Analysis of quantitative proteomic data generated via multidimensional protein identification technology

We describe the analysis of quantitative proteomic samples via multidimensional protein identification technology (MudPIT). Ratio amounts of the soluble portion of the S. cerevisiae proteome from cultures of S. cerevisiae strain S288C grown in either 14N minimal media or 15N-enriched minimal media were mixed and digested into a complex peptide mixture. A 1 x 14N/1 x 15N complex peptide mixture was analyzed by single-dimensional reversed-phase chromatography and electrospray ionization quadrapole time-of-flight mass spectrometry in order to demonstrate the replacement of 14N by 15N under the growth conditions used. After conformation of the incorporation of 15N into the labeled sample, three separate samples consisting of a 1 x 14N/1 x 15N complex peptide mixture, a 5 x 14N/1 x 15N complex peptide mixture, and a 10 x 14N/1 x 15N complex peptide mixture were analyzed via MudPIT. We demonstrate the dynamic range of the system by analyzing a 1:1, 5:1, and 10:1 data set using the soluble portion from S. cerevisiae grown in either 14N or 15N-enriched minimal media. The method described provides an accurate way to undertake a large-scale quantitative proteomic study.     

Development of a multiplexed microcapillary liquid chromatography system for high-throughput proteome analysis

Comprehensive proteome analysis requires the identification (and quantification) of the proteins in samples consisting of thousands of proteins spanning a range of abundance of several orders of magnitude. The currency of proteome analysis by mass spectrometry is the peptides generated by protein proteolysis. The high sample complexity of such samples requires a large separation capacity, which is commonly achieved by fractionation of the mixture followed by further serial separations of each fraction. The sample throughput of proteome analysis is therefore limited by the need to sequentially process large numbers of samples. We have developed a novel four-plexed microcapillary liquid chromatography system for automated, high-throughput separation of complex peptide samples. The system supports the concurrent separation of four different samples by directing identically split solvent-gradient flows into four microcapillary C18 columns. The simple design of the system achieves multiplexed separation without the need for extra solvent pumps. Peak resolution, reproducibility, and parallel separating capacity of the system were investigated using standard peptides. The applicability of the system to high-throughput protein expression profiling was demonstrated in qualitative and quantitative analyses of protein expression in S. cerevisiae grown on two different carbon sources using the isotope-coded affinity tag (ICAT) reagent and matrix-assisted desorption/ionization quadrupole time-of-flight mass spectrometry.     

Correlation of relative abundance ratios derived from peptide ion chromatograms and spectrum counting for quantitative proteomic analysis using stable isotope labeling

In this study, S. cerevisiae crude membrane fractions were prepared using the acid-labile detergent RapiGest from cells grown under rich and minimal media conditions using 14N and 15N ammonium sulfate as the sole nitrogen source. Four independent MudPIT analyses of 1:1 mixtures of sample were prepared and analyzed via quantitative multidimensional protein identification technology on a two-dimensional ion trap mass spectrometer. Using the method described in this study, low-abundance integral membrane proteins with up to 14 transmembrane domains were identified and their protein expression determined when sufficient spectrum counting and ion chromatogram information was generated. We demonstrate that spectrum counting and mass spectrometry derived ion chromatograms strongly correlate for determining quantitative changes in protein expression. Spectrum counting proved more reproducible and has a wider dynamic range contributing to the deviation of the two quantitative approaches from a perfect positive correlation.     

Multidimensional protein identification technology-selected reaction monitoring improving detection and quantification for protein biomarker studies

The targeted analysis of proteins in complex biological samples is best achieved using selected reaction monitoring (SRM). To maximize the sensitivity of this approach, sample fractionation or enrichment is still required, particularly to detect less abundant proteins in clinically relevant biofluids. Here, we report the development of multidimensional protein identification technology (MudPIT)-SRM, taking advantage of the robust online strong cation exchange chromatography for tryptic peptide fractionation and combining it with the multiplexed, quantitative attributes of SRM. The classical MudPIT method has been modified with an in-line strategy to introduce reference peptides onto the analytical column to enable quantitation at each salt step. Applying the MudPIT-SRM approach to profile abundant plasma proteins, we demonstrated mean increases in peak areas of almost 90% compared to conventional SRM. MudPIT-SRM analyses of low abundant proteins present in human wound fluid exudates similarly demonstrated increased peak areas and enabled the detection of proteins which were below the lower limit of detection when analyzed by conventional SRM. The MudPIT-SRM method is relatively facile to conduct and offers performance advantages to enhance sensitivity for biomarker studies.     

