Expanding the subproteome of the inner mitochondria using protein separation

Transcription

1 MCP Papers in Press. Published on September 25, 2006 as Manuscript T MCP200 Expanding the subproteome of the inner mitochondria using protein separation technologies: one and two-dimensional liquid chromatography and two-dimensional gel electrophoresis. Todd McDonald* 1,5, Simon Sheng* 1, Brian Stanley* 1,5, Dawn Chen 2,4, Young Ko 1, Robert N. Cole 2,4, Peter Pedersen 2, and Jennifer E. Van Eyk 1-3,5 Departments of Medicine 1, Biological Chemistry 2 and Biomedical Engineering 3 and The Technical Implementation and Coordination Core of NHLBI Proteomics Center 4, Johns Hopkins University, Baltimore, Maryland, USA 21224, and Departments of Physiology 5 Queen s University, Kingston, Ontario, Canada K7L 3N6 Running title: complementary protein separation of the inner mitochondrial membrane Key words: inner mitochondrial membrane, protein separation, 2-dimensional gel electrophoresis, reversed phase chromatography, one-dimensional liquid chromatography, 2- dimensional liquid chromatography * equal contribution Corresponding author: Jennifer E. Van Eyk 602 Mason F. Lord Bldg., center tower, 5200 Eastern Ave Johns Hopkins University, Baltimore, MD Tel Fax jvaneyk1@jhmi.edu Copyright 2006 by The American Society for Biochemistry and Molecular Biology, Inc. 1

2 Abbreviations ASB-14: Amidosulfobetaine-14 SB3-10: N-decyl-N-N'-dimethyl-3-ammonio-1-propane sulfonate IMM: Inner Mitochondrial Membrane MudPIT: Multidimensional Protein Identification Technology PF2D: ProteomeLab TM PF 2D Protein Fractionation System (Beckman Coulter) 1-DLC: Reversed phase HPLC separation of proteins by hydrophobicity 2-DE: Two-dimensional gel electrophoresis 2-DLC: Two-dimensional liquid chromatography RP-HPLC: Reverse phase high performance liquid chromatography CF: Chromatofocusing PTM: Post-translational modification 1D SDS-PAGE: One dimensional sodium-dodecyl sulfate polyacrylamide gel electrophoresis 2

3 Summary Currently no single proteomic technology has sufficient analytical power to allow for the detection of an entire proteome of an organelle, cell or tissue. One approach that can be used to expand proteome coverage is the use of multiple separation technologies especially if there is minimal overlap in the proteins observed by the different methods. Using the inner mitochondrial membrane subproteome as a model proteome, we compared for the first time the ability of three protein separation methods (two-dimensional liquid chromatography using the PF2D system from Beckman Coulter, 1-D reverse-phase high performance liquid chromatography and twodimensional gel electrophoresis) to determine the relative overlap in protein separation for these technologies. Data from these different methods indicated that a strikingly low number of proteins overlapped, with less than 24% of proteins common between any two technologies, and only 7% common amongst all three methods. Utilizing the three technologies allowed the creation of a composite database totaling 348 non-redundant proteins. Eighty two percent of these proteins had not been previously observed in proteomic studies of this subproteome, whereas 44% had not been identified in proteomic studies of intact mitochondria. Each protein separation method was found to successfully resolve a unique subset of proteins with the liquid chromatography methods being more suited for the analysis of transmembrane domain proteins and novel protein discovery. As well, we demonstrated that both the 1- and 2-DLC allowed for the separation of the α-subunit of F 1 F o ATP synthase which differed due to a change in pi or hydrophobicity. 3

4 Introduction The eukaryotic proteome is a compilation of proteins which represents the integration of numerous cellular processes that begin with the variable transcription of genes to mrna. These products are then translated to proteins which may in turn be potentially co- and/or post translationally modified (PTMs) to produce an array of proteins (1,2). Due to the large number of unique protein species produced, coupled with differences in their relative abundance, there is as of yet no single proteomic technology that has the analytical capacity or sensitivity to realize the goal of complete proteome coverage. One strategy to maximize proteome coverage is to combine synergistic proteomic technologies, particularly if each technology reveals a unique subset of proteins. Using the inner mitochondrial membrane subproteome as a model subproteome, we compared the ability of three protein separation methods (two-dimensional liquid chromatography (2-DLC, PF2D), 1-dimensional reversed-phase high performance liquid chromatography (1-DLC, RP-HPLC), and two-dimensional gel electrophoresis (2-DE) in order to determine the relative overlap in protein separation for these technologies. 2-DE, a classical proteomic technology that separates proteins based on their pi and Mw, has a practical dynamic range of 10 4 orders of magnitude (Reviewed in 3, 4). This restricts the analysis of a proteome to the most abundant proteins and so it often under-represents proteins with extreme hydrophobicity, mass or isoelectric point. Another common separation method is 1-DLC which separates proteins based on hydrophobicity. In proteomics, 1-DLC has primarily been used for peptide separation prior to mass spectrometry (MS), but it can be used for protein separation prior to enzymatic digestion and analysis by MS (Reviewed in 5; e.g. 6,7). 2-DLC traditionally couples a charge based (e.g. isoelectric, strong cation exchange) as a first dimension with RP-HPLC as the second dimension thereby increasing the extent of protein fractionation 4

