Mass Spectrometry of Cardiac Calsequestrin Characterizes Microheterogeneity Unique to Heart and Indicative of Complex Intracellular Transit*

Cardiac calsequestrin concentrates in junctional sarcoplasmic reticulum in heart and skeletal muscle cells by an undefined mechanism. During transit through the secretory pathway, it undergoes an as yet uncharacterized glycosylation and acquires phosphate on CK2-sensitive sites. In this study, we have shown that active calsequestrin phosphorylation occurred in nonmuscle cells as well as muscle cells, reflecting a widespread cellular process. To characterize this post-translational modification and resolve individual molecular mass species, we subjected purified calsequestrin to mass spectrometry using electrospray ionization. Mass spectra showed that calsequestrin glycan structure in nonmuscle cells was that expected for an endoplasmic reticulum-localized glycoprotein and showed that each glycoform existed as four mass peaks representing molecules that also had 0–3 phosphorylation sites occupied. In heart, mass peaks indicated carbohydrate modifications characteristic of transit through Golgi compartments. Phosphorylation did not occur on every glycoform present, suggesting a far more complex movement of calsequestrin molecules in heart cells. Significant amounts of calsequestrin contained glycan with only a single mannose residue, indicative of a novel post-endoplasmic reticulum mannosidase activity. In conclusion, glyco- and phosphoforms of calsequestrin chart a complex cellular transport in heart, with calsequestrin following trafficking pathways not present or not accessible to the same molecules in nonmuscle.

Cardiac calsequestrin concentrates in junctional sarcoplasmic reticulum in heart and skeletal muscle cells by an undefined mechanism. During transit through the secretory pathway, it undergoes an as yet uncharacterized glycosylation and acquires phosphate on CK2-sensitive sites. In this study, we have shown that active calsequestrin phosphorylation occurred in nonmuscle cells as well as muscle cells, reflecting a widespread cellular process. To characterize this post-translational modification and resolve individual molecular mass species, we subjected purified calsequestrin to mass spectrometry using electrospray ionization. Mass spectra showed that calsequestrin glycan structure in nonmuscle cells was that expected for an endoplasmic reticulum-localized glycoprotein and showed that each glycoform existed as four mass peaks representing molecules that also had 0 -3 phosphorylation sites occupied. In heart, mass peaks indicated carbohydrate modifications characteristic of transit through Golgi compartments. Phosphorylation did not occur on every glycoform present, suggesting a far more complex movement of calsequestrin molecules in heart cells. Significant amounts of calsequestrin contained glycan with only a single mannose residue, indicative of a novel post-endoplasmic reticulum mannosidase activity. In conclusion, glyco-and phosphoforms of calsequestrin chart a complex cellular transport in heart, with calsequestrin following trafficking pathways not present or not accessible to the same molecules in nonmuscle.
Both CSQ isoforms are substrates for protein kinase CK2 in vitro (16,23), and phosphorylation sites have been determined for canine cardiac and rabbit fast-twitch isoforms (16). The fast-twitch isoform is phosphorylated on Thr 373 , whereas the cardiac isoform is phosphorylated on a cluster of three serine residues that reside in the cardiac-specific tail (Ser 378,382,386 ). These three serine residues were previously shown to be partially phosphorylated in the purified cardiac isoform, whereas no phosphate appeared in the rabbit fast-twitch isoform (16). A function for CSQ phosphorylation by CK2 has not been determined; however, a mechanism for sorting of resident ER and Golgi proteins by CK2 phosphorylation on cytosolic sites has been characterized (24 -30).
In this study, we report definitive structural findings for CSQ revealed by mass spectrometry, reflecting its cellular transport in heart and muscle cells. CSQ glycosylation and phosphorylation, although highly similar in nonmuscle cells as diverse as human embryonic kidney (HEK) and insect Sf21 cells, were distinctly different in muscle cells, reflecting a pathway in muscle not present or not accessible in nonmuscle, one in which phosphorylation on CK2 sites exists in a complex compartmental relationship with glycan-modifying reactions.
Animals-Canine tissues were obtained from three separate mongrel dogs under anesthesia. Animals were obtained from authorized suppliers and maintained in accordance with National Institutes of Health guidelines. The Division of Laboratory Animal Resources of Wayne State University is fully equipped and licensed by the appropriate agencies.
