Probing of the Membrane Topology of Sarcoplasmic Reticulum Ca2+-ATPase with Sequence-specific Antibodies

The topology of Ca2+-ATPase in sarcoplasmic reticulum (SR) vesicles was investigated with the aid of sequence-specific antibodies, produced against oligopeptides corresponding to sequences close to the membranous portions of the protein. The antisera in competitive enzyme-linked immunosorbent assays only reacted with intact SR vesicles to a limited extent, but most epitopic regions were exposed by low concentrations of nondenaturing detergent, octaethylene glycol dodecyl ether (C12E8) or after removal of cytosolic regions by proteinase K. In particular, these treatments exposed the loop regions in the C-terminal domain, including L7–8, the loop region located between transmembrane segments M7 and M8, with a putative intravesicular position, which had immunochemical properties very similar to those of the C terminus with a documented cytosolic exposure. In contrast to this, the reactivity of the N-terminal intravesicular loop regions L1–2 and L3–4 was only increased by C12E8 treatment but not by proteinase K proteolysis. Complexation of Ca2+-ATPase with β,γ-CrATP stabilized the C-terminal domain of Ca2+-ATPase against proteinase K proteolysis and reaction with most of the antisera, but immunoreactivity was maintained by the L6–7 and L7–8 loops. Immunoelectron microscopic analyses of vesicles following negative staining, thin sectioning, and the SDS-digested freeze-fracture labeling method suggested that the L7–8 epitope, in contrast to L6–7 and the C terminus, can be exposed on either the intravesicular or cytosolic side of the membrane. A preponderant intravesicular location of L7–8 in intact vesicles is suggested by the susceptibility of this region to proteolytic cleavage after disruption of the vesicular barrier with C12E8 and in symmetrically reconstituted Ca2+-ATPase proteoliposomes. In conclusion, our data suggest an adaptable membrane insertion of the C-terminal Ca2+-ATPase domain, which under some conditions permits sliding of M8 through the membrane with cytosolic exposure of L7–8, of possible functional significance in connection with Ca2+translocation. On the technical side, our data emphasize that extreme caution is needed when using nondenaturing detergents or other treatments like EGTA at alkaline pH to open up vesicles for probing of intravesicular location with antibodies.

P-type ATPases comprise a family of homologous membrane proteins involved in the active transport of inorganic cations via an aspartate-phosphorylated intermediate. Common for all types is the presence of an N-terminal and a C-terminal membrane-bound domain, separated by a cytosolic domain that contains the ATP binding site and the catalytically important aspartate residue. Most eucaryotic ATPases, comprising the Ca 2ϩ -, Na ϩ ,K ϩ -, and H ϩ ,K ϩ -ATPases and the H ϩ -pump ATPases present in plants and fungi, are distinguished by a long C-terminal domain (including transmembrane segments 5 and 6) that, together with the central membranous region (transmembrane segment 4), presumably is directly involved in the cation translocation through the membrane (1)(2)(3)(4). While there is consensus concerning the presence of four membrane traverses in the N-terminal domain, the exact number and disposition of membrane-spanning segments in the C-terminal domain is still a matter of debate. For SR 1 Ca 2ϩ -ATPase, a 10 helical transmembrane topology with six membrane-traversing segments in the C-terminal part is favored (5)(6)(7)(8). However, for Na ϩ ,K ϩ -ATPase, different models, focusing on three (9,10) and either four or six membrane traverses in the C-terminal domain, have been proposed (11)(12)(13)(14)(15)(16). In addition, the possibility of other kinds of membrane association than the classical ␣helical traverse has been considered for the C-terminal domain of P-type ATPases (17,18).
ATPases, in particular those of Ca 2ϩ -ATPase and Na ϩ ,K ϩ -ATPase, are strikingly similar, suggesting that a consensus model should be sought (19,20). With this in mind, it may first be noted that the model with three C-terminal membrane segments proposed by Ovchinnikov (21) for Na ϩ ,K ϩ -ATPase, in which the confluent P5-P6 peaks were assumed to represent only one membrane traverse, is now virtually ruled out in favor of two membrane traverses of this segment on the basis of unequivocal proteolytic evidence for a cytoplasmic location of the following L6 -7 loop in both Na ϩ ,K ϩ -ATPase (14), SR Ca 2ϩ -ATPase (22,23), and H ϩ ,K ϩ -ATPase (24). The existence of M7 is based on the presence of a long hydrophobic segment, combined with immunological evidence for an extracytosolic location of the following L7-8 loop in SR Ca 2ϩ -ATPase (8,25) and Na ϩ ,K ϩ -ATPase (13, 26 -28). Furthermore, in H ϩ ,K ϩ -ATPase cysteine residue(s) in the L7-8 loop seem to participate in the acid-catalyzed reaction with omeprazole (24). An extracellular disposition of the L7-8 loop is also suggested on the basis of evidence for interaction of this segment with the ectodomain of the ␤-chain in H ϩ ,K ϩ -ATPase (29) and Na ϩ ,K ϩ -ATPase (30,31). Concerning the remaining C-terminal part of the polypeptide chain, immunological and protein-chemical evidence has been interpreted in favor of the existence of only one further full membrane traverse at the C terminus (15,32,33). However, other data point to the cytosolic presence of the intervening L8 -9 region (34 -37). This conclusion is also supported by data obtained with molecular biology techniques that generally have yielded results in favor of the presence of two additional membrane traverses in the C-terminal part of H ϩ ,K ϩ -ATPase (38), the Mg 2ϩ transporter present in Salmonella typhimurium (39), Na ϩ ,K ϩ -ATPase (16), and yeast proton ATPase (40).
In the present study, we have analyzed the folding of SR Ca 2ϩ -ATPase in the membrane with the aid of sequence-specific antisera. These were produced by immunization of rabbits with oligopeptides, corresponding to sequences in SR Ca 2ϩ -ATPase close to the putative transmembrane segments (Fig. 1). The ability of SR vesicles to react with these antibodies was then probed in competitive ELISA assays, and their localization was established by immunoelectron microscopy. The experiments involved pretreatment of the SR vesicles with proteolytic enzyme or the addition of nondenaturing detergent to expose epitope regions close to the membrane (8,41). The use of oligopeptide antibodies was tested in separate experiments, which involved detailed characterization of the antisera employed. Overall, our data are consonant with a 10-transmembrane model as the basis for a description of the topology of eucaryotic ATPases. However, it is apparent that SR Ca 2ϩ -ATPase exhibits deviations from this topology that probably reflect a lability of some portions of the C-terminal part of the polypeptide chain with respect to insertion into the membrane. While most of the membranous polypeptide chain has a well defined location, being exposed on either the cytosolic or the extracytosolic side, the positions of some portions of the polypeptide chain with respect to the membrane appear to be more adaptable and, depending on experimental conditions, may become exposed on either side of the membrane. This phenomenon, which we term plasticity, in particular concerns evidence for cytosolic interaction of an antibody, directed against the L7-8 region with a putative intravesicular location. Possible implications of this and other findings for the structure and function of the C-terminal domain are discussed.

EXPERIMENTAL PROCEDURES
Materials-Oligopeptides corresponding to selected sequences at the N terminus, C terminus, and putative loop regions (see Fig. 1 and Table  I) of rabbit fast twitch Ca 2ϩ -ATPase were synthesized by Fmoc (N-(9fluorenyl)methoxycarbonyl) solid phase chemistry by Neosystem A.S. (Strasbourg, France). Biotinylated oligopeptides were synthesized by the pin technology of Chiron Mimotopes (Victoria, Australia). The vector pGEX-KT (42) was obtained from J. E. Dixon. Oligonucleotides were from Oligo-Express (Montigny le Bretonneux, France). Restriction and DNA-modifying enzymes were from Boehringer (Mannheim, Germany). Dioleoylphosphatidylcholine was from Avanti Polar Lipids.
Membrane Preparations-Tight SR vesicles were prepared from rabbit skeletal muscle as described by de Meis and Hasselbach (43). Proteolytic treatment to remove cytosolic domains and regions (23) was performed by the addition of proteinase K (0.06 mg/ml) to SR vesicles (2 mg of protein/ml), suspended in a medium containing 50 mM bis-Tris (pH 6.5), 10 mM Mg 2ϩ , 0.1 mM Ca 2ϩ , and 50 mM NaCl. The reaction was terminated at the desired times by the addition of PMSF (1 mM), and the digests were used for subsequent ELISA and EM experiments. In some experiments, the immunoreactivity of SR vesicles was studied after solubilization by C 12 E 8 at a 2.5:1 (w/w) detergent:protein weight ratio. The serial dilutions of these samples were performed in the presence of 0.1 mg of C 12 E 8 /ml to maintain the detergent level above the critical micellar concentration (cmc) (0.05 mg/ml). In other series of experiments, the effect of lower detergent concentrations was studied by the addition of graded amounts of C 12 E 8 to produce perturbation (44) or fragmentation (45) of the membranes as described in the legend to Fig. 2. As an alternative to C 12 E 8 treatment, SR vesicles were in some experiments permeabilized by incubation with 10 mM Tris (pH 9.0), 1 mM EGTA, and 2 mM dithiothreitol for 30 min.
