The endoplasmic reticulum in PC12 cells. Evidence for a mosaic of domains differently specialized in Ca2+ handling.

Velocity and isopycnic gradient centrifugation were employed to fractionate post-nuclear supernatants rapidly prepared from PC12 cells in order to characterize areas of the endoplasmic reticulum involved in various aspects of intracellular Ca2+ homeostasis. The endoplasmic reticulum Ca2+ pumping activity, defined by three properties studied in parallel in the isolated fractions; thapsigargin-sensitive uptake of 45Ca2+, Ca2+-dependent, thapsigargin-sensitive protein phosphorylation and Western blotting of sarcoplasmic reticulum calcium ATPase (SERCA) 2b and putative SERCA3 ATPases, was concentrated primarily in a few fractions located at the top and toward the bottom of velocity and isopycnic gradients, respectively. The endoplasmic reticulum Ca2+ release channel, the inositol 1,4,5-trisphosphate receptor, was concentrated in the same fractions as the Ca2+ pumps, and additionally in a few fractions distinctly poor in SERCAs. In contrast, two lumenal markers (protein disulfide isomerase and calreticulin, the major Ca2+ storage protein of non-muscle endoplasmic reticulum) were enriched in the middle fractions of the velocity gradients while calnexin, a Ca2+-binding membrane protein, was more widely distributed throughout the gradients. These results document a considerable degree of functional and compositional heterogeneity in the endoplasmic reticulum of neurosecretory PC12 cells. Even in the limited areas that appear specialized for rapid Ca2+ uptake and release the ratio between pumps and channels varies considerably. Within the rest of the system, insulated from short-term fluctuations of Ca2+ concentration, Ca2+-binding proteins appear to be extensively distributed, in agreement with the idea that the Ca2+ content of the endoplasmic reticulum serves multiple functions.

In all eukaryotic cells studied, the cytosolic concentration of free calcium ion, [Ca 2ϩ ] i is maintained at a low (about 100 nM) level compared with the millimolar concentrations present in the extracellular environment, and is tightly regulated by a variety of pumps, exchangers, and channels. In contrast, most or all intracellular organelles contain relatively high calcium (bound plus free), and many are believed to maintain lumenal free Ca 2ϩ concentrations closer to that of the extracellular medium than to that of the cytosol (1). At least in theory, this stored Ca 2ϩ may play a dual role: (i) as a source of Ca 2ϩ for intracellular release, and (ii) as a participant in the characteristic physiology of each organelle. Such a duality of function has been clearly demonstrated for the endoplasmic reticulum (ER). 1 This structure is known to be the source of the Ca 2ϩ available for rapid release by intracellular second messengers like IP 3 , while a high Ca 2ϩ concentration seems to be a requirement for its protein processing activities, which take place in the ER lumen (2). To a degree, these two types of function present conflicting requirements in Ca 2ϩ handling. On the one hand, Ca 2ϩ for signaling purposes should be highly mobile, rapidly available, and quickly replaced. On the other hand, the intra-lumenal roles of Ca 2ϩ might expected to call for a more stable, or slowly-changing, availability of the ion. One way in which the ER may reconcile these requirements is by a degree of regional specialization, with only some parts, distributed at appropriate places in the cell, being responsible for rapid uptake and release of Ca 2ϩ in response to a signal, leaving other regions to carry on with basal, yet also Ca 2ϩ -dependent, functions.
The idea of ER domains differently specialized in Ca 2ϩ handling is not new (3). The weight of supporting evidence is quite different, however, in various cell types. It is generally accepted in striated muscle fibers, where the sarcoplasmic reticulum, extraordinarily rich in both Ca 2ϩ pumps (SERCA-type Ca 2ϩ -ATPases) and Ca 2ϩ channels (ryanodine receptors), has been shown to be an extension of the endoplasmic reticulum (4), and also in smooth muscle and in cerebellar Purkinje neurones, where the high levels of expression of the IP 3 receptors could be investigated by high resolution immunocytochemistry (5,6). Results obtained in other types of cells have also been interpreted in terms of at least partial segregation among pumps, channels, and lumenal Ca 2ϩ -binding proteins (see, for example, Refs. 7 and 8). Words of caution about this possible heterogeneity have, however, also been advanced, suggesting biochemical artifacts to be responsible for many such results (9,10).
