Anal. Chem. 2004, 76, 86-97
Synthesis/Degradation Ratio Mass Spectrometry for Measuring Relative Dynamic Protein Turnover Benjamin J. Cargile,† Jonathan L. Bundy,† Amy M. Grunden,‡ and James L. Stephenson, Jr.*,†
Mass Spectrometry Research Group, Research Triangle Institute, 3040 Cornwallis Road, Research Triangle Park, North Carolina 27709-2194, and Department of Microbiology, North Carolina State University, Gardner Hall, Room 4615, Raleigh, North Carolina 27695
One of the major unanswered questions in quantitative proteomics is that of dynamic protein turnover in the cell. Here we present a new approach to quantitative proteomics that measures the relative dynamic turnover of proteins in cellular systems. In this approach, termed synthesis/degradation ratio mass spectrometry, stable isotope labeling is employed to calculate a relative synthesis/ degradation ratio that reflects the relative rate at which 13C is incorporated into individual proteins in the cell. This synthesis/degradation ratio calculation is based on a Poisson distribution model that is designed to support high-throughput analysis. Protein separation and analysis is accomplished by utilizing one-dimensional SDS-PAGE gel electrophoresis followed by cutting the gel into a series of bands for in-gel digestion. The resulting peptide mixtures are analyzed via solid-phase MALDI LC-MS and LC-MS/MS using a tandem time-of-flight mass spectrometer. A portion of the soluble protein fraction from an E. coli K-12 strain was analyzed with synthesis/degradation ratios varying from approximately 0.1 to 4.4 for a variety of different proteins. Unlike other quantitative techniques, synthesis/degradation ratio mass spectrometry requires only a single cell culture to obtain useful biological information about the processes occurring inside a cell. This technique is highly amenable to shotgun proteomicsbased approaches and thus should allow relative turnover measurements for whole proteomes in the future. Recent growth in the field of quantitative proteomics has resulted in a wide variety of mass spectrometry-based techniques for measuring gene expression. The most widely used quantitative approach coupled to mass spectrometry is that of isotope-coded affinity tags (ICAT) as originally described by Gygi et al.1 In this approach, protein profiling can be performed by comparing the amount of proteins present in two different cell states by designating one state as the reference with a light isotope label and adding a heavy isotope (2H or 13C) label to the other.1 The reference and labeled samples are then combined where individual peptides have * Corresponding author. E-mail:
[email protected]. Phone: (919) 316-3978. Fax: (919) 541-7208. † Research Triangle Institute. ‡ North Carolina State University. (1) Gygi, S. P.; Rist, B.; Gerber, S. A.; Turecek, F.; Gelb, M. H.; Aebersold, R. Nat. Biotechnol. 1999, 17, 994-999.
86 Analytical Chemistry, Vol. 76, No. 1, January 1, 2004
the same amino acid sequence but differ in mass by the weight of the isotopic label. An affinity tag is then utilized to separate the labeled and reference peptides from a complex proteolytic digest mixture, although this separation is far from complete. The intensity ratios obtained from the simultaneous elution and analysis via mass spectrometry of these peak pairs then provides a measure of the relative abundance of each peptide. This information can then be used to determine relative expression levels of the original proteins if both the heavy and light forms of the peptides are observed. This basic approach (as well as several variations) has been applied successfully to quantitate proteins from cellular organelles such as microsomes,2 to compare mRNA and protein expression levels,3,4 to evaluate quantitative levels of posttranslational modifications in N-linked glycopeptides from prostate epithelial cells,5 and more recently to perform quantitative studies on the phosoproteome of camptothecin-treated cortical neurons.6 Although the ICAT technique represents a significant advance in quantitative proteomics, there are a large number of problems with the approach that include using a single peptide to quantitate (the number of proteins with multiple cysteines is limited), reliance on the ionization of one to a few peptides in a complex mixture, and an inefficiency at identifying cysteinecontaining peptides that do not appear in pairs. Furthermore, cost considerations associated with the ICAT technique have led to the development of several alternative labeling schemes.7-12 A summary of these alternative strategies and the most recent advances in stable isotope tagging quantitation methodology has recently been reviewed.3 (2) Han, D. K.; Eng, J.; Zhou, H. L.; Aebersold, R. Nat. Biotechnol. 2001, 19, 946-951. (3) Tao, W. A.; Aebersold, R. Curr. Opin. Biotechnol. 2003, 14, 110-118. (4) Ideker, T.; Thorsson, V.; Ranish, J. A.; Christmas, R.; Buhler, J.; Eng, J. K.; Bumgarner, R.; Goodlett, D. R.; Aebersold, R.; Hood, L. Science 2001, 292, 929-934. (5) Zhang, H.; Li, X. J.; Martin, D. B.; Aebersold, R. Nat. Biotechnol. 2003, 21, 660-666. (6) Yu, L. R.; Johnson, M. D.; Conrads, T. P.; Smith, R. D.; Morrison, R. S.; Veenstra, T. D. Electrophoresis 2002, 23, 1591-1598. (7) Kindy, J. M.; Taraszka, J. A.; Regnier, F. E.; Clemmer, D. E. Anal. Chem. 2002, 74, 950-958. (8) Regnier, F. E.; Riggs, L.; Zhang, R. J.; Xiong, L.; Liu, P. R.; Chakraborty, A.; Seeley, E.; Sioma, C.; Thompson, R. A. J. Mass Spectrom. 2002, 37, 133145. (9) Yao, X. D.; Freas, A.; Ramirez, J.; Demirev, P. A.; Fenselau, C. Anal. Chem. 2001, 73, 2836-2842. (10) Moseley, M. A. Trends Biotechnol. 2001, 19, S10-S16. (11) Goodlett, D. R.; Yi, E. C. TrAC, Trends Anal. Chem. 2003, 22, 282-290. (12) Cagney, G.; Emili, A. Nat. Biotechnol. 2002, 20, 163-170. 10.1021/ac034841a CCC: $27.50
© 2004 American Chemical Society Published on Web 11/25/2003
