~L 190 h P K A 4 b RAlEPPlESLLKRRPIGRNKPLlSRl~NILW(IFI~IVIFILVlI\GEKFFDlDSGRKAPLHSPPSMlYllVFNlFVLlKXFNElNSRKlHG
The sequences determined for the intron-exon junctions of the exons that constitute the variably spliced region of PMCA4 are shown in Fig. 8. All of the region that constitutes exon B was sequenced in the genomic DNA and was found to contain no additional introns. The sequences upstream of exons B and C have the requisite structures that distinguish intron sequences at the acceptor site of a splice junction. The 2 intronic residues immediately adjacent to the exons are AG, and sequences upstream from that point are very pyrimidinerich (Breathnach and Chambon, 1981). Also, several sequences that could serve as internal splice recognition sequences (Keller and Noon, 1984), where "lariat" intermediates are formed, were found upstream of each exon (underlined sequences).
The sequences downstream of exons A and B contained sequences that would distinguish them as intron donor sites. The consensus sequence for intron donor sites is GT(A/ G) AGT (Breathnach and Chambon, 1981). The residues immediately downstream of exon A were GTACCT. The obligate GT residues at the 5'-end of the intron were present, but two of the next four nucleotides deviated from the consensus sequence. The sequence of the intron immediately adjacent to the 3'-end of exon B was GTGAGT. This sequence is an exact match for the intron donor site consensus sequence.

DISCUSSION
The recent cloning of several cDNAs corresponding to the Ca2'-ATPases has indicated that there is great diversity in the mRNAs that encode different forms of the enzyme. It should have been possible to use exon-specific oligonucleotide probes in standard Northern blot analysis to determine the

Exon C
tissue-specific distribution of the various Ca2'-ATPase mRNAs. However, we have found by several different methods that these mRNAs appear to be very low abundance messages; and their detection, even with high specific activity cRNA or DNA probes, has been extremely difficult (data not shown). Even though the methodology we employed for Northern blotting was nearly identical to that used by others to examine the PMCA mRNAs, our results have been more comparable to those reported by Strehler et al. (1990) than to those of Greeb and Shull (1989) or, more recently, Kuo et al. (1991). We have determined that the integrity of the RNA on our Northern blots is intact by hybridization with other probes (e.g. amyloid precursor protein cDNA). Therefore, it is difficult to assess the reason for the discrepancies between our observations and other reports. Based on these results, we judged that hybridization of Northern blots with endlabeled oligonucleotide probes would not be feasible for determining the tissue-specific distribution of the various Ca2+-ATPase isoforms. Therefore, we have used the combination of PCR and Southern blotting to examine the distribution of various isoform mRNAs in tissues. The results presented in this report further extend the range of this mRNA diversity by demonstrating that the tissue distribution of each of the alternately spliced Ca2+-ATPase mRNAs is unique and that other spliced forms may exist. The tissue distribution of PMCA4 mRNAs has not been previously reported. The results in Fig. 4 ( C and D) show that there are at least two forms of the PMCA4 mRNA that are alternatively spliced in the region encoding the regulatory domain of the Ca2+-ATPase. One of these PMCA4 variants has been previously published (Strehler et al., 1990) and is designated PMCA4a. The characterization of another PMCA4 isoform mRNA (PMCA4b) and its gene is reported herein. PMCA4a is found in all tissues examined, and its tissue distribution is identical to that of PMCAlb. The PMCA4b mRNA was found primarily in the excitable tissues skeletal muscle, small intestine, heart, spinal cord, and brain. The ability of this technique to detect the previously unknown PMCA4b mRNA demonstrates its utility in examining mRNA expression in complex systems such as the Ca2+-ATPase mRNA families. Comparison of the deduced protein sequences of bovine and human PMCA4b shows a high degree of similarity. The fact that there is more homology between the same isoform in different species than among the different isoforms in the same species suggests that the Ca2+-ATPase genes diverged and their protein products developed specialized functions early in mammalian evolution. This high degree of conservation across species also suggests that there is strong selective pressure to maintain the primary structure of the various Ca2+-ATPase isoforms and that, by extension, the tissue-specific splicing of the mRNAs encoding these sequences would also be maintained.
