Conservedin VivoPhosphorylation of Calnexin at Casein Kinase II Sites as Well as a Protein Kinase C/Proline-directed Kinase Site

Calnexin is a lectin-like chaperone of the endoplasmic reticulum (ER) that couples temporally and spatially N-linked oligosaccharide modifications with the productive folding of newly synthesized glycoproteins. Calnexin was originally identified as a major type I integral membrane protein substrate of kinase(s) associated with the ER. Casein kinase II (CK2) was subsequently identified as an ER-associated kinase responsible for the in vitro phosphorylation of calnexin in microsomes (Ou, W-J., Thomas, D. Y., Bell, A. W., and Bergeron, J. J. M. (1992) J. Biol. Chem. 267, 23789-23796). We now report on the in vivo sites of calnexin phosphorylation. After 32PO4 labeling of HepG2 and Madin-Darby canine kidney cells, immunoprecipitated calnexin was phosphorylated exclusively on serine residues. Using nonradiolabeled cells, we subjected calnexin immunoprecipitates to in gel tryptic digestion followed by nanoelectrospray mass spectrometry employing selective scans specific for detection of phosphorylated fragments. Mass analyses identified three phosphorylated sites in calnexin from either HepG2 or Madin-Darby canine kidney cells. The three sites were localized to the more carboxyl-terminal half of the cytosolic domain: S534DAE (CK2 motif), S544QEE (CK2 motif), and S563PR. We conclude that CK2 is a kinase that phosphorylates calnexin in vivo as well as in microsomes in vitro. Another yet to be identified kinase (protein kinase C and/or proline-directed kinase) is directed toward the most COOH-terminal serine residue. Elucidation of the signaling cascade responsible for calnexin phosphorylation at these sites in vivo may define a novel regulatory function for calnexin in cargo folding and transport to the ER exit sites.

Calnexin was originally identified and purified as a constituent of a complex of four co-isolated integral membrane proteins, two of which were phosphorylated in microsomes by ER 1 -associated kinase(s) (1). This phosphorylation was exclu-sively on serine residues and by controlled protease digestion found to be cytosolically oriented (1,2). As a consequence of the cDNA cloning of this phosphoprotein, it was predicted and subsequently confirmed to be a type I integral membrane protein with extensive sequence similarity in its luminal domain to the ER luminal resident protein, calreticulin (1,2). Calnexin and then calreticulin were found to be novel molecular chaperones of the ER. These chaperones act as lectins to couple oligosaccharide modifications to newly synthesized N-linked glycoproteins with productive glycoprotein folding. The lectin specificity of these chaperones has been identified as the recognition of high mannose oligosaccharides terminating in monoglucosyl residues linked ␣1-3 (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13).
Purification of the ER-associated kinase that phosphorylated calnexin in microsomes led to the identification of CK2 (14). The properties of this kinase were consistent with the conditions that originally revealed calnexin phosphorylation (1,14). Furthermore, purified CK2 has been found to phosphorylate calnexin on putative CK2 sites found within the cytosolic domain of calnexin (14,15).
Calnexin is phosphorylated in vivo (16 -18). Phosphorylated calnexin has been shown to associate with the null Hong Kong mutant of ␣-1-antitrypsin, coinciding with retention of this misfolded glycoprotein within the lumen of the ER (16). Phosphorylated calnexin was also found in association with newly synthesized major histocompatibility complex class I allotypes, which egressed from the ER at slow rates. Those allotypes transported to the Golgi apparatus at more rapid rates were associated preferentially with nonphosphorylated calnexin (17). Prolonged association of newly synthesized major histocompatibility complex class I heavy chains with calnexin was found in a B lymphoblastoid cell line transfected with HLA-B701 after incubation with the phosphatase inhibitor cantharidin or okadaic acid (19). Furthermore, when human synovial epithelial (McCoy) cells were treated with okadaic acid, the major cellular protein whose phosphorylation was shown to increase (based on two-dimensional gels followed by protein microsequencing) was calnexin (18). Remarkably, calnexin phosphorylation also increased 3-fold when McCoy cells were treated with Clostridium difficile cytotoxin B (18), a protein that glucosylates Rho proteins of the Ras superfamily (20).