One alternative to the in vitro ICAT labeling technique for protein quantitation is in vivo labeling of proteins using heavy isotope incorporation. In an original publication by Chait and coworkers,13 protein expression profiles were quantitated through the use of whole-cell stable isotope labeling in conjunction with 2-D gel electrophoresis and mass spectrometry. This technique using a non-gel-based format was described by Smith and co-workers where 13C-, 15N-, and 2H-“depleted” media were used to determine the protein expression profile of Escherichia coli grown under normal and cadmium stress conditions.14 In these experiments, two separate cell cultures were grown under normal (standard media) and stressed conditions (isotopically labeled media). The intact proteins were then extracted and combined, and the resulting mixture was analyzed using capillary isoelectric focusing coupled to Fourier transform ion cyclotron resonance mass spectrometry (FTICR). Expression ratios were then calculated for the 200 most abundant proteins to obtain relative quantitative values ranging from 0.1 to 30 for a 45-min time point, although the identity of all the proteins for which the ratios had been calculated was not known.14 Recently, this technique has been coupled to a shotgun proteomics-based approach employing 15N labeling in conjunction with the commercially available Cys affinity tag, (+)-biotinyliodoacetamidyl-3,6-dioxactanediamine (iodoacetyl-PEO-biotin).15 This method has also been applied to the global proteome analysis of Saccharomyces cerevisiae using ion trap technology and the cross-correlation-based algorithm SEQUEST for protein identification of the 14N/15N pairs.16 A recent review of the 14N/15N labeling approach summarizes the most relevant advances in the field.17 Another in vivo approach to quantitative proteomics is that of stable isotope labeling by amino acids in cell culture.18,19 The basic premise of this approach involves growing one culture of cells on a medium lacking an essential amino acid and supplementing the medium with a stable heavy isotopic amino acid (either Leu-d3 or 13C-labeled Arg). A control culture is grown with a normal medium formulation and is used as a comparison standard. After enough cell doublings to allow for complete label incorporation, cells are then harvested, and the protein extracts are combined and then analyzed via mass spectrometric methods.18,19 The advantages of this technique include minimal sample manipulation, quantitation of proteins independent of size and cysteine content, and uniform protein labeling for better precision measurements associated with the extent of up- or downregulation measurements.18 One major disadvantage is that only peptides containing the labeled amino acid moieties can be used for quantitation. This can limit the adaptability of the technique to shotgun proteomics. (13) Oda, Y.; Huang, K.; Cross, F. R.; Cowburn, D.; Chait, B. T. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 6591-6596. (14) Pasa-Tolic, L.; Jensen, P. K.; Anderson, G. A.; Lipton, M. S.; Peden, K. K.; Martinovic, S.; Tolic, N.; Bruce, J. E.; Smith, R. D. J. Am. Chem. Soc. 1999, 121, 7949-7950. (15) Conrads, T. P.; Alving, K.; Veenstra, T. D.; Belov, M. E.; Anderson, G. A.; Anderson, D. J.; Lipton, M. S.; Pasa-Tolic, L.; Udseth, H. R.; Chrisler, W. B.; Thrall, B. D.; Smith, R. D. Anal. Chem. 2001, 73, 2132-2139. (16) Washburn, M. P.; Ulaszek, R.; Deciu, C.; Schieltz, D. M.; Yates, J. R. Anal. Chem. 2002, 74, 1650-1657. (17) Goshe, M. B.; Smith, R. D. Curr. Opin. Biotechnol. 2003, 14, 101-109. (18) Ong, S. E.; Blagoev, B.; Kratchmarova, I.; Kristensen, D. B.; Steen, H.; Pandey, A.; Mann, M. Mol. Cell. Proteomics 2002, 1, 376-386. (19) Ong, S. E.; Kratchmarova, I.; Mann, M. J. Proteome Res. 2003, 2, 173-181.
Although the aforementioned methodologies have advanced the field of quantitative proteomics into the laboratory of the cell biologist, these techniques represent only a relative measure of protein quantitation. In addition, protein dynamics, or the ability to monitor a synthesis/degradation cycle, has largely been ignored. The amount of any protein present in a cell at a certain time point depends on two rate-dependent events. These are the synthesis and degradation rates associated with a given protein. Insight into these processes can help to provide mechanistic data about how the cell adapts at the protein level to a given perturbation of the cell cycle. Early studies into measuring synthesis and degradation rates in prokaryotic systems employed radioactively labeled amino acids20 or the use of 35S21,22 to calculate degradation rates associated with particular classes of proteins. Recent work involving protein conjugation to ubiquitin23 and the enhancement of the protein degradation process, along with characterization of proteosome function,24 represent examples of the importance of understanding the synthesis/degradation cycle. Also, it has been recently suggested that protein turnover measurements can be employed to “fill in the gaps” between some of the discrepancies observed when transcriptome and quantitative proetomics data are compared.25 In a current publication by Pratt and co-workers, an isotopically labeled amino acid approach combined with two-dimensional gel (2-D gel) electrophoresis is employed to determine protein degradation rates in a steady-state population of S. cerevisiae.25 Peptide mass mapping is utilized to determine the breakdown rates of proteins by examining the mass shifts associated with the analysis of tryptic peptides. Although results from these experiments are promising, this approach does suffer from several inherent disadvantages. Due to the high dependence on the 2-D gel step, high-throughput analysis is limited. In addition, steadystate analysis requires the use of a chemostat, which further elongates sample analysis time because a large amount of effort is needed in order to design an experiment whereby the cell count remains constant, so that degradation rates can be calculated. Also, as spent media and cells are drained off the culture, the cell loss incurred contains labeled proteins, which makes the exact degradation rates difficult to measure. This loss of labeled protein must be accounted for in order to accurately determine degradation rates. Another potential challenge associated with labeled cell loss that would be exacerbated in the 2-D gels would be the inability to measure degradation ratios of low-abundance proteins. Here we report the initial implementation of synthesis/ degradation ratio mass spectrometry (SDMS) for the determination of relative dynamic protein turnover rates. In these experiments, we describe the use of 13C-labeled glucose for monitoring the relative synthesis/degradation ratio of proteins obtained from E. coli. Proteins can easily be separated by SDS-PAGE in the first dimension, followed by in-gel digestion and second-dimension separation using nanobore reversed-phase high-performance liquid (20) Larrabee, K. L.; Phillips, J. O.; Williams, G. J.; Larrabee, A. R. J. Biol. Chem. 1980, 255, 4125-4130. (21) Grunenfelder, B.; Rummel, G.; Vohradsky, J.; Roder, D.; Langen, H.; Jenal, U. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 4681-4686. (22) Gottesman, S. Annu. Rev. Genet. 1996, 30, 465-506. (23) Varshavsky, A. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 12142-12149. (24) DeMartino, G. N.; Slaughter, C. A. J. Biol. Chem. 1999, 274, 22123-22126. (25) Pratt, J. M.; Petty, J.; Riba-Garcia, I.; Robertson, D. H. L.; Gaskell, S. J.; Oliver, S. G.; Beynon, R. J. Mol. Cell. Proteomics 2002, 1, 579-591.
