Definition of the mitochondrial proteome by measurement of molecular masses of membrane proteins

Abstract

The covalent structure of a protein is incompletely defined by its gene sequence, and mass spectrometric analysis of the intact protein is needed to detect the presence of any posttranslational modifications. Because most membrane proteins are purified in detergents that are incompatible with mass spectrometric ionization techniques, this essential measurement has not been made on many hydrophobic proteins, and so proteomic data are incomplete. We have extracted membrane proteins from bovine mitochondria and detergent-purified NADH:ubiquinone oxidoreductase (complex I) with organic solvents, fractionated the mixtures by hydrophilic interaction chromatography, and measured the molecular masses of the intact membrane proteins, including those of six subunits of complex I that are encoded in mitochondrial DNA. These measurements resolve long-standing uncertainties about the interpretation of the mitochondrial genome, and they contribute significantly to the definition of the covalent composition of complex I.

Approximately one-third of the proteins encoded in genomes are hydrophobic membrane proteins, and yet the mass spectrometric analysis of such proteins remains technically challenging. The difficulties of fractionating membrane proteins on 2D gels to allow their proteomic analysis are well known. Many of them fail to enter the first-dimension isoelectric focusing gel and are lost. Others are not detected by staining. Nonetheless, many membrane proteins have been identified by mass spectrometric analysis of peptides allowing the proteins to be associated with specific gene sequences, although few constituent peptides are analyzed usually and therefore the coverage of the sequence is low. However, because of posttranslational modifications, the covalent structure of a protein is incompletely defined by its gene sequence. Sometimes the modifications can be deduced from the precise molecular mass of the protein measured by MS and further analysis of protein ions (1, 2), but often more detailed analyses of constituent peptides are needed to identify exact sites of modification. To carry out intact molecular mass measurements on membrane proteins, they have to be removed from their phospholipid environment and kept in a soluble state while being purified. The most common way of achieving this end is to extract and fractionate the membrane proteins in detergent, and to analyze peptides in protease digests of the fractionated detergent solubilized proteins by MS. Alternatively, membrane proteins have been extracted in organic solvents, fractionated by SDS/PAGE, and identified by MS analysis of peptide digests (3, 4). Unfortunately, detergents are incompatible with the MALDI and electrospray MS techniques used for measuring accurate masses of intact proteins, and so the detergents have to be removed first while the protein is kept soluble in an organic solvent. This objective has been achieved by chromatography in 60% formic acid, allowing protein masses of detergent-purified proteins to be measured (5). Other proteins have been purified by gel filtration in organic solvents before measurement of intact protein masses (6–8).

We have devised a different approach that avoids the use of detergents altogether, where the proteins are extracted from membranes in organic solvents and then fractionated in the same solvents by a variant of normal phase chromatography known as hydrophilic interaction chromatography (HILIC). HILIC is performed with a partially aqueous mobile phase and decreasing gradients of organic solvent, and so the components in the mixture being fractionated elute in order of increasing polarity (9). Many of the proteins recovered from the HILIC column bind irreversibly to reverse-phase columns. Then the molecular masses of the fractionated membrane proteins were determined by electrospray MS. As part of our attempts to help characterize the mitochondrial membrane proteome, we have applied these techniques to membranes from bovine mitochondria. The inner membranes contain >100 intrinsic membrane proteins involved in oxidative phosphorylation, transport of small molecules in and out of the organelle, import of nuclear encoded proteins and ion channels. Many additional peripheral membrane proteins are bound to the intrinsic membrane proteins, for example in the respiratory complexes. In the present work, the molecular masses of ≈30 of these hydrophobic proteins have been measured. The procedures were also shown to be useful for the analysis of hydrophobic proteins in the membrane-bound protein complex NADH:ubiquinone oxidoreductase (or complex I) that has been purified from inner membranes of bovine mitochondria in the presence of dodecylmaltoside. Bovine complex I is an assembly of 45 proteins with a combined molecular mass of ≈1 MDa (10, 11). They are organized into an L-shaped assembly with a peripheral arm projecting into the mitochondrial matrix attached to an intrinsic membrane arm (12, 13). Among others, the molecular masses of six extremely hydrophobic proteins have been measured in this way. These six proteins are all encoded in mitochondrial DNA (14). They form most of the membrane arm of the complex by contributing ≈50 transmembrane α-helices (15). These measurements resolve some residual ambiguities about the interpretation of the sequence of mammalian mitochondrial DNA (14, 16), and also they are another significant step toward defining the precise chemical composition of complex I.

