The Phosphorylation of Subunits of Complex I from Bovine Heart Mitochondria*

In bovine heart mitochondria and in submitochondrial particles, membrane-associated proteins with apparent molecular masses of 18 and 10 kDa become strongly radiolabeled by [32P]ATP in a cAMP-dependent manner. The 18-kDa phosphorylated protein is subunit ESSS from complex I and not as previously reported the 18 k subunit (with the N-terminal sequence AQDQ). The phosphorylated residue in subunit ESSS is serine 20. In the 10 kDa band, the complex I subunit MWFE was phosphorylated on serine 55. In the presence of protein kinase A and cAMP, the same subunits of purified complex I were phosphorylated by [32P]ATP at the same sites. Subunits ESSS and MWFE both contribute to the membrane arm of complex I. Each has a single hydrophobic region probably folded into a membrane spanning α-helix. It is likely that the phosphorylation site of subunit ESSS lies in the mitochondrial matrix and that the site in subunit MWFE is in the intermembrane space. Subunit ESSS has no known role, but subunit MWFE is required for assembly into complex I of seven hydrophobic subunits encoded in the mitochondrial genome. The possible effects of phosphorylation of these subunits on the activity and/or the assembly of complex I remain to be explored.

The reversible phosphorylation of proteins is one of the major ways of regulating cellular activities (1). Up to 30% of eukaryotic proteins are phosphorylated (2), and there are more than 500 human genes encoding protein kinases including subunits of cyclic AMP-dependent protein kinases (protein kinase A) (3). Protein kinase A regulates, for example, the functions of many enzymes, membrane receptors, ion channels, and transcription factors that are involved in a variety of cellular processes. Protein kinase A is localized at various cellular sites by protein kinase anchor proteins (AKAPs), and two forms, D-AKAP1 and D-AKAP2, have been identified in mammalian mitochondria (4 -6). Protein kinase A activity associated with the matrix side of the inner membrane of the organelle appears to be responsible for the phosphorylation of three inner membrane proteins of 29, 18 and 6.5 kDa (7)(8)(9). The so-called "18 k subunit" of complex I (NADH:ubiquinone oxidoreductase) with the N-terminal sequence AQDQ has been proposed to be the 18-kDa protein that becomes phosphorylated (10,11).
Mammalian complex I is an L-shaped assembly with one arm in the inner membrane and the other lying orthogonal to it protruding into the matrix of the mitochondrion. It is made of 46 different proteins, and the 18 k subunit (AQDQ) is a component of the matrix arm (12)(13)(14)(15). The identification of this subunit as a phosphorylated protein was based upon the separation of the subunits of 32 P-phosphorylated complex I by SDS-PAGE followed by N-terminal sequencing of the radiolabeled band with an apparent molecular mass of 18 kDa (10). The site of phosphorylation was not determined experimentally, but it was deduced from sequence motifs. 26 subunits of bovine complex I migrate on SDS-polyacrylamide gels in the molecular mass range of 10 -20 kDa, and many of them are not resolved into protein bands containing unique subunits (13,14,16). Moreover, 14 of these subunits have N-␣-amino groups that are rendered inaccessible to N-terminal sequencing by post-translational modifications. Therefore, we have reanalyzed the phosphorylation of bovine complex I. Two predominantly phosphorylated subunits with molecular masses of 18 and 10 kDa were detected. By a combination of Edman sequencing and mass spectrometric analysis, the band migrating at 18 kDa on an SDS-polyacrylamide gel was shown to contain three proteins, namely the 18-kDa subunit, subunit ESSS and the N-␣-acetylated subunit, B17. The site of phosphorylation is in subunit ESSS, and neither the 18-kDa subunit nor subunit B17 was phosphorylated. The phosphorylated protein with a molecular mass of 10 kDa is the complex I subunit, MWFE. In both proteins, the exact site of phosphorylation was identified experimentally.

EXPERIMENTAL PROCEDURES
Materials-n-Dodecyl-␤-D-maltoside was purchased from Anatrace. Immobilon polyvinylidene difluoride membranes were obtained from Millipore, and bicinchoninic acid protein assay kits were from Pierce. Sequelon-AA kits and all other protein sequencing materials were purchased from Applied Biosystems. Cantharidin was obtained from Calbiochem, and protease inhibitor mixture tablets were from Roche Diagnostics. ATP (10 mCi/ml) was purchased from Amersham Biosciences, and Ultima Gold-AB liquid scintillation fluid was from Packard Bioscience. Pancreatic trypsin and chymotrypsin and endoproteinases Arg-C from Clostridium histolyticum, Asp-N from Pseudomonas fragi, Glu-C from Staphylococcus aureus V8, and Lys-C from Lysobacter enzymogenes were obtained from Roche Diagnostics, who also supplied calf intestine alkaline phosphatase.