Construction of Recombinant CSQ Viruses-Replication-deficient adenoviruses containing wild-type canine cardiac CSQ cDNA (Ad.CSQ) or a triple point S378A,S382A,S386A mutant (Ad.nPP) that removed CK2 phosphorylation sites (16) were amplified from the gt10 clone IC3A (14) by PCR (31) and primers containing restriction sites for directional cloning. The forward primer contained 63 bp of 5Ј-untranslated sequence, the reverse primer, 5 bp of 3Ј-untranslated sequence. The * This work was funded by Grant HL62586 from the NHLBI, National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. mutating reverse primer (67-mer) contained 3 single base changes necessary for Ser to Ala conversion. PCR products were cloned into pBluescript (Stratagene), sequenced by the dideoxy method (32), and then subcloned into a transfer plasmid (pAbl.CMV) containing the cytomegolovirus major immediate-early promoter. Recombinant viruses were identified by restriction endonuclease digestion and isolated by CsCl density centrifugation, with titers of viral stocks determined by plaque assay using HEK 293 cells (33). Plaque-purified clones were expanded in HEK 293 cells, and virus preparations were purified by CsCl density centrifugation. Titers of viral stocks were determined by plaque assay using HEK 293 cells. Recombinant baculovirus encoding canine cardiac was prepared as previously described (34).
Infection of Cells with Viruses-HEK 293 cells were infected with recombinant adenoviruses at a multiplicity of infection of ϳ1.0 for 4 h in Dulbecco's modified Eagle's medium without serum, then incubated under our normal conditions for 48 -72 h. Sf21 cells were infected with recombinant baculovirus for 48 h in ESF 921 medium. All recombinant adenovector work was carried out in accordance with National Institutes of Health guidelines for research involving recombinant DNA molecules and the policies of the Wayne State University Office of Environmental Health and Safety.
Purification of Recombinant Cardiac CSQ from Nonmuscle Cells-Purification of CSQ from cultures of HEK 293 and Sf21 cells was carried out 2 d post-infection. Cell pellets were resuspended at 1 mg/ml in buffer A containing 20 mM MOPS, pH 7.5, 250 mM NaCl, 1% CHAPS, 0.5 mM EGTA, and 0.5% of a protease inhibitor mixture (Sigma). 10 mM NaF and 10 mM ␤-glycerophosphate were included in the buffer to inhibit protein phosphatases. Extracts were centrifuged at 50,000 ϫ g ϫ 20 min, bound to DEAE-Sephacel (Amersham Biosciences), and then washed extensively in buffer B (buffer A without CHAPS) until detergent was removed. CSQ was eluted in buffer B with 750 mM NaCl. Eluate was loaded onto phenyl-agarose and purification carried out as previously described (35) with calsequestrin eluted in purified form by adding 20 mM CaCl 2 to the elution buffer. SDS-PAGE was carried out according to Laemmli (36). Protein assays were carried out using a Lowry protocol (37).
Purification of Native CSQ from Canine Cardiac and Skeletal Muscle Tissue-Native CSQs from canine left ventricles and hind leg muscles were also purified as previously described (35), with minor modifications. Homogenization was carried out in a buffer containing 20 mM MOPS, pH 7.5, 50 mM NaCl, 0.1 mM dithiothreitol, 0.5% protease inhibitor mixture (Sigma), and phosphatase inhibitors as described above. Membrane vesicles were isolated from homogenates following removal of only a 500 ϫ g pellet to recover a larger portion of CSQcontaining membranes.
Dephosphorylation and Phosphorylation of Purified CSQ-CSQ phosphorylation and dephosphorylation were carried out as previously described (16). For phosphorylation, 0.2 g of calsequestrin was incubated in 20 mM MOPS, pH 7.5, 150 mM NaCl, 0.5 mM EGTA, 10 mM MgCl 2 , 0.1% Triton X-100, 20 M [␥-32 P]ATP, and 10 ng of purified CK2 (Promega). When samples were pretreated with acid phosphatase, 1 g of calsequestrin was incubated for 20 min in 30 l of 30 mM MES buffer, pH 5.8, 0.1 mM EGTA with or without the addition of 0.05 units of acid phosphatase (Sigma), determined to be the minimal amount capable of removing all radioactivity resulting from CK2. Phosphatase-or controltreated samples were diluted 10-fold to neutral pH in the presence of sufficient kinase to label calsequestrin to high stoichiometry.