Unilamellar liposomes of dioleoylphosphatidylcholine, produced by cholate dialysis (46), were used for reconstitution of Ca 2ϩ -ATPase, according to the procedure described by Levy et al. (47), as modified by Cornelius and Møller (48). In this procedure, Ca 2ϩ -ATPase in SR vesicles, solubilized by C 12 E 8 at a protein:detergent weight ratio of 2.5:1, was added to the dioleoylphosphatidylcholine liposomes (13-15 mg of phospholipid/ml), partially solubilized by ␤-octylglucoside and suspended in the bis-Tris medium mentioned above, at a 1:50 -1:20 protein: lipid weight ratio. After complete removal of detergent with Biobeads, the reconstituted sample was fractionated by centrifugation overnight at 150,000 ϫ g on a 3-12% (w/v) continuous sucrose gradient. This resulted in the formation of two bands at the upper end of the gradient. The lower band was used for experiments involving the susceptibility to proteolytic degradation by proteinase K and trypsin; although both fractions transport Ca 2ϩ , the best reconstitution, in terms of coupled Ca 2ϩ transport, was present in the lower band, which had a transport stoichiometry of close to 2 Ca 2ϩ /ATP split and a low ATP hydrolysis rate (0.7 mol/mg/min at 23°C), which was activated about 8-fold after solubilization of the lipid with C 12 E 8 . Furthermore, as evidence of a symmetric insertion, ATP hydrolysis of both fractions was increased by the addition of ionophore to about half the level observed after solubilization with C 12 E 8 .
ELISA Procedures-In these assays, polystyrene plates (Nunc-Maxisorp, Life Technologies) were coated for 2 h at room temperature with denatured Ca 2ϩ -ATPase, prepared by mixing SR vesicles (5 g of protein/ml) with 0.5% SDS and boiling for 3 min. This was followed by blocking with 3% BSA, solubilized in the buffer used in the following ELISA procedure (either the bis-Tris buffer mentioned above or PBS). Competitive ELISA procedures were carried out by mixing SR vesicles (in the presence of 0.1% BSA) or oligopeptides with suitable dilutions of antisera (final volume, 100 l) and incubation for 2 h at room temperature in the coated plates. Dilutions of antisera (500 -20,000 ϫ) were chosen so as to make the subsequent color development by the immobilized Ca 2ϩ -ATPase-antibody complex approximately proportional to the concentration of free (unreacted) antibody in the reaction medium, i.e. it was performed under conditions where there remained an excess of adsorbed Ca 2ϩ -ATPase, uncombined with antibody, on the solid phase. After rinsing with antibody-free incubation medium, bound antibody was assayed by treatment with peroxidase-conjugated goat anti-␥-rabbit antibody (Zymed, San Francisco, CA) for 1 h, followed by reaction with o-phenylenediamine and 0.003% H 2 O 2 . After stopping the reaction with sulfuric acid, readings were performed by dual wavelength spectrophotometry (495/405 nm) on a model 3550 microplate reader (Bio-Rad).
Binding properties of the antibodies by the denatured Ca 2ϩ -ATPase immobilized on the ELISA wells were measured at different antisera dilutions. The determinations were performed by a double plate technique in which the 2-h incubate from the first plate, instead of being discarded, was transferred to a second plate, which was then processed in the same way as the first plate. After the second incubation, there was a decrease in color development, resulting from the removal of part of the antibody from the mobile phase during the first incubation. The amount adsorbed by the immobilized phase during the first incubation could then be estimated from a calibration curve relating color devel-opment to antiserum dilution. The fraction of bound antibody removed from the solution was subtracted to estimate the concentration of free antibody at the end of the first assay, and apparent dissociation constants were evaluated from plots of bound antibody (measured on the basis of the color development) versus free antibody (antisera dilutions, corrected for antibody removal). Semiquantitative estimations of antibody concentrations in terms of weight units were calculated from measurements of the amount of denatured Ca 2ϩ -ATPase present on the sides of the wells, determined with radiolabeled Ca 2ϩ -ATPase as described below, assuming that all of this participated in a binding reaction with a 1:1 stoichiometry (Table I).
Indirect Immunofluorescence Microscopy-Sarcoplasmic reticulum vesicles were analyzed by indirect immunofluorescence microscopy before and after treatment with proteinase K and C 12 E 8 . After fixation with 4% paraformaldehyde for 10 min on ice, 20-l aliquots of each sample were applied to gelatin-coated glass slides and heated to room temperature for 5 min to allow the vesicles to remain on the glass slide. The samples were labeled by applying suitable dilutions with fluorescein isothiocyanate-conjugated swine anti-rabbit immunoglobulins (DAKO, Glostrup, Denmark). In control samples, the specific antibodies were omitted or substituted by preimmune sera. The glass slides were observed with a 50 ϫ oil objective lens in a Leica Laborlux S microscope and recorded on Kodak TMY400 film (Eastman Kodak Co.).
Immunoelectron Microscopy on Fixed Vesicles-Ultrastructural detection of antibody labeling was performed on detergent or proteinase K-treated SR vesicles, fixed with ice-cold 4% paraformaldehyde in PBS for 5 min. For examination of the medium exposed (cytosolic aspect) of the SR vesicles by negative staining (49), 5-l sample aliquots were applied to a carbon film supported by a 300-mesh nickel grid. The samples were incubated with PBS containing 0.1% dry milk for 10 min and then incubated with specific antisera in PBS, 0.1% dry milk for 30 min at room temperature. The labeling was probed by incubation with protein A conjugated to 5-or 10-nm gold particles (Bio-cell, Cardiff, United Kingdom) in PBS containing 0.1% dry milk and 0.1% cold fish skin gelatin for 20 min at room temperature. The grids were negatively stained with 1% uranyl acetate and air-dried before observation with a Philips 208 EM electron microscope.
For examination of antibody distribution in thin sections, untreated or proteinase K-treated vesicles were freeze-substituted into Lowicryl HM20. Briefly, vesicle samples were first fixed with 8% paraformaldehyde in 0.1 M cacodylate buffer and sedimented at 45,000 rpm in a Beckman Ti-70.1 rotor for 45 min. Fragments of the vesicle pellet were then encapsulated in 15% gelatine and infiltrated in 2.3 M sucrose containing 2% paraformaldehyde in 0.01 M phosphate buffer. The samples were then frozen in liquid nitrogen and freeze-substituted with methanol containing 0.5% uranylacetate at Ϫ85°C, using a previously described protocol (50). After increasing the temperature and rinsing in pure methanol at Ϫ70°C, the samples were infiltrated with Lowicryl HM20 at Ϫ45°C and UV-polymerized at the same temperature. Ultrathin sections (50 nm) were placed on 300-mesh nickel grids, preincubated with PBS containing 1% BSA and 0.05 M glycine and incubated with sequence-specific sera in the same buffer at 5°C overnight. The primary antibodies were detected with goat anti-rabbit IgG conjugated to 5-nm colloidal gold particles.
To deduce the position of the epitope on the membrane, the exact position of each colloidal gold particle was classified on equatorially sectioned vesicles where the membrane is sharply defined. The location of the colloidal gold particle was taken as its center and scored as being exactly over the membrane, within 10 nm outside the vesicle membrane, or 10 nm within the vesicle membrane. For each sequencespecific antibody, the first 150 colloidal gold particles that fulfilled the above criteria were included in the analysis. SDS-digested Freeze-fracture Replica Immunolabeling-The principle and procedures of this new technique for immunolocalization have previously been described (28,51,52). In brief, the method allows discrimination between the two sides of the membrane on the basis of the distribution of label among the central and peripheral portions of the fracture faces. In the adaptation of the method to the present purpose, a droplet of SR vesicles, sandwiched between two copper foils, was frozen in liquid helium and freeze-fractured by separation of the copper foils. After thorough washing with PBS, the pieces of the platinum/carbon replicas were treated with 2.5% SDS, which removes membrane protein not directly in contact with the replicas. The replicas were then rinsed with 5% BSA-PBS for 1 h, followed by treatment with sequence-specific antibody diluted 1:200 in 5% BSA-PBS buffer for 12 h at 4°C. Finally, the replicas were washed three times with BSA-PBS and incubated with secondary goat antirabbit IgG antibody conjugated with 10-nm gold (dilution 1:20 in 5% BSA-PBS) for 1 h at 37°C.
Sequence-directed Antibodies-All peptides used for production of antisera contained a single N-terminal or C-terminal cysteine residue as part of, or an addition to, the Ca 2ϩ -ATPase sequence, for coupling to keyhole limpet hemocyanin with m-maleimidobenzoyl N-hydroxysuc-cinimide ester according to Green et al. (53). Rabbits were boosted four times with injections of 0.2-0.3 mg of protein conjugate, given subcutaneously at 2-week intervals, together with Freund's incomplete adjuvant. Then, for collection of antisera, injections were extended to 3-4week intervals, followed by collection of blood 1 week after each injection. The development of a titer against Ca 2ϩ -ATPase antibody was screened by ELISA procedures, using SDS-denatured Ca 2ϩ -ATPase or peptide adsorbed to the ELISA wells. Maximal titers were usually obtained after 6 -12 immunizations.
