Expression, purification, and properties of the plasma membrane Ca2+ pump and of its N-terminally truncated 105-kDa fragment.

Isoform 4b of the human plasma membrane Ca2+ pump was expressed in COS cells and in the baculovirus system (Sf9 cells). A 105-kDa pump fragment lacking the first two transmembrane domains and the so-called transduction domain was also expressed. The expression level was 2-4 times the background in COS cells and at least 7 times in the baculovirus system. Tests on membranes from both systems showed that the expressed pump was active. The expressed pump and the 105-kDa fragment were isolated from Sf9 cell membranes by calmodulin affinity chromatography. The pump had Ca(2+)-dependent ATPase activity with a calmodulin stimulation factor of 3, formed a La(3+)-stabilized phosphoenzyme, and had a KM (Ca2+) in the presence of calmodulin of about 1 microM. The 105-kDa fragment, assayed by the phosphoenzyme test on COS or Sf9 cell membranes or by ATPase measurements after isolation from Sf9 cells, proved inactive. Laser confocal microscopy on Sf9 cells showed that both the pump and the 105-kDa fragment were apparently associated with the plasma membrane. The expressed pump in COS and Sf9 cells and the endogenous pump in a number of other cell lines had a slower gel mobility (i.e. a higher apparent molecular mass) than the erythrocyte pump.


Isoform 4b of the human plasma membrane Ca2+ pump was expressed in COS cells and in the baculovirus system (Sf9 cells). A 105-kDapump fragment lacking the first two transmembrane domains and the so-
The plasma membrane Ca2+ ATPase is the largest of all Ptype (Pedersen and Carafoli, 1987a;Pedersen and Carafoli, 1987b) ion motive pumps and has a molecular mass averaging 134 kDa. It is apparently present in all eucaryotic cells, and its properties have been summarized in a number of recent reviews (Carafoli, 1991;Carafoli, 1992;Strehler, 1991). The enzyme has now been cloned from several cell types, including brain (Shull and Greeb, 1988), porcine smooth muscle (De Jaegere e t al., 19901, and human teratoma cells (Verma et al., 1988). The pump is the product of a multigene family: the four genes so far identified give rise to additional isoforms by alternative splicing of primary transcripts (Greeb and Shull, 1989; Heim et al., 1992;Shull and Greeb, 1988;Strehler et al., *The work was made possible in part by the support of Grants 31.28772.90 and 31.30858.91 from the Swiss National Science Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Recipient of START Research Career Development Award 31.27103.89 from the Swiss National Science Foundation.
1) To whom correspondence should be addressed Biochemie 111, ETH-Zentrum, Universitatsstrasse 16, CH-8092 Zurich, Switzerland. Tel: 41-1-256-30-11;Fax: 41-1-252-63-23. 1989). Although some of the important functional domains/ sites of the pump have now been located, other important sites, e.g. the catalytic and the regulatory Ca2+ binding sites have not yet been assigned site-directed mutagenesis would be a powerful tool in identifying them. No successful expression of any of the isoforms of the pump has been reported so far, and this has prompted the work described in this contribution: the expression of the pump in the COS cell system and in the baculovirus-infected Spodoptera frugiperda ceiI line will be described. Both the COS cell system and the baculovirus system produced an active pump. Confocal microscopy experiments on the latter system have shown that the expressed pump was correctly targeted to the plasma membrane. The baculovirus system produced amounts of the pump sufficient to isolate it on calmodulin (CaM)' columns: the present contribution thus describes the first successful purification of an expressed P-type pump.
Early proteolysis work (Zurini et al., 1984) had suggested that a 40-50-kDa portion of the molecule could be removed leaving behind a 90-kDa fragment which was still active. However, the level of activity of the f r a~e n t isolated on a CaM column was very low, leaving open the possibility that a trace contamination of the preparation by the intact enzyme was responsible for the ATPase activity observed. Further work (Zvaritch et aL, 1990) has shown that the N-terminal trypsin cut leading to the formation of the 90-kDa fragment occurred between the main body of the pump and a 30-40-kDa N-terminal portion. The problem of whether the Nterminally truncated pump is still active is important, since the portion missing from the N terminus contains a region that has been defined as the transducing domain (Green and MacLennan, 1989), i.e. a domain that should couple ATP hydrolysis to Ca2+ translocation. The 30-40-kDa missing portion of the molecule contains other domains that are highly conserved among P-type pumps. Even if no known role has so far been conclusively assigned to them, their conservation suggests their importance to the function of the enzyme. The missing N -~r m i n a l fragment, which contains the two first transmembrane domains of the pump, is relatively hydrophobic (Zurini et al., 1984) and tends to associate strongly after proteolysis with the remainder of the pump molecule, i.e. with the 90-kDa portion. For this reason, the purification of the 1att.er has proven extremely difficult: expression in amounts sufficient for purification purposes has thus been successfully attempted in the present study. Functional tests have shown the N-terminally truncated pump to be inactive.

EXPERIMENTAL PROCEDURES
Materials-The baculovirus AcNPV and the transfer vector pVL 1393, a derivative of pVL941 (Luckow and Summers, 19891, were obtained from Dr. Max Summers (Texas A & M University). The S. frugiperda cell line (Sfs) as well as purified AcNPV DNA were kindly provided by Dr. Salima Mathews (Hoffmann-La Roche, Basel, Switzerland). Other cell lines were obtained from American Type Culture Collection. The TNM-FH medium was from Sigma; Grace's insect medium, DMEM (Dulbecco's modified minimal Eagle's medium), nutrient mixture Ham's F-12, antibiotics, and FCS (fetal calf serum) were purchased from GIBCO-BRL (Life Technology AG, Basel, Switzerland). The monoclonal antibody 5F10 was described previously (Borke et al., 1987(Borke et al., , 1989. the polyclonal antibody directed against the pig endoplasmic reticulum Ca2+-ATPase isoform 2b was a kind gift of Dr. F. Wuytack and was described by Eggermont et al. (1989).