Mass Spectrometry-For mass spectrometry, salts and buffers were removed from calsequestrin samples by repeated centrifugations through a Centricon-30 concentrator (Amicon) after the addition of 20 mM EGTA to phenyl-agarose-purified samples to chelate Ca 2ϩ . Electrospray ionization mass spectrometry was carried out at the Biotechnology Resource Facility of the HHMI Biopolymer Facility/W. M. Keck Foundation, Yale University, using a Q-ToF mass spectrometer (Micromass, Altrincham, UK). Prior to analysis, samples were desalted using C-4 ZipTips (Millipore Corp., Bedford, MA). The eluted samples in 50% acetonitrile/0.1% formic acid were analyzed using the nanospray technique in positive ion mode. Masses were calculated using Q-Tof Mass-Lynx software. Spectra were calibrated using either sodium iodide or the fragment ions from the MS/MS spectrum of (Glu)fibrinogen (Sigma).

RESULTS
Cardiac CSQ purified from dog heart contains significant levels of phosphate (Ͼ1 mol/mol) on a carboxyl-terminal cluster of serine residues that are in vitro substrates of protein kinase CK2. To determine whether this reaction is unique to heart cells, we analyzed the purified CSQ from HEK and Sf21 cells, following heterologous expression in nonmuscle cells using a recombinant adenovirus or baculovirus.
Cardiac CSQ was phosphorylated by purified CK2 either under control conditions or following treatment with acid phosphatase to remove endogenous phosphate. Pretreatment with phosphatase led to increases in 32 P incorporation into CSQ, showing that CK2 sites in CSQ existed predominantly in a phosphorylated form in nonmuscle cells. Levels of endogenous CSQ phosphorylation in ER were comparable with that observed in cardiac SR (Fig. 1A). Phosphatase treatments increased subsequent CK2 phosphorylation by 2-4-fold, depending upon the cell type.
CSQ purification from canine skeletal muscle led to isolation of both CSQ isoforms, with the fast-twitch form (63 kDa) being the predominant one, as previously reported (38). Application of our phosphatase/CK2 kinase assay to the skeletal muscle preparation showed that, as in heart tissue, cardiac CSQ (55 kDa) was present mainly as the phosphoprotein (Fig. 1B), whereas the canine fast-twitch isoform did not accumulate in a phosphorylated form (Fig. 1B, upper band). To try and resolve individual forms of phosphorylated CSQ, we subjected purified CSQ to mass spectrometry using electrospray ionization. Mass spectrometry of purified wild-type CSQ from HEK cells showed a series of mass peaks differing by about 81 Da, consistent with a mixture of molecules differing by a single phosphate moiety ( Fig. 2A). In contrast, the phosphorylation site mutant nPP showed a series of mass peaks that differed by ϳ162 Da, consistent with a mixture of glycoforms differing by a single mannose (Fig. 2B). Mass peaks for CSQ versus the nPP mutant also reflected the ϳ47 Da difference between three serines and three alanines. The mass peaks for wild-type CSQ and the nPP mutant in Fig. 2 are offset by 47 Da to align equivalent mass peaks.
Comparison of the aligned spectra for wild-type CSQ and the nPP mutant from HEK cells (compare Fig. 1, A and B) shows that wild-type CSQ consists of two glycoforms (Man8-and Man9GlcNAc) along with three additional phosphoforms. For example, the mass expected for unmodified cardiac CSQ tran- script is 45,269 Da, and CSQ containing a single high mannose oligosaccharide is 47,135 Da (Man9GlcNAc2 ϭ 1865 Da). As predicted, this exact mass peak (47,135 Da) appears in the CSQ spectrum along with 3 peaks of higher mass at 81-Da intervals, corresponding to protein molecules (with the oligosaccharide Man9GlcNAc2) in unphosphorylated, singly, doubly, and triply phosphorylated states. All CSQ phosphate resided on the cardiac-specific carboxyl-terminal serine cluster Ser 378,382,386 , because the latter three mass forms were absent from the nPP mutant.