5 compared to 1-DLC. As with 1-DLC, this method has been used primarily in proteomics for peptide separation; however, it is increasingly being applied to the separation of complex intact protein mixtures. (Reviewed in 5; e.g. 8-10). This increased use is (in part) due to the commercialization of 2-DLC systems, including the PF2D (Beckman Coulter), which is based upon the system developed by Ludman and colleagues (e.g.11-14). The PF2D system uses chromatofocusing in the first dimension (separating proteins based on their pi) and reversed phase chromatography in the second dimension. Except for a single report utilizing chromatographic isoelectric focusing (first dimension) to separate peptides prior to a MudPIT experiment (15) the PF2D system has been used exclusively for protein separation. To date, the only comparison of the PF2D with any other protein separation technology examined the rice proteome through a limited comparison between the PF2D and 2-DE (16). Thus, both the scope of proteome coverage by PF2D alone and its synergy with other proteins separation method is not clear. Mitochondria generate the majority of ATP in the cell and their dysfunction has been implicated in many different diseases. Mitochondria have both an outer and an inner membrane structure with the components of the oxidative phosphorylation pathway located in (or associated with) the inner mitochondrial membrane (IMM). In order to understand these diseases, a determination of the members of this subproteome (including PTMs) is of great interest (17-19). Although intact mitochondria have been studied using different proteomic technologies (20-30), these databases comprise only part of the estimated 697 to 4532 total mitochondrial proteins (31). However, because these estimations can have a false discovery rate of up to 68% (31), the absolute number of mitochondrial proteins is not currently known. A problem with the existing mitochondrial databases derived from proteomic analysis has been the bias toward proteins 5

6 localized to the matrix and outer membrane and the lack of IMM-associated proteins (21,25). To increase the coverage of the IMM subproteome, Da Cruz et al (29,30) used an enriched IMM preparation and demonstrated that there are novel proteins within this subproteome. Using the same well-characterized IMM preparation (29-32) we tested the hypothesis that there would be a minimal overlap of observed proteins when using three different separation technologies (2-DE, 1-DLC and 2-DLC) thereby expanding proteome coverage. Experimental Procedures IMM preparation: Mitochondria were isolated from frozen rat liver and the IMM subproteome was isolated according to the protocol of Pedersen et al. (32). Purity was assessed as described previously (30). For 2-DE, the IMM proteins were solubilized by incubation in either 5% (w/v) CHAPS, 5% (w/v) SB3-10, or 5% (w/v) ASB-14 in ddh 2 0 for 15 minutes at room temperature, prior to the addition IEF buffer (8 M urea, 2 M thiourea, 4% detergent, 1% (w/v) DTT, and 0.25% (v/v) carrier ampholites). For 2-DLC, the IMM proteins were solubilized in 1% (w/v) SDS followed by precipitation with ice-cold acetone, and resuspended in 2 ml of PF2D chromatofocusing start buffer (Beckman Coulter, Carlsbad, CA). For 1-DLC the IMM subproteome was solubilized in 0.1% (v/v) triflouroacetic acid with 20% (v/v) acetonitrile, ph DE analysis: The IMM subproteome (200 to 750 µg) was resolved on IPG Ready Strips (17 cm, ph 4-7 or 3-10 linear gradient (BioRad)). Strips were actively rehydrated with solubilized protein in 350 µl of IEF buffer at 50 V for 10 h, then a rapid voltage ramping method was applied as follows: 100 V for 1 h, 500 V for 1 h, 1000 V for 1 h, linear gradient to V over 1 h, and finally V for 40 kvh using a Protean IEF cell (BioRad). To separate proteins 6

7 with basic pis, the sample was cup loaded into IPG Ready Strips (18cm, ph 6-11, GE Healthcare) in 100 µl of IEF buffer as follows: 150 V for 8 h, 500 V for 1 h, 1000 V for 1 h, linear ramping to V for 30 min, and finally V for 30 kvh over 10 h. A Peltier temperature control platform maintained gels at 20 º C. Focused IPG strips were stored at -80 º C until SDS-PAGE at which time they were thawed to room temperature, incubated for 15 min in equilibration buffer (50 mm Tris-HCl, ph 8.8, 6 M urea, 30% (w/v) glycerol, 2% (w/v) SDS) with 1% (w/v) DTT, followed by 15 minutes in equilibration buffer with 2% (w/v) iodoacetamide (reduction/alkylation). IPG strips were embedded in a 5% acrylamide stacking gel and proteins were resolved by 8%, 10%, 12%, or 15% SDS-PAGE (Protean II XL, 200 x 220 x 1 mm, BioRad). For 1-D SDS-PAGE sample buffer containing protein was boiled for 10 min in 3X SDS Sample Buffer (New England Biolabs, Beverly, MA, USA) prior to loading into 5% acrylamide stacking gel, and resolved by 6%, 12% or 15% acrylamide. 2-DE gels were silver stained according to Shevchenko et al. (33). Stained gels were scanned with a Powerlook II scanner (UMAX Data Systems, Fremont, CA, USA) on a Sun Ultra5 computer (Sun Microsystems, Palo Alto, CA, USA). Gels were vacuum-dried between cellophane sheets until protein spots/ bands were manually excised. 2-DLC(PF2D) analysis: 2-DLC analysis of solubilized IMM proteins (3 mg) was carried out on a ProteomeLab PF2D (Beckman Coulter, CA, USA) (12). The first dimension separated proteins on the basis of their isoelectric point (pi) using a ph gradient generated on the column (see below). Fractions from the first dimension were collected (FC/I Module) and sequentially injected onto the 2 nd dimension RP-HPLC column (see below). The fractions were collected into 96-deepwell plates for subsequent digestion with trypsin and mass spectrometry analysis (see below). With the PF2D, the 1 st and 2 nd dimension occur sequentially in an automatic manner. 7

8 Chromatofocusing: Proteins were separated in the 1 st dimension on a chromatofocusing (CF) column by mixing two buffers differing in their ph; Start Buffer (ph 8.5), Eluent Buffer (ph 4.0) to create a linear ph gradient from ph 8.5 to 4.0 (see Figure 2A). Initially, The CF column was equilibrated for 130 minutes with Start Buffer at a flow rate of 0.2 ml/min. 3 mg of solubilized IMM protein was injected onto the equilibrated column. After a stable baseline was established (20 minutes), the ph gradient was started by introducing the eluent buffer (flow rate 0.2 ml/min) for 75 minutes. Finally the column was washed with 1 M NaCl. Fractions were collected at 0.3-pH intervals during the ph gradient portion of the run and every 5 minutes before and after 2nd dimension reversed phase chromatography/ 1-DLC: Each fraction from the 1 st dimension CF (200 µl) or IMM protein was solubilized in 2% (v/v) TFA/ 20% (v/v) acetonitrile and sequentially analyzed by RP-HPLC kept constant at 50 º C. Proteins were resolved using 3.33% B/minute linear gradient in which solvent A was 0.1% aqueous TFA and solvent B was 0.08% TFA acid in acetronitrile with a flow rate of 0.75 ml/min. Protein elution was monitored at 214 nm (See Figure 2B for an example of a 2-DLC elution profile and Figure 2C for a composite intensity map across all pi ; See Figure 4 for a1-dlc elution profile. The reversed-phase fractions were collected at a rate of 0.25 min/fraction and stored at -80 º C until further analysis. Mass Spectrometry: Protein identification for 2-DE and both LC separations (1-D and 2-DLC) was carried out based on MALDI-TOF MS and ESI MS/MS, respectively. MALDI-TOF: Gel spots excised from silver stained 2-D gels were de-stained according to Gharahdaghi et al. (34), reduced and alkylated prior to in-gel enzymatic digestion with sequence grade modified trypsin (Promega, Madison, WI, USA). Tryptic peptides were extracted with 1% (v/v) formic acid, 2% 8