[y3'P]ATP was obtained from Amersham International, England. All other reagents were of the highest purity grade commercially available. Densitometric measurements of autoradiograms were performed on a densitometer (CD 50; Desaga, Heidelberg, Germany).
Construction of Recombinant Vectors-A full-length cDNA coding for the human plasma membrane Ca2+ pump isoform 4b (hPMCA4b) has been assembled from overlapping partial cDNA clones (Strehler et al., 1990) introducing unique artificial SalI and KpnI restriction sites adjacent to the 5'-and 3'-ends of the coding region, respectively, using standard methods of recombinant DNA technology including the polymerase chain reaction. The cDNA was cut with SalI and KpnI and with a blunt-ended SalI site cloned into a pVL 1393 transfer vector precut with S m I and KpnI. A deletion mutant of the hPMCA4b cDNA coding for a 105-kDa fragment of the pump (replacement of the first 942 base pairs by an artificial start codon and a preceding BamHI site), but extending to the regular C terminus of the pump, was introduced in a BarnHI/KpnI-treatedpVL 1393 vector. The 105-kDa fragment has the N terminus of the previously described 90-kDa fragment (Zurini et al., 1984;Zvaritch et al., 1990), but extends to the C terminus of the intact pump. The exact boundaries of the resulting plasmids pVL-PMCA and pVL-PMCAlO5 are displayed in Fig. 1. The inserts of the final constructs were double-strand sequenced using gene-specific primers. The endoplasmic reticulum Ca2+-ATPase (ERCA) transfer vector construct was obtained by cutting out the cDNA coding for the pig ERCA-isoform 2b ) from a pGEM7-Zf(+) vector with EcoRI and cloning it into the EcoRI site of a pVL 1393 vector. The two possible orientations yielded both the correct construct pVL-ERCA and the nonsense vector pVL-Inf, the source of the control virus used for some negative control infections.
The plasmid construct used for the transfection of COS cells was composed of the full-length cDNA of the plasma membrane Ca2+-ATPase 4b isoform (PMCA4b) excised from pVL-PMCA with BamHI and KpnI (Fig. l), ligated into the vector pSG5 at BamHI and KpnI restriction sites. The KpnI restriction site was introduced into the vector by blunt ligation of the KpnI-linker to the blunt-ended BglII restriction site of the vector. DMEM with 10% FCS, and 100 pg of gentamicin per ml (DMEM/ Cell Culturing and Transfection-COS-7 cells were grown in FCS) in a 6% C02, 37 "C incubator. 60-70% confluent cells in 10-cm Petri dishes were transfected for 2 h at 37 "C with 10 fig of CsCl BamH I gradient-purified DNA and 1 mg of DEAE-dextran in 1 ml of 20 mM Tris-HC1, pH 7.4, 140 mM NaC1, 0.5 mM MgCI2, 0.5 mM CaC12, and 0.5% D-glucose. The cells were then incubated for 2.5 h at 37 "C in 5 ml of DMEM/FCS, containing 100 p~ chloroquine, followed by treatment with 10% dimethyl sulfoxide in PBS for 2 min at room temperature. 48-60 h later the cells were harvested by scraping.
cultures in a humidified incubator at 29 f 1 "C using TNM-FH S. frugiperda (SKI) cells were grown in monolayer or suspension medium supplemented with 10% FCS and 100 pg/ml gentamicin. The culturing of Sf9 cells and all procedures involving them including routine subculturing, transfections, production of high titer viral described in the "Manual of Methods for baculovirus Vectors and stocks, and production of recombinant proteins were performed as Insect Cell Culture Procedures" (Summers and Smith, 1987). Cells other than the Sf9 or COS-7 were grown to 90% confluence in an humidified incubator at 37 "C, 6% COP for 3 days before collection and membrane preparation. All cell lines minus the Ptkl were maintained in DMEM, 10% FCS, and 100 pg/ml gentamicin. For Ptkl cells Ham's F-12 was used in the place of DMEM.
Production and Isolation of the Recombinant AcPMCA Virus-2.5 pg of CsC1-purified recombinant transfer vector DNA and 1 pg of wild-type AcNPV viral DNA were used for cotransfection of Sf9 cells by the calcium phosphate-precipitation technique (Summers and Smith, 1987). Recombinant viruses, the result of in vivo homologous recombination between the polyhedron sequences of the wild type viral DNA and the recombinant transfer vectors, were amplified by serial dilutions of transfection mixtures and by infecting Sf9 cells in 96-well microtiter plates. After 5 days of incubation, the supernatants were transferred to new plates, and the cells were lysed and transferred to Zeta-probe blotting membranes (Bio-Rad) using a slot blot apparatus. The membranes were hybridized with gene-specific DNA probes according to published procedures (Sambrook et al., 1989) to identify wells containing recombinant viruses. Supernatants from wells infected with the maximally diluted transfection mixture which were still positive, were assumed to be enriched in recombinant viruses and were used for subsequent plaque purification by screening for occlusion-negative phenotypes or, if recombinant viruses were unsufficiently abundant, by plaque hybridization (Summers and Smith, 1987). One or two rounds of amplification and two to three viruses AcPMCA and AcPMCA105. The control viruses AcERCA rounds of plaque assays were necessary to obtain pure recombinant and AcInf were obtained in a similar way.