A correlation algorithm for the automated quantitative analysis of shotgun proteomics data

Quantitative shotgun proteomic analyses are facilitated using chemical tags such as ICAT and metabolic labeling strategies with stable isotopes. The rapid high-throughput production of quantitative "shotgun" proteomic data necessitates the development of software to automatically convert mass spectrometry-derived data of peptides into relative protein abundances. We describe a computer program called RelEx, which uses a least-squares regression for the calculation of the peptide ion current ratios from the mass spectrometry-derived ion chromatograms. RelEx is tolerant of poor signal-to-noise data and can automatically discard nonusable chromatograms and outlier ratios. We apply a simple correction for systematic errors that improves the accuracy of the quantitative measurement by 32 +/- 4%. Our automated approach was validated using labeled mixtures composed of known molar ratios and demonstrated in a real sample by measuring the effect of osmotic stress on protein expression in Saccharomyces cerevisiae.     

Application of capture-recapture models to estimation of protein count in MudPIT experiments

MudPIT is an automated shotgun proteomics approach that enhances the separation of peptides for sequencing by mass spectrometry analysis.We here adapt a mathematical model from ecology, namely, the capture-recapture model with a closed population and time-varying and heterogeneous individual probabilities of capture, to model the number of peptide identifications across the various cycles of a typical MudPIT experiment. In the absence of any prior information on abundance levels, the model can be used to estimate the total number of proteins in the experimental sample. We apply the model to a recent MudPIT-based experiment to estimate the total number of rat lung endothelial cell surface proteins. The model provides some practical guidelines for planning MudPIT experiments.     

Comparison of fluorescence, laser-induced fluorescence, and ultraviolet absorbance detection for measuring HPLC fractionated protein/peptide mixtures

Proteomics is the study of all proteins in a biological sample. High-pressure liquid chromatography coupled online with mass spectrometry (HPLC/MS) is currently the method of choice for proteomic analysis. Proteins are extracted, separated at the protein or peptide level (after enzymatic digestion), and fractions are analyzed by HPLC/MS. Detection during off-line fractionation is generally conducted using UV-vis, which is not sensitive enough to distinguish fractions having the largest concentration of proteins/peptides and should not be combined prior to HPLC/MS. To overcome this deficiency, we utilize fluorescence or UV-laser induced fluorescence (UV-LIF) detection for measuring proteins/peptides during the off-line fractionation. Fluorescence detection allows low-abundance proteins/peptides that contain aromatic amino acids to be measured. In this study, peptide/protein samples fractionated using ion-exchange chromatography were detected using UV absorbance, fluorescence, and UV-LIF. The results indicated that fluorescence and UV-LIF were able to detect the lower abundance proteins/peptides to give a more representative chromatogram, allowing the analyst to decide which fractions should be combined prior to HPLC/tandem mass spectrometry (MS/MS) analysis.     