CSQ molecules in which the oligosaccharide has been trimmed to Man8GlcNAc2 (and contains no phosphate) have an expected mass of ϳ46,974 Da, in agreement with the first mass peak observed in Fig. 1A. Peaks corresponding to phosphorylated Man8GlcNAc2 are obscured by the fact that the difference in mass due to a single mannose residue (162 Da) is very nearly the same as the mass change from two phosphates (2 ϫ 81 Da). Thus, for example, the peak of 47,135 Da corresponds both to the Man8GlcNAc2 glycoform with two phosphates and Man9GlcNAc2 with no phosphates. Therefore, wild-type CSQ from HEK cells existed in only two glycoforms, and both glycoforms appeared to exist in each of four phosphoforms (0 -3 sites occupied, summarized in Fig. 4). The nPP form of CSQ existed as five glycoforms, indicating more extensive glycan processing than occurred for the wild-type protein. The mass spectrum for wild-type CSQ overexpressed in Sf21 insect cells was very similar to that from HEK cells, yielding a spectrum containing the identical six peaks (data not shown, but see Fig. 4).
To validate our interpretation of the spectra for CSQ and nPP in HEK cells, we analyzed CSQ and nPP from HEK cells treated with 0.5 g/ml tunicamycin over the entire time course of overexpression, a treatment known to prevent N-linked glycosylation (39). The mass spectrum for CSQ treated with tunicamycin yielded a mass peak corresponding to the amino acid backbone (deduced mass ϭ 45,269 Da) and three higher mass peaks separated by roughly 80 Da, corresponding to molecules The putative identification of glyco-and phosphoforms for the cardiac isoform is indicated above, and glycoforms that do not show phosphate distributed on the serine cluster (0 -3 sites occupied) are indicated by arrows. The mass peak of 46,398 Da (asterisk) probably contains components of both Man3 (plus 2 phosphates) and Man4 glycoforms (see Fig. 4). having one, two, and three phosphates (Fig. 2C). The vast majority of CSQ molecules contained two or three phosphates, showing that glycosylation per se was not necessary for phosphorylation to occur. If CK2 phosphorylation sites were also removed, then all of the CSQ was synthesized as a single mass species of 45,269 (Fig. 2D), which is the deduced mass of the expressed canine CSQ clone without any covalent modifications.
Mass spectrometry of native CSQ purified from canine ventricular tissue gave a mass spectrum that was more complex than that from HEK and Sf21 cells (Fig. 3, upper panel). As for HEK cells, tissue CSQ showed a succession of mass peaks separated by 81 Da, representing a similar combination of glyco-and phosphoforms. The masses of native CSQ molecules, however, were lower by 70 Da than the recombinant protein from HEK and Sf21 cells. This difference may be because of a polymorphism within the animal source of our native CSQ sample, compared with the protein encoded by the well characterized cardiac CSQ cDNA IC3A (14) used in both virus constructs (and from which the deduced mass was calculated). Thus, for example, assuming that the lowest peak seen for tissue CSQ corresponded to a glycoform with no phosphate, the peak of 45,916 Da represents the structure Man1GlcNAc2 but with an additional 79 Da moiety attached to the protein backbone. Therefore, it does not affect any of the differences in mass observed for CSQ isoforms. The nature of this mass discrepancy was not further investigated.
Compared with nonmuscle, molecules of CSQ from heart tissue exhibited a greater degree of mannose trimming and a larger range of glycoforms (Man1 through Man6), and glycoforms were less uniformly phosphorylated. For example, for one of the highest mass peaks observed (46,559 Da, Man5), no protein peak was found having a mass 81 Da higher. A much smaller peak at 162 Da higher, therefore, represents an additional mannose (Man6), which also lacked additional phosphate-containing mass peaks. On the other hand, the lowest glycoform masses (45,916 and 46,238 Da, Man1,3) appeared to exist in all 4 phosphoforms (0 -3 sites occupied). Again, glycoforms and phosphoforms are schematically shown in Fig. 4. The mass peak at 46,398 Da represents Man3 plus 2 phosphates but may also contain molecules of Man4 with no phosphate, in which case it appears that these molecules are also not phosphorylated in vivo, because there is no peak corresponding to Man4 plus 3 phosphates. In summary, Man1,3 are partially to fully phosphorylated in vivo, whereas Man4 -6 remain unphosphorylated.