The specificity of the antibodies produced was documented by various tests. (i) One test was for the ability of the antisera in Western blots only to react with proteolytic fragments of Ca 2ϩ -ATPase, containing the immunopeptide sequence, and for the absence of such interaction with other fragments. Proteolytic fragments of Ca 2ϩ -ATPase were produced by treatment with proteinase K (23), V8 protease (4,54), trypsin, Asp-N-endoproteinase, and elastase. The methods used for detection and identification of immunoreactive proteolytic fragments by N-terminal sequencing and mass spectrometry were the same as described previously (23). In these tests we found that the C-terminal immunopeptide antibodies in a consistent manner reacted with the appropriate C-terminal membranous fragments but not with N-terminal membrane fragments or soluble peptides from the cytosolic domain. Conversely, we only observed reactivity of the N-terminal antisera with a 28-kDa 1-243 N-terminal fragment of Ca 2ϩ -ATPase (23). The evidence for the correct location of the epitope of the 1-12 antiserum against the N terminus has previously been discussed (23). (ii) Another test was the demonstration of specific and competitive displacement of antibody from Ca 2ϩ -ATPase by the peptide used for immunization. This was tested in ELISA experiments with different concentrations of oligopeptide, using denatured Ca 2ϩ -ATPase as the immobile phase. In all cases,thepresenceofpeptideinthemobilephaseresultedinaconcentrationdependent displacement of the corresponding antibody from the solid phase. The specificity of these reactions was demonstrated by the complete absence of cross-reactions with any of the other immunopeptides, except as expected by the antisera produced against the overlapping immunopeptides 985-994 and 979 -994, where the latter peptide interfered with the 985-94 antiserum ( Table I). The affinity of the antisera for the SDS-denatured and immobilized Ca 2ϩ -ATPase, as measured with the aid of the double plate technique (except for the 951-61 antiserum) was high and occurred with an affinity corresponding to, or even higher than the affinity for the reaction with the corresponding immunopeptide (compare the two last columns of Table I).
Epitope Identification-To map in a more precise manner the location of the C-terminal epitopes, we scanned the whole C-terminal region of Ca 2ϩ -ATPase, from Glu-748 to Gly-994, for immunochemical reactivity. For this purpose, we had 47 pentadecameric peptides synthesized by the pin technology of Chiron Mimotopes with an offset of 6 or 3 amino acid residues, so that there would be an overlap of 9 or 12 amino acid residues, respectively, with the preceding peptide. The peptides were provided with a biotinyl-Ser-Gly-Ser-Gly linker arm at the N-terminal end and a blocked (amidinated) C terminus. They were bound at a concentration of 0.1 mM to streptavidin-coated plates, and their antibody-combining properties were tested in a direct ELISA reaction, using a peroxidase-conjugated rat-antirabbit monoclonal antibody for detection of the reaction. The tests were performed at different antibody dilutions, and included as controls were preimmune sera and samples only containing secondary antibody. The use of the monoclonal antibody for detection of the reaction was important, since we found that the goat polyclonal antibody used for the ELISA assays described above when present alone in the mobile phase would react with some of the peptides, producing a false positive reaction.
Molecular Biology Experiments-A C-terminal (955-994) fragment of SERCA 1a Ca 2ϩ -ATPase was expressed in a bacterial system as a fusion protein with glutathione S-transferase. For this, the Ca 2ϩ -ATPase cDNA fragment was extended with a BamHI restriction site (5Ј-GCG-GATCCATGATCTTCAAGCTCAAGGCC-3Ј) at the 5Ј-end and an EcoRI restriction site (5Ј-CGGAATTCTTATCCCTCCAGGTAGTTC-3Ј) at the 3Ј-end. The construct was amplified by polymerase chain reaction and, after digestion with BamHI and EcoRI, ligated to the pGEX-KT vector (42). The ligation product was used to transform Escherichia coli strain JM109 by electroporation. Two positive clones, GST-M10/6 and GST-M10/8, and one control, glutathione S-transferase (no insert in the expression vector), were grown at 37°C in a Luria-Bertani medium supplemented with 50 g/ml ampicillin. At an absorption of 0.2 A 600 /ml, 0.5 mM isopropylthio-␤-D-galactoside was added, followed by harvesting 5 h later by centrifugation for 10 min at 5,000 rpm and washing in STEP buffer (25 mM Tris-Cl, pH 8.0, 150 mM NaCl, 1 mM EDTA, and 1 mM PMSF). The immunochemical properties were investigated by the
For expression of Ca 2ϩ -ATPase in yeast, wild type and deletion mutants of rabbit SERCA 1a were inserted into the yeast expression vector PyeDP1/8 -10 (56) under the control of the inducible GAL10-CYC1 hybrid promoter and the phosphoglycerate kinase terminator and expressed in Saccharomyces cerevisiae, strain FKY282 MAT␣ SRP40 pep4::LEU2, ura 3-1, leu 2-3, his 3-11,-15 trp 1-1, ade 2-1 (CanR), as described previously (57). In a typical Western blot experiment, the yeast was grown for 15 h at 30°C in 10 ml of 2% galactose, 0.1% bactocasamino, 0.7% yeast nitrogen base (all from Difco), and 20 g/ml adenine (Sigma), starting at 10 5 cells/ml and stopping at 2.5⅐10 7 cells/ml. Cells were centrifuged for 5 min at 5,000 ϫ g and resuspended in 100 l of 10 mM NaN 3 . After the addition of 400 l of 2% (w/v) trichloroacetic acid, the cells were homogenized by vigorous vortexing for 4 min in the presence of 1 ml of glass beads (0.5-mm diameter). The solution was pooled, stored on ice for 10 min to allow protein to precipitate, and centrifuged for 15 min at 13,000 ϫ g at 4°C. The pellet was solubilized directly in 200 l of SDS-polyacrylamide gel electrophoresis sample buffer, and heated for 1 min at 100°C, and a 10-l sample (corresponding to expression of approximately 50 ng of Ca 2ϩ -ATPase) was loaded onto a gel. Immunodetection was performed with the ECL system after transfer to the polyvinylidene difluoride membrane (Radiochemical Center, Amersham).
Other Procedures-Bidentate (␤,␥)-CrATP, prepared as described by Dunaway-Mariano and Cleland (58), was used for preparation of CrATP-bound ATPase with occluded Ca 2ϩ (7-8 nmol/mg protein) by overnight incubation in 0.1 mM Ca 2ϩ and 100 mM bis-Tris (pH 6.5), following the procedure described by Vilsen and Andersen (59). 14 C-Labeled ATPase was prepared by incubation overnight of SR vesicles (10 mg of protein) with 25 Ci of 14 C-labeled iodoacetamide (Radiochemical Center, Amersham) in a 10 mM Tris (pH 8.0), 100 mM KCl, and 0.15 M sucrose medium, followed by removal of unreacted iodoacetamide after dilution and repeated centrifugation. The labeled ATPase was used to estimate the amount of immobilized SDS-denatured Ca 2ϩ -ATPase, which was found to be 0.10 g/well.

RESULTS
Reaction of the Antisera with Native and Detergent-treated SR Vesicles-Detergents are frequently used as a convenient means to open membranes to reveal epitopes with an intracellular or intravesicular location. The legitimacy of this procedure was tested in a series of experiments with graded amounts of C 12 E 8 to produce leaky vesicles or solubilization of membranes. The detergent was added to intact SR vesicles, incubated together with antiserum in PBS at various dilutions (from 0.001 to 15 g/well). Unreacted antibody was then detected with denatured Ca 2ϩ -ATPase coated onto the stationary phase of the ELISA wells ( Fig. 2). In the absence of detergent (E), the SR vesicles only weakly interacted with the C-terminal 985-94 antiserum and to a somewhat higher extent with the N-terminal (AS 1-12) and other C-terminal (AS 809 -27 and 877-88) antisera. The addition of C 12 E 8 to produce full membrane solubilization (Ⅺ) gave rise to a large increase in affinity of the C-terminal antibodies but only had little effect on the reaction of Ca 2ϩ -ATPase with the N-terminal 1-12 antiserum. Small amounts of C 12 E 8 were added according to a protocol designed to result in free detergent concentrations of 0.01-0.02 mg/ml (q) and 0.02-0.03 mg/ml (‚), too far below the CMC (0.05 mg/ml C 12 E 8 ) to result in solubilization of the membranes (44,60). This treatment also gave rise to an increased reactivity of Ca 2ϩ -ATPase with the C-terminal antisera but not to the same extent as that caused by full detergent solubilization. The addition of C 12 E 8 at the start of solubilization (OE, 0.25-0.3 g of bound detergent/g of protein) produced almost the same immunochemical reactivity as full solubilization for the 809 -27 antiserum and produced intermediate degrees of reactivity with the two other C-terminal antisera, whereas the reactivity of the N-terminal antiserum remained unaffected.