In order to re-investigate this problem, we have chosen the PC12 cell line, a well known model of neurosecretory and neuronal cells (11). Previous results (12,13) showing that a single rapidly-exchanging Ca 2ϩ pool could be discharged when the cells were challenged with either IP 3 -generating receptor agonists or thapsigargin, a blocker of the SERCA-type Ca 2ϩ pumps, were interpreted as an indication of ER homogeneity in these cells. An additional large pool, however, was not discharged by these treatments, and its intracellular location in other organelles and/or in IP 3 -and thapsigargin-insensitive domains of the ER was not established (13). The present approach is based on subcellular fractionation, by both velocity and equilibrium gradients, with analysis of the isolated fractions carried out by Western blotting against a panel of ER markers, in parallel with measurements of 45 Ca 2ϩ uptake and SERCA phosphoenzyme formation. The results obtained document in PC12 cells the molecular and functional heterogeneity of ER domains, in line with the concepts of general Ca 2ϩ physiology discussed above.

MATERIALS AND METHODS
Cell Culture and Subcellular Fractionation-PC12 cells were grown in Dulbecco's modified Eagle's medium, supplemented with 10% fetal calf serum, 5% horse serum, and 100 units/ml penicillin and streptomycin, at 37°C in a 5% CO 2 atmosphere. Cells were grown to 80 -90% confluence before being removed from the Petri dishes by scraping, and washed once in cold Tris-buffered saline (137 mM NaCl, 4.5 mM KCl, 0.7 mM Na 2 HPO 4 , 1.6 mM Na 2 SO 4 , 25 mM Tris-HCl, pH 7.4), and once in cold Hepes-buffered sucrose (0.25 M sucrose, 1 mM EDTA, 1 mM magnesium acetate, 1.6 mM Na 2 SO 4 , 10 mM Hepes-KOH, pH 7.2), supplemented with a mixture of protease inhibitors (10 g/ml each of aprotinin, chymostatin, leupeptin, 1 mg/ml o-phenanthroline, 1.6 mg/ml benzamidine, 17 mg/ml phenylmethylsulfonyl fluoride in dimethyl sulfoxide; normally used at 1000-fold dilution). Washed cells were pelleted by centrifugation at 800 ϫ g and resuspended in cold Hepes-buffered sucrose-protease inhibitors at a concentration of about 1 ϫ 10 8 cells/ml. Cells were lysed by 7-8 passes through a "cell cracker" (14) (EMBL, Heidelberg, Germany) with an operating clearance of 18 m. The cell lysate was centrifuged at 200 ϫ g for 10 min at 4°C, and 1 ml of the resulting post-nuclear supernatant was layered on top of a continuous sucrose gradient (10.5 ml, 0.3-1.2 M in 10 mM Hepes-NaOH, pH 7.2, with protease inhibitors) and centrifuged for 15 min at 25,000 rpm, using a Kontron SW 41 swing-out rotor in a Beckman L7-55 ultracentrifuge. Centrifugation time was counted from the moment that the rotor reached the selected speed, until the start of (maximum) braking (15). In some experiments, centrifugation was continued for 22 h, in order to reach equilibrium conditions. The gradients were unloaded in 1-ml aliquots from the top, corresponding fractions were pooled from 2 to 3 gradients, diluted with 10 mM Hepes-NaOH, and centrifuged again at 30,000 rpm (Ti 50 rotor) for 30 min. Pellets were resuspended in a small volume of an appropriate buffer (see below), supplemented with protease inhibitors, and either used immediately or frozen at Ϫ80°C. For electron microscopy, pellets were embedded in Epon. Sections were cut perpendicularly to the centrifugation axis, so as to include the whole depth of the pellet, thus ensuring that representative images were obtained. Protein concentrations were determined by the Cu 2ϩ /bicinchoninic acid method, using bovine serum albumin as a standard. RNA content was estimated by the absorbance at 260 nm.