The tissue-specific distribution of the known Ca2+-ATPase mRNAs is summarized in Table 11. These results show that the expression of the Ca2+-ATPase mRNAs exhibits a great deal of diversity. Comparison of the data presented here to those of Greeb and Shull(1988) from adult rat tissues suggests that there are also developmentor species-specific differences in mRNA expression. In the case of PMCA1, the isoform seen by Greeb and Shull (1988) in all tissues was PMCAlb. The methodology they used could not distinguish among the alternately spliced forms of the PMCAl mRNAs, so PMCAlb masked any differences that might have been seen in the other PMCAl mRNAs. These authors found the PMCA2 mRNA in adult rat brain and liver, which agrees with the data presented here for human fetal tissues. However, we detected no PMCA2 mRNA in human fetal kidney, whereas Greeb and Shull found it in adult rat kidney. PMCA3 mRNAs had previously been found primarily in brain and skeletal muscle, with minor amounts in tissues of the digestive tract of adult rat. PMCA3a and PMCA3b mRNAs also were expressed in human fetal brain, but no PMCA3 mRNAs were detected in human fetal skeletal muscle or small intestine.
With regard to developmental regulation of mRNA expression, it is clear that the expression of different PMCA1-PMCA3 mRNAs changes from embryonic to adult brain.3 This may also occur in other tissues and could explain our inability to detect the PMCA2 mRNA in fetal human kidney or the PMCA3 mRNA in human fetal skeletal muscle since these genes may not be expressed until adulthood in these tissues. The use of different Ca2+-ATPases to perform the same function in human and rat cannot be precluded. For example, human kidney might have evolved using the PMCAlb or PMCA4a Ca2+-ATPase, whereas rat kidney may have evolved the use of the PMCAQ Ca2+-ATPase for the same function.
During the course of the PMCA mRNA distribution analysis, we also isolated a cDNA encoding an alternately spliced form of PMCA4 (PMCA4b). The existence of the PMCA4b mRNA was predicted by the PCR-based detection method discussed above. PMCA4b differs from PMCA4a by the addition of a single exon inserted between the exons homologous to exons A and C of PMCA1. The inserted exon from human PMCA4 is 178 bp in size, which is -30 bp longer than the combined exons B, B', and B" of PMCA1, to which it has some homology. Despite this homology, the data presented here show no evidence for mRNA processing that would lead to PMCA4 mRNAs equivalent to PMCAlc and PMCAld We have previously proposed that the carboxyl terminus of PMCA4a is a calmodulin homolog that binds to the calmodulin-binding site and inhibits enzyme activity until it is displaced by interaction with Caz+-calmodulin (Brandt et al., 1988). This proposal is in agreement with studies of the human erythrocyte CaZ+-ATPase that showed that enzymatic removal of the carboxyl-terminal 9-kDa region results in activation and loss of calmodulin sensitivity (Benaim et al., 1984Sarkadi et al., 1986). The addition of PMCA4b exon B causes a frameshift in the PMCA4 mRNA, resulting in a protein that does not contain the putative inhibitory calmodulin homolog sequence. The resultant PMCA4b enzyme may be insensitive to calmodulin and have activity equal to the fully calmodulin-stimulated erythrocyte enzyme. Recent studies have shown that a partially purified liver Ca2+-ATPase is insensitive to calmodulin, but can bind it after denaturation of the enzyme (Kessler et al., 1990).
We have presented evidence here demonstrating a new form of the PMCA4 class of Ca2+-ATPases that may have lost portions of the sequences important for regulation of activity. PMCAl also exhibits a similar phenomenon wherein isoforms PMCAlb-PMCAld contain the phosphorylation site for CAMP-dependent protein kinase, but PMCAla does not because an exon inserted in the mRNA at a position homologous to that of PMCA4b changes the reading frame. The data presented here suggest that two other CaZ+-ATPase families, (PMCAZ and PMCAB) may also exhibit similar alternate splicing in this region of the coding sequence. This further supports the proposal that the use of alternate splicing to change the mRNA sequences encoding the regulatory domains is common in the differential expression of Ca2+-ATPase isoforms. There are mRNAs for at least nine different isoforms of the CaZ+-ATPases. The extent of this diversity and the observation presented here that these isoforms are expressed in a highly specific manner suggest that the very precise control of cytosolic Ca2+ at the plasma membrane may play an even greater role in cellular regulation than previously thought.