SDS-Polyacrylamide Gel Electrophoresis and Immunoblotting-Proteins were separated by SDS-polyacrylamide gels essentially according to Laemmli (1970). The sample buffer ( 2~) was composed of 60 mg of DTT, 50 mg of SDS, 5 mM EDTA, and 0.005% bromphenol blue per ml of stacking buffer. In addition, it contained 400 mg of urea per ml. For immunoblotting analysis the proteins were transferred to nitrocellulose (Towbin et al., 1979). Nonspecific binding was blocked by 5% bovine serum albumin in TBST (10 mM Tris-HC1, pH 8.0,150 mM NaC1,0.05% Tween 20). Primary antibodies were applied in TBST for 1 h at room temperature. After washing the nitrocellulose sheets three times in TBST, the blots were incubated with alkaline phosphatase-coupled secondary antibody (1:7500; Promega) for 40 min, rinsed three times in TBST and developed with the ProtoBlot system (Promega).
Immunofluorescence Experiments on Sf9 Cells-Sf9 monolayer cells were collected 48 h after the infection, washed once in PBS, and placed on Alcian Blue-coated slides. The cells were then fixed in methanol at -20 "C for 15 min and air-dried. The glass slides were rehydrated in PBS and blocked for unspecific binding by incubating them in blocking buffer (5% FCS, 0.1% bovine serum albumin, 5% glycerol, 0.04% azide in PBS) for 45 min at 37 "C. The cells were then incubated with primary antibody diluted in blocking buffer (1:lOOO for monoclonal antibody 5F10, 1:200 for the polyclonal antibody against the endoplasmic reticulum ATPase) for 2 h at 37 "C, and washed three times in PBS and once in blocking buffer. Secondary fluorescein-conjugated goat anti-mouse or goat anti-rabbit antibodies (TAGO, Inc., Burlingame, CA) were applied at 1:lOO dilutions in blocking buffer for 1.5 h at 37 "C. The preparations were exhaustively washed in PBS prior to mounting in a medium containing 70% glycerol, 0.33 M Tris-HC1, pH 9.5, and 5% n-propyl gallate as antibleaching agent. Laser scanning confocal microscopy was performed on a Zeiss Axioplan fluorescence microscope equipped with a Bio-Rad MRC 600 confocal laser scanning unit and a Silicon Graphics workstation computer. Pictures were taken with Ilford FP4 black and white film.
Preparation of Crude Membranes from COS and Sf9 Ceh-crude membranes deficient in CaM were prepared according to the protocol described by Scully et al. (1982) with some modifications. Infected Sf9 cells were harvested at about 61 (AcPMCA) or 46 h (AcPMCA105) after infection by centrifugation a t 1600 rpm for 10 min (4 "C) and were washed once with washing buffer (20 mM Tris-HC1, 130 mM NaCl, pH 7.4, at room temperature), once with washing buffer containing 2 mM EDTA, and once again with washing buffer. The cells were swollen in 5 mM Tris-HC1, pH 7.4, 0.2 mM PMSF at a density of 4.2 X lo' cells/ml for 15 min on ice, then supplemented with 3 mM DTT and homogenized with 25 strokes in a tightly fitting Dounce homogenizer. The homogenate was diluted five times with 5 mM Trisand centrifuged for 20 min at 500 X g to sediment nuclei and unbroken cells. The pellet was resuspended once again in half the initial homogenization volume by applying the same procedure. The two supernatants were combined and centrifuged at 20,000 X g for 20 min. This pellet was defined as the crude membrane fraction and was washed once with membrane buffer (10 mM Hepes, pH 7.2, 130 mM KCl, 0.5 mM M&12, 0.05 mM CaC12, 3 mM DTT, 15% glycerol), recovered at 30,000 X g, and finally resuspended in membrane buffer at a protein concentration of 6-8 mg/ml. The membranes were frozen in liquid nitrogen and stored at -70 "C. The procedure for COS cells was essentially the same as described above except that the storage buffer for the final membrane fraction did not contain Ca2+ and Mg2'.
Preparation of Total Membranes from Different Cell Lines-The cells were collected by scraping, rinsed once in PBS and once in TBS (IO mM Tris-HC1, pH 8.0, 150 mM NaCl), and resuspended in 50 mM Tris-HC1, pH 7.0,500 mM NaCl, 75 pg/ml PMSF, 1 mM DTT at 1-2 X 10' cells/ml. The cells were disrupted by four cycles of freezing and thawing at -70 "C. The pellet was spun down and resuspended in 50 mM Tris-HC1, pH 7.0,l mM DTT, 75 pg PMSF at 2-4 mg/ml protein.
Calmodulin Affinity Chromatography-Solubilization of crude Sf9 membranes and CaM affinity chromatography was performed according to Niggli et al. (1987) except that the solubilization was performed for 20 min at a protein concentration of 4 mg/ml and that all column buffers contained 3 mM DTT and 15% glycerol. The column was washed with 9 bed volumes prior to the EDTA elution step. In the case of the purified 105-kDa product, a second purification step was applied by washing the CaM column with 100 bed volumes of the Ca2+ buffer prior to the elution step with EDTA. Phosphatidylcholine was used as the stabilizing lipid. The control erythrocyte ATPase was isolated as described (Niggli et al., 1987).
