Identification, expression, function, and localization of a novel (sixth) isoform of the human sarco/endoplasmic reticulum Ca2+ATPase 3 gene.

Understanding of Ca(2+) signaling requires the knowledge of proteins involved in this process. Among these proteins are sarco/endoplasmic reticulum Ca(2+)-ATPases (SERCAs) that pump Ca(2+) into the endoplasmic reticulum (ER). Recently, the human SERCA3 gene was shown to give rise to five isoforms (SERCA3a-e (h3a-h3e)). Here we demonstrate the existence of an additional new member, termed SERCA3f (h3f). By reverse transcriptase-PCR using monocytic U937 cell RNA, h3f mRNA was found to exclude the antepenultimate exon 21. h3f mRNA expression appeared as a human-specific splice variant. It was not found in rats or mice. h3f mRNA gave rise to an h3f protein differing in its C terminus from h3a-h3e. Of particular interest, h3f diverged in the first amino acids after the first splice site but presented the same last 21 amino acids as h3b. Consequently, we further investigated the structure-function-location relationships of the h3b and h3f isoforms. Comparative functional study of h3b and h3f recombinant proteins in intact HEK-293 cells and in fractionated membranes showed the following distinct characteristics: (i) resting cytosolic Ca(2+) concentration ([Ca(2+)](c)) and (ii) ER Ca(2+) content ([Ca(2+)](er)); similar characteristics were shown for the following: (i) the effects of the SERCA inhibitor, thapsigargin, on Ca(2+) release ([Ca(2+)](Tg)) and subsequent Ca(2+) entry ([Ca(2+)](e)) and (ii) the low apparent Ca(2+) affinity and the enhanced rate of dephosphorylation of the E(2)P phosphoenzyme intermediate. Subcellular location of h3b and h3f by immunofluorescence and/or confocal microscopy using the h3b- and h3f-specific polyclonal and the pan-h3 monoclonal (PL/IM430) antibodies suggested overlapping but distinct ER location. The endogenous expression of h3f protein was also proved in U937 cells. Altogether these data suggest that the SERCA3 isoforms have a more widespread role in cellular Ca(2+) signaling than previously appreciated.

Cell Ca 2ϩ signaling is a dynamic oscillatory process regulating a variety of important cellular functions such as secretion, contraction, metabolism, neuronal plasticity, and gene transcription (1). Accordingly, Ca 2ϩ signaling is strictly controlled in space, time, and amplitude. This very tight control is necessary to enable cells to extract relevant information from the Ca 2ϩ signal but also implies that disturbances in the intracellular Ca 2ϩ signaling can lead to a plethora of consequences.
Understanding Ca 2ϩ signaling requires the basic knowledge of structures involved in this process. Among these structures are the Ca 2ϩ transport proteins inserted in the membranes of the endoplasmic reticulum (ER), 1 the intracellular reservoir of Ca 2ϩ ions. These include Ca 2ϩ channels (ryanodine and inositol 1,4,5-trisphosphate receptor channels) and Ca 2ϩ pumps (sarco/endoplasmic reticulum Ca 2ϩ -ATPases (SERCAs)). Although much data are available on the role of the channels in the elaboration of the Ca 2ϩ signal, little is known concerning how the signal may be modified by expression of SERCAs with different functional properties.
In recent years, enormous advances in our understanding of SERCA structure-function relationships were made, including new insights regarding their three-dimensional structure (2,3). Furthermore, relevant physiological functions of SERCAs have been defined through their potential role in various human diseases (4)(5)(6). However, some of the results contrast with those obtained in mice using SERCA1-3 knock out models (7)(8)(9)(10)(11)(12). Such distinct phenotypes point to differences between humans and mice and/or to various compensatory phenomena (13).
In this context the existence of a SERCA isoform diversity is important. The SERCA family includes the products of three genes, named SERCA1 (ATP2A1), -2 (ATP2A2), and -3 (ATP2A3), each giving rise to alternatively spliced mRNA and protein isoforms. For a while, the first two genes, SERCA1 and SERCA2, have been known to have two 3Ј-end splice variants encoding isoforms differing in their C termini, mainly expressed in adult (SERCA1a) and neonatal (SERCA1b) skeletal muscles, in cardiac muscle (SERCA2a), and in all cell types (SERCA2b). The third gene, SERCA3, was found to express the unique so-called non-muscle type SERCA3 isoform (SERCA3a according to the new nomenclature). These isoforms were essentially found by molecular cloning. To date, a third SERCA2c mRNA has been described (14), and the SERCA3 genes were shown to possess a higher degree of complexity thanks to molecular cloning added to RT-PCR techniques (15)(16)(17). We and others have found that mice, rats, and humans express a variety of species-specific SERCA3 isoforms, in addition to the species-unspecific SERCA3a mRNA and protein (15)(16)(17). SERCA3b and -3c mRNAs and proteins were first described in mice and humans (18 -21). These mRNAs were issued from the partial or complete insertion of a new exon 21, respectively; this exon is 15 nucleotides shorter in mice than in humans in its 5Ј-end (18,22). Next, rats were found to be devoid of SERCA3b and SERCA3c mRNAs, which are replaced by a so-called SERCA3b/c isoform, which is due to an extension of exon 21 (23). Finally, we found that the insertion of an additional exon 22 in human SERCA3b and SERCA3c mRNAs led to SERCA3d and SERCA3e isoforms (24). Although the rationale for such a diversity of novel SERCA3 isoforms is still not understood, functional differences in the human SERCA3a, -3b, and -3c isoforms were recently pointed out, particularly one step of their catalytic cycle, the dephosphorylation of the E 2 P intermediate, varying with the length of the alternatively spliced C terminus (25).