Preparation of Mitochondria and Membranes-Mitochondria were prepared from fresh bovine hearts as described before (17). Submitochondrial particles were prepared from mitochondria by sonication and centrifugation (17). A sample of complex I, purified from bovine heart mitochondria by chromatography (14), was a gift from Dr. J. Hirst.
Phosphorylation of Mitochondria and Membranes-A portion of washed mitochondria (200 g) recovered from a Percoll gradient or of submitochondrial particles (200 g) was resuspended in 33 l of buffer consisting of 40 mM Tris-HCl, pH 7.4, 0.25 M sucrose, 10 mM magnesium chloride, 20 g/ml oligomycin, and protease and phosphatase inhibitors. A 1-l solution containing 10 mCi/ml [␥-32 P]ATP, 2.5 mM ATP in 0.8 l of Tris-HCl, pH 7.4, and 0.6 l of 3 mM cAMP was added to the suspension of mitochondria or membranes. The mixture was incubated * 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. at 37°C for 15 min, when 7 l of a solution of 30% SDS was added to terminate the reaction.
Phosphorylation of Complex I-A 16.2-l sample of a solution of 1. 6 mg/ml bovine complex I dissolved in buffer consisting of 20 mM Tris-HCl, pH 7.4, 100 mM sodium chloride, and 0.1% n-dodecyl-␤-D-maltoside was mixed with 8 l of 0.2% aqueous n-dodecyl-␤-D-maltoside, 0.22 l of 10 g/ml oligomycin, 0.44 l of protease inhibitors, and 0.22 l of 10 M cantharidin. This mixture was divided into two aliquots. To each were added 1 l of 1 unit/l protein kinase A solution, 0.25 l of 3 mM cAMP, 0.75 l of 2 mM ATP containing 0.33 M [␥-32 P]ATP (1 mCi/ml). The reaction mixture was incubated at 37°C for 15 min, and then 10 l of 5% SDS was added. The reaction products were fractionated by SDS-PAGE as described below.
Protein Fractionation-The products of phosphorylation reactions were fractionated by SDS-PAGE on 12-22% polyacrylamide gradient gels in the system of Laemmli (18) and also on 13-18% polyacrylamide gradient gels by Tricine 1 SDS-PAGE (19). The proteins were stained for at least 15 min with a 0.25% solution of Coomassie Blue R-250 dye dissolved in 50% aqueous methanol containing 7% acetic acid. Then the gel was destained in 20% aqueous methanol containing 7% acetic acid, dried, and subjected to autoradiography. Subunits of complex I that had been 32 P-phosphorylated in vitro were fractionated by reverse phase HPLC on a column of C8 Aquapore RP-300 (PerkinElmer Life Sciences; 100 ϫ 2.1-mm inner diameter) equilibrated in 0.1% trifluoroacetic acid. The proteins were eluted with a gradient of acetonitrile at a flow rate of 0.1 ml/min as follows: 0 -15 min, 0 -24% acetonitrile; 15-85 min, 24 -60% acetonitrile; 85-95 min, 60 -80% acetonitrile. The UV absorbance of the eluate was monitored at 225 nm, and 100-l fractions were collected. The radioactivity of a 30-l portion of each fraction was measured by scintillation counting. Fractions were dried down, redissolved in 1% SDS, and analyzed by SDS-PAGE. The gel was dried and autoradiographed.
Proteolytic Digestion of Protein Bands in Gels-Stained protein bands in SDS-polyacrylamide gels were excised, and the proteins were digested "in-gel" with various proteolytic enzymes, essentially as described for trypsin digests, except that ammonium bicarbonate buffer was replaced with 20 mM Tris-HCl, pH 8.0 (14,20). Digestions were carried out overnight at 37°C, except that those with chymotrypsin and endoproteinase Glu-C were performed at 25°C. The digestion buffers for trypsin and chymotrypsin, but for no other protease, contained 5 mM calcium chloride. Usually a portion of the tryptic digest was analyzed in a MALDI-TOF mass spectrometer (see below). The remaining digests (and other digests) were stored at Ϫ20°C before being subjected to various analyses (see below).