Analytical Chemistry, Vol. 76, No. 1, January 1, 2004
87
chromatography (HPLC). To adapt the method for high-throughput analysis, the HPLC effluent was spotted directly on a matrixassisted laser desorption/ionization (MALDI) plate for subsequent analysis using a MALDI tandem time-of-flight mass spectrometer (MALDI TOF/TOF). Protein identifications were accomplished via sequence tag, probability-based matching (MASCOT), highenergy collision-induced dissociation (CID), or parallel off-line liquid chromatography tandem mass spectrometry employing a cross-correlation-based protein identification algorithm (SEQUEST). Taking advantage of the 200-Hz laser repetition rate of the MALDI TOF/TOF instrument further supports high-throughput analysis. A simple method for enumerating the relative synthesis/degradation ratios (S/DRel) through fitting the data to a Poisson distribution model in order to determine an overall turnover rate is provided. Finally, the utility of this technique is shown through the analysis of a single 1-D gel slice, which produced S/DRel ratios that ranged from approximately 0.1 to 4.4, for 27 different proteins. EXPERIMENTAL SECTION Reagents. Acetonitrile (ACN) and water were purchased from Burdick and Jackson (Muskegon, MI). All other chemicals were obtained from Sigma-Aldrich (St. Louis, MO) and were the highest purity and quality available. Sample Preparation. E. coli strain K-12 bacteria were obtained from the American Type Culture Collection (ATCC, Rockville, MD) and were cultured in M9 minimal media supplemented with glucose (0.2 mL of 1 M solution in 100 mL of culture) and MgSO4 (0.2 mL of 1 M MgSO4 in 100 mL of culture). The 100-mL E. coli cultures were grown in 125-mL flasks at 37 °C with shaking (∼50 rpm) and were grown to an OD600 of 0.4. Then, 1 mL of 1 M 13C6-labeled glucose was added to 100 mL of cultured cells. The cells were allowed to continue growing for 30 min after the addition of labeled glucose. The cells were then pelleted at 5000g. Protein was extracted by suspending the pellet in a buffer consisting of 8 M urea in 25 mM Tris, pH 8.0, and vortexing for 10 min followed by three freeze-thaw cycles. The concentration of protein was determined using the BCA assay (Pierce, Rockford IL). Fifty micrograms of protein was loaded on a one-dimensional gradient (10-20% T) SDS-PAGE gel (Invitrogen, Carlsbad, CA). The gel was run at 125 V with voltage being constant for 1.5 h. Staining was performed using Coommassie Blue to visualize the protein bands. The gel lane was cut from the gel and cut into 25 pieces with numbering beginning at the bottom of the gel for a first dimension of separation. In-gel digestion was then performed using a modified version of a standard in-gel digest protocol (http://donatello.ucsf.edu/ingel.html). In brief, the Commassie Blue was removed from the gel with 3 washes of 25 mM ammonium bicarbonate/50% acetonitrile. The gel slices were then dehydrated by placing them in a Speedvac for 20 min. A 100-ng sample of trypsin (Promega, Madison, WI) was added to each gel slice in 25 mM ammonium bicarbonate and allowed to rehydrate on ice for 10 min. Enough 25 mM ammonium bicarbonate was added to each gel to completely cover the gel slice, and the resultant mixture was incubated at 37 °C overnight. Peptides were extracted using 50% ACN in 0.1% TFA. The ACN was then evaporated off with a Speedvac to obtain a final sample volume of ∼20 µL. 88
Analytical Chemistry, Vol. 76, No. 1, January 1, 2004
Off-Line LC-MALDI TOF/TOF. An Agilent 1100 binary pump was used to generate a 1 µL/min flow using a 300:1 split, which was connected to a M-435 microinjection valve (Upchurch, Oak Harbor, WA). A 100 µm i.d. × 360 µm o.d. reversed-phase column packed as described above was connected to the microinjection valve. Five microliters of peptide extract from a gel band was injected onto the column. The effluent of the column was mixed 1:1 with 10 mg/mL R-cyano-4-hydroxycinnamic acid (in 70% ACN, 0.1% TFA) and spotted directly onto a 100-well MALDI target using a LC Packings Probot (Dionex Corp., Sunnyvale, CA). The gradient was 50 min in length going from 5 to 35% ACN in 0.1% TFA, and spots were made every 30 s. The approximate volume of each MALDI spot was 500 nL. The MALDI plate was than analyzed using the ABI 4700 MALDI TOF/TOF (Applied Biosystems, Framingham, MA) mass spectrometer. Each mass spectrum reported here represents 3000 summed laser shots. All MS/MS spectra were acquired without collision gas present in the cell using a (3 Da mass isolation window. Tandem mass spectra were analyzed by MASCOT and manual de novo sequencing. LCQ Analysis. An LC Packings Ultimate Pump, Switchos column switching device and Famos autosampler (Dionex Corp.) were interfaced to a LCQ DECA XP ion trap mass spectrometer (ThermoFinnigan, San Jose, CA) using a 100 µm i.d. × 360 µm o.d. reversed-phase column. The column was 7 cm in length and was made similarly to columns detailed by Martin et al.,26 except the same material was used for both the frit/plug and the column. The material used in the column was a monodisperse 5-µm polymeric small-bead RPC medium column packing material (Source 5RPC, gift from Amersham BioSciences, Piscataway, NJ). A capillary trap was made out of the same packing material and was used to facilitate the loading of samples injected via the Famos onto the 100-µm-i.d. column. A 5-µL sample of the peptide gel band extract was loaded onto the capillary trap and washed briefly with 0.1% aqueous formic acid (5 min) before switching in-line with the column and mass spectrometer. A 90-min elution gradient was used from 15 to 50% B (A, water with 0.1% formic acid; B, 70% ACN with 0.1% formic acid) at a flow rate of 250 nL/min. The mass spectrometer was set up to take one full-scan MS from the mass range of 300-2000 m/z followed by three MS/MS spectra of the three most intense peaks. All MS/MS spectra were analyzed by TurboSEQUEST in Bioworks 3.0. This allowed rapid identification of the peptides in a gel band before LC-MALDI TOF analysis of the same band. Data Processing. Three different data processing schemes were tested for their ability to accurately and rapidly identify the peptides from the MALDI TOF spectra. The first method employed manual de novo sequencing and accurate mass to derive partial sequence information from which to search the E. coli proteome database. The proteome consisted of all the theoretical tryptic peptides with up to one missed cleavage. The second MS/ MS method employed the database search program MASCOT to search tandem mass spectra against the E. coli proteome database. The third, and most rapid, method was to split the samples in half and run half by LC-MS/MS ion trap mass spectrometry to identify major protein components of the SDS-PAGE gel slice. These proteins were then used to make an E. (26) Martin, S. E.; Shabanowitz, J.; Hunt, D. F.; Marto, J. A. Anal. Chem. 2000, 72, 4266-4274.