The lack of a detergent effect on the reaction of Ca 2ϩ -ATPase with the 1-12 antiserum is in accordance with the cytosolic localization of the N terminus. However, it should be noted that enhanced antibody reactivities were observed with both the 809 -27 and 985-94 antisera, corresponding to the L6 -7 loop and the C terminus, for which there is from proteolytic studies good evidence for a cytosolic localization (14,22). This activation of epitopes with probable cytosolic localization cannot be ascribed to a denaturing effect of the detergent, since we found that the ATPase in the ELISA wells kept full enzyme activity FIG. 3. Effect of C 12 E 8 treatment on sarcoplasmic reticulum morphology and reactivity with antiserum 877-88 as revealed by immunoelectron microscopy. SR vesicles (0.4 mg of protein/ml) were treated with 0.12 mg/ml (a), 0.15 mg/ml (b), and 0.18 mg/ml (c) C 12 E 8 and immunolabeled with 877-88 antiserum, followed by protein A gold labeling and contrast by negative staining with uranyl acetate. Note that a majority of SR vesicles appear intact and do not react with immunolabel after treatment with 0.12 mg/ml of C 12 E 8 but that there also is a population of small immunolabeled membrane fragments, while extensive vesicle disintegration with formation of both small and large gold-labeled membrane fragments is seen after the addition of 0.15 and 0.18 mg/ml C 12 E 8 , corresponding to binding of detergent close to the CMC. Magnification, ϫ 80,000. after incubation with antibody and C 12 E 8 , even after complete solubilization (data not shown). To understand the pronounced detergent effect, we further investigated how this treatment affected vesicle morphology (Fig. 3). After the addition of 0.12 mg/ml C 12 E 8 to a suspension of SR vesicles, containing 0.4 mg of protein/ml (estimated to result in a free C 12 E 8 concentration of 0.030 mg/ml, close to the threshold level for solubilization), the large vesicles with a diameter Ӎ 100 nm, characteristic of our native SR preparations, appeared largely intact (Fig. 3a). Nevertheless, we found that the preparation also contained a population of very small (12-25-nm) vesicles or membrane fragments not present in the original SR preparation. The addition of 0.15 (Fig. 3b) and 0.18 mg/ml C 12 E 8 (Fig. 3c), calculated to result in free detergent concentrations closer to the CMC, gave rise to extensive vesicle fragmentation and formation of many small membranes with a hazy outline. After detergent addition, immunostaining could be observed with all three C-terminal antisera both by indirect immunofluorescence (Fig. 4) and immunoelectron microscopy as demonstrated in Fig. 3 for the 877-88 antiserum, from which it can be seen that the label predominantly is localized to the small vesicles or membrane fragments. Accordingly, it must thus be concluded that the addition of detergent, even in low nonsolubilizing concentrations, produces distinct effects on vesicle morphology that significantly affect their immunoreactivity. The fragmentation of SR vesicles by a nonsolubilizing concentration of C 12 E 8 is probably dependent on the "shock" effect (i.e. transitory exposure to a high detergent concentration), arising from the addition of a stock solution of detergent before full mixing has taken place (see Ref. 45).
ELISA Reactions on the Proteolyzed SR Membranes-The data shown above indicate that the addition of detergent to produce permeabilization, as it is often done in immunochemical experiments to expose inside epitopes, is highly questionable. This experience in our hands also applies to the use of weaker detergents such as Tween 20, which is employed as a blocking agent (instead of, for example, BSA) in ELISA tests.
We have therefore considered other ways of uncovering the membranous epitopes without changing their immunochemical properties. Previously, permeabilization of SR vesicles with EGTA at an alkaline pH has been used to uncover intravesicular epitopes (25). However, we found that this treatment led to a pronounced increase in the immunoreactivity of all epitopes, including that against the N terminus (data not shown). This increase could be related to the modest degree of Ca 2ϩ -ATPase inactivation (approximately 10 -20%) that always accompanies even the most gentle alkaline EGTA permeabilization procedure.
We therefore, as the next step, investigated the usefulness of proteolytic treatment with proteinase K to uncover in a specific way cytosolic regions of the Ca 2ϩ -ATPase, situated close to the lipid membrane, while protecting intravesicular ones. As shown in Fig. 5, proteinase K treatment, according to previous data (23), can be used to selectively remove the large cytosolic domain with the phosphorylation and ATP binding site plus the "hinge" domain and the ␤-strand region, located between M2 and M3. This leaves the remaining ATPase membranous attached sector of the ATPase intact, apart from a few cleavage sites at the N terminus and the L6 -7 loop, as indicated by the wavy lines in Fig. 5. Immunochemical data obtained by proteinase K treatment or combined proteinase K and C 12 E 8 treatment are shown in Fig. 6. 2 It can be seen that for the 809 -827 region, even a short period of proteolytic treatment (25 min), which mainly gives rise to cleavage in the N-terminal region, results in exposure of the epitope to the same extent as observed after solubilization with C 12 E 8 . An increased exposure is also observed for the 877-88 and 985-94 antisera, although in these cases proteolysis does not increase immunochemical activity to the same extent as C 12 E 8 . At the N terminus, proteinase K treatment enhances the immunochemical reactivity with the 1-12 antiserum, whereas proteolysis does not affect the reactivity of the SR vesicles with two other N-terminal antisera, directed against the 77-88 (L1-2) or the 278 -291 (L3-4) region. These findings are consistent with a cytosolic localization of the N terminus and of the C-terminal regions, whereas they suggest an intravesicular localization of the L1-2 and L3-4 epitopes.
For the 77-88 and 278 -91 antisera, their intrinsic ability to react with Ca 2ϩ -ATPase is clearly revealed after solubilization with C 12 E 8 . However, we found that even in the absence of C 12 E 8 the 278 -91 antiserum weakly reacted with the SR vesicles at the highest concentration used (0.15 mg of protein/ml). This reactivity could possibly indicate an unsuspected crossreaction of the antiserum with an epitope other than the 278 -291 Ca 2ϩ -ATPase region, or it could indicate the presence of a small fraction of membrane fragments or vesicles with an outside-in exposure. However, the pattern of reactivity of all Nterminal antisera is clearly different from that of the C-terminal antibodies, including the 877-88 antiserum with a putative intravesicular localization, in that there is no effect of the proteinase K treatment.
The data obtained with all antibodies investigated are quantitatively analyzed and summarized in Table II. As detailed under "Experimental Procedures," our immunochemical assays were performed in such a way that there is approximate proportionality between color development and free antibody con-2 Note that instead of using PBS we performed these experiments in a pH 6.5 bis-Tris medium, since we previously found such a medium to be optimal for retention of vesicle structure (4). However, the bis-Tris medium apparently also stabilized the Ca 2ϩ -ATPase against reaction with antibody as can be seen from the absence of a reaction even at the highest concentrations of SR that had not been pretreated with proteinase K. On the other hand, data obtained after treatment with proteinase K and detergent solubilization were similar in the two media. There is no fluorescence observed in the control (a), and only a faint fluorescence can be seen after the 5-min proteinase K treatment in b and by the addition of 0.12 mg/ml C 12 E 8 (d). However, strong labeling is observed in c after 30-min proteinase K treatment and in e and f after the addition of 0.15 and 0.18 mg/ml C 12 E 8 . Note that detergent treatment leads to the formation of large fluorescent dots, which presumably reflect aggregration of Ca 2ϩ -ATPase membranes that occur in addition to the ultrastructural changes seen by EM (see Fig. 3). centration in the mobile phase. From the concentration of SR vesicles required to produce a 50% decrease in color development ([P] free 50% ), we are therefore in a position to estimate an apparent dissociation constant (K Ј d ) for the reaction of the SR vesicles with antibody according to the equation,  5. Proteolytic digestion by proteinase K and increase in reactivity of Ca 2؉ -ATPase with various antisera after solubilization of proteinase K-treated SR vesicles with C 12 E 8 . By comparison with Fig. 1, it can be seen which portions of the polypeptide chain remain attached to the membrane after treatment with proteinase K for 1 h. Proteolytic cleavage points in the remaining membrane-bound portions are indicated by wavy lines; these occur at amino acid residues 808 and 818 (23) and at amino acid residues 37 and 40. 4 Note that the split shown at position 808 actually may indicate cytosolic exposure of the putative C-terminal end of M6. Increases in immunochemical reactivity of the proteolyzed preparations after C 12 E 8 solubilization are indicated by italicized numbers. Note that the increase in reactivity of loop regions L1-2 and L3-4 is considered to be infinite, since for these epitopes there is no effect of proteinase K treatment alone on immunochemical reactivity.
ATPase interaction, on the basis of the estimated content of specific antibody in the sera (Table I).
The data of Table II confirm that proteolytic treatment with proteinase K does not result in exposure of the L1-2 and L3-4 regions. On the other hand, C 12 E 8 solubilization or combined proteinase K and C 12 E 8 treatment exposes the epitopes of these two loops to about the same extent as the 1-12 N-terminal epitope. Remarkably, the 760 -774 region, representing the N-terminal part of the putative M5 segment is clearly exposed by proteinase K treatment to a higher extent than by solubilization with C 12 E 8 . After proteolysis for 75 min, the 809 -27 antiserum exhibited a slightly lower affinity for Ca 2ϩ -ATPase than after C 12 E 8 treatment or by proteolysis for only 25 min. This decrease is probably caused by the one or two splits taking place in the 809 -827 region as indicated in Fig. 5, resulting in a lowering of the affinity of the antibody for Ca 2ϩ -ATPase. Otherwise, there was no indication that the intactness of the epitopic regions was affected by the proteolytic treatment. This can be seen from the ability of the proteolyzed and intact membranes fully to react with the antisera after solubilization with C 12 E 8 in the competitive ELISA experiments. The intactness of the epitopes could also be documented in direct ELISA experiments, which indicated that antibody reactivity was re-tained or slightly (up to 2-fold) increased after coating of ELISA wells with the 75-min proteolytic digests (data not shown).