Results and Discussion

Extraction of Proteins from Mitochondrial Membranes and Complex I.

In previous work (17), ≈15–20 membrane proteins were extracted from bovine mitochondria in mixtures of chloroform and methanol and fractionated by gel filtration in the solvent. However, many intrinsic membrane proteins were not in the extracts, and the selectivity for membrane proteins was lost when the samples contained detergents. Therefore, in the present work, a number of solvents that might be compatible with other kinds of chromatography were used to extract proteins from mitochondrial membranes. Buffered mixtures of propan-2-ol and acetonitrile proved to be the most suitable. The optimal pH of extraction for achieving maximal selectivity and yield of membrane proteins was 3.7. At lower pH, the extracts contained a wider range of proteins, including globular proteins (Fig. 1A). Above pH 3.7, progressively fewer membrane proteins were extracted, the quantities diminishing with increasing pH. The composition of the solvent mixture was investigated also, and mixtures of acetonitrile containing >60% (vol/vol) propan-2-ol provided conditions for selective extraction (Fig. 4, which is published as supporting information on the PNAS web site). The extracts were analyzed by SDS/PAGE and tandem mass analysis of tryptic digests of the stained bands (Fig. 1B). In this way, 29 proteins were identified in the extract (Table 2, which is published as supporting information on the PNAS web site).

Fig. 1.

Extraction of membrane proteins from mitochondria and isolated complex I with organic solvent. The solvent contained 70% propan-2-ol and 25% acetonitrile, 0.56% hexafluoro-isopropanol, and aqueous buffer at various pH values (see Experimental Procedures). (A and C) Extraction of bovine mitochondria and complex I at pH 2.7, 3.7, 4.7, 6.0, and 10.8 (lanes b–f, respectively). M, molecular mass markers (kDa); lanes a, total mitochondrial proteins and complex I, respectively. (B and D) Identities of proteins in pH 3.7 extracts of mitochondria and complex I, respectively, determined by tandem MS of tryptic digests of the bands. In B and D, the positions of proteins that previously had not been characterized from either mitochondria or complex I, respectively, are shown on the left of the gel, and those of known components are shown on the right. The position of subunit ND4L was determined with an antiserum.

As with mitochondrial membranes, the number of proteins extracted from complex I and the specificity of extraction were both influenced by the pH and composition of the solvent. Again, the most suitable pH value was 3.7 (Fig. 1C). From the 45 different subunits that form the complex, 22 were detected in the extract (see Fig. 1D and Table 3, which is published as supporting information on the PNAS web site). All of them, except subunits B14.5a, B14, B13, B8, and 10 kDa, are authentic membrane proteins with the potential to form at least one transmembrane α-helix. Fifteen of them have been characterized before (18, 19). As in the mitochondrial extract, 6 of the 7 hydrophobic ND subunits that are encoded in mitochondrial DNA were found. Subunit ND4L was detected in both extracts with an antibody (see Fig. 5 and Supporting Experimental Procedures, which are published as supporting information on the PNAS web site). Not only is ND4L hydrophobic, it contains only a single tryptic cleavage site, and so it has not been detected by mass mapping of tryptic digests, although it has been identified by tandem MS analysis of a cyanogen bromide peptide (19). However, subunit ND6 was not detected in either extract.

Fractionation of Solvent Extracts by HILIC.

Almost all of the proteins in the extract of mitochondrial membranes were bound to the column. They were eluted by a linear gradient of decreasing organic solvent (Fig. 2A), analyzed by SDS/PAGE (Fig. 2B), and identified by tandem MS analysis of tryptic peptides (Table 4, which is published as supporting information on the PNAS web site). Thus, transmembrane protein 14C was found, as well as 25 of the 29 components that had been identified by tandem MS analysis of the unfractionated extract.

Fig. 2.

Fractionation of hydrophobic proteins by HILIC. Solvent extracts of bovine mitochondria and complex I were made at pH 3.7 and fractionated by HILIC. (A and C) Elution profiles (solid lines) monitored by UV absorbance at 225 nm. The column was equilibrated in buffer 1 and eluted with a gradient (dotted lines in A and C) of buffer 2. The flow rates in A and C were 0.1 and 0.05 ml/min, respectively. For additional details, see Experimental Procedures. (B and D) Analysis of fractions by SDS/PAGE. Lanes Ex, extracts of mitochondria and complex I, respectively; M, molecular mass markers. The identified proteins are indicated on the right with fraction numbers and peak names in parentheses.