ATPase Activity-The Ca2+-ATPase activity was followed by the coupled enzyme assay as described by Vorherr et al. (1991). In the measurements on the membranes, 5 pg/ml of oligomycin was included to eliminate the interference by ATPases in the mitochondrial membranes, and the Ca2+-ATPase was assumed to be the fraction of the activity which was sensitive to 3 mM EGTA. In the Ca2+ affinity experiments, the release of inorganic phosphate from ATP was measured by the colorimetric method described by Lanzetta et al. (1979). The reaction was performed at 37 "C for 30 min with 250 ng of purified hPMCA in 30 mM Hepes, pH 7.2,120 mM KCl, 1 mM ATP, 2.42 mM MgC12, 1 mM EDTA, 1 mM EGTA and various amounts of Ca2+ to produce the desired free Ca2+ concentration as calculated by the computer program described by Fabiato and Fabiato (1979).
Experiments on the Phosphorylated Intermediate of the Reaction Cycle-The phosphorylation of membrane samples was carried out at 4 "C in 100 pl of reaction buffer containing 20 mM MOPS, pH 6.8, and 100 mM KCl, in the presence of either 50 p M CaC12, 50 p M CaC12 plus 50 p M Lac&, or 4 mM EGTA. The samples purified by CaM HCl, pH 7.4,0.5 mM MgC12,5 mM EGTA, 3 mM DTT, 0.2 mM PMSF, affinity chromatography contained EDTA, and therefore the Ca2+ and La3+ concentrations were increased to 400 p~. The reaction was started by the addition of [y3'P]ATP to a final concentration of 0.1 p~ under vigorous stirring and was stopped after 10 s with 1.2 ml of ice-cold 8% trichloroacetic acid. The precipitates were spun down in a Microfuge and washed once with 800 pl of distilled water. For the hydroxylamine treatment, the washed pellets were resuspended in 0.5 ml of 0.6 M hydroxylamine, pH 5.2, and kept for 10 min at room temperature prior to precipitation with 300 pl of cold 24% trichloroacetic acid. Control samples were treated in the same way, but 0.6 M sodium acetate, pH 5.2, replaced hydroxylamine. The washed precipitates were suspended in the acidic sample buffer containing urea and separated by acidic SDS-polyacrylamide gel electrophoresis (Sarkadi et al., 1986).

Expression of the PMCA4b Isoform cDNA in the COS Cell
System-The activity levels of the expressed PMCA4b cDNA in COS cells were judged from experiments on the catalytic phosphorylation of the ATPase by ATP. In a series of experiments of the type shown in Fig. 2 the highest levels of expressed PMCA, as estimated from densitometric tracings, were 2-4 times higher than those of the arbitrary endogenous pump (densitometric values for the experiment shown in Fig.  2: control, 12.7, expressed, 33.9). Two other phosphorylated protein bands can be seen in Fig. 2 at about 200 and 90 kDa: they most probably represent aggregated (dimerized) forms of the plasma membrane Ca2+-ATPase (Niggli et al., 1979) and the catalytic subunit of the Na/K-ATPase, respectively. In these experiments Na+ ions in the phosphorylating medium were substituted for K+, because separate experiments had indicated that the former reduced the extent of phosphorylation of the ER Ca2+-ATPase in the COS cell membranes and improved the overall resolution of the gels.
The catalytic phosphorylation experiments on the expressed pump revealed that its mobility in SDS-gels coincided with that of the endogenous Ca2+ pump of COS cells, but was significantly lower than that of the erythrocyte Ca2+-ATPase (Fig. 2). This is of interest, since partial protein sequencing of the erythrocyte pump has unambigously shown it to consist mainly of the PMCA4b isoform originally cloned from a human teratoma cDNA library (Strehler et al., 1990).
Expression of the 105-kDa Fragment in the COS Cell System and Study of Its Activity-The previous trypsin proteolysis work mentioned in the Introduction section on the purified erythrocyte Ca2+ pump had suggested that a tryptic fragment of 90 kDa identified in SDS gels retained the basic properties of the pump, i.e. ATP hydrolysis coupled to Ca2+ transport. However, all attempts to isolate the fragment in a functionally active state from the complex mixture of other tryptic fragments have so far failed. Expression of the fragment in COS cells was thus attempted to establish whether the fragment acts as an active Ca'+-transporting ATPase.
A fragment of the PMCA4b cDNA coding for the sequence starting at the N terminus of the 90-kDa polypeptide (Zvaritch et aL, 1990) and extending to the C terminus of the pump was cloned into the pSG5 vector at BamHI and KpnI restriction sites. The molecular mass of the designed polypeptide was expected to be 105 kDa. Since attempts were planned to isolate the expressed, truncated version of the pump by CaM affinity chromatography (see below), the presence of the Cterminal portion containing the CaM binding domain was essential. The expression of the cDNA coding for the 105-kDa fragment was monitored by immunoblotting experiments: total lysates of cells expressing cDNAs of either the PMCA4b or the 105-kDa fragment were subjected to gel electrophoresis, and the blotted proteins were stained with affinity-purified polyclonal antibodies against the human erythrocyte membrane Ca'+ pump (Fig. 3a). As judged from these experiments, the levels of the expression of the 105-kDa fragment were comparable to those of the full-length pump.