Utility of accurate mass tags for proteome-wide protein identification

An enabling capability for proteomics would be the ability to study protein expression on a global scale. While several different separation and analysis options are being investigated to advance the practice of proteomics, mass spectrometry (MS) is rapidly becoming the core instrumental technology used to characterize the large number of proteins that constitute a proteome. To be most effective, proteomic measurements must be high-throughput, ideally allowing thousands of proteins to be identified on a time scale of hours. Most strategies of identification by MS rely on the analysis of enzymatically produced peptides originating from an isolated protein followed by either peptide mapping or tandem MS (MS/MS) to obtain sequence information for a single peptide. In the case of peptide mapping, several peptide masses are needed to unambiguously identify a protein with the typically achieved mass measurement accuracies (MMA). The ability to identify proteins based on the mass of a single peptide (i.e., an accurate mass tag; AMT) is proposed and is largely dependent on the MMA that can be achieved. To determine the MMA necessary to enable the use of AMTs for proteome-wide protein identification, we analyzed the predicted proteins and their tryptic fragments from Saccharomyces cerevisiae and Caenorhabditis elegans. The results show that low ppm (i.e., approximately 1 ppm) level measurements have practical utility for analysis of small proteomes. Additionally, up to 85% of the peptides predicted from these organisms can function as AMTs at sub-ppm MMA levels attainable using Fourier transform ion cyclotron resonance MS. Additional information, such as sequence constraints, should enable even more complex proteomes to be studied at more modest mass measurement accuracies. Once AMTs are established, subsequent high-throughput measurements of proteomes (e.g., after perturbations) will be greatly facilitated.     

Use of artificial neural networks for the accurate prediction of peptide liquid chromatography elution times in proteome analyses

The use of artificial neural networks (ANNs) is described for predicting the reversed-phase liquid chromatography retention times of peptides enzymatically digested from proteome-wide proteins. To enable the accurate comparison of the numerous LC/MS data sets, a genetic algorithm was developed to normalize the peptide retention data into a range (from 0 to 1), improving the peptide elution time reproducibility to approximately 1%. The network developed in this study was based on amino acid residue composition and consists of 20 input nodes, 2 hidden nodes, and 1 output node. A data set of approximately 7000 confidently identified peptides from the microorganism Deinococcus radiodurans was used for the training of the ANN. The ANN was then used to predict the elution times for another set of 5200 peptides tentatively identified by MS/MS from a different microorganism (Shewanella oneidensis). The model was found to predict the elution times of peptides with up to 54 amino acid residues (the longest peptide identified after tryptic digestion of S. oneidensis) with an average accuracy of approximately 3%. This predictive capability was then used to distinguish with high confidence isobar peptides otherwise indistinguishable by accurate mass measurements as well as to uncover peptide misidentifications. Thus, integration of ANN peptide elution time prediction in the proteomic research will increase both the number of protein identifications and their confidence.     

ProRata: A quantitative proteomics program for accurate protein abundance ratio estimation with confidence interval evaluation

A profile likelihood algorithm is proposed for quantitative shotgun proteomics to infer the abundance ratios of proteins from the abundance ratios of isotopically labeled peptides derived from proteolysis. Previously, we have shown that the estimation variability and bias of peptide abundance ratios can be predicted from their profile signal-to-noise ratios. Given multiple quantified peptides for a protein, the profile likelihood algorithm probabilistically weighs the peptide abundance ratios by their inferred estimation variability, accounts for their expected estimation bias, and suppresses contribution from outliers. This algorithm yields maximum likelihood point estimation and profile likelihood confidence interval estimation of protein abundance ratios. This point estimator is more accurate than an estimator based on the average of peptide abundance ratios. The confidence interval estimation provides an "error bar" for each protein abundance ratio that reflects its estimation precision and statistical uncertainty. The accuracy of the point estimation and the precision and confidence level of the interval estimation were benchmarked with standard mixtures of isotopically labeled proteomes. The profile likelihood algorithm was integrated into a quantitative proteomics program, called ProRata, freely available at www.MSProRata.org.     

Proteolytic 18O labeling for comparative proteomics: model studies with two serotypes of adenovirus