Edman Degradation of Radiolabeled Phosphopeptides-The specificities of cleavage of several proteases, combined with the positions of 32 P-phosphorylated residues in peptides in digests with those enzymes, determined by release of radioactivity at a particular cycle in a sequential Edman degradation, are often sufficient to allow the precise site of phosphorylation in a protein of known sequence to be determined (21)(22)(23). This approach was tried out on the 42 kDa radiolabeled band. It proved to be successful, and so it was employed with the 18 kDa and 10 kDa radiolabeled bands from mitochondrial membranes, and also with subunits ESSS and MWFE from complex I. The total radioactivity in the extracts from in-gel proteolytic digests of bands from gels was measured. Then a sample of the digest, usually containing (100 -1,000 cpm), was transferred to half a disc of a Sequelon-AA membrane, which has amino-phenyl groups available for attachment of peptides via their C-terminal carboxyls (24). The membrane was dried at room temperature and then kept at 55°C for 15 min. It was wetted with 5 l of water and dried again. The membrane was cooled to room temperature and a 5-l solution of 10 mg/ml N-(3-dimethylaminopropyl)-NЈ-ethylcarbodiimide hydrochloride in MES coupling buffer, supplied as part of a kit by Millipore, was applied to the membrane. The membrane was allowed to dry, and the process of adding 5 l of coupling reagent and drying was repeated. The radioactivity on the membrane was estimated by counting the Cerenkov radiation. The membrane was rinsed sequentially with 80% aqueous trifluoroacetic acid, 0.1% trifluoroacetic acid, 50% acetonitrile containing 0.1% trifluoroacetic acid, and finally 90% acetonitrile containing 0.1% trifluoroacetic acid. The yield of radioactive peptide(s) coupled to the membrane was estimated by remeasurement of the Cerenkov radiation. Usually, it was at least 70%. Then the membrane was introduced into a vertical sequencing cartridge of an Applied Biosystems Procise 494 protein sequencer, modified to deliver 90% aqueous methanol to extract anilinothiazolinones released by the acid cleavage step in the Edman cycle and to collect the extracts in a fraction collector. The radioactivity released at each cycle was measured by scintillation counting. Before attachment to the membrane of the Asp-N digest of the 18 kDa radiolabeled band from mitochondria or from complex I, glutamate 1 was removed by Edman degradation (22).
Mass Spectrometric Analysis of Peptides-Samples (0.5 l) of tryptic and chymotryptic digests, respectively, of the 18 kDa and 10 kDa radiolabeled bands of complex I obtained by in-gel proteolysis (20) were mixed in a MALDI target plate with ␣-cyano-4-hydroxycinnamic acid (10 mg/ml in 60% acetonitrile containing 0.1% trifluoroacetic acid). The sample was dried, and MALDI-TOF spectra were measured in a TOF Spec 2E instrument (Micromass, Altrincham, Cheshire, UK) in the reflectron and positive ion modes. Spectra from the tryptic digest were calibrated internally with two autolysis products of trypsin (m/z values, 2273.1600 and 2163.0570) and with a product from the matrix (m/z, 1060.0482). Spectra from the chymotryptic digest were calibrated with peptides from subunit B9 (m/z values, 1918.5050 and 1165.6040, from subunit SDAP (m/z, 1887.0228) and from subunit MWFE (m/z, 1605.8138). The measured peptide masses were used to search data bases of masses of predicted proteolytic fragments of proteins using the MASSLYNX program with an error tolerance of 30 ppm. Once candidate phosphopeptides had been identified, 2 l of 40 mM ammonium bicarbonate was deposited on the peptide-matrix cocrystal on the MALDI plate. The crystals were mixed into the solution, and the samples were allowed to dry. Then 2 l of 0.2 unit/ml alkaline phosphatase was mixed into the same samples (25), and the plates were kept for 1 h at 37°C in a humidified vessel. The samples were dried at room temperature and acidified with 60% acetonitrile containing 0.1% trifluoroacetic acid. Then they were reanalyzed by MALDI-TOF mass spectrometry.
The tryptic digests of the 18 kDa and the 10 kDa radiolabeled bands from mitochondrial membranes and from complex I, and the chymotryptic digest of the 10 kDa radiolabeled band from complex I were fractionated by liquid chromatography on capillary reverse phase C18 columns (either 100 ϫ 0.18-mm inner diameter or 150 ϫ 0.075-mm flow rates 1 l/min or 0.15 l/min, respectively). The eluate was introduced directly into a Q-TOF (Micromass) or a Q-STAR hybrid tandem mass spectrometer (MDS Sciex, Concord, Ontario, Canada). A sample of the tryptic digest of the 18 kDa radiolabeled band from complex I was desalted in a microtip containing the reverse phase chromatography support, Poros R2 (Perseptive Biosystems Inc., Framingham, MA). The bound peptides were eluted in 1-2 l of 60% acetonitrile in 2% formic acid directly into a nanoelectrospray needle. Then the loaded needle was attached to a Q-STAR mass spectrometer. In the initial "survey" scan of samples introduced both by liquid chromatography and nanoelectrospray, m/z values of peptides in the range 400 -1,200 were measured. Selected ions were fragmented by CID with argon, and spectra were recorded in the second mass analyzer. The spectra were searched against an NCBI nonredundant data base with the search tool MAS-COT MS/MS Ions search (www.matrixscience.com). Peptide sequence tags were deduced from the data and compared with protein sequence data bases with PeptideSearch (26).