The remaining data of Table II indicate that proteinase K treatment also exposes the other C-terminal epitopes but to a lesser extent than C 12 E 8 solubilization. It is evident that using detergent solubilization as a criterion to decide whether a given epitope has a cytoplasmic or intravesicular location is impossible. For instance, the antibody reactivity of the C terminus is activated by detergent treatment to about the same extent as the L7-8 loop. In most respects, the data obtained with our panel of antisera are consistent with the 10-transmembrane hypothesis but with two exceptions. The first unexpected finding concerns an absence of reactivity of SR vesicles, also after C 12 E 8 solubilization, with AS 951-61, which is directed against the putative luminal loop region L9 -10. Nevertheless, the antiserum reacted satisfactorily with the immunopeptide used for its induction (Table I). The other unexpected finding is the relatively high reactivity of the 877-88 antiserum with Ca 2ϩ -ATPase after proteinase K treatment (Fig. 6), since the 877-888 region, according to the 10-transmembrane model and most other proposed membrane topologies (see "Discussion"), should be located toward the intravesicular space as was previously reported (8,25). Since this is a very critical point, we  h Antiserum only reacted with high affinity against biotinylated peptides 946 -960 and 951-965, not with Ca 2ϩ -ATPase, probably because this region is not exposed in Ca 2ϩ -ATPase. The inhibition constant refers to the effect of immunopeptide on the reaction of antiserum with these biotinylated peptides bound to the stationary phase.
i Cross-reactivity with antiserum against overlapping peptide 985-994. j Solubility problem of peptide suspected to be involved in apparent low affinity.
have performed a number of additional experiments to examine the reactivity of the 877-88 antiserum under conditions that were designed to stabilize the membranous sector of Ca 2ϩ -ATPase. This we did either by performing the ELISA experiments at a low temperature (4°C) or a high Ca 2ϩ concentration (10 mM) (see Ref. 23). However, these changes in procedure only produced a slight general decrease in the reactivity of Ca 2ϩ -ATPase with antibody (data not shown). We also treated Ca 2ϩ -ATPase with CrATP (Fig. 7), which has been shown to stabilize the enzyme so that it retains bound Ca 2ϩ in an occluded state (59,61,62). With these preparations, proteolysis proceeded at a very slow rate, only resulting in partial cleavage of the peptide bond between Thr-242 and Glu-243. Thus, at the end of the 75-min proteolysis period only the N-terminal cleavage product p28N and the C-terminal cleavage product (p83C) could be detected together with undegraded CrATP-Ca 2ϩ -ATPase (Fig. 8). By contrast, unliganded Ca 2ϩ -ATPase was extensively degraded under the same conditions. Furthermore, the proteinase K-treated CrATP membranes retained the ability to occlude Ca 2ϩ (7-8 nmol/mg protein). As can be seen from a comparison of Figs. 6 and 7, CrATP treatment did not affect the reaction with the 877-88 antiserum but reduced the reactivity of Ca 2ϩ -ATPase with the 809 -27 antiserum. Even more striking, the reactivity of Ca 2ϩ -ATPase with all N-terminal antisera and the 985-94 antiserum was abolished. This suggests that N-terminally cleaved CrATP-Ca 2ϩ -ATPase, with an intact C-terminal domain, is capable of reacting with the 877-888 region in the L7-8 loop, whereas CrATP treatment leads to less exposure and/or a frozen conformation of the membrane protein as a whole, consistent with the ability of this preparation to occlude Ca 2ϩ as in the unproteolyzed membranes. Immunoelectron microscopy on Intact and Proteolyzed SR Membranes-Ultrastructural investigation after negative staining (Fig. 9) showed that following proteinase K treatment the SR vesicles are intact and retain a smooth and sharp periphery (compare with the appearance after detergent treatment, Fig. 3). Intactness of the membranes, preventing access of the antibodies to the vesicle interior, is also supported by our previous finding that intravesicular proteins (calsequestrin and the M55 glycoprotein) are retained inside the vesicles (23). By gold labeling, we were able to observe distinct immunoreactivity on the outside of the vesicles after pretreatment of SR vesicles with proteinase K, for the 877-88 antiserum as well as for the 809 -27 and 985-94 antisera (Fig. 9, b, d, f). In agreement with the ELISA data, gold labeling was sparse and only slightly above the background in vesicles not treated with proteinase K (Fig. 9, a, c, e).
Epitope localization was also analyzed on thin sections of vesicles labeled with sequence-specific antibodies and detected with goat anti-rabbit IgG conjugated to 5-nm colloidal gold (Fig. 10). Analysis was confined to equatorially sectioned vesicles with well defined membranes, and the position of each The experiments were performed according to the same procedure as described in Fig. 6 on SR vesicles that had been pretreated for 18 h at room temperature with 1 mM CrATP, 0.1 mM Ca 2ϩ , and 100 mM bis-Tris (pH 6.5) and then subjected to proteinase K treatment for various times, as indicated on the graph. Symbols are the same as in Fig. 6, and broken lines refer to data in which the vesicles had additionally been exposed to C 12 E 8 solubilization after proteinase K treatment. gold particle, defined by its center, was scored as being either directly over the membrane, just outside the membrane or just inside the membrane. As seen from the histograms, there is after proteinase K treatment a predominantly cytosolic location of AS 877-88, similar to that of AS 985-94. Due to the relatively high immunoreactivity of AS 877-88, it was feasible to perform an analysis of the distribution of this antibody also before proteinase K treatment. In this case, we found an intravesicular excess in the distribution of gold label, similar to that of the 278 -91 antiserum. Thus, the proteinase K treatment has led to a redistribution of epitope activity, which by comparison with the control 278 -91 and 985-94 antisera may correspond to a change from an intravesicular to a cytosolic location.
To further define the epitope location, we employed the SDSdigested freeze-fracture labeling method for examination of immunoreactivity on unfixed membranous material (51,52). In this technique, freeze-fracture is performed at liquid helium temperature, and the bulk of the membranous material in a freeze-fracture replica is removed by SDS. However, SDS treatment does not solubilize the layer of membrane proteins that is directly attached to the replica, which thus is left free to react with antibody. With this technique, we were recently able to localize the homologous L7-8 region in Na ϩ ,K ϩ -ATPase on the extracellular side of renal basolateral membranes (28). Fig. 11 shows the adaptation of the technique to vesicular SR preparations. In the freeze-fracture replicas, the familiar contours of concave particle-rich P-faces and convex, smooth, and largely particle-deficient E-faces are seen (Fig. 11A). For antisera 1-12, 809 -27, and 985-94 (Fig. 11A, a, c,and e) immunolabeling is present on both P-faces and at the periphery or rim of E-faces. On the other hand, with antisera 278 -91 (Fig. 11A, b) and 877-88 (Fig. 11A, d), labeling is sparse and confined to the rim or periphery of the fracture faces. After pretreatment with proteinase K, immunolabeling with AS 877-88 is stronger, and the labeling pattern is different in that now also the central regions of the P-faces are immunoreactive (Fig. 11A, f). By contrast, proteinase K pretreatment did not affect the labeling pattern with any of the other antisera (not shown).
In the interpretation of these data (Fig. 11B), it should be noted that in the central regions of the P-faces only the cytosolic portions of the Ca 2ϩ -ATPase polypeptide chain are exposed, the intravesicular portions of the polypeptide chain being hidden within the replicas (see lower part of Fig. 11B). Therefore, the labeling of central regions obtained with the N-terminal, 809 -27, and C-terminal antisera is diagnostic of a cytosolic localization of the immunoreactive Ca 2ϩ -ATPase regions of these antibodies. On the other hand, as shown by the upper part of Fig. 11B at the periphery of the E-fracture faces, both cytosolic and intravesicular polypeptide chain is exposed and therefore capable of reacting with antibody. The immunoreactivity at the periphery of E-faces suggests that the freezefracture conditions here also may result in exposure with antibody in both orientations. Therefore, the exclusive labeling of the periphery and the absence of immunoreactivity from central regions in native vesicles after treatment with antiserum 877-88 (Fig. 11A, d) are consistent with an intravesicular position of the label that is switched toward the cytosolic side after proteinase K proteolysis (Fig. 11A, f).
Exposure of the C-terminal Domain to Proteolysis-To further pinpoint the location of the L7-8 loop, we studied the ability of proteinase K to cut the Ca 2ϩ -ATPase polypeptide chain in the loop regions of the C-terminal domain. As can be seen from Fig. 12, treatment of intact vesicles with proteinase K, in the absence of C 12 E 8 (lane 4), results in the formation of a number of prominent degradation products (revealed by Coomassie Blue staining in Fig. 12A and by Western blot with the 877-88 antiserum in Fig. 12B). Many of these have previously been identified (23) (viz. p28N and p83C, which represent the N-terminal and C-terminal cleavage product, respectively, of a cut between Thr-242 and Glu-243; p29/30, soluble peptides from the middle third of the ATPase polypeptide chain, with the N terminus starting at the phosphorylation site and the C terminus extending to around Ser-610; p54, Ca 2ϩ -ATPase peptide 243-610; and a number of C-terminal peptides (p19C, homologous to the 19-kDa fragment produced by tryptic digestion of Na ϩ ,K ϩ -ATPase (63); p21C; p27C; and p28C)). In addition, lane 4 shows the presence of calsequestrin (CS) and the M55 glycoprotein as well as of a number of peptides with molecular masses below 19 kDa. It is notable that none of these peptide components with low molecular masses react with the 877-88 antiserum (Fig. 12B), i.e. the C-terminal domain is resistant to proteolytic digestion. The low molecular mass peptides that are formed under these conditions mainly represent peptide fragments, resulting from cuts in the cytosolic region of the N-terminal Ca 2ϩ -ATPase polypeptide, as will be reported elsewhere. 3 As can be seen from Fig. 12A, lanes 5-7, the degradation pattern of SR vesicles is not affected by the addition of small amounts of C 12 E 8 (0.02-0.08 mg/ml of reaction mixture). However, with the addition of 100 g of C 12 E 8 /ml, distinct changes in the proteolysis pattern arise. Under these conditions, both the calsequestrin and the M55 glycoprotein bands disappear (Fig. 12A, lane 8) as evidence of exposure of the vesicle interior to degradation by proteinase K. Furthermore, there is now the appearance of a new band (p14) that is strongly immunoreactive with the 877-88 antiserum (Fig. 12B, lane 8). The effect of adding still larger amounts of C 12  tides with low molecular masses and a decrease in the concentration of some of the other proteolytic degradation products (p27C, p19C, and p54). From sequence analysis data obtained on the excised p14 band, it is possible to deduce that it represents a peptide formed by a split between Leu-870 and Thr-871. Since p14, like the other C-terminal fragments (p19C, p21C, etc.), strongly reacts with the 985-94 antiserum, we conclude that p14 represents the 871-994 Ca 2ϩ -ATPase peptide. The N terminus of p14 is placed in the L7-8 loop, which thus is the only region close to the C terminus of Ca 2ϩ -ATPase that is susceptible to degradation by proteinase K in the presence of a low detergent concentration.