coli database of limited size containing only the tryptic peptides identified (up to one missed cleavage). This limited number of theoretical masses was then compared with the masses of the peaks in the spectra obtained from running the same sample by LC-MALDI TOF. This method proved to be the fastest and most reliable for identification of the most proteins, although not all proteins identified by LC-MS/MS ion trap mass spectrometry were found in the LC-MALDI experiment. Since multiple peptides were found from a limited number of proteins rather than a single peptide per protein (from all the proteins), this lends confidence that the identifications were real and not random matches. Analysis of the synthesis/degradation ratio was accomplished by comparing real data with a theoretical modeling of the synthesis/ degradation ratio using a Poisson model as described in the text. RESULTS AND DISCUSSION Definition of the Relative Synthesis/Degradation Ratio. The SDMS technique described here measures how quickly a newly made protein accumulates as a function of the rate of protein degradation relative to other proteins in the cell. Thus, there are two fundamental values involved in calculating such a ratio. The first is that of protein synthesis, which will be in part driven by the amount of mRNA, the amount of appropriately charged tRNA, the growth state of the cell, the rate at which the mRNA in question is being made and removed, and other biological variables. The second fundamental value is that of degradation rate, which will be modified by the removal of protein by the cell using the principle of the N-terminal rule, the specific proteolysis of key regulator proteins, certain conditions that denature the protein, the presence of protein chaperones for help in refolding, perturbations such as heat shock, and a host of other biologically meaningful processes. Although being able to measure the exact kinetics of both protein synthesis and degradation individually would be the most useful, being able to clearly define their relationship or ratio can also provide valuable insight into the current mechanisms operating in a cell under a given set of conditions. Finally, one has to consider the amount of information gained versus the cost and time required to perform an experiment. Here, a single facile doping of the cell culture, and potentially one shotgun proteomic analysis, can provide novel insight into the mechanisms occurring in the cell under a given set of conditions. Such useful information is gained both quickly and inexpensively with most of the additional cost coming from the labeling reagent and most of the extra time incurred during data analysis. Techniques that use very small quantities of protein for mass spectrometric analysis, as well as computer-automated data analysis, will eliminate even these hurdles in the future. Advantage of Culture Doping. To link the mRNA directly with the protein expression profiles obtained by shotgun proteomics and isotopic labeling methods, a way to analyze the relative expression rate or relative synthesis/degradation ratio must be found. The method described above uses culture doping to measure the relative ratio of new protein synthesis to protein degradation for each specific protein in a cell. Because the measurement made is relative, problems associated with the use of chemostats to establish continuous cell doubling rates for calculation of an exact kinetic turnover rate constant are avoided. The use of a chemostat that establishes the conditions of a continuous culture to determine turnover kinetic rate constants
Scheme 1. Experimental Design of the Synthesis/Degradation Ratio Mass Spectrometry Experimenta
a The separation step and the Poisson modeling used for the S/DRel ratio calculation are the keys for high-throughput analysis.
is admirable; however, the time required to develop the necessary conditions for different growth rates of the cells, much less to account for culture perturbations such as heat shock, makes this technique cumbersome at best in high-throughput proteomics. On the other hand, culture doping requires that the cells have enough time to take up and start incorporating the substrate, in this case [13C6]-glucose, into proteins (note: the precursor from which all proteins are being synthesized is the same). For most prokaryotic organisms, this problem is solved by waiting approximately one cell doubling time because this measurement is relative, and an exact time is not required outside of waiting long enough for the cells to start using the labeled glucose. Even differences in the rate of amino acid synthesis, tRNA charging, and other biological phenomena that would allow one amino acid to accumulate 13C more quickly than the other amino acids can be ignored as long as the newly synthesized protein can be distinguished from the original peptide isotope distribution. Preferably all amino acids would accumulate 13C at the same rate, and that rate would coincide perfectly with the percentage of [13C]glucose addition. However, the elegance of this methodology is that accumulating 13C at the same rate is not required for the experiment to work. Only the ability to distinguish new-made polypeptide from the original protein is required for this methodology to work. Finally, by being adaptable to shotgun proteomics, the relative ratio measurement greatly speeds up monitoring of the relative ratios of synthesis/degradation by avoiding the need for multiple two-dimensional gel electrophoresis and in-gel digestions of every spot. Utility of [13C6]Glucose to SDMS Analysis. In Scheme 1 is shown the experimental design for the SDMS approach. The experiments discussed here were designed to generate data to show proof of principle for the SDMS approach in measuring dynamic turnover. To a growing E. coli culture in mid-log phase, 13C-labeled glucose was added and the culture was harvested after a 30-min time period to ensure that approximately one generation of cells could incorporate the 13C-labeled carbon source. An important part of this separation scheme is the fact that [13C]Analytical Chemistry, Vol. 76, No. 1, January 1, 2004
89
Figure 1. Type of data obtained by SDMS approach using the MALDI TOF/TOF. All peptides derived from yfiD found in E. coli strain K12 are shown. SDMS mass spectrum of peptides (a) LGDIEYR, (b) VEGGQHLNVNVLR, (c) ETLEDAVK, and (d) FNSLTPEQQR. Although there is some overlap between the newly synthesized protein isotope distribution and the original peptide isotope distribution at low mass, the majority of each distribution is well-resolved, making modeling and calculation of the S/Drel ratio easy. Also note the newly synthesized peptides’ Poissonlike distribution around a central mean. This suggests that the uptake and conversion of glucose to amino acids is either fast or there exists small metabolite pools. If it were slow, or large metabolite pools existed at any step, one would expect the distribution to smear from the monoisotopic mass to the mean of the current distribution as the pool equilibrated.
carbon labeling is employed. By using 13C as opposed to a dueterated label, the separation behavior of the peptides is equivalent on a reversed-phase column since the labeled and unlabeled peptides exhibit the same chemical properties. The coelution of the labeled and unlabeled peptides is critical in order to calculate a synthesis/degradation ratio for a given peptide pair and hence determine some measure of the intracellular stability of a particular protein. Another important feature of this approach is that all peptides are labeled. This removes the restriction of finding a peptide with a specific amino acid and thus allows for multiple measurements of the synthesis/degradation ratio and better accuracy in determining that ratio. It also allows for a partial shotgun proteomics approach to be used for the separation scheme. It should be noted that 15N could also have been used for doping of the culture. However, a substantially higher quantity of 15N would be needed to increase the percentage of labeled nitrogen in each peptide to adequately separate the newly synthesized peptide isotope distribution from the naturally occur90 Analytical Chemistry, Vol. 76, No. 1, January 1, 2004
ring one. Thus, 13C represents the optimal isotope for labeling in this experiment, but in future experiments, such as those looking at different carbon substrates, 15N will need to be considered as an alternative label. Type of Data Obtained by SDMS Measurements and Benefits of Comprehensive Amino Acid Labeling. Shown in Figure 1 are typical SDMS mass spectra of peptides from YfiD (a gylcyl radical donor annotated as a putative formate acetyltransferase) obtained after a 30-min incubation period with 13C-labeled glucose (protein derived from gel band 5). Here the use of the fifth gel band is arbitrary and represents simply the use of moderately small proteins in the analysis. It should be noted that an attempt of the same experiment on a instrument with much lower resolution, such as an ion trap, would make distinguishing of the individual isotopes impossible and, thus, later data analysis difficult. Although higher resolution could be obtained with slower scan rates, there is no guarantee that the instrument could fit all the isotopic peaks in a single zoom scan when operated in data-