9 (v/v) acetonitrile, followed by two extractions with 50% acetonitrile. In some cases, the extracts of the same protein spot from multiple gels were combined. Proteins were spotted by mixing 0.5 µl of reconstituted extract from 2-DE gel spot with 0.5 µl matrix (10 mg/ml α-cyano-4- hydroxy-trans-cinnamic acid, 50% acetonitrile, 0.1% TFA), on a stainless steel 100 well MS plate, and then air dried. Samples were analyzed using a Voyager DE-Pro or Voyager-DE STR MALDI-TOF mass spectrometer (PerSeptive Biosystems, Framingham, MA, USA) operated in the delayed extraction/reflector mode with an acceleration voltage of 20 kv, grid voltage setting 72%, and a 120 ns delay. Minimum of five spectra (200 laser shots) were obtained for each sample. External calibration was performed using a Sequazyme peptide mass standard kit (Perseptive Biosystems). ESI MS/MS: Samples from LC were handled as previously described (12). Essentially, the reversed-phase fractions ( µl), were concentrated using a Speedvac concentrator (ThermoSauvant, NY) to 5-10 ul. 1 M NH 4 HCO 3 was added to neutralize samples to ph 8.0. Modified trypsin (Promega) at an enzyme-to-substrate ratio of 1:50 was added and incubated at 37 o C for overnight (>12 hours). 10% TFA was added to stop the digestion. This was performed at the Johns Hopkins TICC proteomics core using an LTQ ion trap MS (ThermoFinnigan) interfaced with a Suveryor HPLC system (ThermoFinningan) or a QSTAR/Pulsar MS (Applied Biosystems/MDX Sciex) interfaced with an UltiMate TM Capillary/Nano LC system (LC Packings, Peptides were fractionated by RP-HPLC on a 75 µm x 100 mm C18 PepMap column with a 10 µm emitter using 0-60% acetonitrile/0.5% formic acid gradient over 30 min at ~250 nl/min. Ionized peptides were analyzed in the mass spectrometers using scan modes consisting initially of a survey spectum (MS only) from which the eight (LTQ) or three (QSTAR) most abundant ions were determined. The instruments were tuned and calibrated according to manufacturer s instructions. The 9

10 resulting MS/MS spectra were used to search the NCBI non-redundant database using MASCOT (Matrix Science, UK). Database search parameters included the variable modification of oxidized methionine with 2 miscleavable tryptic digestions. The mass tolerance for QSTAR was 0.1 Da for both peptide and MS/MS; The mass tolerance for LTQ was 1.5Da for peptide and 0.8 for MS/MS. Protein identification to create a non-redundant protein database: Protein identification by peptide mass fingerprinting was conducted with the database search tool MS-Fit in the program Protein Prospector ( Mass tolerance was limited to 25 ppm after internal calibration to trypsin mass peaks. Identifications required a minimum of 5 mass peaks corresponding to the major peaks in the spectra, a minimum MOWSE score of 10 5 with no other scores for different proteins within 2 orders, and no other mass peaks above 50% intensity that could not be attributed to the identified protein or known contaminants, and greater than 40% sequence coverage. Proteins identified with amino acid sequences obtained from ESI MS/MS had a minimum of 2 peptide matches (see online supplement Figure 1 for representative MS spectra) with a minimum Mascot Score of 40 for each peptide. When protein identification was made with 2 peptide matches, the fragments had to be unique to the protein (and not matched to another other potential identification (i.e. 2 non-redundant peptides). Further stringency was added by eliminating any peptide that could be assigned to more than one protein. To create a non-redundant database, protein identifications were manually examined in the database for possible redundancies including multiple names and homologies because numerous instances were found where the same protein was contained in multiple database protein identifications. Redundancy was eliminated through the use of a command line version of BLAST (blastclust ; 10

11 ftp.ncbi.nih.gov) which clustered accession numbers according to a 90% protein sequence similarity over 90% of their length. Those proteins which matched these criteria were considered to be the same protein. An estimate of hydrophobicity was calculated from a grand average of hydrophobicity (GRAVY). This score along with the theoretical pi and mass were obtained using the proteomics package in BioJava ( on the intact (nonprocessed) amino acid sequence. In an attempt to estimate the effect of mitochondrial signal sequence by calculating the theoretical pi, Mw and GRAVY score for mitochondrial signal sequence for 20 well characterized mitochondrial proteins (See Table S1). The average was reported as a footnote in Table 1. The prediction of transmembrane domains was carried using DAS-transmembrane prediction server ( (35). The extent of amino acid sequence homologies for all unknown or homology proteins were determined by ( and if identity was over 95% homology at amino acid sequence it was considered to be the same protein (See the online supplements Table S1 and for MS spectracharacterization). If BLAST search could not find a similar protein, the protein was left as undefined or theoretical. Comparison between protein databases: The overlap extent between our IMM protein database and previously published databases was determined using the command line version of blast (as described above). Previous databases were first screened for redundancy and valid accession numbers. Protein sequences from these databases were clustered against the proteins in Table 1 based on 90% similarity over 90% of the protein length. Proteins from different databases were defined as the same protein if they met these criteria. 11