In the present study we searched for a SERCA3 mRNA excluding exon 21, but including exon 22, in order to see whether the human SERCA3 gene gives rise to all the theoretical combinations of alternative splicings of exons 21 and 22. We report an additional SERCA3f (h3f) mRNA issued from this new alternative splicing of the human SERCA3 gene and its distribution pattern in human cell lines of different origins and normal human tissues. By using a newly developed h3fspecific polyclonal antibody, we demonstrated the existence of the endogenous h3f protein, which exhibits the same last 21 amino acids as h3b protein. Next, we stably expressed the h3f protein and compared its functional property with h3b by studying the resulting cytosolic and endoplasmic reticulum Ca 2ϩ contents, the sensitivity toward the SERCA inhibitor, thapsigargin, three steps of the catalytic cycle, as well as the location and conformational state of the new Ca 2ϩ pump.
Preparation of U937 and Recombinant HEK-293 Cell Membrane Proteins-Isolation of membrane fractions enriched in intracellular membranes from U937 and HEK-293 cells was as described in Refs. 23 and 24. Protein concentration of membrane fractions was determined by using BSA as a standard.
Generation and Characterization of a Novel Human SERCA3f (h3f)-specific Polyclonal Antibody-The h3f-specific antibody was generated by immunizing SPF rabbits with the mixture of the peptides (P1 and P2) indicated in Fig. 3B (double XP peptide antibody production, Eurogentec, Herstal, Belgium). The sensitivity and the cross-reactivity of the antibodies were analyzed by Eurogentec. Affinity purification of antiserum against each of the peptides was performed by Eurogentec. Purified anti-h3f (P1) and anti-h3f (P2) antibodies were tested by immunoblotting using recombinant h3f protein (data not shown). The anti-h3f (P2) presenting the highest immunoreactivity was used in immunoblotting, immunofluorescence, and immunoprecipitation experiments.
Electrophoresis of Proteins and Western Blottings-Electrophoresis was performed on 8% SDS-PAGE, and Western blots were done as in Ref. 24 for the IID8, PL/IM430, N89, and anti-hSERCA3b antibodies. For the h3f protein, nitrocellulose membranes were incubated with a 1:2000 dilution of the anti-h3f antibody, in Tris-buffered saline (pH 7.4), 5% non-fat milk, and 0.1% Tween 20 for 2 h at room temperature. After washing, blots were treated with a 1:10000 dilution of horseradish peroxidase-conjugated anti-rabbit IgG for 1 h. Antibody binding was revealed using enhanced chemiluminescence Western blotting reagents according to the manufacturer's instructions (Amersham Biosciences).
Immunoprecipitation-Human U937 cell and SERCA3f-transfected HEK-293 cell membranes were solubilized for 2 h at 4°C by stirring in a modified RIPA's buffer: 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1% Nonidet P-40, 1% sodium deoxycholate, 0.1% SDS, 0.1 mM dithiothreitol, and the following inhibitors: 0.002 mg/ml aprotinin, 0.002 mg/ml leupeptin, 0.05 mM phenylmethylsulfonyl fluoride, 0.1 mg/ml soybean trypsin inhibitor, 0.05 mg/ml Bowman-Birk trypsin-chymotrypsin inhibitor. Binding of the anti-h3f antibody to protein A-Sepharose (50 mg/ml) was achieved after 1 h of incubation with shaking at 4°C. The solubilized membranes were then added to this mixture, and incubation was followed for 1 h 30 min at 4°C. Immunoprecipitates were sedimented by centrifugation at 10,000 ϫ g for 10 min, and a second round of immunoprecipitation was performed by adding fresh anti-h3f antibody-coupled protein A-Sepharose to the supernatant. The pooled immunoprecipitates were then washed once with 10 mM Tris-HCl (pH 8), 5 mM EDTA, 1% Triton X-100, 1% sodium deoxycholate, 0.5 M NaCl, and four times with the same buffer but without NaCl. The immunoprecipitates were submitted to 10% SDS-PAGE, electrotransferred onto nitrocellulose membranes, and treated for Western blottings as above.