dependent mode. The S/Drel ratio calculated for each of the peptides is shown in Figure 1. Four separate peptides from differentially separated MALDI spots were compared directly and represented 30% sequence coverage of the intact protein. The average S/Drel ratio value for the four peptides was 4.37 with a standard deviation of 0.20. By calculating the S/Drel ratio for multiple (>2) peptides from the target protein, precision data can be obtained for the protein turnover rate. Each of the four peptides identified from YfiD were subjected to high-energy CID tandem mass spectrometry (MS/MS) on the ABI 4700 TOF/TOF. To verify that there was no spectral overlap of other peptides present in the isotopic distribution envelopes of the labeled/unlabeled peptide pairs, MS/MS data were obtained on the peaks associated with the 12C- and 13C-incorporated portion of the isotopic distribution and were then compared to MS/MS data derived from the 12C peak of the distribution. In Figure 2 are shown the MS/MS data of the peptide with sequence LGDIEYR for both the 12C (m/z 865.442) and the entire distribution (maximum allowable mass isolation width of 10 Da). Prominent y1, y2, y4, and b3 ions are found in both spectra, with a shift of several daltons higher in mass units observed for the 13C-incorporated MS/MS data in Figure 2b. As expected, the higher the m/z value for the MS/ MS data of the 13C-incorporated portion, the higher the mass shift observed. This type of MS/MS analysis can be easily employed to help answer questions of unlabeled/labeled peptide pair purity. This figure demonstrates the type of data obtained by SDMS, the requirement for high resolution of the precursor ion, and the ability to empirically prove that only a single peptide is contributing to the isotope distribution. Modeling SDMS Data for Partial Automation of Data Analysis. One of the key aspects to high-throughput SDMS will be the ability to automate the data analysis. This means that a model capable of calculating theoretical S/Drel ratios needs to be developed. Use of the program IsoPro to calculate a theoretical distribution based on the estimated percent incorporated 13C label resulted in an ∼25% error when compared with the experimental data, as shown in Figure 3a. This suggests that the doped labeled glucose does not have rapid equilibration kinetics in some pool of biological molecules, such as during the conversion of the glucose to amino acids. Thus, some other theoretical model will need to be used for more accurately estimating the isotope distribution from the newly made proteins. Another method to use is a Poisson-based model to try and better account for the natural flux of the added labeled glucose, since Poisson models are known to represent natural phenomena better than normal distributions typical of other isotope calculation programs. First, a quick method to calculate the normal isotope distribution is used. Modeling of the 12C distribution presented here is based on the automated reduction and interpretation of highresolution electrospray mass spectra as originally described by Horn et al..27,28 The area of the 12C distribution is modeled based on where the nominal mass value falls in five defined mass ranges: 500-1000, 1000-1500, 1500-2000, 2000-2500, or >2500. For each of these mass ranges, the approximate isotopic distributions of the monoisotopic 12C peak and the first six 13C peaks are (27) Horn, D. M.; Zubarev, R. A.; McLafferty, F. W. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 10313-10317. (28) Horn, D. M.; Zubarev, R. A.; McLafferty, F. W. J. Am. Soc. Mass Spectrom. 2000, 11, 320-332.
Figure 2. Tandem mass spectral data taken using the MALDI TOF/ TOF to show that the observed distribution comes from a single peptide. (a) MS/MS data of the 12C peak from the peptide LGDIEYR from a putative acetyltransferase. (b) MS/MS data of the 12C and 13C distribution of the peptide in Figure 1a. The fact that fragmentation peaks from the normal isotope distribution are observed in the tandem mass spectrum of the newly synthesized isotope distribution (and only broaden) suggests that there are no other contaminating peptides hidden in the Poisson-like distribution. These spectra also demonstrate that the same peptide is present in both distributions.
defined for each of the limiting mass range values stated previously. For example, at m/z 500, the abundance ratios of the isotopes are as follows: 12C ) 0.750 47, 13C ) 0.205 95, 213C ) 0.037 58, 313C ) 0.005 12, and 413C ) 4.13 × 10-5, 513C ) 0. Next an interpolation factor (IF) is calculated that allows for interpolation of the 12C isotope distribution across all range limiting values as shown in eq 1,
IF(500,1000,1500,2000,2500) ) 1 - [(12Cmass - R500,1000,1500,2000,2500)/500] (1) where 12Cmass is the nominal mass of the monoisotopic peak and R is the value for each limiting mass range. Intensity correction values (ICV) are then calculated based on the intensity value for the 12C monoisotopic peak obtained experimentally and the Analytical Chemistry, Vol. 76, No. 1, January 1, 2004
91
defined abundance ratios for each of the mass limits using equations defined for the lower and upper mass range values that bracket the monoisotopic mass:
ICVLOWER LIMIT(12C,13C...513C) ) [isotopic abundance ratio(12C,13C...513C)/12C abundance ratio] [12C intensity][IFRvalue] (2) and
ICV UPPER LIMIT(12C,13C...513C) ) [isotopic abundance ratio(12C,13C....513C)/12C abundance ratio] [12C intensity][1 - IFRvalue] (3) The values for the lower and upper limits are then summed to get the corrected intensity value for each of the six original isotopic distribution masses. These six corrected intensity values are then summed to obtain the area of the 12C isotope distribution described in eq 1. In Figure 3a is shown an expanded view of the 12C isotope distribution associated with the peptide AAFDFAVEHQSVER derived from membrane-bound ATP synthase, F1 sector, delta subunit after addition of 13C-labeled glucose (data normalized to the 12C peak). Also in Figure 3b is shown the calculated isotopic distribution using the intensity correction values summed from eqs 2 and 3 (represented by the gray bar graph). All calculated values agree well within approximately 5-15% of the true intensities derived from the experimental data. Since these calculations are much simpler and faster than complex mathematics associated with other isotope distribution programs, incorporation of these calculations into a fully automated data analysis program like THRASH should be relatively facile and not CPU intensive. The distribution of the 13C-labeled peptide isotopes is calculated using the standard form of the Poisson distribution defined here as
P(x) ) λxe-λ/x!
Figure 3. Comparison of the relative merits of using true isotope distribution from the IsoPro program to that of the Poisson model. (a) This figure shows the IsoPro distribution (bar graph) overlaid on the experimental data for the AAFDFAVEHQSVER peptide. The comparison S/Drel ratio using the experimental peaks to the IsoPro model gives an ∼25% error. (b) Corrected isotope values for the predominantly 12C distribution are shown as bar graphs plotted against the experimental data to show that the interpolation works to within (10%. (c) Poisson model (bar graph) and experimental data for the AAFDFAVEHQSVER peptide that gives an ∼15% error. This Poisson model is employed in the calculations in order to streamline the data processing associated with S/Drel ratio calculations. Although the Poisson model is an approximation of the distribution, it fits the data better than the expected isotope distribution, but some additional variation to the newly synthesized peptide distribution is occurring and further modeling will be needed to perfectly fit the real data and the theoretical model. 92
Analytical Chemistry, Vol. 76, No. 1, January 1, 2004
(x ) 0, 1, 2, 3...30)
(4)
where λ is the mean value of the 13C distribution from the labeled peptide and x represents a total of 31 isotope peaks total (0 represents the 12C monoisotopic peak) for a unlabled/labeled peptide pair. The Poisson function for each of the 31 distribution points is then calculated based on the input mean 13C value. Starting with the calculated 12C isotopic distribution values calculated by summing eqs 2 and 3, intensity corrections are then calculated for the 31 isotopic distribution components based on the original intensity of the mean 13C value. For each of the 31 isotopic distribution components, Poisson values are calculated for each corresponding 13C mass shift. These Poisson values are then summed and normalized to the highest summation value. Normalized Poisson values are then multiplied by the intensity of the mean 13C distribution value and are then added to the original six 12C isotopic distribution values calculated by summing eqs 2 and 3. The remaining 35 calculated values are used as is and are shown in combination with the summed first 6 data points plotted for the peptide AAFDFAVEHQSVER derived from membrane-bound ATP synthase, F1 sector, delta subunit after addition
of 13C labeled glucose in Figure 3c. The S/Drel ratio is then calculated via eq 5 to obtain a value of 0.271.