For the analysis of phosphopeptides in the tryptic digest of the 18 kDa radiolabeled band of complex I, the digest was desalted in a nanotip containing Poros R3. The bound peptides were eluted with 50% methanol directly into a nanoelectrospray needle and then introduced into a Q-STAR mass spectrometer. Phosphopeptides were identified in tandem MS experiments by examining the data for precursor ions that lose 79 Da by CID (27,28). Spectra were recorded in negative ion mode from 400 -1000 m/z with a step size of 0.5 Da and a dwell time of 50 ms, yielding a spectrum that contained only phosphopeptides. Phosphopeptide precursor ions were selected, fragmented by CID, and positive ion fragment ion spectra recorded.

Phosphorylation of Proteins in
Mitochondria-Purified mitochondria were incubated with [␥-32 P]ATP in the presence and in the absence of cAMP, and the proteins were analyzed by SDS-PAGE and autoradiography (see Fig. 1). Three radiolabeled bands with apparent molecular masses of 29, 18, and 10 kDa were phosphorylated strongly in the presence of cAMP, but much more weakly so in its absence ( Fig. 1, lanes (d) and (c), respectively). In contrast, a band at about 42 kDa was radiolabeled heavily in both the presence and the absence of cAMP. This band is likely to contain the E1␣ subunit of pyruvate dehydrogenase (see below), which is phosphorylated in a cAMP-independent manner by mitochondrial pyruvate dehydrogenase protein kinase (29). A large amount of radioactivity was detected also at the foot of the gel below stained bands, and a lesser amount at the top. The intensities of these radiolabeled bands were not influenced noticeably by the presence of cAMP ( Fig. 1, lanes (c) and (d)). The lower band probably arises from radioactive substrate and contains phospholipids, but without further investigation, the presence of a small radiolabeled protein obscured by large amounts of radioactive substrate cannot be excluded.
To investigate the properties of the phosphorylated amino acids in the radiolabeled proteins, the proteins were fractionated by SDS-PAGE, transferred to membranes, and treated with either acid or alkali (Fig. 1, lanes (e)-(h)). The radioactivity associated with the 42, 18, 16, and 10 kDa bands (and also of a much weaker band at 29 kDa) was resistant to acid treatment but was removed by alkali ( Fig. 1, lanes (g) and (h)). Because phosphotyrosine is stable in both acid and alkali, and phosphohistidine is stable above pH 8.0 but labile under acidic conditions, and phosphoserine and phosphothreonine residues are resistant to acid but labile under alkaline conditions (30), these experiments suggest that the 42-, 18-, 16-, and 10-kDa proteins (and the 29-kDa protein) are all phosphorylated on either serine or threonine residues.
In these experiments, there was also evidence for the presence of a phosphorylated protein at 39 kDa (partially resolved from the 42 kDa band in Fig. 1, lane (e)), which had the properties of a protein phosphorylated on a histidine residue ( Fig. 1(h). This protein is likely to be the ␣-subunit of succinyl-CoA ligase, which is autophosphorylated on histidine 259 (31). In the present work, no attempt has been made to confirm the identity of this band.
Analysis of 18-and 10-kDa Proteins Phosphorylated in Mitochondria-The protein contents of the 18 kDa and 10 kDa bands that were radiolabeled in mitochondria were analyzed by peptide mass mapping and tandem MS sequencing of peptides in tryptic digests of bands excised from SDS-polyacrylamide gels. By this means, evidence was obtained for the presence in the 18 kDa band of three subunits of bovine complex I, namely   subunits ESSS, B17, and the 18 k subunit (N-terminal sequence AQDQ) and subunit IV of cytochrome c oxidase. The 10 kDa band contained subunits SDAP, B9, MWFE, and MNLL from complex I, subunit VIII of complex III, subunit VIb of complex IV, and subunits F 6 and g of the ATP synthase complex (see Supplemental data). Although subunits ESSS and B17 and the 18 k subunit are not resolved in the SDS-PAGE system of Laemmli, as is known (14), it is also known that in Tricine-SDS-PAGE (or "Schä gger") gels, the 18 k subunit is resolved as a band with higher apparent molecular mass than subunits ESSS and B17. Therefore, the radiolabeled 18 kDa band was excised from a Laemmli gel and incorporated into a Schä gger gel. Two stained bands were observed after the second gel, and only the lower one was radioactive (Fig. 2). By peptide mass mapping, the upper band was shown to contain the 18 k subunit of complex I, and the lower band, subunits ESSS and B17 of the same complex. The gels were also run in reverse order, and essentially the same result was obtained (data not shown). Therefore, these experiments do not support the attribution of the 18 k (AQDQ) subunit as a site of phosphorylation in complex I (10).