Immunofluorescence and Confocal Microscopy-For immunofluorescence and confocal microscopy, stably transfected HEK-293 cells with h3b and h3f cDNAs were grown on glass coverslips in 6-well plates. The cells were fixed with 3.8% paraformaldehyde for 10 min and permeabilized in PBS, 0.2% Triton X-100. Nonspecific antigens were blocked with a saturation solution (PBS, 2% BSA) for 30 min at room temperature. Cells were stained with a 1:500, 1:200, and 1:500 dilution of the anti-h3b, anti-h3f, and PL/IM430 antibodies (PBS, 2% BSA), respectively, for 1 h 30 min at room temperature. After three washing steps in PBS, 2% BSA, the cells were treated with a 1:200 dilution of Alexa green-conjugated anti-rabbit IgG or Alexa fluor 568-conjugated antimouse IgG (Molecular Probes Europe BV, Leiden, The Netherlands) for 1 h 30 min at room temperature. Cells were washed again three times in PBS before mounting using Dako fluorescent mounting medium (DakoCytomation, Glostrup, Denmark). Nuclei were stained with 250 ng/ml DAPI (Molecular Probes Europe BV, Leiden, The Netherlands) that was added for 5 min during the second washing step. Control experiments were performed using empty vector pcDNA3-transfected cells. Images were captured with Pullnix CCD camera and Nikon E 600 microscope using perfect image software (Clara Vision, Orsay, France) or by confocal microscopy (Bio-Rad MRC 1024).
RNA Extraction and RT-PCR-Total RNA extraction from human platelets and cell lines was as described (24). For RT-PCR experiments, essentially identical protocols as in Ref. 24 were used. The primers used to amplify hSERCA2b and hSERCA3b mRNAs are described in Ref. 24. The primers used (Genosys and Sigma) to amplify h3f are indicated in Table I. PCR was performed as in Ref. 24 except for h3f, where a Touch Down-PCR was performed for 10 cycles with annealing temperature decrements from 70 to 60°C. PCR was then conducted for different cycles, each consisting of successive periods of denaturation at 95°C for 1 min, annealing at 68°C for 1 min, and extension at 72°C for 1 min. GAPDH amplifications were used as internal RNA controls. Absence of DNA contamination was assessed by amplifying h2b in the absence of RT (data not shown). PCR products were visualized on ethidium bromide-stained 1.5 and 2% agarose gels as described in Ref. 24. The gels were scanned using Adobe Photoshop and, where indicated, quantified by Molecular Analyst, version NIH Image 1.62b7.
Plasmid Construction-For expression construct, the cDNA encoding the h3a in pcDNA3 was used (24). 3Ј-End variant-specific hSERCA3f cDNA was generated by overlap extension of two variant-specific PCR products covering the region encompassing exons 18 -23 (nt 2674 -3189 of h3f mRNA). Specific products were subcloned in pCR2.1 and then excised by EcoRV/XbaI after subcloning in pUC vector. h3f cDNA was created by switching the EcoRV/XbaI fragment of the active h3a construct by the 3Ј-end variant-specific product. The sequence of the construct was verified by automated dye terminator sequencing (Applied Biosystems AB1 100 model 377, Genome Express, Grenoble, France) across the junctions and through the modified region (not shown).
Stable Transfection in HEK-293 Cells-cDNA for transfection was purified using a plasmid purification kit (Macherey-Nagel, Dü ren, Germany). Cells were transfected with 10 g of the cDNA using the transfection agent ExGen 500 (Euromedex, Souffelweyersheim, France) according to Ref. 24 Measurements of Ca 2ϩ Release and Ca 2ϩ Influx-Confluent HEK-293 cells were loaded with 1.75 M Fura-2-AM as above. Cells were rinsed with phosphate-buffered saline, detached, collected by centrifugation, and resuspended in 8 ml of Hepes buffer (pH 7.45) containing 136 mM NaCl, 10 mM Hepes, 2.7 mM KCl, 2 mM MgCl 2 , 1 mg/ml D-glucose, and 1 mM CaCl 2 . Fluorescence measurements were performed with cuvettes containing 2 ml of cell suspensions using the same fluorimeter. Levels of [Ca 2ϩ ] were calculated as above.
After 1 min at 37°C, 2 mM EGTA was added for 1 min. Then 15 M of the SERCA inhibitor thapsigargin (Calbiochem) was added to the cells from concentrated stock solutions in dimethyl sulfoxide (Me 2 SO) (Sigma). The same amount of Me 2 SO vehicle added to the cells, which did not exceed 0.1%, was included in control experiments and did not interfere with the assays. Three min later, 20 mM CaCl 2 was added to the cell suspensions for 2 min.
Study of Phosphoenzyme Intermediates-Phosphorylation from [␥-32 P]ATP was carried out for 15 s at 0°C in a medium containing 40 mM MOPS/Tris (pH 7.0), 80 mM KCl, 5 mM MgCl 2 , 1 mM EGTA, 5 M [␥-32 P]ATP, and various concentrations of CaCl 2 to set the free Ca 2ϩ concentration as indicated. Phosphorylation from inorganic phosphate was performed for 10 min at 25°C in the presence of 0.5 mM 32 P i , 100 mM MES/Tris (pH 6.0), 2 mM EGTA, 10 mM MgCl 2 , and 30% (v/v) Me 2 SO. Dephosphorylation was studied at 25°C by a 19-fold dilution of the phosphorylated sample into 100 mM MES/Tris (pH 6.0), 2 mM EGTA, 10 mM EDTA (removing Mg 2ϩ and thereby terminating phosphorylation), 15% (v/v) Me 2 SO, and 0.5 mM non-radioactive P i (25,27). In all cases, the phosphoenzyme was quenched with 0.5 volume of 25% (w/v) trichloroacetic acid containing 100 mM H 3 PO 4 . The acid-precipitated protein was washed by centrifugation and subjected to SDS-PAGE in a 7% polyacrylamide gel at pH 6.0, and the radioactivity associated with the separated Ca 2ϩ -ATPase band was quantified by imaging, using a Packard Cyclone TM Storage Phosphor System (28,29). Appropriate background phosphorylation levels (obtained in the presence of excess Ca 2ϩ for phosphorylation from 32 P i ) were subtracted, and data analysis was performed as in Ref. 25.