S/D ratio )
∑ intensities of ∑ intensities of
13
C Poisson distribution
12
C isotope distribution
(5)
Comparison of the theoretical S/Drel ratio calculation to that from the experimental data shows an ∼15% error, which is significantly better than using theoretical distribution obtained by IsoPro. Thus, the Poisson model was used to make a spreadsheet that helped to partially automate the rest of the analysis. Although the Poisson model is likely just an approximation of the labeled distribution, the interest here is for a fast and automatable way to calculate the S/Drel ratio rather than to understand the biology behind the shape of the isotopic distribution. Future work to find a better model for these distributions is needed, but the current error from the Poisson model is not much more than twice that which can be accounted for by the MALDI process and biological variability, as shown from the standard deviation value of the four peptides from YfiD. Benefits of SDS-PAGE Followed by Off-Line LC-MALDI TOF Analysis for SDMS. To make the method amenable to highthroughput analysis, cellular protein extracts were subjected to 1-D SDS-PAGE gel electrophoresis, followed by in-gel digestion. This use of SDS-PAGE to simplify or limit the number of proteins in the mixture before shotgun proteomics is most comparable to the directed shotgun approach to proteomics. The extracted digested peptides were then subjected to HPLC using 0.1% TFA as the mobile-phase modifier. The effluent from the HPLC was then diverted to a microfraction collector and was spotted directly on the MALDI target plate. Just prior to spotting, matrix was added at a tee junction to the LC effluent. This methodology is an adaptation of the original report on MALDI LC-MS/MS (solidphase analysis of peptides) by Karger et al.29 There are three features inherent in this approach that are significant for SDMS analysis. The first is the fact that signal averaging can be accomplished with several thousand laser shots in a very short period of time (laser repetition rate of the instrument is 200 Hz). This signal averaging phenomenon lends itself well to low-intensity signals that would otherwise have poor ion statistics with direct on-line LC-MS experiments where scanning instruments are used. A second advantage of this approach is the ability to archive a particular LC-MS analysis for a given time period. Here one could use a particular reverse-phase run as a performance standard or reanalyze a particular spot that contains a low-intensity peptide where increased signal averaging is needed. Perhaps the most important advantage of solid-phase LC-MS is the ability to optimize the reversed-phase chromatography using traditional ion (29) Rejtar, T.; Hu, P.; Juhasz, P.; Campbell, J. M.; Vestal, M. L.; Preisler, J.; Karger, B. L. J. Proteome Res. 2002, 1, 171-179. (30) Cargile, B. J.; Talley, D. L.; Stephenson, J. L., Jr. Electrophoresis, in press. (31) Blankenhorn, D.; Phillips, J.; Slonczewski, J. L. J. Bacteriol. 1999, 181, 2209-2216. (32) Green, J.; Baldwin, M. L. Mol. Microbiol. 1997, 24, 593-605. (33) Perrot, F.; Hebraud, M.; Charlionet, R.; Junter, G. A.; Jouenne, T. Electrophoresis 2000, 21, 645-653. (34) Kirkpatrick, C.; Maurer, L. M.; Oyelakin, N. E.; Yoncheva, Y. N.; Maurer, R.; Slonczewski, J. L. J. Bacteriol. 2001, 183, 6466-6477. (35) Stancik, L. M.; Stancik, D. M.; Schmidt, B.; Barnhart, D. M.; Yoncheva, Y. N.; Slonczewski, J. L. J. Bacteriol. 2002, 184, 4246-4258.
Figure 4. (a) Benefit of a high-resolution instrument to the SDMS method showing that some overlapping peptide interferences can be resolved to ensure proper S/Drel ratio calculations. (b) Limitations of the technique employing the gel band-LC-MALDI TOF/TOF separation approach as shown by numerous overlapping peptides, which make it impossible for accurate S/Drel ratio calculations to be made. It is expected that an even higher resolving power instrument such as an FTICR will remove the majority of the issues associated with incomplete resolution of peptides of similar mass in the first and second dimensions.
pairing reagents such as TFA.29 Since the chromatographic system is not directly coupled to an electrospray interface, one can maximize column efficiency and hence the number of peptides that can be analyzed using SDMS. This is a crucial part of the SDMS approach, since the ability to effectively calculate the S/Drel ratio depends on the ability of the separation system to effectively isolate labeled/unlabeled peptide pairs from one another. In some instances, the high resolving power of the MALDI TOF can deconvolute interfering peptides as shown in Figure 4a. In Figure 4b is shown an example of the current limitations caused by overlapping peptides associated with this separation approach. Given this potential problem and the inherent limited range of MALDI TOF alone, one should consider the potential peak capacity and dynamic range of this directed-shotgun approach. SDS-PAGE has a large resolving power and high peak capacity. Even on a 10-cm-long gel, several hundred distinct bands can be Analytical Chemistry, Vol. 76, No. 1, January 1, 2004
93
seen and could provide a peak capacity close to that number, if large numbers of gel slices were taken. Here, however, only 25 gel slices were taken, thus limiting the peak capacity to 25. Another benefit of SDS-PAGE gels is the ability to resolve proteins that differ greatly in concentration, with the maximum upper limit for a single protein on a 10-cm gel being ∼2 µg. The lower limit of this technique is well below detectable levels using Commassie Brilliant Blue stain, but for the sake of argument, it will be considered to be 100-fold less. This gives the SDS-PAGE step a dynamic range of ∼100. Capillary chromatography is the most common method for coupling other separation techniques to mass spectrometry. Reports of the peak capacity of the technique in proteomic applications have ranged from 100 to 1000. The peak capacity in this experiment has a maximum of 100 because only 100 spots from the LC run are spotted on the MALDI target. Most peptides that are observed in this experiment elute in one to two wells only, as can be seen in the single ion chromatograms. Thus, a peak capacity of between 50 and 75 is possible. The other consideration is dynamic range. Most columns we have made possess a minimal dynamic range of 100 (data not shown). Thus, when combined with the SDS-PAGE, an overall peak capacity of 1250 (25 × 50) and dynamic range of 10 000 are possible. The MALDI TOF/TOF has a peak capacity of roughly 200 (2000 m/z/ 10 m/z peak widths). Here, the 10 m/z peak widths refers to the need to completely resolve S/Drel ratio isotope clusters to facilitate automated data