Although some progress has been made on the kinases and sites of phosphorylation of calnexin after in vitro phosphorylation of intact microsomes (14,15) little is known of the sites of calnexin phosphorylation in vivo. Here we report on their identification in two mammalian cell lines, HepG2 cells (human) and Madin-Darby canine kidney MDCK cells. Both cell lines revealed phosphorylation of the cytosolic domain of calnexin exclusively on serine residues within CK2 motifs as well as a protein kinase C (PKC) and/or proline-directed kinase (PDK) motif.

EXPERIMENTAL PROCEDURES
Materials-Rabbit antibodies raised against a synthetic peptide, corresponding to amino acid residues 487-505 of canine calnexin, described previously were used (3). 32 4 Labeling of Cultured Cells and Immunoprecipitation-Cells were radiolabeled as described previously (16,17) with the following modifications. Briefly, cells were washed in phosphate-free Dulbecco's modified Eagle's medium supplemented with 1% dialyzed FBS followed by incubation in the same medium for 1 h at 37°C. Cells were then labeled by the addition of 2 mCi/ml [ 32 P]orthophosphate for 3 h. At the end of labeling, the cells were lysed as described previously (3). Briefly, cells were washed twice with ice-cold phosphate-buffered saline (20 mM NaPO 4 , pH 7.5, 150 mM NaCl) and once with ice-cold HEPES-buffered saline (50 mM HEPES, pH 7.6, 200 mM NaCl) before harvesting. Cells were then lysed in 2% CHAPS/HEPES-buffered saline lysis buffer (2% (w/v) CHAPS, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml each leupeptin and aprotinin, 10 mM NaF, 10 mM NaPP i , 0.4 mM NaVO 4 , and 5 mM NaMoO 4 ) for 30 min on ice. The same procedures were used to immunoprecipitate calnexin from both 32 P-labeled and cold phosphate-labeled cellular extracts as described previously (3). Immunoprecipitated calnexin was isolated by SDS-PAGE. The gel was dried for 2 h at 80°C under vacuum, or the proteins were transferred onto PVDF membrane (1). Radioactive bands were visualized by radioautography at room temperature.
Phosphoamino Acid Analysis-Phosphoamino acid analysis was performed as described (21). Briefly, in vivo phosphorylated calnexin was resolved by SDS-PAGE and electroblotted onto a PVDF membrane. 32 P-Labeled calnexin was detected by radioautography, and the corresponding PVDF bands were excised. The membrane containing phosphorylated calnexin was washed extensively with distilled water and subjected to acid hydrolysis, immersed in 6 N HCl at 110°C for 60 min. The hydrolysate was transferred to a microcentrifuge tube, lyophilized, and dissolved in pH 1.9 buffer (88% formic acid, glacial acetic acid, H 2 O; 2.5:7.8:89.7 (v/v/v)). Phosphoamino acid analysis was by two-dimensional electrophoresis on TLC plates in the presence of phosphoamino acid standards; phosphoserine, phosphothreonine, and phosphotyrosine. First dimension electrophoresis was carried out in pH 1.9 buffer for 20 min at 1.3 kV employing a Hunter thin layer electrophoresis system (C.B.S. Scientific, Del Mar, CA). Second dimension electrophoresis was carried out in pH 3.5 buffer (pyridine, glacial acetic acid, H 2 O; 0.5:5:94.5 (v/v/v)) at 1.5 kV for 20 min. The standards were visualized by spraying a 0.25% (w/v) ninhydrin acetone solution followed by incubation at 65°C for 10 min. The radiolabeled amino acids were detected by radioautography with an enhancing screen at Ϫ70°C. Recovery from each step was monitored by Cerenkov counting.