In disrupted cells the expressed polypeptides were found associated with the membranous fractions. However, phosphorylation experiments with [y3'P]ATP in the presence of Ca2+, and of Ca2+ plus La3+, failed to reveal any activity in the 105-kDa fragment (Fig. 3b). A phosphorylated band in the molecular mass region of the 105-kDa fragment was observed when Ca'+ ions were present in the phosphorylating medium. However, the intensity of phosphorylation was the same in the controls and in the samples from the cells which had expressed the 105-kDa fragment, suggesting that the phosphorylation was due to the activity of the endogenous ER Ca'+-ATPase. Indeed, in the presence of Ca'+ and La3+, which specifically favor the accumulation of the phosphorylated intermediate of the plasma membrane Ca'+-ATPase, while degrading that of the ER pump, no phosphorylation was observed in the molecular mass region around 105 kDa.
Expression of PMCA4b in the Baculovirus System-To achieve the higher expression levels of the PMCA4 pump necessary for studies on the isolated expressed enzyme,  lanes 1 and 3) and of cells transfected with a control plasmid (lanes 2 and 4 ) . Membrane suspensions (20 pg of total protein) were incubated with [y3'P]ATP in the presence of 50 pM cas+ (lanes 1 and 2 ) or of 50 pM cas+ and 50 pM La3+ (lanes  3 and 4 ) . The proteins were separated on SDS gels under acidic conditions and subjected to autoradiography. Other experimental details are found under "Experimental Procedures.'' AcPMCA and AcPMCAlO5, containing the cDNA coding for the full-length or the truncated version of the human ATPase isoform 4b, respectively, were isolated as described under "Experimental Procedures.'' Infection of Sf9 monolayer cells with AcPMCA (multiplicity of infection of 10) resulted in the time-dependent expression of a protein of about 142 kDa that was well visible on a Coomassie Blue-stained gel and that was found to bind the erythrocyte Ca'+-ATPase-specific monoclonal antibody 5F10 in Western blots. As is typical for the baculovirus system, the expression product appeared after 22 h, reached a maximum level at 70 h, and decreased thereafter, coinciding with the appearance of degradation products due to increased cell lysis. Interestingly, as the expression times became longer, an increasing fraction of the expression product only entered the gel when a sample buffer containing urea (see "Experimental Procedures") was used. The time course for the expression of the mutant virus AcPMCA105 was found to be somewhat faster. Sf9 suspension cultures generally yielded smaller expression levels, but were nevertheless useful to produce large amounts of recombinant protein.
Targeting of the hPMCA Protein in Sf9-To establish whether the expressed hPMCA protein was targeted to the plasma membrane, laser scanning confocal microscopy was carried out in combination with immunofluorescence staining of AcPMCA-and AcPMCAlO5-infected Sf9 cells with monoclonal antibody 5F10, directed against the erythrocyte Ca'+ pump (Fig. 4, a and b ) . Control infected cells (AcInf) yielded only very faint signals with the antibody (Fig. 4c) showing that the endogenous Ca'+ pump is apparently very scarce in Sf9 cells. Alternatively, the antibody did not cross-react with the endogenous insect cell pump: this seems a likely possibility, since the antibody failed to detect any protein in Western blots with 40 pg of uninfected Sf9 cell membranes. Because Sf9 cells contain very large nuclei, which became even enlarged after infection (Currie et al., 1991), it was difficult to assign signals to specific subcompartments in the very small region of the cell left free by the nucleus under the cell surface. To validate the interpretation of the hPMCA expression signals, Sf9 cells were thus infected with a virus containing the cDNA coding for the pig endoplasmic reticulum Ca'+-ATPase isoform 2b (AcERCA) , and a polyclonal antibody directed against the SERCA2b pump protein (Wuytack et a l , 1989) was used to detect and locate the expressed protein (Fig. 4d). Control infected cells (AcInf) yielded no signal (Fig. 4e), confirming the specificity of the antibody. On comparing the three expressed pumps, it is clear that the full-length and the truncated version of the hPMCA protein accumulated in the cell periphery. By contrast, in the case of the ERCA2b pump, strong reaction was observed around the nucleus of Sf9 cells and only a slight reaction in the cell periphery. Membrane preparations of the cells infected with the AcERCA virus yielded high amounts of expressed active pump.' Isolation of the Expressed hPMCA Proteins-The relatively high expression level of the recombinant hPMCA pump in insect cells permitted experiments aimed at isolating the expressed enzyme. The CaM affinity chromatography procedure initially developed for the isolation of the erythrocyte ATPase was used. The first step was the preparation of crude membranes from infected Sf9 cells as described under "Experimental Procedures.'' "Crude" membranes were separated from the nuclei and the cytosol, but no other subfractionation was applied. In  AcPMCA105 ( b ) , AcERCA ( d ) , or with the nonsense virus AcInf (c, e), fixed in methanol and incubated with the following antibodies: a monoclonal antibody against the erythrocyte ATPase (5F10) was used for the cells displayed in a, b, and c, and a polyclonal antibody against the endoplasmic reticulum Ca2+-ATPase was used for the cells displayed in d and e. Fluorescein isothiocyanate-conjugated secondary antibodies were used to produce the fluorescence visualized in the laser scanning confocal microscope. proteins. a, 20 pg of total membrane proteins of noninfected ( l a n e I), AcPMCA-infected ( l a n e 2 ) , and AcPMCA1054nfected ( l a n e 3 ) Sf9 cells were separated on 4-12% SDS gel and stained with Coomassie Blue. The full-length (gel mass, -142 kDa) and the 105-kDa expression products are not seen under these conditions in noninfected cell membranes. b, 200 ng of purified, expressed full-length PMCA ( l a n e 2), of the 105-kDa product ( l a n e 3 ) and of the purified erythrocyte ATPase ( l a n e 1 ) were electrophoresed and visualized by silver staining. The double band detected at about 50-kDa was a buffer artefact, since it was also present in lanes containing no protein. c, a gel run as in b was used for electroblotting. The proteins were detected by a polyclonal antibody directed against the erythrocyte ATPase as described under "Experimental Procedures." expression products of about 142 and 105 kDa (arrows) were absent in the control cell lane. The membranes of infected cells were then solubilized and used for the CaM column chromatography step. EDTA eluates of the Ca2+-washed column were subjected to gel electrophoresis and silver-stained as shown in Fig. 5b: the expression products from AcPMCA ( l a n e 2)-and AcPMCA105 ( l a n e 3)-infected cells were clearly evident as single bands. Interestingly, under the same conditions, no intact pump could be isolated from AcPMCA105infected Sf9 cells, indicating that the plasma membrane Ca2+ pump, if present in Sf9 cells, is extremely scarce (not shown).