It should be noted that in Fig. 12 C 12 E 8 was added as a stock solution to the reaction mixture, and the sudden exposure to a high concentration of micellar detergent is probably responsible for the "shock" effect observed as a function of detergent concentration. This we propose to be the case because in experiments, where the detergent had been diluted in the reaction medium before the addition of SR vesicles, the changes in proteolysis occurred more gradually and at slightly higher final detergent concentrations (not shown, but demonstrated for V8 in Fig. 2B of Ref. 4 (q)). It should be noted that the concentrations of C 12 E 8 required to induce permeabilization of the vesicles and exposure of the vesicle interior by this procedure are far below those required to obtain membrane solubilization (see Ref. 44).
The striking coincidence between formation of p14 and degradation of the intravesicular proteins calsequestrin/M55 demonstrated by Fig. 12 can be taken as evidence of an intravesicular location of L7-8 whose exposure is dependent on breakdown of the vesicular barrier. However, it is difficult rigorously to rule out other and more subtle effects of added detergent. We have therefore also examined the effect of proteinase K on proteolysis of Ca 2ϩ -ATPase reconstituted into liposomes. The advantage of this kind of membrane preparation is that by reconstitution tight vesicles can be obtained that contain Ca 2ϩ -ATPase molecules both in outside-in and insideout orientations. As seen in Fig. 13B, lanes 1 and 5, the formation of p14 in this preparation can be detected without the addition of any detergent to the preparation. At the same time, lane 2 in Fig. 13A shows that only half of the reconstituted Ca 2ϩ -ATPase is susceptible to cleavage by trypsin, employing reaction conditions where Ca 2ϩ -ATPase is completely split to A and B fragments in intact SR vesicles (Fig. 13A, lane 6). The resistance of half of the Ca 2ϩ -ATPase in the reconstituted preparation to tryptic degradation is maintained both after the use of longer incubation times and higher trypsin concentrations (data not shown) as evidence of the symmetric incorporation of the Ca 2ϩ -ATPase into the vesicles. On the other hand, opening of the vesicles by the addition of C 12 E 8 progressively results in further tryptic degradation (Fig. 13A, lanes 3 and 4) as well as an increased formation of p14 (Fig. 13B, lanes 2, 3, 5,  and 6).
The Specificity of the Sequence-directed Antisera-A point of importance in relation to the immunochemical properties of Ca 2ϩ -ATPase sera is their stringent specificity toward the regions against which they have been produced. To check in more detail for the presence of potential epitope regions other than those already indicated by Western blots and competitive oligopeptide data (Table I), we also mapped antibody reactivity according to the method proposed by Geysen et al. (64). This was done by the use of pentadecameric peptides, covering the whole C-terminal domain. For all antisera tested, we observed a peak region localized in the sequence corresponding to the FIG. 9. Immunoreactivity before and after treatment with proteinase K on negatively stained sarcoplasmic reticulum vesicles. Samples were immunolabeled with Ca 2ϩ -ATPase antisera followed by gold-conjugated protein A and contrasted by negative staining with uranyl acetate. Micrographs show immunolabeling with antisera 809 -27 (a, b), 877-88 (c, d), and 985-94 (e, f), before (a, c, e) and after (b, d, f) treatment with proteinase K for 30 min, respectively. Note that little labeling with gold particles is observed before proteinase K treatment. The labeling with gold particles after proteinase K treatment is located on vesicles with a sharp and smooth periphery; in contrast to detergent treatment, no broken vesicles and fragments are observed in the background. Magnification, ϫ 80,000. peptide used for immunization (see Fig. 14). In general, reactivity outside the antigenic sites was either absent or very low. Only in one case, viz. AS 809 -27 (Fig. 14B), did we detect a number of subsidiary peaks, surrounding the major 809 -827 peak. It is difficult to find a common denominator for these secondary peaks except that they are predominantly characterized by hydrophobic sequences located in putative membrane segments. However, we did observe reactivity with one hydrophilic peptide (880 -894), but cross-reactivity of the 809 -27 antiserum with this region was not substantiated in competitive ELISA experiments with the 872-892 peptide (Table I), nor do we have any evidence from Western blot experiments of reaction of AS 809 -27 with the M5-M6 region.
The above observations are but one of a number of examples where the test seems to "overreact" by indicating the presence of potential epitopes that cannot be detected in Ca 2ϩ -ATPase by the use of other immunological procedures. Thus, it may be mentioned that purification of the ␥-fraction of the antisera by protein A affinity chromatography significantly changes the reactivity pattern, although the use of such preparations did not seem to affect our other immunochemical procedures. We propose that the use of small peptides may entail higher reactivity, probably because of a higher degree of conformational flexibility and unspecific (hydrophobic interaction) when used as solid support. As a consequence, this would suggest that purified antisera can be used for detection of specific reactivities, but as a precaution we have in this paper only presented data with antisera and not purified IgG.
The absence of reactivity of the 877-88 antiserum with the C terminus in a protein environment could also be demonstrated after bacterial expression of a fusion protein, comprising Met-955 to Gly-994 of Ca 2ϩ -ATPase linked to the C terminus of glutathione S-transferase (Fig. 15). As can be seen from Fig. 15, this construct only reacted with the 985-94 and not with the 877-88 antiserum. In a similar way, we tested the specificity of antisera 809 -27 and 877-88 after bacterial expression of peptide Gly-808 to Gly-831 or Ala-859 to Ala-893 linked to the C terminus of glutathione S-transferase. We found that only the first construct (Gly-808 to Gly-831) reacted with AS 809 -27, while conversely the second construct (Ala-859 to Ala-893) only reacted with AS 877-88 (data not shown).
We also tested immunospecificity by the use of rabbit SR Ca 2ϩ -ATPase mutants with deletions in the C-terminal region. The truncated proteins were expressed in yeast, following the protocol previously described (57). As expected for the behavior of specific antibodies, we were able to demonstrate loss of immunochemical reactivity of a SERCA 1 ⌬885-994 deletion mutant with the 985-94 antiserum, whereas the deletion mutant retained the ability to react with the 887-88, 809 -27, and other Ca 2ϩ -ATPase antisera. The specificity of the 809 -27 antiserum was highlighted by the finding that a double mutant D813A/D818A (80) in Western blot reacted with the following antisera: AS 1-12, 877-88, and 985-94 but not with AS 809 -27. This suggests that D813 and D818 are critically involved in the immunoreactive response. Furthermore, a shorter Ca 2ϩ -ATPase version, SERCA 1a ⌬754 -994, did not react with the C-terminal antisera 759 -74, 877-88, and 985-94 but retained the ability to react with antisera directed against peptides within the sequence 1-753 (data not shown). DISCUSSION In the present study, we have characterized the reactivity of antisera directed against defined epitope regions of the SR Ca 2ϩ -ATPase localized at the N terminus, C terminus, and putative loop regions close to the membrane. In many respects, the results obtained are consistent with current models for the membrane topology of P-type ATPases and with data obtained by proteolytic digestion of intact SR vesicles (22,23). This concerns the evidence for cytosolic localization of the N terminus and C terminus and the finding that the immunological reactivity of the N-terminal L1-2 and L3-4 loops, with a putative intravesicular location, was unaffected by proteinase K treatment. For the C-terminal domain, our data indicate a high degree of exposure of the L6 -7 loop in agreement with the susceptibility of this region for proteolytic cleavage (22,23), whereas other loop regions (L7-8 and L8 -9) and the C terminus were less accessible. Only in the case of antiserum 951-61 were we unable to detect any reactivity after proteinase K and C 12 E 8 treatment, suggesting that this region is embedded within the Ca 2ϩ -ATPase structure; however, the 951-61 antiserum was not very strong (Table I), and the possibility cannot be excluded that even after denaturation in SDS the 951-961 region may be shielded from reacting with the detergent.