Electrophoresis and Western Blotting-Protein suspensions were solubilized by 2-fold dilution in SDS sample buffer (150 mM Tris-HCl, pH 7.4, 8 mM Na 2 EDTA, 3% SDS, 20% sucrose, 0.14 mg/ml bromphenol blue, 1% ␤-mercaptoethanol) and heated on a boiling water bath for 5 min. Solubilized samples for Western blotting experiments were either frozen at Ϫ80°C, or immediately separated by Laemmli-type SDS-PAGE mini-gels, typically using 4% stacking gels and 5, 6, or 7% separating gels. Equal amounts of protein were loaded from each fraction. Gels were blotted on 0.2-m pore nitrocellulose membranes for 16 h in 25 mM Tris, 192 mM glycine, pH 8.3, containing 20% methanol at a constant 25 V, in a cold room. Blots were briefly stained with 0.2% (w/v) Ponceau Red in 3% trichloroacetic acid to reveal the molecular weight standards, blocked with 5% (w/v) low-fat dried milk in phosphate-buffered saline-Tween (PBS-T: 100 mM NaCl, 100 mM Na 2 HPO 4 / NaH 2 PO 4 , pH 7.5, 0.1% (v/v) Tween 20) for 1 h, then exposed to the primary antibodies for 1 h, washed in 5% milk in PBS-T (3 ϫ 5 min), incubated with horseradish peroxidase-linked anti-rabbit or anti-mouse IgG (1:5000 in 5% milk in PBS-T) for 1 h, and finally washed in PBS-T (5 ϫ 5 min). Immunostaining was recorded photographically, using an enhanced chemiluminescence kit from Amersham, and the films analyzed by scanning densitometry.
Phosphorylation-Ca 2ϩ -dependent phosphorylation by [␥-32 P]ATP (Amersham, 3000 Ci/mmol), was carried out on ice. Briefly, samples (30 -100 g of protein) were resuspended in 100 mM KCl, 20 mM Hepes-NaOH, pH 6.8, in the presence of 50 M CaCl 2 or 1 mM EGTA, in a final volume of 100 -200 l. Where applicable, thapsigargin was preincubated with the membranes for at least 5 min before addition of Ca 2ϩ . For the proteolysis experiments with trypsin, 8 g of trypsin (chymotrypsin-free) was added up to 5 min before the start of phosphorylation. The reaction was started by addition of 1 Ci of [␥-32 P]ATP (final concentration 20 -40 nM) and quenched after 30 s by addition of 1 ml of ice-cold 10% trichloroacetic acid, containing 50 mM NaH 2 PO 4 and 0.5 mM ATP. Samples were allowed to stand on ice for 15 min before the precipitated material was collected by centrifugation, and washed once with 1 ml of ice-cold distilled water. Precipitates were solubilized in SDS sample buffer as above, but without heating. Solubilized samples were subjected to acidic SDS-PAGE, using the method of Sarkadi et al. (24), typically running overnight in a cold-room at a constant 50 V. Gels were blotted onto 0. Calcium Uptake and Release (Fluorescence Assay)-Vesicular uptake and release of Ca 2ϩ was measured using the fluorescence of extravesicular Calcium Green-2, as described previously (25). PC12 post-nuclear supernatant was prepared in Ca 2ϩ uptake buffer (100 mM KCl, 5 mM NaCl, 2 mM MgCl 2 , 25 mM Hepes-KOH, pH 7.4) and diluted to 8 -10 mg/ml. 750 l of membrane suspension were incubated at 37°C, with constant stirring, in a Perkin-Elmer LS50 spectrofluorimeter. Ca 2ϩ uptake was initiated by addition of Na 2 ATP (final concentration 2 mM) in the presence of 10 mM creatine phosphate, 30 units of creatine kinase, and 0.25 M Calcium Green-2 octapotassium salt (Molecular Probes, Eugene, OR). Ca 2ϩ uptake and release were measured by the decrease and increase, respectively, of fluorescence excited at 490 nm and monitored at 530 nm. At the end of each experiment, aliquots of EGTA and CaCl 2 were added for calibration, and free Ca 2ϩ calculated using a K d of 530 nM for the Ca 2ϩ -Calcium Green-2 complex. 45 Ca 2ϩ Uptake Measurements-Resuspended subcellular fractions (30 -80 g of protein) were incubated at 30°C, in 200 l of 100 mM KCl, 5 mM MgCl 2 , 5 mM sodium oxalate, 30 mM Hepes-NaOH, pH 7.2, in the presence of 1.3 mM CaCl 2 containing 2 Ci of 45 Ca 2ϩ (Amersham International, Amersham, UK), at a final specific activity 80 -100 Ci/mol. Ca 2ϩ uptake was initiated by addition of Na 2 ATP to a final concentration of 2.5 mM, and stopped after 20 min by rapid filtration through a 0.45-m pore nitrocellulose filter (Millipore HAWP), followed by rapid washing with 50 volumes of ice-cold 100 mM KCl, 5 mM CaCl 2 , 5 mM LaCl 3 , 30 mM Hepes, pH 7.2. Radioactivity retained on the filters was counted by liquid scintillation. Where indicated, thapsigargin was added to the membranes 5 min before addition of Ca 2ϩ .