In Gel Digestion and Mass Spectrometry (DE-MALDI MS and nano-ESI MS)-Calnexin immunoprecipitates from nonradiolabeled HepG2 and MDCK cell lysates were resolved by SDS-PAGE, visualized by Coomassie Blue staining (stain was 0.2% (w/v) Coomassie Brilliant Blue R250 in 50% (v/v) methanol in water containing 2% (v/v) acetic acid; destain was 50% (v/v) methanol in water containing 2% (v/v) acetic acid); and the corresponding gel slice was excised. Coomassie stain was removed by extraction (twice) with 50% (v/v) acetonitrile/H 2 O, followed by two cycles each of extracting with acetonitrile and swelling with 100 mM NH 4 HCO 3 . Calnexin was in gel reduced and alkylated with 15 mM dithiothreitol and 1.3 mM iodoacetamide and then in gel digested with 13 g/ml bovine trypsin in the presence of 5 mM CaCl 2 as described (22). Tryptic peptides were first extracted with acetonitrile, followed by two cycles each of swelling with 100 mM NH 4 HCO 3 and extracting with acetonitrile and then two cycles each of swelling with 5% formic acid and extracting with acetonitrile. The efficiency of the extraction of calnexin tryptic phosphorylated fragments from gel pieces was evaluated with radiolabeled calnexin and Cerenkov counting. Greater than 70% of the radioactivity was extracted and recovered (data not shown). For DE-MALDI-ToF MS, dried peptide extracts were redissolved in 5% formic acid containing 5% methanol (10 l) (22). An aliquot (0.4 l) was spotted onto a stainless steel target precoated with ␣-cyano-4-hydroxycinnamic acid. The target was allowed to air dry before being washed with an aqueous solution containing 1% trifluoroacetic acid. Excess wash solution was blown off the target using compressed air. The target was loaded into the mass spectrometer for analysis by DE-MALDI-ToF Calnexin is indicated on the right, and molecular weight standards are indicated on the left. 32 P-Labeled PVDF membrane-bound calnexin was subjected to phosphoamino acid analysis as described under "Experimental Procedures." B, phosphoamino acid analysis of calnexin from both HepG2 and MDCK cells (left and right, respectively) by TLC; first dimension electrophoresis in pH 1.9 buffer and second dimension electrophoresis in pH 3.5 buffer. The time of exposure was 68 h at Ϫ70°C with an intensifying screen. The direction of electrophoresis toward the anode is indicated. The dashed circles indicate the positions of nonradiolabeled phosphoamino acid standards as visualized by ninhydrin staining. o, origin; P-S, phosphoserine; P-T, phosphothreonine; P-Y, phosphotyrosine; P i , free phosphate and partials, products of incomplete peptide hydrolysis. mass spectrometry using a Bruker Reflex instrument fitted with a 337-nm nitrogen laser. All spectra were acquired with the instrument in reflector mode. Acquisition parameters were set as follows: sampling frequency, 500 MHz; gain, 50 mV; source voltage, 24,000 V; reflector voltage, 24850 V. DE-MALDI-ToF mass spectra were acquired for each digest and compared with mock in gel trypsin digest. Nano-ESI experiments were performed as described (23). Briefly, in gel trypsin-digested calnexin was desalted via a POROS R2 capillary column (22,23) (PerSeptive Biosystems, Farmingham, MA); dried in a vacuum centrifuge; and resuspended in 10 l of spraying solution (50% (v/v) methanol, 5% (v/v) ammonia in water for negative ion mode or 50% (v/v) methanol, 1% (v/v) formic acid in water for positive ion mode). 