In Vitro Phosphorylation of Inner Membranes and of Complex I-Submitochondrial particles prepared from bovine heart mitochondria were subjected to in vitro phosphorylation by [ 32 P]ATP in the presence of cAMP, and the proteins were fractionated by SDS-PAGE (Fig. 3). Radioactivity was detected in two major regions around 18 and 10 kDa. As expected, the radioactive bands at 42 and 39 kDa (representing two mitochondrial matrix components, the E1␣ subunit of pyruvate dehydrogenase and the ␣-subunit of succinyl CoA ligase, respectively) were not present, nor was the unidentified 29-kDa component, detected in phosphorylation experiments with mitochondria (lane (c) and Fig. 1). Incorporation of radioactivity into both the 18 kDa and 10 kDa bands was stimulated by the addition of purified complex I (compare lanes (a) and (b) in Fig.  3). In complex I alone (lane (e)), the 18 kDa and 10 kDa bands were radiolabeled heavily, and no phosphorylation of the complex was observed in the absence of exogenous protein kinase A. In complex I (lane (e)), three weakly radiolabeled bands were also observed. At present, no attempt has been made to identify them, but they do not include the 42-kDa subunit, which has been reported to be phosphorylated in mitochondria (32). 32 -P-phosphorylated complex I were fractionated by reverse phase HPLC, and the UV absorbance and the radioactivity of the eluant were monitored. Two major radioactive peaks eluting at 46 and 58% acetonitrile were found (Fig. 4, (a) and (b)). They both contained more than one component detected by staining with Coomassie Blue dye, but only single radioactive species with molecular masses of 18 and 10 kDa, respectively, were present. From previous HPLC analyses conducted on subcomplexes of complex I (14), it is known that these elution positions correspond to the MWFE and ESSS subunits of complex I. Subunits B9, SDAP, and MNLL, which were all detected by tandem MS peptide sequencing of the radiolabeled 10 kDa band in mitochondria, elute at 57, 54, and 38% acetonitrile, respectively, and subunit B17 and the 18 k subunit, found in the same way in the 18 kDa band of radiolabeled mitochondria, elute at 54 and 32% acetonitrile, respectively. Taken with other experiments described above, these data are sufficient to identify the phosphorylated subunits of bovine complex I as subunits ESSS and MWFE and to eliminate the 18 k subunit as a site of phosphorylation. These conclusions are borne out by experiments described below in which the phosphorylation sites have been identified directly by Edman sequencing and mass spectrometry.

Identities of the Phosphorylated 18-and 10-kDa Subunits of Complex I-The subunits of in vitro
Identification of Phosphorylation Sites by Edman Sequencing-The procedure was tested with the 42 kDa radiolabeled band from mitochondria. It confirmed that the E1␣ subunit of pyruvate dehydrogenase is phosphorylated on serine 264, at the so-called "site 1" (33). Details of these experiments are presented in the Supplemental data. Then the procedure was used to analyze the 18 and 10 kDa radiolabeled bands from mitochondria and from complex I. Radioactivity was released by Edman degradation at cycles 3, 6, and 19 of the tryptic, chymotryptic, and endoproteinase Asp-N digests, respectively, of the 18 kDa radiolabeled band both in mitochondria and in complex I (Figs. 5 and 6). These results are consistent with the phosphorylation of serine 20 of the ESSS subunit. They cannot be reconciled with the sequences of either the 18 k subunit of complex I or of subunit B17 (see Supplemental data). In the analysis of the 18 kDa band of complex I, in addition to the main release of radioactivity from tryptic peptides at cycle 3, lesser amounts were also released at cycle 14. The release at cycle 14 arises from a partial tryptic peptide (residues 7-26) present in relatively low yield. Both phosphopeptides (18 -26 and 7-26) were observed by mass spectrometric analysis of the same digest (see below). The same patterns of radioactive release were obtained from the 18 kDa band from mitochondria and the 18 kDa band from complex I.
Similarly, release of radioactivity at cycles 3 and 9 in the tryptic and chymotryptic digests, respectively, of the 10 kDa bands from mitochondria and complex I is consistent with the phosphorylation of serine 55 of subunit MWFE (Figs. 5 and 7).