RT-PCR-Total RNA was extracted from PC12 (10 7 cells), rat liver homogenate, and platelet-enriched rat blood (3 ml of rat blood in 50 ml of Hepes-buffered sucrose, 1 mM EGTA was centrifuged for 15 min at 1300 ϫ g at room temperature; cells remaining in the supernatant were pelleted and used as a platelet-enriched fraction) with 1 ml of TRIzol reagent (Life Technologies, Inc.). RNA was treated with RNase-free DNase prior to use. Complementary DNA was generated from 3 g of RNA by random hexamer priming in the presence of Moloney murine leukemia virus reverse transcriptase, and SERCA2b or SERCA3-specific messages were amplified by 40 cycles of PCR (1 min at 94°C, 30 s at 55°C or at 65°C, 1 min at 72°C) using Taq DNA polymerase and the primers (Primm, Milan) previously described by Bobe et al. (26); the SERCA3 primer was slightly modified to ensure 100% matching with the published sequence of the rat enzyme (27). Primers directed to the sequence of ␤-actin served as an internal control for the PCR reaction. Amplification products were separated by agarose gel electrophoresis and visualized by ethidium bromide fluorescence.

Ca 2ϩ Uptake Activity in the Post-nuclear Supernatant Preparations
Post-nuclear supernatant from PC12 cells, prepared after gentle homogenization by the cell cracker procedure (14), contains membrane-bound vesicles which can accumulate Ca 2ϩ from the medium in the presence of ATP, and release it in response to IP 3 (Fig. 1). Rapid Ca 2ϩ accumulation occurs even in the presence of inhibitors of mitochondrial respiration (oli-gomycin, antimycin, and azide). It is blocked by millimolar concentrations of vanadate, an inhibitor of all P-type transport ATPases, and by micromolar concentrations of thapsigargin, a specific inhibitor of SERCA-type calcium ATPases (data not shown). Ca 2ϩ accumulation ceases when the free concentration of Ca 2ϩ in the medium approaches 100 nM. This condition may represent a steady-state between uptake and release; addition of a saturating concentration of IP 3 results in rapid release of Ca 2ϩ , which is subsequently taken up again. Addition of thapsigargin results in a slower release of Ca 2ϩ from the vesicles. The subsequent addition of the Ca 2ϩ ionophore, ionomycin, induces the release of a further quantity of Ca 2ϩ (approximately 2-fold more than that released in response to thapsigargin), presumably originating from organelles insensitive to the blocker (see Fig. 1 and Ref. 13). In view of the well known changes in Ca 2ϩ distribution which occur on disruption of the cell's internal structure, we should point out that the relative sizes of the Ca 2ϩ pools present in this type of preparation do not necessarily correspond with those observed in intact cells (13).
Previous studies in a variety of cells and tissues, including platelets and hepatocytes, have indicated the presence of multiple SERCA isoforms (26 -29). To investigate, at the molecular level, which SERCA types might be responsible for the thapsigargin-sensitive Ca 2ϩ uptake illustrated in Fig. 1, we performed RT-PCR assays on total RNA, using primers specific for SERCA2b and for SERCA3. As shown in Fig. 2, amplification products of the expected size were detected, indicating the presence of both of these isoforms in PC12 cells. Positive results were also obtained using RNA extracted from platelet-enriched rat blood, and from rat liver (Fig. 2). Controls where reverse transcriptase was omitted were negative (not shown).

Velocity Gradient Fractionation
In order to investigate the nature of the Ca 2ϩ -accumulating vesicles, post-nuclear supernatant was subjected to various procedures of subcellular fractionation. Most experiments were carried out using velocity gradient centrifugation, in which particles are separated primarily by size rather than by density. Fig. 3 shows the typical distribution of protein and RNA in such a gradient while Fig. 4 illustrates the ultrastructure of three representative fractions from different zones of the gra-dient. As can be seen, the distribution of protein is bimodal; considerable amounts of sedimentable protein remained near the interface between the loading volume and the top of the gradient (fraction 1), and even larger amounts were recovered in fraction 2 (10 and 15%, respectively, of the total recovery of protein), together with RNA. Fraction 3 was still rich in protein, although less so than fraction 2. The protein content of the middle fractions (4 -8) was relatively low, on average only 30% of that of fraction 2, increasing progressively in the lower fractions (9 -11) of the gradient. The pellet (resuspended in an equal volume) was again rich in protein, and contained the bulk of the RNA. Electron microscopic analysis of thin sections cut through the pellets obtained from the gradient fractions revealed small vesicles scattered among polysomes in fraction 3; larger cisternae, part of which were covered with polysomes, together with a few recognizable mitochondria in fraction 8; packed membrane sheets together with both rough-surfaced and flat, smooth cisternae and mitochondria in fraction 10 (Fig. 4).