analysis. This capacity could easily double or triple with the development of software that can handle overlapping peaks. The dynamic range of the MALDI-TOF/TOF is easily over 100. This gives a total dynamic range of 1 000 000 that could easily be extended, with optimization of the technique, to 10 000 000 by taking smaller gel bands, using higher resolution chromatography, sampling the chromatograph more frequently, and using software to compensate for overlapping peaks. Although this peak capacity and dynamic range should be large enough for E. coli, larger more complex organisms will likely require separation schemes with great analytical merit. One approach that shows great promise in resolving this issue is the use of isoelectric focusing in the first dimension of shotgun proteomics.30 Toward High-Throughput Analysis Using SDMS. Identification of the peptides present in a reasonably high-throughput fashion is a critical aspect of this approach. Three separate identification strategies were employed to ensure correct peptide/ protein assignment. First, peptides were evaluated with highenergy CID using the MALDI TOF/TOF instrument utilizing the probability-based algorithm MASCOT. In the second strategy, the CID data obtained from the MALDI TOF/TOF instrument was evaluated using the sequence tag approach. Although this approach was much more successful than using MASCOT to analyze the high-energy CID data, the need for manual interpretation of the data removes the possibility of high-throughput automation. The third strategy employed analysis of peptides derived from the gel fraction using LC-MS/MS on a quadrupole ion trap mass spectrometer and identification using the SEQUEST algorithm. By identifying the proteins in the gel band, the peptides with both similar retention time and the same mass could be correlated with the LC-MALDI data. In addition, the conversion of these proteins to a peptide database of limited size allowed the mass accuracy 94
Analytical Chemistry, Vol. 76, No. 1, January 1, 2004
of the MALDI TOF/TOF, ∼10 ppm with internal calibration from other identified peptides, to identify a few more peptides from each protein based on mass measurement alone. Comparing the S/Drel ratio between the accurate mass-based identified peptides and other previously identified peptides from the same protein helped to validate some of these identifications. The most successful strategy employed to date was the SEQUEST-based methodology because this method was both rapid and allowed complete automation of the identification of the proteins present in each gel band. At this point in time, it is not certain what algorithm or combination of algorithms is the most compatible method for SDMS-based experiments since the SEQUEST-based method required the LCMS data to be correlated with off-line LCMALDI TOF/TOF data. A summary of these results is shown in Table 1. Note that a number of peptides identified by mass alone were left out of the table since they could not be verified via tandem mass spectrometry. Examination of the Data from SDMS Measurements Compared to Known Biology. A summary of all calculated S/Drel values for the proteins identified in gel fractions 5 and 6 are shown in Table 1. The number of peptides used in the S/Drel ratio calculations was determined by two factors. First, the signal-to-noise level had to be high enough to allow for correct model prediction using the Poisson distribution. Second, no interfering peptides could coelute with a given unlabeled/labeled peptide pair. The differences observed in Table 1 between the number of peptides used to identify the protein and the number of peptides used for S/Drel ratio calculations are attributed to these two factors. One interesting aspect of the summarized data in Table 1 is the presence of the protein YfiD at a high turnover rate (∼1 order of magnitude higher than any other protein identified). These data indicate that the turnover (or production) of YfiD was significantly induced upon addition of the 13C-labeled glucose, and it should be noted that upon addition of the glucose, a microaerobic (lowoxygen tension) culture condition was established. Previous studies have shown that expression of E. coli yfiD is maximally induced under conditions of acidic pH and microaerobicity.31-36 Recently it has been suggested that YfiD functions to restore activity to the oxygen-sensitive enzyme, pyruvate formate-lyase (PFL).36,37 PFL is a central carbon catabolism enzyme that is responsible for cleaving pyruvate to form acetyl-CoA and formate during anaerobic growth, and it has been shown that PFL requires an intact gylcyl radical domain for activity and that this glycyl radical domain is readily cleaved upon oxygen exposure.38 Therefore, given its function, it is expected that the highest turnover rate (or levels) of YfiD would be required under microaerobic conditions, when the possibility of oxygenolytic cleavage of PFL is greatest.36 Therefore, the results obtained with the SDMS technique are consistent with an increase in either turnover rate or expression associated with previous studies involving yfiD expression. This example shows that SDMS can potentially provide biological insight into the mechanics of cellular life. (36) Wyborn, N. R.; Messenger, S. L.; Henderson, R. A.; Sawers, G.; Roberts, R. E.; Attwood, M. M.; Green, J. Microbiology 2002, 148, 1015-1026. (37) Wagner, A. F.; Schultz, S.; Bomke, J.; Pils, T.; Lehmann, W. D.; Knappe, J. Biochem. Biophys. Res. Commun. 2001, 285, 456-462. (38) Wagner, A. F.; Frey, M.; Neugebauer, F. A.; Schafer, W.; Knappe, J. Proc. Natl. Acad. Sci. U.S.A. 1992, 89, 996-1000.
Table 1. Summary of S/DRel Ratio Calculations for Gel Slices 5 and 6 peptide ID method
no. of peptides IDed per protein
peptide sequences used in S/Drel ratio calculationsa
S/Drel ratio
gi|16131187|ref NP_417767.1| 50S ribosomal subunit protein L5[E. coli K12]
SEQUEST
7
MWEFFER EQIIFPEIDYDK
0.155 0.139
0.147
gi|16128411|ref NP_414960.1| orf, hypothetical protein [E. coli K12]
SEQUEST
5
VQAQIQGDEIR
0.310
n/a
gi|16131812|ref NP_418409.1| component in transcription antitermination [E. coli K12]
SEQUEST
5
LQQVGDKPRPK TLFEPGEMVR
0.252 0.284
0.268
gi|16128536|ref NP_415085.1| outer membrane porin protein; locus of qsr prophage [E. coli K12]
SEQUEST
2
HYFSSNDADDGDTTYAR
1.23
n/a
gi|16130343|ref NP_416912.1| PTS system, glucose-specific IIA component [E. coli K12]
SEQUEST
7
IVGDGIAIKPTGNK VGDTVIEFDLPLLEEK STLTPVVISNMDEIK LSGSVTVGETPVIR
0.340 0.322 0.347 0.312
0.330 ( 0.016
gi|16129784|ref |NP_416344.1| carboxyterminal protease for penicillin-binding protein 3 [E. coli K12]
SEQUEST
7
IAKDPEFQNIMK LEKARPAEQPAPVK
0.202 0.213
0.207
gi|16130599|ref NP_417172.1| orf, hypothetical protein [E. coli K12]