The yield of purified PMCA in the purification experiments on the infected Sf9 cells was estimated to be about 1.1% of the total membrane protein of the cells (93 pg of purified pump protein from 8.3 mg of membrane protein in the experiment shown in Fig. 5). Lane 1 shows the control purified erythrocyte Ca2+-ATPase, which often shows, in SDS gels, degradation products of 125 and 90 kDa. As already observed in the COS cell system (see above), also in the baculovirus system the expressed PMCA apparently run with lower mobility than the erythrocyte enzyme. The immunoreactivity of the isolated expressed proteins with a polyclonal antibody against the erythrocyte pump is shown in Fig. 5c. It is important to mention at this point that the expression time was critical for the successful isolation of the recombinant pump protein. Long expression times (see "Experimental Procedures") resulted in higher losses of the hPMCA protein into the nuclear pellet and in the inefficient solubilization of the protein from the crude membrane fraction, even if larger amounts of Coomassie Blue-stainable, immunoreactive protein were clearly present in the membranes. This is consistent with the previously mentioned observation that an urea containing sample buffer was required at late stages of infection to force the protein to enter the SDS gel. This indicates that the hPMCA pump formed stable aggregates as its concentration in the cell increased as a result of the prolonged expression times. ATPase Activity of the Expressed Proteins-To establish whether the pump expressed by Sf9 cells was active, Ca2+-ATPase activity measurements and the formation of the phosphorylated intermediate from ATP were studied. Table I summarizes the ATPase results (see "Experimental Procedures"). Infection of Sf9 cells with AcPMCA resulted in the increase of the membrane-associated ATPase activity in the presence of Ca2+, but in the absence of CaM, from 27 to 71 nmol of ATP hydrolyzed per min and mg protein: this activity level included contributions from other ATPases different from the Ca2+-ATPase of the plasma membrane, chief among them the Ca2+-ATPase of endoplasmic reticulum. The activity increased from 34 nmol in control membranes to 216 nmol in those from infected cells if 5 pg/ml CaM were included in the medium: the stimulation factor by CaM was thus 3.06 (as compared to 1.25 in the membranes from noninfected cells). Evidently, the specific expression of large amounts of the PMCA pump decreased the relative contribution of the non-CaM-stimulated ATPase(s) to the overall measured activity level. By contrast, expression of the truncated AcPMCA105 pump caused no significant changes in the basal and in the CaM-stimulated ATPase activity. The CaM-stimulated activity of the purified expressed hPMCA was 1,810 nmol x min"

TABLE I ATPase activiw of membranes from PMCA-infected Sf9 cells and of the isolated PMCA protein
The ATPase activity was measured as described under "Experimental Procedures." The values are expressed as nmoles ATP hydrolyzed X mg-l X min". The samples were either freshly prepared total cell membranes or CaM affinity-purified PMCA and PMCA105 proteins expressed by the infected cells. The activity of the membrane fractions declined by 20-27% upon freezing. The experiment was carried out on two different DreDarations. whose results were verv similar. The table shows one of them. x mg" as compared to 525 in the absence of CaM, i.e. the stimulation factor was 3.45. The expressed 105-kDa product was completely inactive. The Ca2+ affinity of the isolated expressed hPMCA, estimated by following the release of inorganic phosphate from ATP (Fig. 6), yielded a K M value of about 1 and 5 PM in the presence and absence of CaM, respectively. The values were thus similar to those of the erythrocyte pump. The findings of Table I and Fig. 6 were further supported by the experiments of Fig. 7. The expressed PMCA protein was able to form the hydroxylamine-sensitive phosphorylated intermediate (Fig. 7a). However, the extremely small amounts of protein available made the visualization of the pump band in the absence of the stabilizing agent La3+ very difficult: as expected of the plasma membrane Ca2+ pump (Schatzmann and Burgin, 1978;Szasz et al., 1978), the steady state concentration of the phosphorylated intermediate was enhanced by La3+. At about 110 kDa, a second hydroxylamine-sensitive, but Ca2+-independent phosphorylation band was visible. It probably derived from small amounts of a contaminant (P-type pump) protein which apparently became strongly phosphorylated since no corresponding band was visible in silver-stained gels (see Fig. 5b). To avoid interference by this contaminant protein with the expressed 105-kDa truncated hPMCA, the latter was subjected to a second round of CaM affinity purification with more extensive Ca2+ washing prior to the EDTA elution. Fig.  7b shows that the repurified expressed PMCA 105-kDa product, now free of contamination, was inactive, i.e. it failed to form the phosphorylated intermediate. The phosphorylation of the erythrocyte ATPase is also shown in Fig. 7 as a control. Electrophoretic Mobility of PMCA-As mentioned above for 400 p~ Ca2+ plus 400 p~ La3+ (lune 2), 4 mM EGTA ( l a n e 3), or 400 pM Ca2+ plus 400 p~ La3+ followed by hydroxylamine treatment ( l a n e 4). Lune 5 represents the control phosphorylation of the purified erythrocyte Ca*+-ATPase in presence of Ca2+ and La3+. The phosphorylated samples were separated on 6% acidic polyacrylamide gels. The dried gels were exposed for 2 days at -70 "C. Additional details are found under "Experimental Procedures." b, 200 ng of the isolated 105-kDa protein (lunes 1-4) and of the erythrocyte ATPase (lune 5) were phosphorylated and electrophoresed as in a. The 105-kDa protein was repurified on a micro-CaM column as discussed in the text using a more exhaustive washing with Ca2+ to eliminate the 110-kDa contaminating protein.