Ca 2ϩ Uptake
The velocity gradient fractions characterized in Fig. 3 were investigated for Ca 2ϩ uptake, using 45 Ca 2ϩ as a tracer, in the presence of oligomycin (6 g/ml), antimycin (6 g/ml), and azide (1 mM) to eliminate the possibility of mitochondrial Ca 2ϩ uptake. As can be seen from Fig. 5, ATP-dependent Ca 2ϩ uptake activity was concentrated in the upper region of the gradient, with the peak uptake per unit protein in fraction 2, and decreasing progressively over the next three fractions. In fractions 6 -11 the uptake was always marginal (Ͻ10% of that in fraction 2) and negligible in the pellet. The involvement of SERCAs in this activity was estimated by including 1 M thapsigargin in a set of samples analyzed in parallel to those without the inhibitor. Fig. 5 shows that most (Ͼ70%) of the Ca 2ϩ uptake was thapsigargin-sensitive, and thus most likely due to SERCA-type Ca 2ϩ pumps. Compared with this activity, the distribution of Ca 2ϩ uptake insensitive to thapsigargin was shifted to the right, peaking in fraction 3 rather than in fraction 2. From fraction 6, however, this thapsigargin-insensitive activity was low or negligible.

Protein Phosphorylation
In order to identify the Ca 2ϩ -dependent pumps responsible for the accumulation of Ca 2ϩ revealed in the preceding section, selected velocity gradient fractions were incubated with [␥-32 P]ATP under conditions where only P-type ATPases are likely to be phosphorylated, i.e. pH 6.8, no Mg 2ϩ , 20-40 nM ATP (24), separated by acidic SDS-PAGE, and blotted onto nitrocellulose at pH 6.8. Phosphorylation was carried out in the presence of 50 M Ca 2ϩ alone, in the presence of excess EGTA, and in the presence of 1 M thapsigargin. The inhibitor was always preincubated with the samples before addition of Ca 2ϩ , since the presence of Ca 2ϩ greatly slows the rate of binding of thapsigargin by SERCA (30).
The central lanes in Fig. 6 show that, in velocity gradient fractions 2 and 3, strong Ca 2ϩ -dependent, thapsigargin-sensitive labeling was revealed in the 97-116-kDa region, i.e. the anticipated position for SERCA-type Ca 2ϩ -ATPases. Some labeling was also observed when fractions 9, 10, and 11 were used, but was much weaker than that observed in the upper fractions (not shown), consistent with the distribution of 45 Ca 2ϩ accumulation (Fig. 5). In several preparations, the labeling was clearly due to more than one phosphorylated species (compare Fig. 6, A and B). Limited digestion with trypsin prior to phosphorylation resulted in the disappearance of labeling in the 97-116-kDa region and the appearance of a 55-kDa phosphorylated fragment, as previously reported for SERCA2b in platelets (28). In some cases, an additional phosphorylated fragment was observed at about 46 kDa (right-hand lanes in Fig. 6, A and B).
In order to identify the phosphoproteins revealed by the above experiments, the radiolabeled blots were immunostained using a widely employed pair of anti-SERCA antibodies. The left-hand lane of Fig. 6A show that a specific antibody against the C-terminal peptide of SERCA2b (16) decorated a band with an apparent molecular mass of 110 kDa, which coincided with the center of the strong 32 P labeling in that preparation. A phosphorylated protein of lower molecular mass (100 kDa, Fig.  6B) was not recognized by this antibody, but by the anti-SERCA1 monoclonal antibody Y1F4 (16), reported to recognize an epitope largely conserved in all members of the SERCA family.