1 l was inserted into a homemade capillary needle. Spraying capillary needles were made from borosilicate glass capillaries (Clark Electromedical Instruments, Pangbourne, Reading, UK) employing a micropipette puller (Sutter Instrument Co., Novato, CA), and gold was sputtered employing a vapor desorption instrument. A PE-Sciex API III triple quadrupole mass spectrometer (Perkin-Elmer) fitted with a nano-ESI source (24,25) was used to acquire all electrospray mass spectra. The ion spray voltage was set to 600 -650 V with an orifice potential of 60 -70 V. Instrument polarity was set appropriate to the detection of positive or negative ions. Q1 scans were used to mass analyze ions relating to the intact peptides in the mixture. These were recorded by scanning the first quadrupole between m/z 400 and 2000 using a 1.0-ms dwell time and a 0.1-Da step size. Peptides containing phosphate produce a char-acteristic PO 3 Ϫ ion fragment at m/z 79 (26,27). Molecular ions for these peptides were recorded by scanning the first quadrupole between m/z 300 and 1400 with the third quadrupole set to transmit m/z 79 only (precursors m/z 79 scans). Argon gas was used in the collision cell (quadrupole 2) at a collision gas thickness of 250 units. Scanning the first and third quadrupoles with an offset of m/z 49 profiled phosphopeptides by loss of H 3 PO 4 (molecular mass of 98) from doubly charged molecular ions (Constant Neutral Loss scans) (28). For MS/MS detection, Q1 was set to transmit a mass window of 2 Da for product ion scans. Product ion spectra were accumulated with a 0.2-Da mass step size. Dwell time was 1.0 ms, and collision energy was optimized to obtain the MS/MS spectra. Spectra interpretation was performed using BioMultiView (Sciex) software.
Computer Analysis-Candidate kinases for the cytosolic serine residues of calnexin were predicted based on consensus sequence motifs employing the PROSITE data base of EXPASY 2 (29). Predicted peptide m/z values were evaluated employing tools provided by ProteinProspector. 3 Calnexin and calmegin sequence alignments were initially generated by BLASTp (30), FASTA (31), and MSA (32) and optimized manually.

Calnexin Is in Vivo Phosphorylated on Serine Residues in
HepG2 and MDCK Cells-Calnexin was originally identified as a major substrate of ER-associated kinase(s) by in vitro phosphorylation of intact microsomes with [␥-32 P]GTP as phosphate donor (1). In order to determine the in vivo sites of phosphorylation for calnexin, both HepG2 and MDCK cells were in vivo labeled with [ 32 P]orthophosphate followed by immunoprecipitation with anti-calnexin antibodies. SDS-PAGE-resolved immunoprecipitates were electroblotted onto PVDF membranes and similar levels of phosphorylated calnexin from both cell types were revealed by radioautography (Fig. 1A, lanes 1 and  2). The bands corresponding to phosphorylated calnexin were excised from the PVDF membranes and subjected to phosphoamino acid analyses. Radioautograms of the two-dimensional TLC plates for both human and canine calnexins revealed only 32 P-labeled serine that comigrated with the nonradiolabeled phosphoserine standard as detected by ninhy-2 Available on the World Wide Web at http://expasy.hcuge.ch. 3 Available on the World Wide Web at http://prospector.ucsf.edu.
, predicted casein kinase II site; PKC, predicted protein kinase C site; PDK, predicted proline-directed kinase site; PKA, predicted protein kinase A site. drin staining (left and right parts of Fig. 1B, respectively). The phosphorylated residues were not altered with longer (up to 24 h) in vivo radiolabeling (data not shown). Hence, calnexin was exclusively in vivo phosphorylated on serine residues in both cell types.
In isolated ER microsomes, in vitro phosphorylation by ERassociated kinase(s) is exclusively on the cytosolic domain of calnexin (1). The cytosolic COOH-terminal domains of canine (MDCK) and human (HepG2) calnexins (1,33) share 94% identity (Fig. 2). The cytosolic domain of canine calnexin contains six serines, and the cytosolic domain of human calnexin contains five equivalent serines and a threonine. The five conserved cytosolic serine residues are in primary sequence motifs that are predicted to be recognized by CK2 (34), PKC (35,36), PDK (37,38), or protein kinase A (39,40) (Table I). To identify the in vivo sites of phosphorylation of calnexin we proceeded to mass spectrometry.