Alternative Splicings Generate Six Human SERCA3a-f mRNAs, Identification of SERCA3f
We and others recently found that the human (h) SERCA3 gene contains 23 exons and gives rise to five 3Ј-end mRNA variants, termed SERCA3a-e (Fig. 1A). hSERCA3a mRNA (h3a) excludes exons 21 and 22. hSERCA3b (h3b) and hSERCA3c (h3c) mRNAs insert partial or complete exon 21, respectively. hSERCA3d (h3d) and hSERCA3e (h3e) mRNAs express partial or complete exon 21, respectively, plus exon 22. This led us to look for a putative additional variant only expressing exon 22 (upper part of Fig. 1B), called hSERCA3f (h3f) to follow the nomenclature. Hence, we PCR-amplified 3Ј-end SERCA3 mRNAs covering exons 18 -23 from U937 cells (lower part of Fig. 1B) by using the sets of primers indicated (middle part of Fig. 1B) and both GAPDH and SERCA2b as internal controls (24). Using primers P1 and P2, amplifying exons 18 -22 (Fig. 1B, lane 1), the expected doublet corresponding to h3d/h3e mRNAs was observed (upper bands) together with an additional shorter PCR product (the difference in size is about 100 bp), pointing to the absence of exon 21 (103 bp). This was confirmed by PCR-amplifying SERCA3 products using two sets of primers, P1 and P3 or P4 and P5, excluding exon 21. Both upstream primer set P1/P3 (  Novel (Sixth) Isoform of Human SERCA3 Gene codon of h3b mRNA and the same polyadenylation signal as h3a-e mRNAs. This splicing mechanism led to an additional expected h3f protein, which differs from the other members of the family in its C-terminal portion. The h3f (1033 aa)-specific C terminus comprising a tail of 40 aa is shown in Fig. 3B. Of particular interest, the first 19 aa of the h3f protein after the first splice site totally differ from those of the h3b-h3e proteins, which exhibit the same amino acids coming from exon 21, and the last 21 aa are shared with h3b and come from exon 23. Hence, at a time when the challenge is to understand the advantage of having a diversity of SERCA3 isoforms, this finding provided the opportunity to examine whether the common C-terminal sequences of h3b and h3f proteins gave rise to any common specific functional characteristics.

Stable Expression of Recombinant Human SERCA3f Isoform
To study the new isoform and compare it with h3b, h3f cDNA was constructed and stably expressed in HEK-293 cells. Recombinants were tested for protein expressions by SDS-PAGE followed by Western blotting using the Pan-hSERCA3 antibody (PL/IM430). About 9 -10 positive clones showed high expression levels compared with HEK-293 cells transfected with empty vector pcDNA3 (data not shown). Fig. 3 characterizes one of these clones at mRNA and protein levels in comparison with HEK-293 cells transfected with h3b or empty vector (pcDNA3). The expression of endogenous h2b mRNA in the absence (ϪRT) and in the presence of RT (h2b) and GAPDH were used as internal controls (24). RNA studies (Fig. 3A) showed no modulation of endogenous h2b and the expected significant, specific, and similar increases in h3b and h3f mRNAs in h3b-and h3f-transfected HEK-293 cells, respectively. Similarly, h3b and h3f recombinants were studied for protein expressions by SDS-PAGE followed by Western blotting using both common N-terminal and different C-terminal isoform-specific antibodies. The N-terminal specific Pan-SERCA2 monoclonal antibody IID8 was used to visualize h2b. The N-terminal specific Pan-hSERCA3 monoclonal antibody PL/IM430 was used to estimate expression levels. Fig. 3B shows the C termini of the h3b and h3f isoforms and the epitopes of the isoform-specific polyclonal antibodies used. The underlined peptide in Fig. 3B initially used to generate the anti-h3b antibody (last 19 aa) was later found to be present in h3f. Consequently, a novel h3f-specific polyclonal antibody was generated by using the two framed P1 and P2 peptides.