Most of the subunits extracted from complex I were retained completely by the HILIC column, except for the ND subunits, which were retained partially (Fig. 2 C and D). Minor variations in elution positions of subunits (although not elution order) from one experiment to another were attributed to differences in the amount of detergent retained in different samples of complex I by concentration of dilute solutions by membrane filtration.

Measurement of Intact Molecular Masses.

The intact molecular masses of 24 proteins in the HILIC fractions from the mitochondrial extract were measured by electrospray ionization (ESI)-MS (Table 1 and Table 5, which is published as supporting information on the PNAS web site). Twenty-two of them were identified also in the unfractionated extract, but two others, complex III subunit XI and cytochrome oxidase subunit VIII, had not been found in the unfractionated extracts. Therefore, the total number of proteins that were found in the extract was 33 (29 plus ND4L in the unfractionated material, plus transmembrane protein 14C, subunit XI of complex III, and subunit VIII of cytochrome oxidase). Thirty-one of them seem to be authentic membrane proteins with the capacity to form at least one transmembrane α-helix. The exceptions are subunit IX of complex III [which is the mitochondrial import sequence cleaved from the Rieske protein that remains associated with the complex III outside the membrane domain (20)] and cytochrome oxidase subunit VIb.

Table 1.

Molecular masses of some hydrophobic proteins extracted from bovine heart mitochondria and complex I

Protein	Mass, Da		Mass difference	Modification
	Observed	Calculated
Transmembrane protein 14C*	11,604.4	11,735.0	−130.6	−Met
Usmg5*†	6,303.5	6,434.6	−131.1	−Met
Brain protein 44-like*	12,299.5	12,388.6	−89.1	−Met + acetyl
Brain protein 44*	14,193.1	14,281.9	−88.8	−Met + acetyl
Phospholamban*	6,122.7	6,080.5	+42.2	+ acetyl
ND1‡	35,699.1	35,670.2	+28.9	+ formyl
ND2*‡	39,283.5	39,254.4	+29.1	+ formyl
ND3‡	13,082.4	13,054.7	+27.7	+ formyl
ND4‡	52,130.0	52,099.4	+30.6	+ formyl
ND4L‡	10,825.4	10,797.3	+28.1	+ formyl
ND5‡	68,319.8	68,286.8	+33.0	+ formyl
B14.7‡	14,667.5	14,758.1	−90.6	−Met + acetyl

Masses were measured by ESI-MS and calculated from protein sequences. The masses of the proteins in the table have not been measured previously. The masses of other proteins in the mixtures that confirm known values are summarized in Tables 5 and 6.

*Isolated from bovine mitochondrial membranes.

^†Up-regulated during skeletal muscle growth protein 5, also known as diabetes-associated protein in insulin-sensitive tissue (DAPIT).

^‡Isolated from bovine complex I.

The intact molecular masses of 18 of 24 of the proteins have been characterized before (Table 5) (2, 18, 19, 21–24). However, the intact masses of six of the proteins had not been measured previously (summarized in Table 1), although they have been identified as components of mitochondria (25, 26). One of them, ND2, is encoded in mitochondrial DNA, and five others are encoded in the nucleus. These data show that, in common with other proteins encoded in mtDNA (17) and with proteins encoded in chloroplast genomes (5, 27), ND2 has retained its translational initiator, formylmethionine, at residue 1, and that it is not modified at any other site. These data demonstrate also that the translational initiator methionine residues of four of the five nuclear-encoded subunits (transmembrane protein 14C, Usmg5, brain protein 44, and brain protein 44-like) have been removed. Two of them (transmembrane protein 14C and Usmg5) are not modified further, whereas brain protein 44 and brain protein 44-like have been acetylated, probably at their N termini. In the remaining nuclear encoded subunit, phospholamban, the translational initiator methionine has been retained and N-α-acetylated (Fig. 6B, which is published as supporting information on the PNAS web site). Phospholamban is a component of the sarcoplasmic reticulum (28), although it has been detected before in mitochondria (26, 29), and so it may be targeted to both compartments. Residue 2 of brain protein 44 and of brain protein 44-like are serine and alanine, respectively, the two most common acetylated N-terminal residues (30), and the N terminus of mature transmembrane protein 14C is proline.

Five additional subunits of complex I that are encoded in mitochondrial DNA (ND1, ND3, ND4, ND4L, and ND5) were detected in the mitochondrial extract (Fig. 1B), but there was insufficient material to permit measurement of their intact masses. However, by fractionation of the extract of complex I, the same subunits plus subunit ND2 were recovered as simplified mixtures in the break-through of the HILIC column in sufficient quantity to allow their molecular masses to be measured (see Fig. 2). The spectra show that all six subunits have retained their translational initiator formylmethionine and that otherwise they are unmodified (Table 1 and Fig. 3).