Analyses of Tryptic Digests of Calnexin by DE-MALDI-ToF MS-Calnexin was immunoprecipitated from nonradiolabeled cell lysates from both cultured MDCK and HepG2 cells. After separation by SDS-PAGE and visualization by Coomassie staining, calnexin was in gel digested with trypsin as described under "Experimental Procedures." DE-MALDI-ToF mass spectra for the tryptic peptides of calnexin from MDCK and HepG2 were collected (Fig. 3, A and B). Peptide masses not observed in the mock in gel digest (data not shown), were employed to confirm the identity of MDCK and HepG2 calnexins. Total coverage for calnexin from MDCK cells was 165 of 573 (28.8%) amino acid residues, and coverage from HepG2 cells was 171 of 572 (29.9%) amino acid residues. With respect to the cytosolic domain of calnexin, however, the degree of coverage was 58.4 and 33.7% from MDCK and HepG2 cells, respectively. This coverage by DE-MALDI-ToF included phosphopeptides observed as peak 1 and peak 7 in Fig. 3A for MDCK calnexin and as summarized in Table II. Coverage of the cytosolic domain was increased by nano-ESI MS and was greater than 90% by combination of both mass spectrometric techniques (see below). Poor coverage of the luminal domain was consistent with the generation of a protease-resistant core in the presence of Ca 2ϩ (2), conditions employed during in gel trypsin digestion.
Analyses of Tryptic Digests of Calnexin by Nano-ESI MS: Identity and in Vivo Sites of Phosphorylation-The Q1 positive ion spectrum between m/z 800 and 1150 of HepG2 calnexin tryptic digests displays four ions (P1-P4) tentatively assigned as tryptic fragments of calnexin by comparison with mock in gel trypsin digest (Figs. 4, A and B). By comparison of the observed masses and the cDNA predicted calculated average masses  Table II. The x axis indicates the mass-to-charge (m/z) values, and the y axis indicates the relative ion intensity.  (Fig. 3, A and B (Fig. 4, inset).
Only two phosphopeptides were detected by DE-MALDI-ToF MS (Table II). Hence, we proceeded to characterize in greater detail the phosphorylated peptides in the desalted total tryptic digests of nonradiolabeled phosphocalnexin from both MDCK and HepG2 cells employing two selective techniques for identification of phosphorylated fragments: scans for precursors m/z 79 in negative ion mode (26,27)  precursor ions that fragment to produce a characteristic product ion of m/z 79 that corresponds to the phosphate anion, PO 3 Ϫ . CNL scans reveal the masses of parent ions that lose the phosphate group, H 3 PO 4 (molecular mass of 98), as a neutral fragment. A loss of m/z 49 would be observed from doubly charged precursor ions, [M ϪH 3 PO 4 ϩ 2H] 2ϩ .
Precursors m/z 79 scans of the tryptic peptides from MDCK calnexin were dominated by three major phosphorylated peptide ions (Fig. 5A). Two of the three m/z values (753.4 and 502.2) correspond to the doubly and triply negatively charged states of the phosphorylated tryptic fragment 555 AEEDEILN-RpSPR 566 (where pS represents phosphoserine; calculated average mass, 1509.5 atomic mass units). The ion at m/z 405.0 is a commonly observed nonspecific background peak. 4 CNL m/z 49 scans of the tryptic fragments of MDCK calnexin (Fig. 6A) confirmed the former assignment, since the doubly charged phosphopeptide 555 AEEDEILNRpSPR 566 (m/z 754.6) was observed.
Mass spectral analyses of the tryptic peptides from HepG2 calnexin by precursors m/z 79 scans revealed a clustered series of multiply charged ions (Fig. 5B) that upon deconvolution identified several large phosphorylated partial tryptic frag-ments derived from the cytosolic domain of HepG2 calnexin (see below). Analyses of this tryptic digest of HepG2 calnexin by CNL m/z 49 scans detected a [M ϩ 2H] 2ϩ peptide ion with an m/z value of 755.2 (Fig. 6B), equivalent to the fragment similarly observed for MDCK calnexin (Fig. 6A). This fragment was not observed as strongly by precursors m/z 79 scans of the trypsinized HepG2 calnexin as it was observed for trypsinized MDCK calnexin (Fig. 5, compare A and B). The selectivity of the CNL m/z 49 scans, however, confirmed the presence of this same fragment 554 AEEDEILNRpSPR 565 in the HepG2 calnexin tryptic digest.