12 Western blotting: For 1-D analysis, 200 µl of LC fractions was evaporated to dryness in a Speedvac and proteins were solubilized in LDS sample buffer (Invitrogen) and resolved by 1-D SDS-PAGE using NuPAGE Novex Bis-Tris Gels with MES running buffer (Invitrogen). For 2-DE analysis, 200 µl fractions were evaporated to dryness in a Speedvac and proteins were solubilized in IEF buffer and resolved in a Protean IEF cell (BioRad). IPG Ready Strips (7 cm, ph 3-10, GE Healthcare), were actively rehydrated with solubilized protein in 115 µl of IEF buffer at 50 V for 10 h, then a rapid voltage ramping method was applied as follows: 100 V for 1 h, 500 V for 1 h, 1000 V for 1 h, linear gradient to 5000 V over 1 h, and finally 5000 V for 10 kvh. The second dimension was performed using NuPAGE Novex Bis-Tris Gels with MES running buffer (Invitrogen). Gels were transferred to a PVDF membrane for Western blotting using a wet transfer apparatus (BioRad) in NuPAGE Transfer Buffer (Invitrogen, Carlsbad, California) at 100 V for 45 h at 4 C. Western blot analysis was carried out using a mouse monoclonal antibody specific to the α-subunit of F 1 F o ATP synthase according to the manufacturer s protocol (clone 7H10, Molecular Probes-Invitrogen). The primary antibody was detected with rabbit anti-mouse antibody conjugated to alkaline phosphatase (Jackson Immuno Research Laboratories) and CDP-Star chemiluminescence reagent (NEN-Mandel). Results and Discussion We had previously shown that pre-incubation with the IEF-compatible zwitterionic detergent SB 3-10 for 10 minutes at room temperature prior to the addition of standard IPG buffer (see methods) improved isoelectric focusing, reduced horizontal streaking, and increased in the number of distinct protein spots visualized in 2-DE from less than 30 to 145 using the same broad range conditions (ph 3-10, 12% SDS PAGE) (20, Figure 1A). Furthermore, extending the ph gradient over a larger distance by using 18 cm ph 4-7 IPG (Figure 1B), ph 6-11 IPG (Figure 12

13 1C) and 12% SDS-PAGE (or 6 and 15% SDS PAGE, data not shown) to collectively expand the proteome allowed for the visualization of over 200 protein spots. Of these, 115 spots were identified corresponding to 77 non-redundant proteins (Table 1). Twenty proteins including the α-subunit of F 1 F o ATP synthase (NCBI accession ; Figure 1 and 3) were visualized as multiple spots by 2-DE due to modifications that altered their pi (Table 1, see *). The optimal condition for the solublization of the IMM preparation differed between 2-DLC and 2-DE as a higher ph starting buffer (ph 8.5) was used for 2-DLC. Therefore, it was necessary to solubilize the IMM extract in SDS, precipitate this in ice cold acetone, and suspend the pellet directly into 2-DLC buffer at ph 8.5 (Figure 2, see methods). Based on ESI MS/MS analysis, an average of 5 proteins were present in each reverse-phase fraction (0.25 minute, range: 1 to 35 proteins per fraction) (data not shown). Thus, for protein mixtures with the same or greater complexity as the IMM, analysis by MALDI-TOF MS is of limited utility (16,36). A more useful approach would be to utilize ESI MS/MS unless there is additional data (e.g. whole protein mass (37-39)) that provides insight into the number of proteins in a given fraction because this would allow decisions to be made regarding the appropriate MS strategy. A total 146 non-redundant proteins were identified by 2-DLC (Table 1) through ESI MS/MS analysis of a total of 106 fractions. Eighteen proteins were identified in more than two nonsequential fractions (in either dimension) suggesting that these proteins may have a potential PTM (40). The α-subunit of F 1 F o ATP synthase separated into two second dimension fractions suggesting it also has a modification that alters its hydrophobicity (Figure 2B). These fractions were analyzed by 1D SDS PAGE followed by western blot for the α-subunit. This showed that the full length α-subunit from F 1 F o ATP sythnase of approximately 59 kda eluted in

14 minutes whereas a 38 kda lower molecular weight form eluted in a fraction at 16.4 minutes (Figure 3). Although the 38 kda form was not identified in the initial silver stained 2-DE analysis (due to its low abundance), its presence was confirmed by 2-DE western blot (Figure 3B). Since the theoretical pi of the α-subunit is greater than 8.5, the multiple pi forms observed by 2-DE eluted in the void volume (ph>8) of the first dimension of the PF2D. The 2-DLC was also useful in separating distinct protein isoforms. For example, acetyl-coenzyme A acetyltransferase also eluted in the void volume of the first dimension but eluted from the second dimension at different retention times (16.3 and 18.4 minutes). ESI MS/MS analysis showed that the less hydrophobic fraction contained isoform 1 of acetyl-coenzyme A, based upon the identification of three unique amino acid sequences (Figure S2). The later eluting fractions contained both protein isoforms (1 and 2) of acetyl-coenzyme A based upon the observation of the same three unique amino acid sequences (Figure S1). The different retention times for isoform 1 suggest that there is a yet uncharacterized hydrophobic modification. For 1-DLC, the IMM proteins were solubilized in 2% TFA with 20% acetonitrile (ph 2.3) and resolved by reversed phase chromatography using a linear gradient from 0 to 100% B over 30 minutes (Figure 4). ESI MS/MS analysis of 38 fractions identified 230 non-redundant proteins with each fraction containing between 1 and 30 proteins (Table 1). The α-subunit of F 1 F o ATP synthase eluted in reversed phase fractions with retention times equivalent to those observed by 2-DLC (arrows Figure 4). Comparison of the protein separation technologies The combined IMM subproteome database obtained from the three protein separation technologies consisted of a total of 348 non-redundant parent proteins (Table 1). There was little 14