Fig. 3.

Analysis by ESI-MS of subunits ND1, ND2, ND3, ND4, ND4L, and ND5 of complex I. (A and C) Analysis of fractions FT2 and FT1, respectively, from Fig. 2C. (A) The four multiply charged series (components A–D) correspond to subunits ND2, ND1, ND4, and ND5, respectively. (C) The two multiply charged series (components A and B) correspond to subunits ND3 and ND4L, respectively. (B and D) The spectra, reconstructed on a true molecular mass scale, show the molecular masses (in daltons) of the subunits (summarized in Table 1).

Tandem mass analysis of the N-terminal peptide of subunit B14.7 of complex I has shown previously that the translational initiator methionine has been removed, and that alanine-2 is acetylated (31). The intact subunit has never been recovered by reverse-phase HPLC fractionation of subunits of complex I (18, 19), but it was obtained by HILIC fractionation of the extract of complex I, allowing its intact molecular mass to be measured (Table 1 and Fig. 7, which is published as supporting information on the PNAS web site). This experiment confirmed the N-α-acetylation of residue 2 as its only posttranslational modification. It is the most hydrophobic of the nuclear encoded proteins of the complex, with the potential to form three transmembrane α-helices. The ESI-MS analysis of the HILIC fractions also confirmed the molecular masses of 11 subunits of bovine complex I that had been characterized before (Table 6, which is published as supporting information on the PNAS web site).

Biological Significance.

The development of methods to permit the fractionation and characterization of membrane proteins by MS has a general significance for proteomic studies. The extraction procedure is rather selective for membrane proteins, but for reasons that are as yet obscure, not all hydrophobic protein components of the mitochondrial membranes are present in the extracts. For example, two notable absentees are the c-subunit of ATP synthase and the ND6 subunit of complex I, which are both abundant components of the membranes. Also, it has the advantages of allowing membrane proteins to be extracted from detergent-purified protein complexes and of being compatible with HILIC. The use of this form of chromatography for purifying membrane proteins is a new development that has allowed some well known, but hitherto poorly characterized proteins to be studied, namely six of the seven hydrophobic proteins that form most of the membrane arm of complex I and are encoded in mammalian mitochondrial DNA.

The measurement of the molecular masses of subunits ND4, ND4L, and ND5 helps to remove some long-standing uncertainties about the interpretation of the sequences of bovine and human mitochondrial genomes (14, 16). The molecular masses of bovine subunits ND4 and ND4L show that the corresponding genes do overlap by 7 bases, as proposed but never proven, and the molecular mass of ND5 is compatible with the proposed 17-base overlap of the corresponding gene with that of ND6 on the opposite DNA strand. The measured mass of ND4 corresponds to the product from the polyadenylated ND4 mRNA rather than to that of ND4X, the proposed “extended” version of the protein that could be made by read-through from ND4 into the following tRNA^His gene, producing a protein 17 aa longer. It now seems less likely that this read-through mechanism operates in either bovine or human mitochondria.

The molecular masses of the ND subunits (excepting ND6) show that, apart from the retention of their N-terminal formyl groups, the proteins are not covalently modified. This finding is significant in the context of the proton-pumping mechanism of complex I. Transfer of two electrons, one at a time, from NADH to ubiquinone, via flavin mononucleotide and nine iron-sulfur clusters, is accompanied by the transfer of four protons from the matrix of the mitochondrion to the inter-membrane space (32). However, the mechanism of proton transfer is not understood. Various proposals have been made, including a conformational coupling mechanism (33–35), a mechanism related to the Q-cycle that operates in complex III (36), and a mechanism depending on the possible presence of an unidentified cofactor in the membrane domain of the enzyme (37, 38). If this cofactor exists, it is not attached to any of the six ND subunits that have been analyzed (assuming that it is stable at pH 3.7), nor to any of the 13 nuclear-encoded subunits that contribute to the membrane domain. It remains possible that such a cofactor could be attached to ND6, or that it could be bound noncovalently to the membrane arm of complex I.

Experimental Procedures

Preparation of Bovine Heart Mitochondria and Purification of Complex I.

Mitochondria were isolated from bovine hearts, and complex I was purified from them in the presence of n-dodecyl-β-d-maltoside (Anatrace, Maumee, OH) as described before (19, 39, 40). The purified complex I was concentrated to ≈5–7 mg/ml on a Vivaspin-4 membrane with a molecular mass cut-off of 100 kDa (Sartorius, Göttingen, Germany), and the concentrate was stored in liquid nitrogen.