Western Blots
Velocity Gradient Fractions-This series of experiments was carried out to establish the relative distributions among the velocity gradient fractions of various markers of the ER, involved directly or indirectly with the uptake and release of Ca 2ϩ (Figs. 7 and 8). Markers of another endomembrane system, the Golgi complex and the trans-Golgi network (GC and TGN), also present in post-nuclear supernatant and possibly involved in cellular Ca 2ϩ homeostasis (31-33), were investi-gated in parallel (Fig. 8, panel c). The distributions of the two SERCA types, as revealed by the anti-C-terminal antibody and the Y1F4 antibody, as well as that of the IP 3 R, are shown in panel a of Fig. 8 on an equal-protein basis, and for comparison in panel aЈ in terms of the total amounts recovered. In good agreement with the results on 45 Ca 2ϩ uptake and phosphoenzyme formation described in the preceding sections, the two SERCAs were both detected in the upper fractions, with the peak in fraction 2. The IP 3 receptor, while also localized to the upper fractions of the gradient, was shifted with respect to the SERCAs, its relative distribution peaking in fraction 4, with appreciable amounts still detected in fractions 5-7. Even more Panel c, sialyltransferase (छ), mannosidase II (ࡗ), TGN38 (å). Distributions are given in terms of equal protein loading for panels a, b, and c, and in terms of total amounts recovered in panel aЈ. Results are presented as the mean Ϯ S.E. of data from three to five independent experiments (SERCA2b, Y1F4, IP3R, n ϭ 5; calnexin, protein disulfide isomerase, n ϭ 4; calreticulin, sialyltransferase, mannosidase II, TGN38, n ϭ 3). different were the relative distributions of ER proteins involved in the storage and lumenal functions of Ca 2ϩ , rather than in its uptake and release (Fig. 8, panel b). The lumenal ER chaperonins, protein disulfide isomerase and calreticulin (the latter also being the major ER Ca 2ϩ storage protein in PC12 cells) were poorly represented in the top, SERCA-rich fractions, showing uni-modal distributions centered on fraction 6. In contrast, the ER membrane protein calnexin, also a Ca 2ϩ -binding chaperonin, was more evenly distributed, at least across the first 9 fractions. With respect to the GC/TGN makers (Fig. 8,  panel c), the medial/trans-Golgi enzyme mannosidase II (21) was found principally in the lower fractions and the pellet, while markers of trans-Golgi compartments and the TGN, sialyltransferase, and TGN38, respectively (22)(23)34), were found almost exclusively in the upper half of the gradients.
Equilibrium Gradient Fractions-The gradient centrifugation results presented so far were all obtained using velocity gradients. To investigate further the composition of the PC12 post-nuclear supernatants, we also studied the distribution of ER markers in identical gradients, centrifuged for much longer times in order to reach equilibrium-density conditions. As can be seen in Fig. 9, SERCA2b (panel b) and IP 3 R (panel c) were recovered in the lower fractions; the bulk of the Ca 2ϩ pump was recovered in fractions 8 and 9, while the channel had a similar but broader distribution, with a significant amount found in the pump-poor fraction 10. Calnexin (panel d), in contrast, still distributed widely across the gradient, with appreciable values in fractions 4 -11.

DISCUSSION
The basic question underlying the present work concerns the possible heterogeneity of the ER with respect to one of its fundamental functions, the rapid uptake and release of Ca 2ϩ . During the last decade, this issue has been investigated in a variety of cell types with the primary aim of identifying specialized structures which, like the sarcoplasmic reticulum of muscle cells, could play a key role in cellular Ca 2ϩ homeostasis. High-resolution immunocytochemistry has provided important evidence in this direction, although because of the exacting nature of the technique in terms of antibody affinity and specificity, good results have only been obtained for a few cell types particularly rich in IP 3 receptors, such as the Purkinje neurones and the smooth muscle fibers of the vas deferens (5,6). More recently, the discovery of spatially discrete Ca 2ϩ release events, sparks, pacemakers, hot spots etc., which may be important in their own right, or as initiators of oscillations and waves, has reinforced the idea of ER heterogeneity (1,35,36). On the biochemical level, however, results have been largely inconclusive. For the present study we have chosen the PC12 cell line, frequently employed as a neurosecretory, and also as a neuronal model. Although Ca 2ϩ homeostasis in these cells has been investigated in depth by our laboratory, over a number of years (1,12,13), discrete sparks and oscillations have never been observed, implying that ER heterogeneity in PC12 cells may correspond to a general principle of cellular physiology, rather than a cell type-specific function.