15 overlap (7%) between the proteins observed by the three different methods and only 24% overlap between any two methods. Twenty one percent of the total proteins were identified only by 2- DLC and 47% identified only by 1-DLC even though pi, mass and GRAVY score calculations were similar (Figure 5). This in-depth comparison highlights the advantages and limitation of the three protein separation methods. Even with optimization of solublization and running conditions, 2-DE is limited with respect to proteins with extreme mass, hydrophobicity and pi, although it has the advantage of distinguishing multiple forms of proteins with differences in molecular mass or isoelectric point (Table 1). Conversely, the 2-DLC and 1-DLC have the advantage of enriching for low mass proteins below 30 kda and basic proteins with pis above 8.5/9.0. With the PF2D the basic proteins are concentrated in the void volume from the first dimension so individual pi information is lost for this class of protein. To date, the majority of studies using the PF2D (or similar LC system) have not analyzed this basic fraction. Since the PF2D requires an approximate 10 fold higher protein load compared to the other two methods (to track profiles by absorbance) one would expect to be able to detect lower abundant protein which leads to greater proteome coverage. As well, despite 2-DLC and 1-DLC sharing the same final resolution step (RP-HPLC), the overlap in protein identifications was surprisingly small (9%). This difference may be due to differences in the solublization buffers used in preparation for each method. For instance, 2-DLC required a ph 8.5 aqueous buffer whereas RP-HPLC contained 2% TFA (ph 2.3) as well as an organic solvent. Membrane proteins are notoriously difficult to solubilize in aqueous buffers (41) and the addition of an organic to the start buffer for 1-DLC may have increased the identification of certain classes of proteins. For instance, translocase of the inner mitochondrial membrane (TIM), solute carrier and cytochrome P450 family members are predicted (35) to have multiple transmembrane domains (ranging from 1 to 15

16 7) (data not shown) which would limit their solubility. These proteins possess a range of GRAVY scores (-0.53 to 0.09), yet GRAVY scores represent an average of hydrophobicity. Highly hydrophobic or hydrophilic protein domains affecting retention may be underrepresented by presenting the data in this manner. The detection of this unique class of proteins is advantageous for experiments in which the goal is to determine which proteins are present; however, for quantitative studies 1-DLC is of limited use due to its reduced ability to resolve individual proteins compared to 2-DLC. Additionally, both LC methods would most likely require additional downstream methods for quantitative analysis. In this respect, 2-DLC has the advantage because image analysis of the elution profiles can limit the number of fractions requiring further analysis. Comparison of existing mitochondrial and IMM protein databases There are several large proteomic-based databases for the intact mitochondrial proteome. This includes the database produced by Taylor and colleagues containing 632 proteins (21) and was later updated (27) by 107 additional proteins. This group used sucrose gradient centrifugation to resolve mitochondrial protein complexes followed by 1D SDS PAGE and LC/MS/MS. In a separate study, Mootha et. al. (28) were able to identify 399 non-redundant mouse mitochondrial proteins isolated in a similar manner from brain, heart, kidney and liver. This database was expanded to 428 non-redundant proteins through the inclusion of mouse and human proteins from the MITOP data base and annotated mitochondrial proteins from the NCBI Locuslink database. Even though it was not exploited by the different groups, the initial separation by density centrifugation provided potential biological information. The PF2D also provides biological information based on protein characteristics (pi, hydrophobicity). Our database contains 154 proteins (44% of the combined database; Table 1) not previously identified in these 16

17 databases. The only IMM subproteome database published to date identified 182 parent proteins using shot-gun type approaches (peptide digestion and separation by multiple chromatographic steps) (29). Because these proteins were proteolysed prior to separation and identification, there is a reduced probability of obtaining biological information. Our database contains 286 proteins (82% of combined database; Table 1) not in this database and 134 proteins not observed in any of these extensive databases (~39% of the combined protein database; Table 1, Figure S3). This difference is likely due to both the enrichment of the IMM proteins prior to analysis (21,25,28) and the increased resolving power of separating proteins based on a variety of physical characteristics. The majority of these proteins were observed only using 1-DLC (96 ; 72%) suggesting that this is a useful technology to use for the discovery of novel proteins in a subproteome. It is important to note that the isolation protocol for IMM is well established (32) and the majority of the proteins which had not been previously reported are known to be associated with the mitochondria. Conclusion: The data shows a low degree of overlap between the protein separation methods described, and very significantly extended our coverage of the IMM proteome by using an integrative approach. Furthermore, using the PF2D, a 2-DLC system, we were able to identify a unique subset of proteins not observed by 2-DE or 1-DLC. This may be due to difference in starting protein quantity (3 mg for the PF2D to 100 µg for 1-DLC). 1-DLC was able to resolve a unique set of proteins, with highly hydrophobic domains. As well, the majority of the proteins not observed by previous databases were only observed using the 1-DLC. Overall the results show the power 17

18 of integrating different separation technologies, and also caution against making the assumption that one can validate findings using one separation method with another. 18

19 Acknowledgements The authors would like to acknowledge the following funding sources that supported this work: Canadian Foundation for Medical Research (JVE), NHLBI Proteomic Initiative (contract N0- HV-28180)(JVE, BC), NIH (grant CA10951, PLP), and the Donald P. Amos Family Foundation (JVE) and the use of the mass spectrometers at the Johns Hopkins TICC proteomic core supported by the NHLBI proteomic initiative and the Johns Hopkins Institute for Cell Engineering. Finally the authors would like to acknowledge David Graham for critical review and editing of the manuscript. 19