the PMCA4 isoform expressed in COS cells (Fig. 2) also the pump expressed in the Sf9 insect cells (Fig. 5) showed a lower gel mobility than the pump of erythrocytes. It comigrated instead with the COS cells endogenous pump (see Fig. 2). To test whether this was a specific property of the PMCA4 isoform or an artifact of the expression system employed, other cell types were analyzed. The ATP-dependent, Ca2+and La3+-stimulated phosphorylated intermediates of three different cell lines are shown in Fig. 8 and 50 p~ Ca2+ as described under "Experimental Procedures" and separated on acidic 5-12.5% gradient SDS-polyacrylamide gels. The dried gels were exposed for 2 days at -70 "C. Other details are found under "Experimental Procedures." which was specifically detected in the presence of Ca2+ and La3+ (Fig. 8, lanes 1-3). The band corresponding to this phosphoenzyme migrated with slower velocity than that of the erythrocyte Ca2+ pump (Fig. 8, lane 4 )

DISCUSSION
The importance of expressing P-type ion pumps in uitro is obvious. Problems like the identification of functionally important sites, the targeting of pumps to different membrane systems, the structural reasons for the specificity of the transported ions all largely depend on the successful in uitro expression of the enzymes. The family of P-type ion-motive ATPases (Pedersen and Carafoli, 1987a;Pedersen and Carafoli, 1987b) has now grown to include enzymes that transport Na+, K+, Ca2+, Mg2+, H+, Cd", either isolated or in an obligatorily concerted reaction, in eucaryotic or procaryotic cells. Three of these pumps have now been expressed in in vitro systems: the Ca2+ pump of sarcoplasmic/endoplasmic reticulum (Campbell et al., 1991;Maruyama and MacLennan, 19881, the H+ pump of the fungal plasma membrane (Holzer and Hammes, 1989;Portillo and Serrano, 1988), and the a and / 3 subunits of the Na+/K+ pump (Price and Lingrel, 1988;Takeyasu et al., 1987). The expression systems have been successfully used for site-directed mutagenesis work that has permitted others to assign with a high degree of probability important sites, e.g. that are responsible for the inhibition by vanadate (Ghislain et al., 1987) and, more generally, a number of sites/residues that are apparently essential to the function of all pumps of this class, even if the reasons for their critical importance were not immediately obvious at the outset (MacLennan, 1990). Very important sites on which mutagenesis work is now beginning to shed light are those linked to the binding and transport of the ions during the catalytic cycle: contrary to earlier expectations (Brand1 et al., 1986) recent work on the expressed sarcoplasmic reticulum Ca2+ pump (Clarke et al., 1989) has provided strong indications that these sites may be located within four of the ten predicted transmembrane domains of the pump.
All of the expression systems so far used have resulted in the enrichment of the P-type pumps in the membranes of the expressing cells. The whole cells, or membrane preparations derived from them, have then been the objects of study. In no case, however, have expression experiments been published in which the system used has permitted the production of amounts of pumps sufficiently large for purification attempts. As a consequence, no protein chemistry work has so far been possible on any of the expressed P-type pumps.