In agreement with RNA data, no modulation of h2b protein was observed, whereas the level of recombinant h3b and h3f proteins (Fig. 3C) showed similar high expression levels compared with HEK-293 cells transfected with pcDNA3. This is shown by the same recognition of h3b and h3f recombinant proteins by the PL/IM430 antibody. Previously, we estimated that the increase in h3b expression was about 150 -200-fold over the endogenous level in HEK-293 cells (24). As expected, both h3b and h3f proteins were recognized by the anti-h3b antibody (21), whereas h3f was specifically recognized by the novel anti-h3f antibody. The signal obtained by this antibody could be abolished in the presence of 10 M synthetic peptide used for immunization. Alternatively, the recognition of h3f protein, regardless of the antibody used, and its correct molecular weight ensured the intact nature of this recombinant protein. These results were also confirmed by immunofluorescence study (Fig. 3D). Control HEK-293 cells transfected with pcDNA3 showed no immunostaining with either the anti-h3b or the anti-h3f antibodies. In contrast, both h3b and h3f recombinant proteins were recognized by the anti-h3b antibody (Fig.  3D, upper panel), whereas only h3f was recognized by the anti-h3f antibody (Fig. 3D, lower panel).

Functional Characteristics of the New Human SERCA3f Recombinant Protein
The comparative Ca 2ϩ pumping properties of the recombinant h3b and h3f proteins were studied by using various techniques (Figs. 4-6). These included studies of Ca 2ϩ mobilization in intact cells (Figs. 4 and 5) as well as specific steps of their catalytic cycle using fractionated membrane proteins (Fig. 6).
Studies in Intact Cells-First, in our work in intact cells, the effect of h3f overexpression on intracellular [Ca 2ϩ ] was studied and compared with that of h3b overexpression (24) (Fig. 4)  Control HEK-293 cells stably transfected with empty vector (pcDNA3) and h3f recombinant selected for its highest SERCA3 protein expression (data not shown) have been analyzed for endogenous h2b and GAPDH (only at RNA level) expressions and compared with those of the h3b recombinant described previously (24). A, comparative RNA study of h3b and h3f transfectants. RT-PCR (n ϭ 5) using 250 ng of RNA and 20 (h2b and GAPDH), and 26 (h3b and h3f) cycles. As a control of the absence of DNA contamination, h2b mRNA amplification was performed in the presence (h2b) or the absence of RT (ϪRT). B, C termini of hSERCA3b and hSERCA3f proteins and epitopes of h3b-(21) and h3f-specific polyclonal antibodies. Slashes mark the first splice sites in these 3Ј-end regions. The underlined sequences represent the amino acid stretches of the peptide used for h3b immunization. The framed sequences represent the amino acid stretches of the peptides (P1 and P2) used for h3f immunization. C, comparative protein study of h3b and h3f transfectants. Membrane proteins were isolated from HEK-293 cells transfected with empty vector pcDNA3 and HEK-cells transfected with h3b and h3f cDNAs. 10 g of membrane proteins were separated by 8% SDS-PAGE and further treated for Western blotting by using the antibodies indicated. When using anti-h3f, the blots were treated either in the absence or presence of 10 M peptide (P2) used for immunization (n ϭ 4). D, digital images obtained from immunofluorescence microscopic study of HEK-293 cells transfected with empty vector pcDNA3, h3b, and h3f cDNAs using the anti-h3b and -h3f (anti-P2) antibodies (n ϭ 4).
Second, the analysis of the Ca 2ϩ homeostasis of HEK-293 cells overexpressing h3b and h3f was completed by measurements of Ca 2ϩ mobilization from the ER Ca 2ϩ pools and the subsequent capacitative Ca 2ϩ influx across the plasma membrane (Fig. 5). To gain insight into the consequence of SERCA inhibition on Ca 2ϩ homeostasis in control pcDNA3-, h3b-, and h3f-transfected HEK-293 cells, the Fura-2-loaded cells were treated with the SERCA inhibitor, thapsigargin (Tg), in the presence of extracellular EGTA, followed by the addition of excess Ca 2ϩ . Fig. 5A   Ca 2ϩ concentration ([Ca 2ϩ ] e ) that reached 123 Ϯ 11, 186 Ϯ 15, and 168 Ϯ 12 nM in empty vector pcDNA3-, h3b-, and h3ftransfected cells, respectively. Transfection of HEK-293 cells with h3b or h3f resulted in a 1.5-or 1.3-fold increase in capacitative Ca 2ϩ influx following thapsigargin treatment, respectively, over control empty vector pcDNA3-transfected cells.