To compare the mass spectrum for the fast-twitch isoform, we purified CSQ from canine hind leg muscle by a procedure identical to that for heart tissue. The mass spectrum for the skeletal muscle protein (Fig. 3, lower panel) consisted of only two protein peaks, separated by 323 Da, a difference of two mannose residues. This pattern appears to be maintained in the cardiac isoform as well (compare arrowheads in both panels). Based upon a deduced mass of 42,216 for the canine fast-twitch skeletal muscle isoform, 2 the 2 peaks in the mass spectrum of 42,810 and 43,133 Da are likely to represent the Man1GlcNAc2 and Man3GlcNAc2 forms but with each population of molecules leaving an additional 25 Da unaccounted for (again most likely representing a polymorphism within the sample studied here, compared with that from which the cDNA clone was derived). Most notable in the fast-twitch isoform was the absence of phosphate in the two observed glycoforms, a finding in agreement with our in vitro phosphorylation data (Fig. 1B).
In Fig. 4, we have summarized the data from the mass spectra, providing a schematic view of the distributions of CSQ glyco-and phosphoforms. In cases where mass peaks probably contain contributions from more than one glycoform, we have approximated relative contributions, as detailed in the figure legend.

DISCUSSION
Cellular trafficking of CSQ in muscle and nonmuscle cells is readily visualized by mass spectrometric analysis of intact CSQ molecules, which is unusually revealing given the nature of CSQ as a soluble ER/SR glycoprotein with a single oligosaccharide. Resolution of CSQ microheterogeneity yields important insights into its intracellular trafficking by charting the actions of intracellular mannosidases and yields insights into CSQ phosphorylation by revealing differences in the degree of phosphorylation among many CSQ glycoforms.
Phosphorylation of CSQ Is an Active Process in All Cells-Cardiac CSQ contained between 1.5 and 2.0 mol P i /mol protein on CK2 sites, whether biosynthesized in canine heart or in nonmuscle cells. Phosphorylation of cardiac CSQ varied from non-phosphorylated to fully (3 mol/mol) phosphorylated. Interestingly, nonmuscle cells contained sufficient calsequestrin kinase to produce high steady levels of the phosphorylated protein, even upon overexpression to levels comparable with that 2 L. Jones, personal communication.

FIG. 4. Schematic summary of CSQ glycoforms and phosphoforms in SR and ER.
Results of mass spectrometry of native dog heart, native dog fast-twitch skeletal muscle, cardiac CSQ (wild-type and nPP mutant) from human HEK cells, and cardiac CSQ from Sf21 insect cells. Amounts of individual glycoforms were derived from relative peak heights and are expressed as number of mannose residues in the oligosaccharide chain. For the two extremes of oligosaccharide trimming (Man1GlcNAc2 and Man9GlcNAc2), the structures are drawn (squares) N-acetyl glucosamine and (circles) one to nine mannose residues. Relative levels of singly, doubly, and triply phosphorylated CSQs (dark gray bars) are indicated next to each non-phosphorylated glycoform (black bars). In some cases, contributions from more than one isoform are predicted to comprise a single peak in the mass spectrum, and individual components are approximated as follows. a, mass peak 46,398 Da in heart CSQ (Fig. 3, upper panel) was assumed to contain Man3 (plus 2 phosphates) and Man4 glycoforms (indicated by connecting bar) in a 2 to 1 ratio. b, mass peak 47,135 Da in HEK CSQ (Fig. 2, panel A) was assumed to contain equal amounts of two isoforms as shown. c, mass peak 47,295 Da (same spectrum) was assumed to contain Man9 and 8 glycoforms in a 2 to 1 ratio. Individual glycoform contributions from overlapping mass peaks for Sf21 CSQ were approximated the same as for HEK CSQ. of heart cells. Mass spectrometry showed similar patterns of phosphorylation on similar glycoforms in human and insect cells, indicating an extraordinary conservation of protein processing and suggesting that phosphorylation of lumenal ER/SR proteins may be a ubiquitous cellular reaction.