The clustered series of multiply charged ions recorded by precursors m/z 79 scans of trypsinized HepG2 calnexin was subjected to deconvolution employing the Hypermass Reconstruction software (41). The reconstruct algorithm was initially carried out over the range of 2,000 -10,000 Da and then focused in on the range of interest. No discernible species could be identified at other masses. The reconstruct profile (Fig. 7) 4 M. A. Ward, unpublished observations. indicated that the clustered series of multiply charged ions was derived from partial digests of the same phosphorylated region of the cytosolic domain of HepG2 calnexin. The partial tryptic fragments (Fig. 7) corresponded to the region K 496 T 497 DAP-QPDVKEEEEEKEEEKDKGDEEEEGEEKLEEKQKSDAEE-DGGTVSQEEEDRKPKAEEDEILNRSPR 565 NR 567 of calnexin (boldface type represents potential sites of phosphorylation). The observed masses and their calculated monoisotopic and average masses are tabulated (Fig. 7). These phosphorylated partial tryptic fragments contained three of the five serine residues, Ser 534 , Ser 544 , and Ser 563 , of the cytosolic domain of human calnexin, and these fragments contained either two or three phosphate groups. Thus, these three serine residues represent the in vivo phosphorylation sites for calnexin. Close inspection of the precursors m/z 79 scans for phosphorylated trypsinized MDCK calnexin revealed a similar clustered series of multiply charged ions (Fig. 5, A and B), albeit these signals were weaker relative to the dominant ions (m/z 405.0, 502.2, and 753.4) but consistent with those observed in Fig. 5B. Partial digestion of phosphorylated MDCK calnexin was probably prevalent but to a lesser extent in this sample. Attempts to deconvolute these clusters of multiply charged ions (Fig. 5A) were unsuccessful due to the strong signals from the three abundant ions. Differences in sensitivity to trypsinization is unclear. As a precaution to autodigestion, trypsinization was performed in the presence of 5 mM Ca 2ϩ (22,23). This precaution may have contributed to an increased proteolytic resistance of the cytosolic domain, since Ca 2ϩ binds to this domain (42). This caveat is similar to that indicated above for poor coverage of the luminal domain (Fig. 3, A and B), resulting from an increased proteolytic resistance induced by millimolar concentrations of Ca 2ϩ (2). Nevertheless, these data are consistent with MDCK calnexin being phosphorylated on the equivalent three serine residues, i.e. Ser 535 , Ser 545 , and Ser 564 (summarized in Fig. 8). DISCUSSION A recently uncovered family of resident ER proteins has revealed properties of novel lectin-like molecular chaperones. These recognize N-linked glycoproteins and couple N-linked oligosaccharide modification with productive glycoprotein folding (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)43). The family is composed of calnexin, a type I transmembrane protein of the ER (1, 2); calreticulin, a KDELterminated soluble ER-resident protein (44); and calmegin, a testis-specific ER transmembrane protein with sequence conservation at the predicted luminal domain (45).