SEQUEST
4
TGFYMSLIGTPDEQR INSNEELALPK
0.600 0.662
0.631
gi|16129785|ref |NP_416346.1| orf, hypothetical protein [E. coli K12]
SEQUEST
4
TEFYADLNR IACVRIPVGR NQIIGVLDIDSTVFG
0.412 0.444 0.411
0.422 ( 0.019
gi|16131143|ref |NP_417721.1| acetylCoA carboxylase, BCCP subunit; carrier of biotin [E. coli K12]
SEQUEST
2
SPMVGTFYR
0.258
n/a
gi|16131603|ref| NP_418191.1| membranebound ATP synthase, F1 sector, delta-subunit [E. coli K12]
SEQUEST
4
AAFDFAVEHQSVER
0.271
n/a
gi|16132194|ref |NP_418793.1| hyperosmotically inducible periplasmic protein [E. coli K12]
SEQUEST
4
VETTDGVVQLSG TVDSQAQSDR
0.232
n/a
gi|16128478|ref |NP_415027.1| acyl-CoA thioesterase I; also functions as protease I [E. coli K12]
SEQUEST
6
MMNFNNVFR GFQPQQTEQTLR WVLVELGGNDGLR YNEAFSAIYPK
0.488 0.432 0.536 0.491
0.442 ( 0.088
gi|16128796|ref |NP_415349.1| putative asparaginase [E. coli K12]
SEQUEST
2
GMERVSPEIFSTSLR
0.288
n/a
gi|16130596|ref| NP_417169.1| regulator of plasmid mcrB operon (microcin B17 synthesis) [E. coli K12]
SEQUEST
4
MDSSFTPIEQMLK GHEFLR DQLEQITRK
0.527 0.351 0.449
0.457 ( 0.110
gi|16129071|ref |NP_415626.1| orf, hypothetical protein E. coli K12]
SEQUEST
2
IDRPEEYADIATK
0.202
n/a
gi|16129733|ref| NP_416293.1| glyceraldehyde-3phosphate dehydrogenase A [E. coli K12]
SEQUEST
3
LEKAATYEQIKAAVK
0.439
n/a
protein
av S/Drel ratiob
Analytical Chemistry, Vol. 76, No. 1, January 1, 2004
95
Table 1 (Continued) protein
peptide ID method
no. of peptides IDed per protein
peptide sequences used in S/Drel ratio calculationsa
S/Drel ratio
av S/Drel ratiob
gi|16128660|ref |NP_415210.1| flavodoxin 1 E. coli K12]
SEQUEST
3
GLADDDHFVGLAIDEDR
0.280
n/a
gi|16128858|ref| NP_415411.1| periplasmic protein effects translocation of lipoproteins from inner membrane to outer [E. coli K12]
SEQUEST
4
NQSSDWQQYNIK DGTIHQFSAVEQDDQR
0.381 0.310
0.345
gi|16129019|ref |NP_415574.1| orf, hypothetical protein [E. coli K12]
SEQUEST
5
EGQHAFVNFR IDKEGQHAFVNFR DFDGTFTFDEK VNVTINTTSVDTNHAER
0.171 0.154 0.152 0.135
0.153 ( 0.015
gi|16128138|ref |NP_414687.1| dnaK suppressor protein [E. coli K12]
SEQUEST
2
TVTHMQDEAANFPDPVD
0.253
n/a
gi|16130220|ref |NP_416788.1| NADH dehydrogenase I chain E [E. coli K12]
SEQUEST
4
MHENQQPQTEA FELSAAER
0.0869 0.117
0.102
gi|16131184|ref |NP_417764.1| 50S ribosomal subunit protein L6 [E. coli K12]
SEQUEST
5
HADNTLTFGPR DGYADGWAQAGTAR
0.163 0.149
0.156
gi|1788933|gb |AAC75632.1| putative formate acetyltransferase [E. coli K12] (YfiD)
Sequence Tag (TOF/TOF) and MASCOT
4
LGDIEYR VEGGQHLNVNVLR ETLEDAVK FNSLTPEQQR
4.44 4.32 4.67 4.05
4.37 ( 0.20
gi|1790169|gb |AAC76754.1| membranebound ATP synthase, F1 stor, epsilon- subuni [E. coli K12]
Sequence Tag (TOF/TOF)
1
GQDLDEAR
0.335
n/a
gi|1789700|gb |AAC76329.1| 50S ribosomal subunit protein L18 [E. coli K12]
Sequence Tag (TOF/TOF)
1
SGFQYHGR
0.185
n/a
gi|1790585|gb |AAC77102.1| GroES, 10 Kd chaperone binds to Hsp60 in pres. Mg-ATP, suppressing its ATPase activity [E. coli K12]
Sequence Tag (TOF/TOF)
2
GEVLAVGNGR VGDIVIFNDGYGVK
0.330 0.343
0.336
gi|1789702|gb |AAC76331.1| 30S ribosomal subunit protein S8, and regulator [E. coli
Sequence Tag (TOF/TOF)
1
AVVESIQR
0.188
n/a
gi|1788984|gb |AAC75679.1| orf, hypothetical protein [E. coli K12]
Sequence Tag (TOF/TOF)
1
YGRIMLDTAK
0.346
n/a
a Peptides used for quantitation met two criteria, a minimum signal-to-noise ratio in order to perform the S/D ratio calculation and no interferences rel from other peptides. The difference between the number of peptides used to identify the protein and the number used for quantitation reflect these criteria. b Average values calculated for a minimum of two peptides. Standard deviation of 1σ calculated for a minimum of three peptides.
CONCLUSIONS Here we describe a new technique for quantitative proteomics that analyzes the dynamic turnover of the proteome of a cell. This technique is unique from other quantitative strategies for a variety of reasons. Perhaps the most important of these is that only a single cell culture is needed to obtain useful biological information about the processes occurring inside a cell. Thus, conditions that require shifts in the absolute amount of protein present at a given 96 Analytical Chemistry, Vol. 76, No. 1, January 1, 2004
time are not necessary. Furthermore, this technique can be used to compare two differing cell states, so that a comparison of the protein relative turnover rates can be obtained. This measurement may be more significant than the relative or absolute quantity of protein present, based on the inconsistencies observed between the transcriptome and proteome abundance measurements. This statement will likely be more applicable when organisms more complex than yeast are analyzed and the mRNA and protein
abundances are compared. It should also be noted that proteins (i.e., enzymes) from an organism that is just starting to express environmentally inducible proteins should have a very high S/Drel ratio for these species, and thus, the need to compare different cellular states to look for environment-specific polypeptides is alleviated. Another benefit is that any peptide observed could be used for measuring the S/Drel ratio, which means that reliance on the ability of one to a few peptides to ionize well is reduced if not removed. In addition, by measuring multiple peptides from a single protein, a very precise number (depending on the number of peptides analyzed) for the S/Drel ratio can be determined. Since a significant number of biological changes are thought to occur by increasing the level of a protein only slightly, an understanding of the biological phenomenology occurring inside a cell on a much finer scale is possible compared to other quantitative techniques (although this will require the comparison of differing cellular states but not require inducing gross environmental changes). For these reasons, we feel that the SDMS method will open new
gateways for the study of biological phenomena and help to illuminate the enzymatic pathways of cellular life. However, to truly be considered high-throughput biology, this technique will need to be implemented in a shotgun proteomics-style fashion and will require software for the automatic interpretation of these data. These are concurrent areas of interest in our laboratory. ACKNOWLEDGMENT The authors acknowledge the Internal Research and Development Program from the Research Triangle Institute for funding of this research. In addition, we recognize James E.H. Powell from Amersham Biosciences for providing the monodisperse polymeric small bead RPC medium column packing (Source 5RPC) for the reversed-phase separation work. Received for review July 23, 2003. Accepted October 10, 2003. AC034841A
Analytical Chemistry, Vol. 76, No. 1, January 1, 2004
97