A prominent deviation of the data from current models is the evidence pointing to a cytosolic location of the L7-8 loop, with immunochemical properties in ELISA tests very similar to those observed for the C terminus. This observation differs from previously published immunological data (25,41). It is FIG. 10. Analysis of the distribution of immunogold label between the cytosolic and intravesicular side on thin sections of SR vesicles. The histograms show the distribution of immunogold labeling over equatorially sectioned SR vesicles. The colloidal gold particles with conjugated goat anti-rabbit IgG, which detect the specific antibodies, were classified as located immediately inside the membrane (IN), directly over the membrane (OVER), or immediately outside the vesicle membrane (OUT). Each histogram shows the distribution after counting of 150 gold particles over well defined membranes labeled with antisera 278 -91 and 985-94, both after protein K treatment, and antiserum 877-88 before and after protein K-treatment (877-88K). The histograms illustrate that there is an excess of labeling of the inner side of the vesicle membrane with antiserum 278 -91, while the opposite is observed for antiserum 985-94. Antiserum 877-88 predominantly labels the inside of the membrane in untreated vesicles but labels the outside (cytosolic side) after protein K treatment. also different from the evidence for extracellular exposure of the corresponding region in Na ϩ ,K ϩ -ATPase, based on similar immunological data (13,26,27), and molecular biology evidence for interaction of L7-8 with the ␤-subunit (30,31). In the evaluation of the present data, it was therefore important, first, to be able to demonstrate the absence of cross-reactivity of the 877-88 antiserum with other parts of the Ca 2ϩ -ATPase molecule. This we did by establishing the specificity of this and the other antisera against a series of C-terminal oligopeptides (Fig.  14) as well as the specific interaction with C-terminal deletion mutants and a fusion protein (Fig. 15). In the similar immunological experiments by Matthews et al. (41), which led to a different conclusion, the criterion for an intravesicular orientation of the 877-888 region was that the reaction of their L7-8 (877-88) antiserum with Ca 2ϩ -ATPase was dependent upon solubilization of the vesicles with C 12 E 8 , whereas this was not required for reaction of Ca 2ϩ -ATPase with antibody against the C terminus. By contrast, in our experiments the reaction of both antibodies with Ca 2ϩ -ATPase had similar characteristics and was dependent on pretreatment of the vesicles with C 12 E 8 or proteinase K. We have ascertained that the divergent results do not reside in different properties of the antisera used, since initially we used antisera supplied by the Southampton group and obtained results that did not differ from the antisera that we subsequently produced ourselves. However, there were differences in the exact protocol being followed in the ELISA procedure; in the set-up used by the Southampton group, Tween 20 was used as a blocking agent, whereas we used BSA. When tested in our laboratory, the use of Tween 20 led to an increased exposure of both regions reacting with the 877-88 and 985-94 antisera. We know from other experiments that the presence of Tween 20 leads to a slow solubilization of Ca 2ϩ -ATPase vesicles (65), and therefore the use of this detergent for immunolocalization purposes is discouraged. It should also be noted that in our rather extensive experience with these antisera we have not been able to design experimental conditions that would clearly show the effect of vesicle permeabilization per se on immunoreactivity. This is, for instance, evident from the experiments involving the addition of graded amounts of C 12 E 8 (Fig. 2). We have also attempted to expose the intravesicular side, avoiding the use of any detergent, by treatment of SR vesicles with EGTA at pH 9, following the procedure used by Clarke et al. (25) to demonstrate the intravesicular location The schematic drawing shows that Ca 2ϩ -ATPase directly attached to the replicas is not removed by SDS treatment, so that exposed epitopic regions can be revealed by treatment with sequencespecific antibody and gold-conjugated secondary antibody. In the central regions of P-fracture faces (lower part of the drawing), only antibody with cytosolic epitopes (q) has access to the ATPase, whereas in the periphery intravesicular regions can also be labeled by antibodies directed against these regions (f, upper part of the drawing; see also Fujimoto et al. (28)). of a monoclonal antibody, directed against the L7-8 loop. However, this treatment increased the reactivity not only of the C-terminal antisera but also of the N-terminal 1-12 antiserum to levels comparable with those seen by SDS denaturation (not shown). We therefore have to conclude that attempts to open the vesicles to expose intravesicular epitopes are likely to result in increased reactivities also of epitopes with an established cytosolic location. Despite these methodological problems, we have direct evidence that in our experiments the 877-88 antiserum did react with SR vesicles from the cytosolic side, since after proteinase K treatment to remove cytosolic regions we were able to locate bound antibody on the outside of the vesicles by EM with the aid of gold-labeled secondary antibody (Figs. 9 -11). The labeling of the vesicles was uniform, i.e. there was no evidence that it was confined to a subpopula-tion of vesicles with deviating properties (such as, for example, an inside-out orientation).
Effect of Ca 2ϩ -ATPase Organization on Immunochemical Reactivity-Most of the experiments reported here involve pretreatment of SR vesicles with proteinase K to uncover membranous epitopes. As previously discussed (23), and supported here by the EM data, this treatment does not destroy vesicle integrity (Fig. 9). Even when we subject SR vesicles to long term (24 -48-h) proteinase K treatment, they retain their calsequestrin and M55 glycoprotein content intact inside the vesicles. On the other hand, the addition of detergent to produce vesicle fragmentation and solubilization usually increased the reactivity not only of the intravesicular, but also of cytosolic, epitotopes. It should be noted that this increase in reactivity occurs by a C 12 E 8 treatment that is sufficiently mild to fully retain Ca 2ϩ -ATPase activity. It thus cannot be attributed to irreversible, structural changes in Ca 2ϩ -ATPase structure. Increased immunoreactivity, following detergent treatment, in addition to exposure of intravesicular epitopes, is likely to result from increased conformational flexibility, caused by such factors as a decrease in protein-protein interactions, substitution of membrane lipid with more fluid-like detergent, and increased accessibility of the antibodies to regions of Ca 2ϩ -ATPase close to both the cytosolic and intravesicular aspects of the membrane. The importance of such factors is underscored by the finding that we do not observe increased immunological reactivity after C 12 E 8 treatment with antisera directed against epitopes in the cytosolic region. 4 Topology of the C-terminal Domain-From the preceding evidence on cytosolic exposure of the 877-888 epitope, what can we deduce or suggest concerning the membrane folding of SR Ca 2ϩ -ATPase? A cytosolic location for the L7-8 loop in Na ϩ ,K ϩ -ATPase was predicted on the basis of the membrane folding model proposed by Ovchinnikov et al. (9,10), in which only the presence of three traverses in the C-terminal domain (corresponding to M5, M7, and M9 in Fig. 1) was assumed. However, for Ca 2ϩ -ATPase, Na ϩ ,K ϩ -ATPase, and H ϩ ,K ϩ -ATPase there is now from protein-chemical evidence little doubt concerning the cytosolic location of the L6 -7 loop (20); a cytosolic location of the 877-888 Ca 2ϩ -ATPase epitope therefore requires either the nonexistence of M7 or the presence of an additional membrane traverse after M7 and before the 877-888 region. M7 is characterized by having a long chain of uncharged, hydrophobic amino acid residues; therefore, we consider an extramembranous position of this segment as unlikely. In Fig. 16, A-C, we compare various topological models for the C-terminal domain. Panel A shows the intravesicular position of L7-8 according to the 10-transmembrane segment model, whereas panels B and C indicate two other models that could account for a cytosolic location of the 877-888 region. In panel B, a return of the polypeptide chain to the cytosolic side is considered to take place via a short, hydrophobic segment (M7a, which presumably should be thought of as a ␤-traverse) located in a hydrophobic region before Thr-870 and including part of the putative 859 AEDPG turn sequence present at the end of M7 (Tyr-858). Another possibility, as shown in panel C, is that the distal part of M7 does not cross the membrane but folds back toward the cytosol. It is here of interest to note that the M7 segment in SR Ca 2ϩ -ATPase has properties that could facilitate these proposed schemes for membrane folding. Thus, the M7 segment in SERCA ATPase is longer than in any other sequenced P-type ATPase (20,66). M7 is also characterized by a high content of short amino acid residues ( 845 GAATV-GAAA 853 ) in its distal part, which would allow a flexible turn in the membranous environment of M7.
A consequence of these proposals is the probability of a cytosolic location of M8 rather than a membrane-inserted location. Otherwise, we cannot account for proteolytic evidence of cytosolic exposure of the following L8 -9 loop in Ca 2ϩ -ATPase (22), 3 a localization that is also supported by the present immunological data (Table II). In this connection, it is of interest to note that there is evidence that M8 may be loosely associated with the membrane (see below) and that it has amphipathic features in helical wheel projections, which would be compatible with a location parallel to, rather than perpendicular to, the membrane.