Studies of Isolated Membrane Proteins-Finally, we examined the catalytic cycle of h3b and h3f recombinant proteins in more detail, and we compared the results with the well characterized SERCA1a isoform (Fig. 6). By using membrane preparations from transiently transfected cells, it was recently shown that the Ca 2ϩ affinity is similar in h3a-h3c and lower compared with SERCA1 (25). In the present study, we used mixed fractionated membranes (100 000 ϫ g) isolated from the stably transfected HEK-293 cells. Fig. 6B shows the Ca 2ϩ dependence of phosphorylation by [␥-32 P]ATP at 0°C, which gave 2-and 3-fold higher K 0.5 values for Ca 2ϩ activation of h3b (2.6 M) and h3f (3.6 M), respectively, as compared with SERCA1a (1.2 M). The value obtained with stably h3b-transfected cells is indistinguishable from that obtained under similar conditions following transient transfection with h3b (data not shown). It is clear that h3f exhibits the same, or even lower, Ca 2ϩ affinity as the other SERCA3 isoforms. Fig. 6C shows that the amount of E 2 P formed by reverse phosphorylation from 32 P i in the absence of Ca 2ϩ (cf. Fig. 6A) was low for control pcDNA3-transfected cells, whereas high and similar phosphorylation levels were found for membranes from cells expressing SERCA1a, h3b, and h3f proteins. We then compared the decay rates for the E 2 P phosphoenzyme intermediate. Previously, the rate of dephosphorylation of E 2 P was found to be higher in h3a, h3b, and h3c relative to SERCA1a (25). Moreover, there was a correlation between the rate of dephosphorylation and the length of the differentially spliced C terminus in these isoforms (5,50, and 36 aa, respectively). As seen in Fig. 6D, the dephosphorylation of E 2 P occurred at rates that were similar in h3b and h3f and about 10-fold higher in these enzymes as compared with SERCA1a (estimated rate constant 0.35 versus 0.03 s Ϫ1 for SERCA1a). Hence, the structural feature imposing the very high dephosphorylation rate in h3b seems to be contained within the part of the C terminus that is conserved in h3b and h3f.

Localization and Conformational State of SERCA3b and SERCA3f Recombinant Proteins
To look further for similarities and differences between both isoforms, we examined their expressions at cellular level by immunofluorescence using various antibodies (Fig. 7). We first refined our study of their expression (Fig. 3D) by immunomicroscopy using the anti-h3b and anti-h3f antibodies. Digital images obtained from planes at middle (Fig. 7A, Middle) and close to bottom (Fig. 7A, Bottom) of HEK-293 cells transfected with h3b or h3f proteins revealed apparent differences in their location (Fig. 7A). Most interesting, h3b showed a net staining of the nuclear envelope and in the endoplasmic reticulum throughout the cell, whereas a more prominent reticular pattern throughout the cell and in cellular tips was observed for h3f. This is consistent with an endoplasmic reticulum distribution of both proteins but suggests that the two proteins are associated with distinct regions of the ER, h3f possibly presenting more connections with plasma membranes.
No data are available on the conformational states of SERCA3 proteins in intact cells. To obtain some information, we compared their locations by confocal microscopy (Fig. 7B) using both the isoform-specific h3b and h3f antibodies and the Pan-hSERCA3 monoclonal antibody, PL/IM430, recently suggested to interact with SERCA3 enzymes in the E1 conformation (26). Most surprising, when merged, images (anti-h3b/PL/ IM430 (Fig. 7B, upper panels) and anti-h3f/PL/IM430 (Fig. 7B, lower panels) did not completely overlap. For h3b protein, the PL/IM430 antibody preferentially recognized a restricted area of the ER very close to the nuclei (Fig. 7B, red), whereas cytoplasmic regions of ER close to plasma membrane are more stained with the anti-h3b antibody (Fig. 7B, green). The intermediate regions appeared to present a similar recognition by both antibodies (Fig. 7B, yellow). In contrast, for h3f protein, the PL/IM430 antibody recognized cytoplasmic regions of ER FIG. 6. SERCA3b and SERCA3f exhibit similarities in the E 1 -P and E 2 P steps of the catalytic cycles. A, schematic representation of the SERCA reaction cycle, positioning the E 1 P and E 2 P intermediates. B, Ca 2ϩ affinity for activation of phosphorylation from ATP. Phosphorylation was carried in the presence of 5 M [␥-32 P]ATP and the indicated concentrations of free Ca 2ϩ as described under "Experimental Procedures." The data were normalized separately by taking as 100% the maximum phosphorylation level reached, and the lines show the best fits of the Hill equation, giving the following K 0.5 values: SERCA1a (1.2 M), SERCA3b (2.6 M), SERCA3f (3.6 M). C, E 2 P formation. Phosphorylation from 32 P i was carried out as described under "Experimental Procedures" in the absence of Ca 2ϩ (Ϫ) or in the presence of 5 mM CaCl 2 instead of EGTA (ϩ). D, E 2 P dephosphorylation. Dephosphorylation was carried out at 25°C as described under "Experimental Procedures," using the dephosphorylation medium containing 100 mM MES/Tris (pH 6.0), 10 mM EDTA, 2 mM EGTA, 0.5 mM non-radioactive P i , and 15% (v/v) dimethyl sulfoxide. The lines represent the best fits of a monoexponential decay function. There was no significant difference between SERCA3b and SERCA3f, both giving a decay constant of 0.35 s Ϫ1 , whereas the decay constant of SERCA1a was 0.03. close to plasma membrane (Fig. 7B, red), whereas the anti-h3f antibody recognized a restricted area of the ER close to the nuclei (Fig. 7B, green). The intermediate regions appeared to present a similar recognition by both antibodies (Fig. 7B, yellow). These observations can indicate a co-existence of SERCA3 isoforms in different conformational states and suggest that, in transfected HEK-293 cells, h3b and h3f proteins would present inverse gradients of the E 1 and E 2 conformational states from nuclei surrounding part of ER to the cytosolic part of ER.