Extraction of Membrane Proteins from Mitochondria and Complex I.

Mitochondria were resuspended at 4°C in a solution consisting of 2 mM Tris·HCl (pH 7.4), 250 mM sucrose, and 1 mM EDTA and then centrifuged (16,000 × g, 6 min, 4°C). The pellet (defined as 1 vol), or the sample of complex I, was vortexed occasionally for 5 min at room temperature in 9 vol of mixtures of propan-2-ol and acetonitrile from 95% (vol/vol) propan-2-ol to 95% (vol/vol) acetonitrile containing 0.56% hexafluoroisopropanol and 4.44% water buffered to a final concentration of 20 mM with formic acid (pH 2.7), ammonium formate (pH 3.7 and 4.7), ammonium acetate (pH 6.0), or ammonium hydroxide (pH 10.8). Insoluble material was removed by centrifugation (16,000 × g, 10 min, 18°C).

SDS/PAGE Analysis of Proteins.

Mitochondrial extracts (≈100 μl) were delipidated by addition of 4 vol of diethyl ether at −20°C. After being kept at −20°C for 30 min, the samples were centrifuged (16,000 × g, 10 min, 4°C). The pellets were resuspended in 100 μl of a mixture of chloroform and methanol (2:1, vol/vol) containing 1 mM DTT and 0.01 vol of a 20% (wt/vol) solution of SDS. SDS (20%) was added also to the same concentration to portions of organic solvent extracts of complex I. The samples were dried in vacuo and resolubilized in sample buffer, and the pH values of any acidic samples were adjusted to neutrality. Samples were analyzed by SDS/PAGE on 12–22% gradient gels and stained with Coomassie Blue R250 (41).

Protein Identification.

Excised protein bands from SDS/PAGE gels were digested with trypsin (Roche, Mannheim, Germany) (31, 42). Peptides were extracted from the gel in 4% formic acid containing 60% acetonitrile. Alternatively, proteins in portions of HILIC fractions were dried completely in vacuo, redissolved in 50 mM ammonium bicarbonate containing 0.25 mM CaCl₂, and digested with trypsin at 37°C. Peptides were analyzed by peptide mass fingerprinting and tandem MS peptide sequencing in a 4700 MALDI-TOF-TOF mass spectrometer (Applied Biosystems, Warrington, U.K.) by using α-cyano-4-hydroxy-transcinnamic acid as the matrix. Sample data were calibrated internally with the autolysis products of trypsin. Monoisotopic peak mass lists and tandem MS data were screened against NCBInr protein sequence databases by using MASCOT from either an internal server or via a web interface (43). Alternatively, tandem MS analysis was performed in a Q-TOF instrument (Micromass, Altrincham, U.K.) as described (41).

Fractionation of Extracts.

Protein extracts of mitochondrial membranes and complex I were fractionated on polyhydroxyethyl-aspartamide columns (100 mm × 2.1 mm i.d. or 100 mm × 1 mm i.d, 300-Å pore size, and 5-μm particle size; PolyLC, Columbia, MD) equilibrated in buffer 1 (63% propan-2-ol/22.5% acetonitrile/0.5% hexafluoro-isopropanol/14.0% water/20 mM ammonium formate, pH 3.7). The proteins were eluted with a linear gradient of buffer 1 with buffer 2 (30% propan-2-ol/0.5% hexafluoro-isopropanol/69.5% water/20 mM ammonium formate, pH 3.7). The absorbance of the eluate was monitored at 225 nm.

Molecular Mass Measurements.

The protein contents of fractions were analyzed by LC-ESI-MS “on-line” to the column (50 × 1 mm i.d.) with a Quattro Ultima triple quadrupole mass spectrometer (Micromass). Alternatively, samples were introduced “off-line” into a stream (flow rate 2–3 μl/min) of 50% aqueous acetonitrile into either an electrospray interface of a Quattro Ultima instrument, or a nano-flow electrospray interface of a Q-TOF mass spectrometer (Micromass). In some cases, the spectra were improved by addition to the sample of formic acid (final concentration 1–5%, vol/vol). Both mass spectrometers were operated in positive ion single MS mode. They were calibrated with horse heart myoglobin and bovine trypsinogen. Protein molecular masses were determined from the series of multiply charged ions by using the component analysis function of MassLynx software, version 3.5 (Micromass).

Abbreviations

HILIC

hydrophilic interaction chromatography

ESI

electrospray ionization

ND

subunits of NADH dehydrogenase encoded in mitochondrial DNA