These data cannot be reconciled with any of the sequences of the complex I subunits B9, SDAP, or MNLL (see Supplemental data). The tryptic digestion data show that cleavage of subunit MWFE occurred after arginine 52 and not after arginine 53. The lack of cleavage of this normally scissile bond is compatible with the phosphorylation of serine 55 because the presence of a phosphate group in this position relative to a tryptic cleavage site is known to impede the enzyme (34). In the data from the sequence analysis of the 10 kDa band from mitochondria, but not from complex I, radioactivity was also released in cycles 3 and 7 (Fig. 7). This release is the result of the presence of another phosphoprotein in this 10 kDa band, namely subunit d of the succinate dehydrogenase complex. However, this subunit was not detected by mass spectrometric analysis of the unfractionated 10 kDa band from mitochondria. 2 Mass Spectrometric Characterization of Phosphorylation Sites-The MALDI-TOF spectrum of the tryptic digest of the 18 kDa radiolabeled band contained ions of peptides from the complex I subunits ESSS, B17, and AQDQ ( Fig. 8(a)). Ions with m/z values of 1035.4908 and 2044.0972 correspond closely to two phosphopeptides from the ESSS subunit. Their sequences are RPSEPTLR (residues 18 -25) and AVIAPSTLAGKRPSEP-TLR (residues 7-25), with each of them bearing one phosphoryl moiety. Their neutral masses are 1034.4910 and 2043.0878, respectively. After treatment with alkaline phosphatase, these ions were replaced with new ions with m/z values of 955.5375 and 1964.2736, the mass differences of 80 in each case corresponding to the loss by enzymic hydrolysis of one phosphoryl group (Fig. 8(b)). Similarly, the MALDI-TOF spectrum of the chymotryptic digest of the 10 kDa radiolabeled band contained ions arising from peptides of the complex I subunits MWFE, 2  SDAP, and B9 (Fig. 8(c)). It also contained an ion with an m/z value of 1917.9144 (MH ϩ ) corresponding to the sequence LMERDRRVSGVNRSY (residues 47-61 of subunit MWFE) bearing one phosphoryl group (calculated MH ϩ 1917.9120). After treatment with alkaline phosphatase, this ion was replaced by an ion with m/z 1837.9621, the mass difference of 80 again corresponding to the enzymic removal of a phosphoryl group (Fig. 8(d)). In the MALDI-TOF spectra of the untreated proteolytic digests of subunit ESSS, there was no evidence of ions arising from nonphosphorylated versions of the phosphopeptides, indicating that the sites are fully phosphorylated in vitro.
Additional evidence for the phosphopeptide AVIAPST-LAGKRPSEPTLR (but not of phosphorylated RPSEPTLR) in the tryptic digest of the 18 kDa band was obtained by electrospray tandem MS experiments involving precursor ion scanning in negative ion mode in the mass range 400 -1,000 Da (see Supplemental data The product ion spectrum of the triple charged phosphopeptide AVIAPSTLAGKRPSEPTLR (m/z 682.36) was dominated by triple and double charged y type fragment ions (labeled y 3ϩ ions and y 2ϩ ions, respectively in Fig. 9). The characteristic loss of 98 Da was observed to arise from some y ions (e.g. y17 3ϩ and y16 2ϩ ) but not from the precursor ion (m/z 682.36). The amino acid sequence determined from this spectrum (AVIAPSTLAGKRPSEPTLR, where the underlined segment was deduced) identified the peptide unambiguously as a phosphorylated fragment of subunit ESSS. Although the modified serine could not be identified directly from this spectrum alone, unmodified serine 12 was identified by the ions y14 2ϩ and y13 2ϩ , and unmodified threonine 13 by y13 2ϩ and y12 2ϩ , leaving serine 20 or threonine 23 as the only possible phosphorylation site.