Although the mechanism and precise cellular compartmentation remain uncertain, it appears that CSQ phosphorylation involves CK2 or a CK2-like protein kinase which co-localizes with CSQ, if only transiently.
CSQ Glycan Processing in ER and SR-In all cells, CSQ glycosylation occurred on only one of its two potential N-glycosylation sites, likely Asn 316 because this site is highly conserved among species and isoforms (40). Prevention of glycosylation with tunicamycin did not prevent phosphorylation on CK2 sites, and besides N-glycosylation and phosphorylation on Ser 378,382,386 no other modifications of CSQ occurred.
Nonetheless, glycan processing was different in muscle and nonmuscle cells. The predominant glycoforms of cardiac CSQ in nonmuscle cells were Man9GlcNAc2 and Man8GlcNAc2, consistent with processing by ER ␣ 1,2-mannosidase and indicative of a protein that does not leave the ER (41)(42)(43). The mechanism for ER retention of CSQ is unknown, as has been previously been discussed (44,45) and investigated (46,47). Although many resident ER proteins contain the carboxyl-terminal peptide retrieval signal -KDEL (48), this sequence is absent from CSQ and other muscle-specific resident SR/ER proteins (44,45).
In contrast to Man8 and 9 forms in nonmuscle ER, maturation of cardiac CSQ in heart led predominantly to Man1, 3, and 4 forms of the glycan with most present as Man1 or 3, indicative of transit through the Golgi and the actions of post-ER mannosidases (41)(42)(43). These data are in agreement with previous biochemical analyses by Jorgensen et al. (21), who reported that fast-twitch CSQ exists predominantly as the Man3GlcNAc2 glycoform. Golgi transit of CSQ was also reported by Thomas et al. (22), who showed that CSQ is transported to terminal cisternae of the SR junction in clathrincoated vesicles in developing chick skeletal myotubes. A mechanism by which CSQ could move through the Golgi complex in muscle but could not move beyond the ER in nonmuscle represents an interesting area for future research.
Cell Biology of Phosphorylated Cardiac CSQ-Cardiac CSQ exhibited a complex pattern of phosphorylation on molecules that are clearly processed within the secretory pathway, whereas the fast-twitch isoform was not phosphorylated, consistent with results of point mutants using this isoform (49,50). Cardiac CSQ existed in a broad range of glycoforms; those which contained phosphate (Man1,3) and those lacking phosphate (Man4 -6). The pattern of phosphorylated glycoforms suggests that transit to specific compartments may require prior phosphorylation or that some compartments may contain a CSQ phosphatase. It could also be that phosphorylated glycoforms of cardiac CSQ do not co-localize with non-phosphorylated glycoforms, a possibility that is currently under investigation. Nevertheless, our data support the idea that there exists a cellular relationship between glycosylation, a modification generally viewed as a marker for cell transport, and the phosphorylation state of the molecule. Among CSQ glycoforms that do undergo phosphorylation (Man1,3), 10 -20% of the molecules contained no phosphate. It is hypothesized that these phosphate-free molecules are part of a dynamic phosphorylation-dephosphorylation cycle.
In nonmuscle cells, deletion of phosphorylation sites from CSQ resulted in a roughly 2-fold decrease in the Man9 glycoform in favor of a mannose-trimmed glycan, reflecting the actions of ER ␣ 1,2-mannosidase and Golgi ␣ 1,2-mannosidase I, the enzymes generally thought to reduce glycans to Man5 (41)(42)(43). This divergence between glycosylation and phosphorylation of cardiac CSQ in nonmuscle as well as muscle cells may reflect divergent pathways that lead to differential glycosylation in a phosphorylation-dependent manner. A similar divergence of phosphorylation and glycosylation was previously reported for the lumenal ER glycoprotein GRP94 (34), where CK2 phosphorylation in intact cells occurred for only one of two distinct pathways for glycoprotein processing.
In conclusion, heart cells process CSQ through two primary processes within the secretory pathway on the way to retention by terminal cisternae, N-glycosylation, and phosphorylation on CK2 sites. The distribution of phosphate among CSQ glycoforms suggests that phosphorylation and glycosylation processes involve both a common and a distinct cellular compartmentation. Details of CSQ transit in cardiac cells may shed light on mechanisms that regulate calcium transients by maintaining levels of component proteins in terminal cisternae.