Major differences among the three proteins are found at their COOH termini. Mammalian calnexins reveal 89 cytosolically oriented residues (1,2,33,42), which were here shown to be phosphorylated at three of the four invariant serine residues. The calmegin deduced protein sequences predict 119 amino acids cytosolically oriented (45,46) with six conserved (human and mouse) potential serine phosphorylation sites. The calmegin conserved potential serine phosphorylation sites are, as with the observed sites of serine phosphorylation in calnexin, also in the COOH-terminal half of the respective cytosolic domain. Five of the six potential serine phosphorylation sites of calmegin are within motifs similar to those observed for calnexin; three are in CK2 motifs, one is in a PKC motif, and another (mouse sequence only) is in a PDK motif (Fig. 9). Thus, calmegin is predicted to be phosphorylated on equivalent serines to those in calnexin (Fig. 9). The alignment of the cytosolic domains of calnexins and calmegins identifies three major (and three minor) loops (boxed in Fig. 9) that are unique to the calmegins. Furthermore, this alignment reveals that the cytosolic domains of both calnexin and calmegin can be divided into four subdomains: a juxtamembrane basic, lysine-rich subdomain; a central acidic, glutamic acid-rich subdomain; a phosphorylation signaling subdomain; and a putative COOH-termi- nal ER retrieval subdomain (47) (Fig. 9). Calreticulin is a luminal ER-resident protein that has no cytosolically oriented sequences but rather a COOH-terminal KDEL ER retrieval signal (48).
For yeast, greater evolutionary divergence has occurred at the cytosolic domain of calnexin as opposed to the intraluminal domain. The 48-amino acid cytosolic domain of the calnexin homologue Cnx1 in the fission yeast, Schizosaccharomyces pombe, is phosphorylated in vivo. 5 The S. pombe calnexin gene is essential for viability, but the cytosolic domain is dispensable for that essential feature (49). The Saccharomyces cerevisiae calnexin homologue Cne1p reveals only one potential cytosolic amino acid (Thr 482 ), and this calnexin gene, CNE1, is nonessential for S. cerevisiae viability (50). The riddle of the evolution of a cytosolic domain for S. pombe calnexin coincident with essential function(s) for viability for which the cytosolic domain appears to be dispensable may be ultimately resolved by an analysis of the signaling cascades that phosphorylate calnexin in mammalian species and in S. pombe. 5 The cytosolic domains of all known mammalian calnexins display 83% identity with four invariant serines (Fig. 9). As a consequence of this high degree of sequence identity, we set out to identify the sequences constitutively phosphorylated in two cell lines, i.e. human HepG2 and canine MDCK cells. By phosphoamino acid analyses, both calnexins were exclusively in vivo phosphorylated on serine residues.
The first cytosolic serine residue, Ser 484 and Ser 485 of HepG2 and MDCK calnexins, respectively, is a juxtamembrane serine residue that is invariant among mammalian calnexins (Fig. 9) and corresponds to a potential PKC phosphorylation site. A tryptic peptide containing the transmembrane domain and thus the juxtamembrane serine residue was not detected by MS analyses. Low recovery of this tryptic fragment may be due to poor extractability (51) of such a peptide and/or due to partial proteolysis generating very large hydrophobic poorly extractable fragments probably linked to the protease-resistant Ca 2ϩ luminal core of calnexin (2). The next COOH-terminal serine residue was conserved in HepG2 (Ser 490 , a potential CK2 phosphorylation site) and MDCK (Ser 491 , a potential CK2/ PDK phosphorylation site) calnexins but not rodent calnexins (Fig. 9). This site was detected only as a nonphosphorylated fragment by MS analysis of the HepG2 tryptic digests (Fig. 3B).
As summarized in Fig. 8, only Ser 534 , Ser 544 , and Ser 563 (human numbering, invariant in mammalian calnexins; see Fig. 9) were in vivo phosphorylated as detected by two selective nano-ESI MS techniques for detection of phosphorylated peptides. Two of these three invariant serine phosphorylation sites, Ser 534 and Ser 544 , are within well recognized CK2 motifs (34). This coincides with earlier observations that calnexin in microsomes was in vitro phosphorylated by CK2 (14,15) and that CK2 was purified as an ER membrane-associated kinase (14). The identification of a third site of calnexin phosphorylation (Ser 563 in HepG2; Ser 564 in MDCK) was not predicted from previous in vitro studies (1,14,15). This site, invariant in mammalian calnexins, is within a motif potentially recognized by either PKC (35,36) or PDK (37,38).