20 References 1. Roberts, G.C., Smith, C.W. (2002) Alternative splicing: combinatorial output from the genome. Curr Opin Chem Biol. 6, Farriol-Mathis, N., Garavelli, J.S., Boeckmann, B., Duvaud, S., Gasteiger, E., Gateau, A., Veuthey, A-L., Bairoch, A. (2004) Annotation of post-translational modifications in the Swiss-Prot knowledge base. Proteomics 4, Van den Bergh, G., Arckens, L. (2005) Recent advances in 2D electrophoresis: an array of possibilities Expert Rev Proteomics 2, Rabilloud, T. (2002) Two-dimensional gel electrophoresis in proteomics: old, old fashioned, but it still climbs up the mountains. Proteomics 1: Neverova, I., Van Eyk, J.E. (2005) Role of chromatographic techniques in proteomic analysis. J Chromatogr B Analyt Technol Biomed Life Sci. 815, Morris, D.L. Jr., Sutton, J.N., Harper, R.G., Timperman, A.T. (2204) Reversed-phase HPLC separation of human serum employing a novel saw-tooth gradient: toward multidimensional proteome analysis. J Proteome Res. 3, Fujii, K., Nakano, T., Hike, H., Usui, F., Bando, Y., Tojo, H., Nishimura, T. (2004) Fully automated online multi-dimensional protein profiling system for complex mixtures. J Chromatogr A. 1057, Garbis, S., Lubec, G., Fountoulakis, M. (2005) Limitations of current proteomics technologies. J Chromatogr A. 1077, Baumann, M., Meri, S. (2004) Techniques for studying protein heterogeneity and posttranslational modifications. Expert Rev Proteomics. 1, Swanson, S.K., Washburn, M.P. (2005) The continuing evolution of shotgun proteomics. Drug Discov Today. 10, Wang, Y., Wu, R., Cho, K.R., Shedden, K.A., Barder, T.J., Lubman, D.M. (2006) Classification of cancer cell lines using an automated two-dimensional liquid mapping method with hierarchical clustering techniques. Mol Cell Proteomics 5, Sheng, S., Chen, D., Van Eyk, J.E. (2006) Multi-dimensional liquid chromatography separation of intact proteins by chromatographic focusing and reversed phased of the human serum proteome: optimization and protein database. Mol Cell Proteomics 5, Zheng, S., Schneider, K.A., Barder, T.J., Lubman, D.M. (2003) Two-dimensional liquid chromatography protein expression mapping for differential proteomic analysis of normal and O157:H7 Escherichia coli. Biotechniques 35, Lubman, D.M., Kachman, M.T., Wang, H., Gong, S., Yan, F., Hamler, R.L., O'Neil, K.A., Zhu, K., Buchanan, N.S., Barder, T.J. (2002) Two-dimensional liquid separations-mass mapping of proteins from human cancer cell lysates. J Chromatogr B Analyt Technol Biomed Life Sci. 782, Chen, E.I., Hewel, J., Felding-Habermann, B., Yates, J.R. 3rd. (2006) Large scale protein profiling by combination of protein fractionation and multidimensional protein identification technology (MudPIT). Mol Cell Proteomics. 5, Komatsu, S., Zang, X., Tanaka, N. (2006) Comparison of two proteomics techniques used to identify proteins regulated by gibberellin in rice. J Proteome Res. 5, Gibson, B.W. (2005) The human mitochondrial proteome: oxidative stress, protein modifications and oxidative phosphorylation. Int J Biochem Cell Biol. 37,

21 18. Pagliarini, D.J., Dixon, J.E. (2006) Mitochondrial modulation: reversible phosphorylation takes center stage? Trends Biochem Sci. 31, Hurd TR, Filipovska A, Costa NJ, Dahm CC, Murphy MP. (2005) Disulphide formation on mitochondrial protein thiols. Biochem Soc Trans. 33, Fukada, K., Zhang, F., Vien, A., Cashman, N.R., Zhu, H. (2004) Mitochondrial proteomic analysis of a cell line model of familial amyotrophic lateral sclerosis. Mol Cell Proteomics 12, Taylor, S.W., Fahy, E., Zhang, B., Glenn, G.M., Warnock, D.E., Wiley, S., Murphy, A.N., Gaucher, S.P., Capaldi, R.A., Gibson, B.W., Ghosh, S.S. (2003) Characterization of the human heart mitochondrial proteome. Nat Biotechnol. 3, Gaucher, S.P., Taylor, S.W., Fahy, E., Zhang, B., Warnock, D.E., Ghosh, S.S., Gibson, B.W. (2004) Expanded coverage of the human heart mitochondrial proteome using multidimensional liquid chromatography coupled with tandem mass spectrometry. J Proteome Res. 3, Rezaul, K., Wu, L., Mayya, V., Hwang, S.I., Han, D. (2005) A systematic characterization of mitochondrial proteome from human T leukemia cells. Mol Cell Proteomics 4, McDonald, T.G., Van Eyk, J.E. (2003) Mitochondrial proteomics. Undercover in the lipid bilayer. Basic Res. Cardiol. 98, Taylor, S.W., Fahy, E., Ghosh, S.S. (2003) Global organellar proteomic. Trends Biotechnol., 21, Gabaldon, T., Huynen, M.A. (2004) Lineage-specific gene loss following mitochondrial endosymbiosis and its potential for function prediction in eukaryotes. Biochim Biophys Acta, 1659, Gaucher, S.P., Taylor, S.W., Fahy, E., Zhang, B. Warnock, D.E., Ghosh, S.S., Gibson, B.W. (2004) Expanded coverage of the human heart mitochondrial proteome using multidimensional liquid chromatography coupled with tandem mass spectrometry. J. Prot. Research, 3, Mootha, V.K., Bunkenborg, J., Olsen, J.V., Hjerrid, M., Wisniewski, J.R., Stahl, E., Bolouri, M.S., Ray, H.N., Sihang, S., Kamal, M., Patterson, N., Landers, E.S.. Mann, M. (2003) Integrated analysis of protein composition, tissue diversity, and gene regulation in mouse mitochondria. Cell, 115, Da Cruz, S., Xenarios, I., Langridge, J., Vilbois, F., Parone, P.A., Martinou, J.C. (2003) Proteomic analysis of the mouse liver mitochondrial inner membrane. J Biol Chem. 278, Da Cruz, S., Parone, P.A., Martinou, J.C. (2005) Building the mitochondrial proteome. Expert Rev Proteomics. 4, Calvo, S., Mohit, J., Xie, X., Sheth, S.A., Chang, B., Goldberger, O.A., Spinazzola, A., Zeviani, M., Carr, S.A., Mootha, V.K.(2006) Systematic identification of human mitochondrial disease genes through integrated genomics. Nature Genetics 38, Pedersen, P,L,, Greenawalt, J.W., Reynafarje, B., Hullihen, J., Decker, G.L., Soper, J.W., Bustamente, E. (1978) Preparation and characterization of mitochondria and submitochondrial particles of rat liver and liver-derived tissues. Methods Cell. Biol. 20, Shevchenko, A., Wilm, M., Vorm, O., Mann, M. (1996) Mass spectrometric sequencing of proteins silver-stained polyacrylamide gels. Anal. Chem. 68,