In the survey scan mass spectrum of the chymotryptic digest of the radiolabeled 10 kDa band of complex I, a triple charged ion ([Mϩ3H] 3ϩ m/z 639.99) was detected. It corresponds to a peptide with a measured mass of 1916.97, close to the calculated value for the sequence LMERDRRVSGVNRSY (residues 47-61 of complex I subunit MWFE) bearing a single phosphoryl moiety (calculated neutral mass of the phosphopeptide, 1916.9040). However, no satisfactory sequence information

Identities of Phosphorylated Proteins in Bovine
Mitochondria-The experiments described above confirm earlier observations that rather few abundant proteins in bovine mitochondria are phosphorylated in vitro in a cAMP-dependent manner (7)(8)(9). The experiments have confirmed the cAMP-independent phosphorylation of the E1␣ subunit of pyruvate dehydrogenase, and data are presented in the Supplemental section that support the earlier identification of its main site of phosphorylation as serine 264, the so-called site 1 (33). There was no evidence for phosphorylation at two other sites known as sites 2 and 3. Evidence was also uncovered that was compatible with the phosphorylation of a histidine residue in the ␣-subunit of succinyl CoA ligase (31). Both pyruvate dehydrogenase and succinyl CoA ligase are components of the mitochondrial ma-trix, and the main emphasis of the present work has been to identify other phosphorylated proteins that are associated with the inner membrane. In SDS-PAGE analyses of radiolabeled membranes, radioactive bands were observed migrating at 29, 18, 16, and 10 kDa, and similar patterns of labeling have been reported in the past. However, there was no evidence in this work for the phosphorylation of the 42-kDa subunit of complex I that has been reported to become heavily phosphorylated in "steady-state" conditions (32). The experiments in this paper are focused on the two relatively intense 18 kDa and 10 kDa bands. In past work the 10 kDa band has been referred to as the "6.5 kDa band" (8). By tryptic mass mapping of the radiolabeled band from mitochondrial membranes, the 18 kDa band was shown to contain three components of complex I (subunits ESSS, B17, and the so-called 18 k subunit) and a subunit of complex IV, and the lower band contained four additional subunits of complex I (SDAP, B9, MWFE, and MNLL), and also subunits of complexes III and IV and of ATP synthase. Bands at 18 kDa and 10 kDa were also radiolabeled by phosphorylation of purified complex I, and the same subunits of the complex were detected by mass spectrometric analysis of the bands as in FIG. 9. Identification of phosphorylation sites in subunit ESSS of complex I by mass spectrometry. Product ion mass spectra of phosphopeptides from a tryptic digest of radiolabeled 18 kDa band of complex I are shown. (a), portion of the spectrum from the fragmentation of an ion (mass 518.2 2ϩ ) corresponding to phosphorylated tryptic peptide T3 (residues 18 -25) from subunit ESSS. The partial sequence pSerGlu from the peptide RPSEPTLR was identified from the ion series y4 -y6. The spectra also contain the y1 ion (mass 175) from the C-terminal arginine residue, and the b2 ion (mass 254.1) corresponding to the N-terminal dipeptide Arg-Pro (not shown) aiding the identification of this peptide. Ions with m/z 469.2 and 684.3, marked with asterisks, arise by loss of phosphoric acid from either the double charged precursor ion or from a y type fragment ion, respectively. A portion of the spectrum has been magnified by a factor of 12. (b), product ion mass spectrum of a 3ϩ ion (mass 682.0) corresponding to the phosphorylated tryptic peptide T2-3 (residues 7-25) from subunit ESSS. The sequences deduced from a mixture of triple and double charged y type fragment ions are indicated above the spectrum. They enabled the sequence of the underlined region of peptide AVIAPSTLAGKRPSEPTLR to be deduced. The spectrum also indicates that the phosphoryl modification is C-terminal to residue 10 and that serine and threonine at residues 6 and 7 in the peptide are unmodified. The ions b2, b3, and b4 also identified the dipeptide sequence Ile-Ala in the same tryptic peptide. Ions marked with an asterisk arise by the ␤-elimination of phosphoric acid from phosphoserine. the analysis of radiolabeled membranes. By the use of two different gel systems to separate the proteins in the 18 kDa band, it was demonstrated conclusively that the 18 k subunit of complex I was not phosphorylated and that radioactivity was associated with either subunit ESSS or B17. By the use of proteases of known specificities and Edman sequencing of the digests of the 18 kDa band from both mitochondrial membranes and from complex I, the radioactivity was shown to be associated with serine 20 of subunit ESSS and not with subunit B17. Independent confirmation of the site of phosphorylation was obtained by tandem mass spectrometry of the tryptic digest of the 18 kDa band from complex I. Similar analyses of the 10 kDa band led to the conclusion that the MWFE subunit of complex I is phosphorylated on serine 55. No other complex I subunit in the 10 kDa band became modified.