Our experimental approach on this occasion was based primarily on subcellular fractionation of post-nuclear supernatant, rather than immunocytochemistry, in part because the anti-SERCA antibodies available to us were found to be unsuitable for this demanding technique. In terms of methodology, three aspects of our study deserve mention. The first is that cell lysis was carried out by the cell cracker technique in order to keep organelle damage to a minimum, rather than by the classical procedures with Potter or Dounce homogenizers. Second, most of the separations were carried out by rapid, velocity gradient centrifugation, utilizing the differences in sedimentation rate between fragments of different sizes rather than the small differences in their intrinsic densities. In some experiments centrifugation was prolonged until the equilibrium condition was reached, in order to compare our velocity gradient results with those obtainable by the more widely employed technique of isopycnic density gradient centrifugation. Finally, our gradients were analyzed not only with functional assays (Ca 2ϩ uptake and phosphoenzyme formation) in which rapidity of fraction isolation could have a bearing, but also by Western blotting. The latter approach on the one hand obviates the risks of major artifacts due to inactivation of proteins during isolation, while on the other it can yield quantitative comparisons on the molecular make-up of the various fractions, provided that the antibodies employed are of high specificity and appropriate affinity.
In terms of specificity, our results with the two anti-SERCA antibodies we have employed requires detailed discussion, given the central role of the Ca 2ϩ pumps in our study. Our results with the anti-SERCA2b antibody (17), raised against the unique C-terminal portion of the enzyme, appear entirely convincing, in terms of apparent size on SDS-PAGE, Ca 2ϩdependent thapsigargin-sensitive phosphorylation, and tryptic digestion pattern (55-kDa phosphorylated fragment), and correlate well with the distribution of Ca 2ϩ uptake in the upper fractions of the velocity gradients. In contrast, the results obtained with the Y1F4 monoclonal, raised against the skeletal muscle sarcoplasmic reticulum form of the enzyme, i.e. SERCA1 (16), appear more problematical. The epitope recog- nized by this antibody lies in the middle of the protein sequence (37), and is conserved, with only one amino acid change (from NKMFVK to SKMFVK) in rat SERCA2a, SERCA2b, and SERCA3. In blots from NIH 3T3 cells (38) and from bovine chromaffin cells (39), Y1F4 had already been shown to label two bands, at apparent molecular masses of 100 and 116 kDa. In our velocity gradient fractions, these two peptides are cleanly separated (Fig. 7), with the smaller protein co-distributing with SERCA2b, and the 116-kDa form present only in the lower fractions, which are almost completely incompetent for Ca 2ϩ uptake (see Fig. 5) and deficient in Ca 2ϩ -dependent phosphoprotein formation. For this reason, we exclude that this peptide is SERCA2b, as was previously suggested (1,38). The reason why the SERCA2b in the upper fractions is not recognized by Y1F4 is not clear. The smaller, 100-kDa band labeled by the monoclonal antibody, on the other hand, was apparently phosphorylated in a Ca 2ϩ -and thapsigargin-dependent fashion, and could correspond to SERCA3, the expression of which in PC12 cells was revealed by RT-PCR. Although we have not observed the SERCA3-derived 80-kDa tryptic fragment described by Wuytack et al. (28), we note that this fragment is reportedly difficult to visualize (40). Interestingly, in preparations were the 100-kDa phosphoprotein is well represented, tryptic digestion yields, in addition to a 55-kDa fragment, a novel 46-kDa phosphorylated fragment (see Fig. 6B). Although the assignment of the Y1F4-reactive 100-kDa band to SERCA3 is not entirely sure, it will not be further discussed here.
Taken together, the results we have obtained with PC12 cells lend support to the concept of ER heterogeneity with respect to uptake and release of Ca 2ϩ . Among the velocity gradient fractions, only the upper five were clearly competent for thapsigargin-sensitive Ca 2ϩ uptake, and the same fractions exhibited SERCA immunoreactivity. This co-distribution suggests that in PC12 cells, contrary to what has been hypothesized for platelets, for example (41), the two SERCAs do not segregate in different structures but remain co-localized. From a biochemical point of view, the kinetic properties of the two pumps are not much different from each other (42), and the functional significance of their co-localization or segregation remains unclear.
In the upper, slowly moving region of the velocity gradients, the 45 Ca 2ϩ experiments revealed the distribution of an uptake activity insensitive to thapsigargin, which was shifted one fraction downwards with respect to the SERCA immunoreactivity. Recently, Caspersen and Treiman (43) also described a thapsigargin-insensitive component of Ca 2ϩ uptake in fractions from bovine chromaffin cells, which partially segregated from the thapsigargin-sensitive component. At the moment this activity remains unexplained. In our experiments, it is probably not due to the plasma membrane Ca 2ϩ pump, since it is unaffected by 10 M vanadate (data not shown). Given that its distribution overlaps with at least some markers of the GC/TGN (see Fig. 8) it is possible that it is linked to that endomembrane system, which has been repeatedly suggested to accumulate Ca 2ϩ (31)(32)(33).