We have presented evidence for diphosphorylated and triphosphorylated (Figs. 7 and 8) calnexins but no conclusive data for a monophosphorylated form, i.e. with only one of the two CK2 sites or only the S 563 PR site being phosphorylated, since no singly phosphorylated fragments encompassing the three sites were identified. Nonphosphorylated peptides encompassing the observed two CK2 sites of serine phosphorylation were identified by DE-MALDI-ToF MS analyses: 532 QKS-DAEEDGGTVSQEEEDRKPK 553 of HepG2 calnexin and 535 SDAEEDGGTASQEEDDRKPK 554 of MDCK calnexin (Figs. 2 and 3, and Table II). The strongest evidence for only one of the two CK2 sites being phosphorylated was observed from MDCK trypsinized calnexin ( Fig. 3A and Table II). The peptide ion, m/z 2130.9 ( Fig. 3A and Table II) corresponds to the partial tryptic fragment 533 pyro-QKSDAEEDGGTASQEEDDR 551 , containing one phosphate group and cyclization of the NH 2terminal glutamine (calculated average mass, 2130.9) (52). The cyclization of the NH 2 -terminal glutamine of this tryptic fragment was suggested also by Cala and co-workers (15). For both Refs. 45 and 46) are shown. CC and CW (italic type), amino acid residues predicted to be the last two residues of the transmembrane domains of calnexin and calmegin, respectively. Calnexin identified (closed symbols) and calmegin putative (open symbols) phosphorylation sites are shown. Potential kinases are as follows: CK2 (arrows), PKC (arrowheads), and PDK (asterisks). Residue numbers refer to the mature calnexins or calmegins. Serine residues are in boldface type; amino acid differences among the calnexins and among the calmegins are indicated; and identities are indicated by dashes. Gaps introduced to optimize alignments are shown as dots. Calnexin alignment identifies two deletions of one amino acid at position 507 and 524 of rodent calnexins; similarly, calmegin alignment identifies two deletions of one (Mo, 510) or two (Hu, 528) amino acids. Calnexin/calmegin alignment identifies six deletions/insertions of 2, 6, 8, 3, and 11 residues and 1 residue (enclosed boxes, calmegin). Consensus sequence is as follows: invariant (boldface uppercase type) and five of six identities (bold lowercase type). Boldface dots indicate differences. The four subdomains (i.e. the juxtamembrane basic, acidic, phosphorylation, and the predicted ER retrieval (47) domains) are indicated. calnexins, the two phosphorylated CK2 sites (Ser 534 and Ser 544 , HepG2 calnexin) are contained within the same tryptic fragment, and thus by our strategies, these two phosphorylation sites could not be characterized individually. Evidence for a nonphosphorylated state of the most COOH-terminal serine residue was not obtained. This may be a consequence of complete proteolytic digestion of this nonphosphorylated form and subsequent loss of the corresponding tryptic tripeptide, (R)S 563 PR, during desalting/washing steps prior to MS analyses. However, the S 563 PR site in the diphosphorylated large partial tryptic fragment (Fig. 7) that contains three potential sites of phosphorylation may correspond to a nonphosphorylated Ser 563 site. On this basis, there are six potentially different (three mono-and three diphosphorylated) partially phosphorylated states of calnexin that probably represent a regulatory mechanism for calnexin action.
The phosphorylation results presented in this paper extend the previous finding by Capps and Zuniga (17) and Le et. al. (16) that a significant proportion of calnexin was phosphorylated in vivo. The identification of phosphorylated calnexin in association with a subset of incompletely folded major histocompatibility complex class I allotypes (17) or of the misfolded null Hong Kong mutant of ␣-1-antitrypsin (16) is suggestive of a coincident and perhaps regulatory role with the action of the luminal domain of calnexin in glycoprotein folding and quality control. Conservation of the three serine targets of protein kinases as elucidated here predicts that this conservation and their phosphorylation are under strict control. Elucidation of the signaling cascades that trigger calnexin phosphorylation at the PKC/PDK site as well as the CK2 sites may lead to new insights in the regulation of cargo folding and transport from the ER. These studies may also lead ultimately to a rationale for the evolution of three distinct genes in mammals that encode this family of molecular chaperones.