Plasticity of the C-terminal Domain-Although the alternative models in panels B and C of Fig. 16 are consistent with many of the immunological findings, at least two major prob-  3 and 4), a concentrated solution of C 12 E 8 (50 mg/ml) was added to the reconstituted vesicles to final concentrations of 1.5 and 4.0 mg/ml, immediately before the addition of trypsin. Proteolysis was stopped by the addition of trypsin inhibitor (0.2 mg/ml). After shaking for 30 min, tube contents were treated with SDS/␤-mercaptoethanol, and samples corresponding to approximately 2 g of protein were loaded onto each lane of a 13.5% acrylamide gel. Note that in this gel system the main ATPase degradation products are present as overlapping A and B bands, which in the SR sample (lane 6) comigrate with calsequestrin (CS) and M55. Other bands are trypsin (Tr), trypsin inhibitor (T.I.), and minor tryptic degradation bands A 1 and A 2 shown by Coomassie Blue staining. B, treatment of reconstituted liposomes or SR vesicles (same concentrations as in panel A) was performed with proteinase K (3 or 30 g of proteinase K/ml as indicated in the figure) in a medium containing 100 mM bis-Tris (pH 6.5), 0.1 mM CaCl 2 , and C 12 E 8 where indicated. Proteolysis was stopped with PMSF (1 mM) and cooling of the samples as described previously (23) before solubilization with SDS/␤-mercaptoethanol and loading of approximately 0.13 g of protein onto each lane of a 13.5% acrylamide Laemmli gel. After transfer to an Immobilon membrane, the blot was reacted with the 877-88 antiserum (dilution 1:15,000) and immunostained with the ECL reagent. The lane indicated with an asterisk (lane 4) represents an attempt to deposit active protease inside the liposomes by freezing and thawing. lems remain. First, as far as can be deduced from available data, these models cannot be applied directly to other ATPase types. For example, for Na ϩ ,K ϩ -ATPase there is from immunolocalization experiments obtained by Mohraz et al. (13) and by us (26 -28) strong evidence for the exposure of the L7-8 loop on the extracellular side of the membrane, a localization that is supported by molecular biology evidence that this region is involved in interactions with the ectodomain of the ␤-subunit in Na ϩ ,K ϩ -ATPase (30,31). Second, our data obtained on the susceptibility of the C-terminal domain of Ca 2ϩ -ATPase to proteolysis (Figs. 12 and 13) and by immunoelectron microscopy (Figs. 10 and 11) suggest that the L7-8 loop of intact ATPase has a preponderant intravesicular location. It therefore remains to be considered to what extent the evidence for cytosolic localization obtained in proteolyzed preparations reflects topological properties of intact ATPase. To address this question, it is first of interest to consider evidence of lability of the C-terminal domain of Na ϩ ,K ϩ -ATPase. After treatment with trypsin to produce "19-kDa membranes," Kaplan and coworkers (76) observed that the membrane association of M5-M6 is weak, leading to release of this membrane segment at 37°C as a water-soluble fragment after removal of occluded K ϩ . The dependence of retention of the M5-M6 fragment in the membrane on concomitantly bound cation was considered as evidence of direct involvement of this segment in cation binding and the conformational membrane changes involved in cation transport. With posttryptic Na ϩ ,K ϩ -ATPase membranes, evidence of transposition of the cytosolic L8 -9 loop toward the extracellular side has also been obtained after removal of occluded cation (34,36). However, this demonstration required elevated temperatures (50 -60°C) and the presence of reducing agent, treatments that may have led to irreversible unfolding of the original structure. In our experiments on Ca 2ϩ -ATPase, cytosolic exposure of the L7-8 loop was observed already at room temperature (23°C), and it persisted at high Ca 2ϩ concentrations (10 mM) where Ca 2ϩ remains bound to the translocation sites (23). The L7-8 segment also could be detected on the cytosolic side after complexation of Ca 2ϩ -ATPase with CrATP, which protects the protein against both proteolytic degradation (Fig. 8) and reaction with most other antisera (Fig.  7). From these findings, it seems safe to conclude that, although lability of the C-terminal domain is a common feature, FIG. 14. Immunoreactivity of various antisera with C-terminal oligopeptides. Pentadecameric peptides, corresponding to the C-terminal Ca 2ϩ -ATPase sequence (Glu-748 to Gly-994), with an offset of 6 or 3 amino acid residues, were bound to streptavidin-coated plates via a biotinyl-Ser-Gly-Ser-Gly linker arm. The antisera were tested at various dilutions (from 1:2,000 to 1:30,000) in direct ELISA experiments, and antibody binding was estimated by a secondary monoclonal mouse anti-rabbit peroxidase-conjugated antibody. Controls included samples with preimmune sera and samples without the addition of oligopeptide antisera. Results concerning antisera 759 -774, 877-88, 918 -30, 951-61, 985-94, and 979 -94 are all shown in panel A with different codes. Antiserum 809 -27 is shown on a separate graph (B), since it gave a number of secondary peaks due to cross-reactivity.
there are basic differences between the behaviors of the Cterminal part of Ca 2ϩ -ATPase and Na ϩ ,K ϩ -ATPase. These as discussed below may reflect differences in the exact way by which they transport cations.
Concerning the evidence for bidirectional exposure of the L7-8 loop, the possibility cannot be completely excluded that this phenomenon is facilitated by the experimental conditions themselves (proteolysis leading to changes in membrane insertion, stabilization of the epitope on the other side of the membrane by interaction with its antibody). However, it is reasonable to assume that the experimental evidence of cytosolic exposure reflects an underlying adaptability in the insertion of M8 and adjoining membrane-spanning segments that is not observed in the N-terminal domain. Such adaptability, for which we use the term plasticity, can be rationalized on the basis of the moderate hydrophobicity of the C-terminal domain (see Fig. 1). With respect to the 877-888 region, cytosolic exposure from an initially intravesicular position could occur by sliding of M8 together with L7-8 through the membrane, as indicated by the arrow in Fig. 16A, resulting in a membrane topology of the type shown in Fig. 16, B and C. In relation to these topologies, it is of interest that Green and Toms (67) previously reported the formation of a water-soluble peptide after cleavage of the Ca 2ϩ -ATPase polypeptide chain by chemical destruction at tryptophan residues (this was the peptide that subsequently was used to synthesize a DNA probe for isolation of Ca 2ϩ -ATPase cDNA (68)). On the basis of the amino acid analysis data, the peptide can be identified as encompassing M8 and the C-terminal part of L7-8 (19). The release of this peptide from the membranous environment also suggests that M8 may not be firmly anchored within the membrane. In a more speculative sense, changes in the insertion of M8 in the membrane may have functional importance in connection with the conformational changes involved in Ca 2ϩ -binding and translocation. Thus, site-directed mutagenesis experiments have implicated a glutamate residue in M8 (Glu-908) as being essential for the Ca 2ϩ translocation process (3), although subsequent experiments have shown that this residue probably does not function as a Ca 2ϩ -occluding ligand (69,70). However, this circumstance does not exclude the possibility of an important role of M8, e.g. as a Ca 2ϩ -channel lining helix, associated with Ca 2ϩ translocation (71).
In connection with our findings for Ca 2ϩ -ATPase, it is of interest to note that the concept of a dynamic interplay of certain membrane segments with respect to the membrane is gradually gaining wider recognition, e.g. as demonstrated in topological studies on colicin (72), on P-glycoprotein (73), on E. coli by an UhpT transporter protein (74), and on the SecG protein component of the preprotein translocase complex (75). With respect to Na ϩ ,K ϩ -ATPase, the exposure of two sulfhydryl groups in the C-terminal domain of Na ϩ ,K ϩ -ATPase has been reported to be susceptible to conformational changes of the same type (E1/E2) as those connected with cation translocation (77). As mentioned above, in functional Na ϩ ,K ϩ -ATPase ␣␤-subunits the L7-8 segment seems to be stably localized toward the extracellular side (13,26,28,30), but after expression of ␤-galactosidase fusion proteins in yeast and E. coli Fiedler and Scheiner-Bobis (16) obtained evidence for a cytosolic localization of L7-8 in C-terminally truncated ␣-Na ϩ ,K ϩ -ATPase units. This may suggest a "pull" of the L7-8 segment toward the cytosolic side also in functioning ␣␤-Na ϩ ,K ϩ -ATPase units, a circumstance that by the authors was inter- . Total bacterial extracts, containing either the pGEX-KT vector alone or vector with the fusion protein construct (pGEX-KT ϩ M10-cDNA), were loaded on a 15% Laemmli gel. As a control, 100 ng of SR Ca 2ϩ -ATPase was also loaded onto the gel. After transfer to polyvinylidene difluoride membrane, immunodetection was undertaken with AS 985-94 and then (after stripping the membrane according to the ECL protocol) with AS 877-88 at the same dilution (1:10,000). Note the presence of a band strongly reacting with AS 985-94, close to the expected position (31 kDa) for the fusion vector, which does not react with AS 877-88 and is absent in the unmodified vector. The faint band of higher molecular mass, reacting with the 985-94 antiserum, probably represents a bacterial component, unrelated to Ca 2ϩ -ATPase, which is revealed both in the presence of modified and in the presence of unmodified vector.
FIG. 16. Topological models for the C-terminal folding of the Ca 2؉ -ATPase polypeptide chain. A, topology with six transmembrane segments in the C-terminal domain according to the standard 10-helix membrane model (79). B, a different six C-terminal transmembrane model, based on a short ␤-traverse (M7a), following immediately after M7 and cytosolic localization of the remaining part of L7-8, and of M8 and L8 -9. C, model in which M7 folds back toward the cytosol, resulting in cytosolic location of L7-8, M8, and L8 -9. f, the location of the epitopic region 877-888. The arrow in A illustrates how sliding of M8 through the membrane can result in a change in the cytosolic location of L7-8 from the intravesicular to the cytosolic side. For further explanation, see "Discussion." preted in terms of the formation of an intramembranous loop, similar to the P-loop in channel proteins.
In conclusion, there seems little doubt that plasticity of the C-terminal domain is an important factor for many of the conflicting data that have been and still continue to be reported on the topology of the C-terminal domain of P-type ATPases. As a consequence, we suggest that instead of trying to pinpoint an exact membrane topology for the C-terminal domain, which may be an unattainable goal, it would be more fruitful to direct attention toward an exploration of the extent of plasticity and the role that this phenomenon may have in the function of P-type ATPases.