Existence of the Endogenous SERCA3f Protein
To prove the existence of native endogenously expressed h3f protein (Fig. 8), studies were performed by immunoprecipitation using U937 cell membrane proteins. Indeed, in agreement with the expected low level of h3f protein (detection of h3f mRNA using 31 PCR cycles, Fig. 2), we did not succeed seeing the protein by Western blotting using up to 500 g of membrane proteins isolated from either human platelets or U937 cells. The endogenous h3f protein was seen after immunoprecipitation of U937 cell membrane proteins by using the anti-h3f (P2) antibody followed by Western blotting using the PL/IM430, N89, and anti-h3b antibodies (Fig. 8). The h3f protein (Fig. 8, upper bands) migrated at the expected molecular weight (112.6 kDa) and at the same position as recombinant h3f protein (from immunoprecipitation or cell membranes), as seen by using a low exposure time for the latter protein. Moreover, in further agreement with its low level of expression, endogenous h3f protein appeared as a comparatively faint band in U937 cell membrane protein, whereas prominent staining was observed in the membrane protein fraction from h3f-transfected HEK-293 cells using the same exposure times (not shown). The lower band visible in the U937 cell immunoprecipitate probed by the PL/IM430 monoclonal antibody (Fig. 8, top panel) may refer to some proteolysis of h3f protein occurring under the experimental conditions used. The prominent, faster migrating band visible only in immunoprecipitated samples probed with the isoform-specific polyclonal antibodies corresponds to the immunoprecipitating IgG detected by the secondary horseradish peroxidase-conjugated anti-rabbit antibodies.

DISCUSSION
Here we report the first evidence for human SERCA3f issued from a novel alternative processing of the penultimate exon 22. This finding demonstrates that a large scale of alternative splicings involving the two antepenultimate and penultimate exons 21 and 22 are used by the human SERCA3 gene to generate five isoforms (15). Hence, the SERCA3 genes give rise to a common SERCA3a isoform (16 -17) and eight speciesspecific isoforms (Fig. 9). The human SERCA3 gene encodes SERCA3b, -3c (21), -3d, -3e (24), and -3f (present work) isoforms in addition to SERCA3a. The mouse gene encodes the SERCA3b and -3c isoforms in addition to SERCA3a, which are similar but differ from the human ones (30). The rat gene encodes one totally distinct SERCA3b/c protein in addition to SERCA3a (24).
Until recently, it was widely believed that mammalian proteomes are almost identical and that species variations are mostly due to the regulation of gene expression. However, there is evidence that protein variation can also result from regula- FIG. 8. The novel SERCA3f protein exists as endogenous protein. Membrane proteins isolated from U937 cells (400 g) and h3ftransfected HEK-293 cells (50 g) were treated for immunoprecipitation (IP) experiments as described under "Experimental Procedures." Immunoprecipitated proteins and h3f-transfected HEK-293 cell membrane proteins (5 g) were then analyzed by 10% SDS-PAGE, electroblotted onto nitrocellulose membranes, and treated for Western blotting using the PL/IM430, N89, and anti-h3b antibodies. For the PL/IM430 antibody, exposure times are as follows: 10 min, 30 s, and 3 min for U937 cell membrane immunoprecipitate, h3f-transfected HEK-293 cell membrane immunoprecipitate, and h3f-tranfected cell membrane proteins, respectively. For N89, exposure times are as follows: 30, 1, and 1 min, respectively. For anti-h3b, exposure times are as follows: 15, 2, and 1 min, respectively (n ϭ 2). tion of alternative splicing (31). Accordingly, a very recent study (32) analyzing 166 pairs of orthologous human and mouse genes demonstrates that about half of the genes have species-specific isoforms. Such a species-specificity will further complicate the use of animal models to understand the biological role of human proteins. Alternatively, concerning the SERCA3 gene, this might explain the phenotypic differences observed in the case of abnormal SERCA3 gene expression in humans and rodents. Indeed, mutations of SERCA3 gene are described in type II diabetic patients (6), whereas knock out of SERCA3 leads to defects in endothelium-and epithelium-dependent relaxation of vascular (11) and tracheal (12) smooth muscles.
The novel human SERCA3f protein is shown here as an endogenous protein in U937 cells, albeit at a low expression level. This was expected as regards the fact that the level of SERCA3f mRNA is fairly low in these cells as well as in the other cells and tissues explored in the present work. However, the SERCA3f protein should be of importance based on the study of h3f mRNA distribution, which demonstrates its expression in almost all human cell and tissue types. Hence, in contrast with previous suggestions, it is more and more evident that human SERCA3 gene products are not only expressed in a restricted number of adult tissues (33)(34)(35)(36)(37).
The last 21 aa of the C terminus are the same in the novel h3f protein as in h3b. This led us to further compare h3f and h3b proteins. Both proteins were found to present similar functionality in our studies of partial reaction steps of the Ca 2ϩ transport cycle. As reported previously for SERCA3a-c isoforms (25), the apparent affinity for Ca 2ϩ activation of phosphorylation from ATP (E 1 P formation) was significantly lower in h3f than in SERCA1a, perhaps even lower than seen for the other SERCA3 isoforms. The fact that the new member of the SERCA3 family displays lower Ca 2ϩ affinity than SERCA1a suggests that this is a common property of all SERCA3 family members. Furthermore, the very high rate of E 2 P dephosphorylation characterizing h3b (25) was also found for h3f and may on this basis be associated with the common C-terminal sequence present in h3b and h3f.