22 34. Gharahdeghi, F., Weinberg, C.R., Meagher, D.A., Imai, B.S., Mische, S.M. (1999) Mass spectrometric identification of proteins from silver-stained polyacrylamide gel: a method for the removal of silver ions to enhance sensitivity. Electrophoresis 20, Cserzo, M., Wallin, E., Simon, I., von Heijne G., Elofsson, A. (1997) Prediction of transmembrane alpha-helices in procariotic membrane proteins: the Dense Alignment Surface method. Prot. Eng. 10, Pirondini, A., Visioli, G., Malcevschi, A., Marmiroli, N. (2006) A 2-D liquid-phase chromatography for proteomic analysis in plant tissues. J Chromatogr B Analyt Technol Biomed Life Sci. 833, Linke, T., Ross, A.C., Harrison, E.H. (2006) Proteomic analysis of rat plasma by twodimensional liquid chromatography and matrix-assisted laser desorption ionization time-offlight mass spectrometry. J Chromatogr A [Epub ahead of print] 38. Zhu, K., Miller, F.R., Barder, T.J., Lubman, D.M. (2004) Identification of low molecular weight proteins isolated by 2-D liquid separations J Mass Spectrom. 39, Betgovargez, E., Knudson, V., Simonian, M.H. (2005) Characterization of proteins in the human serum proteome. J Biomol Tech. 16, Zhu, K., Zhao, J., Lubman, D.M., Miller, F.R., Barder, T.J. (2005) Protein pi shifts due to posttranslational modifications in the separation and characterization of proteins. Anal Chem. 77, Santoni, V., Molloy, M., Rabilloud, T. (2000) Membrane proteins and proteomics: un amour impossible? Electrophoresis, 21,

23 Figure Legends Figure 1. 2-DE analysis of rat liver inner mitochondrial membrane separated using various ph gradients A) Rat inner mitochondrial membrane proteins separated using ph 3-10 IPG followed by 12% SDS-PAGE (panel A), ph 4-7 IPGs/12% SDS-PAGE (panel B), and ph 6-11/12% SDS-PAGE (panel C). Proteins were annotated using NCBI accession number (see Table 1 for protein identification). For more details see the Method section. Inserts show enlargements of two regions (1 and 2) of the gels illustrating improved spot resolution with narrow ph gradient gels. This allowed the identification of 2 forms of ubiquinol-cytochrome-c reductase complex core protein I (inset 1, # ) and 4 forms of the α-subunit of F 1 F o ATP synthase (insert 2, # ). Figure 2. 2-DLC separation of rat liver inner mitochondrial membrane proteins. 3.0 mg of liver inner mitochondrial membrane proteins was separated in the first dimension using chromatofocusing (CF-LC), with proteins eluting with pis greater than ph 8.5 in the void volume, followed by a descending ph gradient from ph 8.5 to 4.0, finally in a salt wash for pis less than ph 4.0 (panel A left). The dotted line represents the change in ph from 8.5 to 4.0. CF fractions with a ph range of 0.3 units from the first dimension were collected individually and subsequently separated using reversed phase high performance liquid chromatography (RP- HPLC) in the second dimension (panel A right). Each RP-HPLC fraction was distinct and unique (examples, panel B). The arrows indicate the fraction where the α-isoform of F 1 F o ATP synthase was present (retention time ~16.4 and 18.8 mins.). The 2-DLC composite map of the intensity of the protein profiles from the second dimension RP-HPLC elutions across all ph ranges is shown in Panel C. Each RP-HPLC fraction was processed prior to ESI MS/MS for protein identification 23

24 which many proteins were identified which were not identified using either 2DE or RP-HPLC alone. For more details see Table 1 and Table S1. Figure 3. Western blotting for the α-subunit of F 1 F o ATP synthase reveals PTM. The α- subunit of ATP synthase was identified in two reversed phase high performance liquid chromatography (RP-HPLC) fractions (at retention time 18.8 and 16.4 mins) by ESI MS/MS. Western blotting of these fractions with monoclonal antibody specific to the α-subunit revealed a 59 kda and a 39 kda form (panel A). The small GRAVY score difference between these two forms was resolved by RP-HPLC ( and for 59 kda and 39 kda, respectively). The intensity of the signal suggests the smaller form was present in the inner membrane at significant levels. Western blotting of inner mitochondrial membrane proteins resolved by 2-DE revealed both the smaller and larger form, but also confirmed the presence of a modification to the larger form which altered its pi, a PTM which was not present in the smaller form. Figure 4. Reversed phase high performance liquid chromatography. 100 µg of inner mitochondrial membrane proteins solubilized in 2% TFA/ 20% acetonitrile was resolved by reversed phase high performance liquid chromatography using a linear gradient of 0 to 100% acetonitrile over 30 minutes (absorbance solid line, gradient dotted line). 38 fractions were processed and analyzed by ESI MS/MS for protein identification, in which many proteins were identified which were not identified using either 2DE or 2D LC. For more details see Table 1. The arrows indicate the fraction where the α-subunit of F 1 Fo ATP synthase was present (retention time ~16.4 and 18.8 mins.). 24