The release limb of the rapidly exchanging Ca 2ϩ store in the ER is constituted by channels; in PC12 these are primarily represented by the IP 3 receptor (12). The distribution of IP 3 receptor protein only partially overlaps that of the SERCAs. Both in velocity and equilibrium gradients a considerable proportion (50 and 30%, respectively) is recovered in fractions relatively poor in Ca 2ϩ pumps. We conclude that among the various regions of the ER of PC12 cell, the ratio between pumps and channels varies considerably, and that the balance between uptake and release activities is likely to follow suit. The major dislocation, however, was found between the distribu-tions of the SERCAs (and, to a lesser extent the IP 3 R), and the other ER markers investigated: calreticulin, protein disulfide isomerase, and calnexin. These markers are now recognized as chaperonins, i.e. participants in the processing and folding of newly synthesized proteins in the lumen of the ER, where free [Ca 2ϩ ] is of the order of 1 mM (44). In non-muscle cells, calreticulin and calnexin are also believed to be the major components of the Ca 2ϩ -buffering system of the ER lumen. The distributions of protein disulfide isomerase and calreticulin in the velocity gradient fractions was unexpected inasmuch as they were found to be very low in the top, SERCA-and IP 3 receptorrich fractions, and enriched in the middle fractions, which were relatively poor in pumps and channels. This suggests an at least partial dissociation of the latter components from those involved in lumenal Ca 2ϩ storage. It should be emphasized, however, that the observed distributions of soluble lumenal proteins like protein disulfide isomerase and calreticulin may be subject to artifacts because of leakage from the ER-derived microsomes during cell cracking and gradient centrifugation. Calnexin, however, is a membrane protein, and its association with microsomes is necessarily preserved. Compared with the SERCAs, this membrane protein was found to be more uniformly distributed in both velocity and equilibrium gradients, where about 70% is recovered in SERCA-poor fractions. The only possible interpretation is that in PC12 cells the areas of ER membrane specialized for rapid uptake and release of Ca 2ϩ are a minority of the total, whereas Ca 2ϩ -binding proteins associated with other Ca 2ϩ -dependent activities appear to be distributed throughout the entire ER system.
In conclusion, our subcellular fractionation approach has revealed that a high degree of ER heterogeneity exists not only in IP 3 receptor-rich cells such as Purkinje neurones and smooth muscle fibers (5, 6), but also in neurosecretory PC12 cells and presumably in other cell types as well. This result strongly implies the existence in vivo of multiple domains which, although certainly part of the ER by virtue of possession of well known markers, participate in Ca 2ϩ homeostasis to different extents and in different manners. This conclusion brings us to the question of the nature of the ER structures most immediately involved in Ca 2ϩ homeostasis. A widely held, although rarely formalized belief (discussed in Ref. 1), is that the smooth ER is responsible for this function. This identification appears unlikely to us because, in our equilibrium gradients, essentially all of the SERCA and IP 3 R were recovered in relatively heavy fractions, whereas a large proportion of the calnexin remained in lighter, presumably smooth-surfaced fractions. The simplest way to account for this difference in distribution is that the SERCAs and the IP 3 R are enriched on relatively small vesicles (and thus migrate slowly on the velocity gradients) which are denser than the average smooth microsomes, either because they are rough-surfaced or because of a particular protein/lipid composition. A similar inference was drawn from isopycnic gradient analysis of canine pancreatic microsomes (45,46), where both SERCA and IP 3 R were predominantly associated with ribosome-bearing fractions, although as in our study, the distribution of IP 3 R was wider than that of SERCA, i.e. IP 3 R was also associated with SERCA-poor membranes (46). Finally, we would like to highlight one aspect of our results for wider consideration; within intact cells, the ER is composed of a continuous and dynamic network of tubules and cisternae, thus it is likely that Ca 2ϩ pumped in one area will diffuse also to the areas of the reticulum which have few or no Ca 2ϩ pumps. The differential distribution of the latter, as well as that of the channels, on the one hand points to strategic cytoplasmic sites where [Ca 2ϩ ] i changes could be executed more rapidly and with greater amplitude than elsewhere; on the other hand, it suggests that many Ca 2ϩ -dependent activities could proceed undisturbed in large areas of the ER which are relatively insulated from short-term fluctuations in Ca 2ϩ concentration.