The h3b and h3f proteins were found to present both differences and similarities when studying their role in the modulation of intracellular [ Another way to look for the activity of the SERCA3 isoforms was to test their sensitivity toward thapsigargin by depleting thapsigargin-sensitive Ca 2ϩ stores and, subsequently, to induce capacitative Ca 2ϩ influx from the extracellular medium. The relatively low increase in thapsigargin-induced Ca 2ϩ mobilization in h3b-or h3f-transfected HEK-293 cells compared with HEK-293 cells transfected with control pcDNA3 (cells only expressing SERCA2b) suggests that the h3b and h3f proteins present modest and similar sensitivity toward thapsigargin. Accordingly, in early studies in human platelets, which coexpress SERCA2b and high levels of SERCA3, we observed that thapsigargin preferentially inhibited the autophosphorylation of SERCA2b isoform (39).
Similarly, Ca 2ϩ influx following thapsigargin-induced Ca 2ϩ release was increased in both h3b-and h3f-transfected HEK-293 cells compared with pcDNA3-transfected ones. This was expected as regards the role of SERCA-depleted ER Ca 2ϩ pools in the store-operated Ca 2ϩ influx mechanism. Although this phenomenon has been studied for a while, it remains largely unexplained. Hence, SERCA3b and SERCA3f-depleted ER Ca 2ϩ pools may have a significant role in the regulation of store-operated Ca 2ϩ channels.
Finally, we observed that h3b and h3f proteins exhibited apparently similar but distinct locations within the ER and seemed to be co-expressed in two conformational states. The distinct location of h3b and h3f proteins was suggested by the results obtained using the anti-h3b and anti-h3f antibodies. This could mean that alternative splicing of the SERCA proteins modifies their membrane targeting. It is tempting to speculate that the different parts of their C-terminal tails carry the information for differential targeting. Interestingly, it was recently described that alternative splicing of the first intra- FIG. 9. The present SERCA3 family. C termini of human, mouse, and rat SERCA3 isoforms. Slashes mark the first splice sites in the 3Ј-end regions of the SERCA3 isoforms. The framed sequences represent the ACLYP sequence absent in the 5Ј-part of mouse and rat exon 21. The underlined sequences represent the amino acid stretches common either to human SERCA3b, -3c, -3d, and -3e or to mouse SERCA3b and -3c isoforms. The italic sequences represent the amino acid stretches common to human SERCA3b and -3f (present work). Numbers indicate the size of proteins (aa). cellular loop of PMCA (plasma membrane Ca 2ϩ -ATPase) 2 isoform alters its membrane targeting (40). The indication that both h3b and h3f proteins might be expressed in two conformational states comes from the results obtained using the PL/IM430 antibody. Although this antibody against platelet intracellular membranes has been known for a while (41), its epitope was only recently described while this work was in progress (26). PL/IM430 recognizes a non-contiguous set of amino acids that are only accessible in the E 1 conformation. Compared with anti-h3b and anti-h3f antibodies, the PL/ IM430 antibody showed roughly similar stainings of h3b and h3f proteins. This would mean that h3b and h3f recombinants are mainly expressed in the E 1 form in stably transfected HEK cells. However, merged images revealed that PL/IM430 recognition did not totally overlap, in term of intensity, with that of anti-h3b or anti-h3f antibodies. This would agree with the following: (i) recent data suggesting that at any time 20% of SERCA3a recombinants are in the E 2 conformation and not compatible with PL/IM430 binding (26); and (ii) in our previous studies (42) performed on human platelet SERCA3 proteins, demonstrating the presence of two tryptic fragmentation profiles, only one of them was recognized by the PL/IM430 antibody.
In addition to the existence of a novel human SERCA3 isoform, the present work brings some advances in understanding of a plurality of SERCA3 gene products by showing some differences in their modulation of Ca 2ϩ homeostasis and location. Other differences between the SERCA3 isoforms can also be expected based on recent data obtained with the isoforms of PMCAs. At least 30 PMCA isoforms encoded by four genes (PMCA1-4) have been known for a while. The C-terminal sequence of the so-called "b" splice variants of PMCA2 and PMCA4 genes has been shown to interact with several PDZ (PSD95/Dlg/ZO-1) proteins (43)(44)(45) to turn their activation. As concerns SERCAs, it was described very recently that insulin receptor substrate 1 (IRS-1) protein can interact with the mouse SERCA3b isoform (30). The growing number of SERCA proteins and the structural differences between the isoforms might predict their distinct involvement in various proteinprotein or multiprotein complexes.
To conclude, at a time when the field of cellular Ca 2ϩ homeostasis is evolving rapidly, the present work describes a novel member of Ca 2ϩ pumps, the human SERCA3f. It increases the extensive Ca 2ϩ signaling toolbox (46) and opens areas of investigation seeking to understand normal and abnormal Ca 2ϩ signaling.