25 Figure 5. Distribution of the biochemical properties of proteins identified with 2-DE, 2- DLC or 1D-LC. Panel A shows the molecular weight distribution over 10 kda increments of the proteins identified by 2-DE, 2-DLC or 1-DLC. Panel B shows the pi of the proteins identifications (from Table 1) over increments of 0.5 units of the proteins separated by either 2- DE (black bars), 1-DLC (dashed) or 2-DLC (grey bars). Identifications following 2-DE reveal clustering of protein pis between and while proteins identified following 2D-LC and 1D-LC tended to be shifted towards higher pis. Panel C shows the hydrophobicity distribution of identified proteins separated by 2-DE, 1-DLC or 2-DLC. Hydrophobic proteins with positive GRAVY scores were identified using either protein separation methods, although there is a slight increase with 1-DLC and 2-DLC. Note that the theoretical values were calculated on the intact non-processed amino acid sequence. The mitochondrial signal sequence from twenty known mitochondrial proteins would have contributed an average mass, pi and GRAVY score of 3.4 ± 0.8 kda, 0.85 ± 0.58 pi, ± 0.08 units respectively (see method section and Table S2 for more details). Panel D shows the overlap of proteins identified by each technology. Only 7% of identifications were common to all three and significant portions were observed by only a single technology. 25

26 Online supplement Figure S1. Representative mass spectrometry spectra obtained from LTQ analysis. Mass spectra obtained using the LTQ MS that correspond to specific (and unique) amino acid sequences for acetyl COA acetyl transferase isoform 2 (panels A-C) and isoform 1 (panels D-E), adenylate kinase isoform 4 (panels H,1), isoform 3 (panels J,K) and isoform 2 (panels L, M), glutamate dehydrogenase (panels N,O), methylmalonate semialdehyde dehydrogenase (panels P,Q) and thiosulfate sulfurtransferase (panels R,S). Figure S2: Detailed analysis of reversed-phase elution of acetyl-coa acetyltransferase isoforms 1 and 2. Panel A shows the reversed phase elution profile of the proteins from the fraction collected during the chromatographic focusing step of the 2-DLC analysis of 3.0 mg of rat liver inner mitochondrial subproteome. Only fractions eluting at 16.3 and 18.4 minutes contain acetyl-coa acetyltransferase. Panel B lists the amino acid sequence of the two isoforms and in italize the amino acid sequence that correspond to the fragments sequenced by MS/MS (see Figure S1 for examples of MS/MS spectra). The amino acid sequences are unique to each isoform. Figure S3: Pie chart illustrating the overlap between various protein databases. The nonredundant proteins (Table 1) were compared to the proteins present in different mitochondrial or inner mitochondrial proteins databases (21, 27-29). Of the 348 nonredundant proteins in the combined mitochondrial database, 134 (39%) were unique to this database and not previously reported. (B) Pie chart illustrating the technologies used to obtain the unique proteins from (A) shows that the vast majority (96%) were obtained using only RP and / or the PF2D. 26

27 A ph 3 ph ph 4 ph B C ph 11 ph Figure

28 A First dimension CF-LC Second dimension RP-LC UV-1 SMP-3.0mg UV-2 SMP-3.0mg _ Absorbance (214nm) AU Minutes Time (mins.) AU ph Minutes Time (mins.) AU B Absorbance (214 nm) ph ph ph ph Dried, Neutralized, enzymatic digest, ESI MS/MS Time (mins.) Absorbance (214 nm) ph ph ph C ph Time (mins.) Mins Figure 2

29 A KDa / B KDa / ph 3 ph 10 Figure 3

30 UV-2 SMP-Pellet uL Absorbance (214 nm ) AU Gradient (% B ----) AU Minutes Time (mins) Figure 4

31 A Mass Distribution 2DE RP PF2D B Isoelectic Point Distribution 2DE RP PF2D % of Proteins Observed % of Proteins Observed < kda > >12 pi C % of Proteins Observed Grand Average of Hydrophocity Distribution 2DE RP PF2D D PF2D 21% 2DE and RP 3% 2DE and PF2D 5% RP and PF2D 9% ALL 7% to to to to 0 0 to to to to +1.5 RP 47% 2DE 8% Figure 5

32 Table 1: Composite Database of Identified Proteins Found in previous databases? 5 Technology Average Accession # 1 Protein Name 2 Observed 3 Mass 4 pi 4 Gravy score 4 Mootha et Taylor et al Da Cruz et al (28) (21,22) al (29) F23Rik protein [Mus musculus] RP ,4-dienoyl-CoA reductase (NADPH2) RP x x ',5'-oligoadenylate synthetase 1 RP x oxoglutarate dehydrogenase E1 component, mitochondrial precursor 2DE x x oxoisovalerate dehydrogenase beta subunit, mitochondrial precursor 2DE x S ribosomal protein L44, mitochondrial precursor RP x hydroxy-3-methylglutaryl-Coenzyme A synthase 2 2DE, RP, PF2D x hydroxyacyl Coenzyme A dehydrogenase (HCDH) RP, PF2D x * 3-hydroxyacyl-CoA dehydrogenase type II 2DE, RP, PF2D x x x hydroxyisobutyrate dehydrogenase RP, PF2D x nitrophenylphosphatase domain and non-neuronal SNAP25-like protein homolog 1 RP, PF2D x G 60 kda heat shock protein, mitochondrial precursor (Hsp60) (60 kda chaperonin) (HSP-60) 2DE, PF2D x x x Aa2-258 PF2D AAA-ATPase TOB3 RP x Ab2-132 PF2D acetyl-coenzyme A acetyltransferase 1 2DE, PF2D x x acetyl-coenzyme A acyltransferase 2 (mitochondrial 3-oxoacyl- Coenzyme A thiolase) 2DE, RP, PF2D x x * aconitate hydratase, mitochondrial precursor 2DE x x actin, alpha 1, muscle-specific PF2D x x actinin alpha 4 RP x acyl-coa dehydrogenase, long-chain specific, mitochondrial precursor 2DE x acyl-coa dehydrogenase, medium-chain specific, mitochondrial precursor PF2D x x acyl-coa dehydrogenase, short-chain specific, mitochondrial precursor 2DE x x acyl-coa dehydrogenase, very-long-chain specific, mitochondrial * precursor 2DE, RP x x acyl-coa synthetase long-chain family member 1 RP, PF2D x x Acyl-CoA synthetase long-chain family member 4 RP acyl-coenzyme A oxidase 1, palmitoyl PF2D x