The work described in this contribution has shown that the human PMCA4 isoform could be successfully expressed in the two systems studied. This isoform is the only human one for which full-length DNA clones have been constructed, and was thus practically an obligate choice among the dozen or so isoforms presently known. Full-length clones of the other isoforms will presumably soon become available, and reasonable hopes can thus be entertained that their expression will rapidly follow. It must be emphasized, however, that the expression of the plasma membrane Ca2+ pump has proven to be unexpectedly difficult: the amounts of active protein expressed, as compared for example to those of the Ca2+ pump of sarcoplasmic reticulum (Maruyama and Machnnan, 19881, have been substantially smaller under all experimental conditions tested in the COS cell system. Densitometric tracing evaluations from a number of experiments of the type shown in Fig. 2 have yielded levels of expression of the active ERCA pump in the COS cell system averaging 12-fold the controls. Expression of the PMCA4b isoform in the COS cell system to levels approximately similar to those reported here have been recently detected by Penniston and co-workers3 in experiments in which the successful expression was attributed to the modification of the consensus sequence of the PMCA4b cDNA around the starting codon. Since in the present experiments the original PMCA4b cDNA sequence was used, the reason for the modest levels of active Ca2+ pump expression in the COS cell system is thus apparently not a "weak" consensus sequence around the starting codon. Still unknown intracellular regulatory mechanisms may prevent the overproduction of the pump protein in COS cells. The baculovirus system compares favorably to the COS cell system, but even in its case the activity tests (i.e. the CaM-stimulated Ca2+dependent ATP hydrolysis) have indicated apparent expression levels less than 10-fold higher than those of noninfected Sf9 cells. However, for the reasons mentioned under "Results," these enrichment factors are most likely underestimated. This is also indicated by the finding mentioned above that no endogenous pump could be isolated from Ac-PMCAlO5-infected Sf9 cells. The reasons for the lower expression levels of the plasma membrane pump, as compared, for example, to those of the Ca2+ pump of sarcoplasmic reticulum, are presently being investigated one possibility for which preliminary supporting evidence has been obtained, a t least for the COS cell system, is the particular susceptibility of the expressed pump to intracellular proteases. It has been shown that the plasma membrane Ca2+ pump is peculiarly sensitive to the action of the intracellular Ca2+-dependent protease calpain (James et al., 1989), as compared for example to the Ca2+ pump of sarcoplasmic reticulum, which is completely resistant to it (Wang et al., 1989). Work on alternative expression systems is currently underway. Despite the difficulties experienced, it has been possible to isolate the pump expressed in Sf9 cells using a CaM affinity chromatography column. As mentioned in the Introduction section, this is thus the first report of a successful isolation of an expressed Ptype pump: suitable scaling-up procedures can now be expected to yield amounts of the expressed protein adequate for protein chemistry work. The expressed protein, embedded in the native membrane system in both the COS and the Sf9 cells, or purified from J. T. Penniston, personal communication.
the latter cells, has proven to be active. Actual tests of ATPdependent Ca2+ transport have been carried out only occasionally, since it is generally assumed that the plasma membrane Ca2+ pump never becomes uncoupled (see Schatzmann (1982) for a comprehensive discussion of the matter). Ca2+dependent ATP hydrolysis (and formation of the phosphorylated intermediate) are thus adequate tests for the functional integrity of the pump. The CaM stimulation factors observed on the purified pump were in the range of those routinely measured in normal pump preparations, e.g. from erythrocytes. Although not shown under "Results" for reasons of space, gel overlay experiments with '251-CaM have yielded the expected high affinity for the modulator (i.e. Kd values in the nanomolar range). In addition to being active, most of the expressed pump was apparently correctly targeted to the appropriate membrane system, i.e. the plasma membrane. The confocal microscopy experiments on the baculovirus system have shown that the expressed plasma membrane pump was concentrated in the extreme periphery of Sf9 cells compared to the expressed endoplasmic reticulum Ca2+-ATPase which occupied the entire space between the nucleus and the plasma membrane. Although the spatial resolution of the technique did not permit the exclusive assignment of the expressed plasma membrane pump and its truncated version to the plasma membrane, it appears very probable that this was indeed the case. Thus, at least a significant fraction of the pump was apparently targeted correctly to the plasma membrane, while the expressed SERCA pump was clearly distributed throughout the perinuclear space, i.e. in the cell region where the endoplasmic reticulum is normally mostly concentrated. One unexpected result with both expression systems has been the slightly higher apparent molecular mass of the expressed protein with respect to that of the erythrocyte pump, so far considered as the reference enzyme for this multigene family of proteins. Clearly, the mass of the erythrocyte protein now appears to be the exception rather than the rule, as also indicated from the results on the apparent mass of the endogenous pump in other cell lines shown in Fig. 8. The reasons for the (slightly) lower apparent molecular mass of the erythrocyte pump are presently being investigated, the working hypothesis being some form of post-translational processing.
An important portion of the work has dealt with the expression of the N-terminally truncated 105-kDa fragment of the pump. The reasons for the importance of expressing this fragment have been discussed in the Introduction section: although the N-terminally truncated fragment had been isolated in the original trypsin proteolysis work by Zurini et al. (1984) and claimed to be active, its very low activity levels could have been due to the contamination of the fragment preparation with traces of the intact ATPase. The decision to use the entire sequence from the N terminus of the 90-kDa fragment to the C terminus of the pump, rather than to the C terminus of the 90-kDa fragment (Zvaritch et al., 1990) in the expression experiments was based on the observation that the portion of the pump downstream of the C terminus of the 90-kDa fragment was certainly removed by trypsin. Since this portion of the pump is not involved in the catalytic cycle, and is entirely extramembranous, its absence or presence most likely does not influence the basic function and membrane architecture of the pump. The results have shown that the Nterminally truncated fragment could be expressed by both systems used. They have also shown that the fragment was inactive, i.e. the activity of the purified fragment preparations observed in the early work by Zurini et al. (1984) was evidently due to traces of contaminant intact pump. This was shown for the fragment purified from Sf9 cells by CaM affinity chromatography, and also for the fragment embedded in the membrane environment for both expression systems. This last point is worth a comment: the bulk of the expressed truncated fragment was apparently still membrane-associated, as also shown by the confocal microscopy experiments on Sf9 cells. Thus, whatever the reasons for the inactivity of the fragment, which are obviously related to the absence of the portion of the pump containing the first two transmembrane domains and the so-called transducing domain, they apparently do not prevent the targeting of the truncated pump to the membrane, at least in Sf9 cells. Whether the insertion process leads to the correct folding of the truncated product in the membrane is, however, an open question.