Evidence was also obtained by fractionating the mitochondrial membranes and then carrying out mass spectrometric and Edman analyses that the 10 kDa radiolabeled band in mitochondria contained not only phosphorylated subunit MWFE, but also phosphorylated subunit d of complex II and also several other nonphosphorylated 10 kDa proteins that had not been detected by mass spectrometric analysis of the unfractionated band. The failure to find these proteins in the unfractionated band probably arises from their being hydrophobic proteins with relatively few trypsin cleavage sites. Suppression of ionization of peptides from these subunits by more abundant species could also be a factor. 2 The demonstration that subunit ESSS of bovine complex I is phosphorylated in a cAMP-dependent manner contradicts published work that concluded that the 18 k subunit with the N-terminal sequence AQDQ contains the site of phosphorylation of the 18 kDa band. This conclusion was based solely upon N-terminal sequencing of the 18 kDa radiolabeled band from which the N-terminal sequence AQDQ was deduced (10). In this experiment it is likely that the band also contained subunits B17 and ESSS. It would not have been possible to detect subunit B17 because it has no free N terminus, and the Nterminal region of subunit ESSS is difficult to sequence, as noted previously (13). Several consecutive residues in the Nterminal region of this subunit yield phenylthiohydantoin-derivatives that are difficult to detect, especially so when they are mixed at each cycle with another prominent phenylthiohydantoin-derivative from the N-terminal sequence of the 18 k subunit. The difficulties of analyzing the N-terminal sequence of subunit ESSS led it being overlooked as a subunit of complex I until recently (13). In contrast to the present work, no attempt was made in the earlier analyses of the 18 kDa phosphorylated band to determine the site of phosphorylation experimentally. Instead, it was deduced (wrongly) from the sequence of the 18 k subunit (10).
The sites of phosphorylation of both the ESSS and MWFE subunits are in sequences that conform to the first of the two canonical motifs R/KXS/T and R/KXXS/T for cAMP-dependent phosphorylation sites, where S (or T) represents the phosphorylated residue (35).
Topologies of Subunits ESSS and MWFE-The hydrophobicity plots of subunits ESSS and MWFE suggest that each contains a single transmembrane ␣-helix from residues 58 -80 and 5-22, respectively (13,16). The presence of a hydrophobic region in each protein is consistent with the presence of ESSS in subcomplex I␤ (which represents part of the membrane arm of the intact complex) and of MWFE in the membrane sector of subcomplex I␣ (13). Thus, the phosphorylation site precedes the transmembrane region in ESSS, and it follows it in subunit MWFE. The uptake of [ 32 P]ATP into mitochondria probably is mediated by the ADP/ATP translocase, and carboxyatractylate, a specific inhibitor of the translocase, has been reported to inhibit completely incorporation of radioactivity into the 18 kDa band, but to inhibit incorporation of radioactivity into the 10 kDa band only partially (by about 50%) (7). The complete inhibition of phosphorylation of the 18 kDa band is compatible with the phosphorylation site for subunit ESSS being in the mitochondrial matrix, which in combination with the hydrophobicity profile would suggest that the N terminus of this protein is in the mitochondrial matrix. The partial inhibition of phosphorylation of the 10 kDa band is compatible with the presence of two phosphorylated proteins in this band, one with its phosphorylation site in the mitochondrial matrix, and the other with its phosphorylation site in the intermembrane space. As mentioned above, two phosphorylated proteins were detected in the 10 kDa band, subunit MWFE from complex I and subunit d from complex II. The phosphorylation site for subunit d appears to be in the mitochondrial matrix. 2 If so, the partial inhibition of incorporation of radioactivity in the presence of atractyloside suggests that the phosphorylation site for subunit MWFE would be in the intermembrane space, and its N terminus would be in the mitochondrial matrix. These preliminary deductions about the topologies of ESSS and MWFE and their phosphorylation sites will require further experimental verification.
Biological Roles of Phosphorylation of Subunits of Complex I-Little or nothing is known about the possible role, if any, of subunit ESSS in the enzymic activity and/or in the assembly of complex I, and neither subunit ESSS nor MWFE appears to be endogenously phosphorylated in the enzyme isolated from bovine heart (14). The observed in vitro phosphorylation of serine 20 of the bovine protein cannot play a general role in either the activity or the mechanism of the enzyme because the residue is not conserved, for example in the mouse and human proteins, despite the proteins being well conserved overall (69 and 86% identity of bovine with mouse and human homologues, respectively) (15). More is known about the role of subunit MWFE in complex I. It has a key role in the pathway of assembly of the complex, and its expression is linked to mitochondrial protein synthesis and to the assembly of the seven hydrophobic subunits of complex I (ND1-ND6 and ND4L) that are encoded in the mitochondrial genome (36,37). It remains to be established how the in vitro phosphorylation reported here may influence its properties.
Finally, it is worth noting that subunits ESSS and MWFE are the only subunits of human complex I to be found on the X chromosome (15,38). The promoter region of the NDUFA1 gene encoding subunit MWFE contains a cAMP response element, suggesting that expression of this subunit is linked to cAMP signaling pathways that regulate cellular energy metabolism (38). It is not known whether the gene for ESSS is regulated in a similar way.