A new method for proteolytic stable isotope labeling is introduced to provide quantitative and concurrent comparisons between individual proteins from two entire proteome pools or their subfractions. Two 18O atoms are incorporated universally into the carboxyl termini of all tryptic peptides during the proteolytic cleavage of all proteins in the first pool. Proteins in the second pool are cleaved analogously with the carboxyl termini of the resulting peptides containing two 16O atoms (i.e., no labeling). The two peptide mixtures are pooled for fractionation and separation, and the masses and isotope ratios of each peptide pair (differing by 4 Da) are measured by high-resolution mass spectrometry. Short sequences and/or accurate mass measurements combined with proteomics software tools allow the peptides to be related to the precursor proteins from which they are derived. Relative signal intensities of paired peptides quantify the expression levels of their precursor proteins from proteome pools to be compared, using an equation described in the paper. Observation of individual (unpaired) peptides is mainly interpreted as differential modification or sequence variation for the protein from the respective proteome pool. The method is evaluated here in a comparison of virion proteins for two serotypes (Ad5 and Ad2) of adenovirus, taking advantage of information already available about protein sequences and concentrations. In general, proteolytic 18O labeling enables a shotgun approach for proteomic studies with quantitation capability and is proposed as a useful tool for comparative proteomic studies of very complex protein mixtures.     

Quantitative proteomic analysis using a MALDI quadrupole time-of-flight mass spectrometer

We describe an approach to the quantitative analysis of complex protein mixtures using a MALDI quadrupole time-of-flight (MALDI QqTOF) mass spectrometer and isotope coded affinity tag reagents (Gygi, S. P.; et al. Nat. Biotechnol. 1999, 17, 994-9.). Proteins in mixtures are first labeled on cysteinyl residues using an isotope coded affinity tag reagent, the proteins are enzymatically digested, and the labeled peptides are purified using a multidimensional separation procedure, with the last step being the elution of the labeled peptides from a microcapillary reversed-phase liquid chromatography column directly onto a MALDI sample target. After addition of matrix, the sample spots are analyzed using a MALDI QqTOF mass spectrometer, by first obtaining a mass spectrum of the peptides in each sample spot in order to quantify the ratio of abundance of pairs of isotopically tagged peptides, followed by tandem mass spectrometric analysis to ascertain the sequence of selected peptides for protein identification. The effectiveness of this approach is demonstrated in the quantification and identification of peptides from a control mixture of proteins of known relative concentrations and also in the comparative analysis of protein expression in Saccharomyces cerevisiae grown on two different carbon sources.     

Performance of a linear ion trap-Orbitrap hybrid for peptide analysis

Proteomic analysis of digested complex protein mixtures has become a useful strategy to identify proteins involved in biological processes. We have evaluated the use of a new mass spectrometer that combines a linear ion trap and an Orbitrap to create a hybrid tandem mass spectrometer. A digested submandibular/sublingual saliva sample was used for the analysis. We find the instrument is capable of mass resolution in excess of 40,000 and mass measurement accuracies of less than 2 ppm for the analysis of complex peptide mixtures. Such high mass accuracy allowed the elimination of virtually any false positive peptide identifications, suggesting that peptides that do not match the specificity of the protease used in the digestion of the sample should not automatically be considered as false positives. Tandem mass spectra from the linear ion trap and from the Orbitrap have very similar ion abundance ratios. We conclude this instrument will be well suited for shotgun proteomic types of analyses.     

Shotgun analysis of integral membrane proteins facilitated by elevated temperature

The beneficial effects on peak selectivity and resolution of conducting liquid chromatography (LC) at elevated temperature (e.g., 30-80 degrees C) are generally well-known; however, its importance for peptide recovery is not nearly as well recognized. This report demonstrates that microLC analysis of membrane proteomic samples significantly benefits from the application of heat. Enriched membrane and membrane-embedded peptides (the latter obtained by membrane shaving) were analyzed by microLC-tandem mass spectrometry (MS/MS) from 20 to 60 degrees C using a standard reversed-phase material. Maximal protein and hydrophobic peptide recovery was obtained at 60 degrees C. The membrane-shaving method employed, a recently optimized version of the high pH/proteinase K protocol, provided significant integral membrane protein enrichment: 98% of identified proteins were predicted to have at least one transmembrane domain (87% to have at least three), and 68% of peptides were predicted to contain transmembrane segments. Analysis of this highly enriched sample at elevated temperature increased protein identifications by 400%, and peptide identifications by 500%, as compared to room-temperature separation. Given that most microLC-MS/MS analyses are currently conducted at room temperature, the findings described herein should be of considerable value for improving the comprehensive study of integral membrane proteins.     

Robust estimation of peptide abundance ratios and rigorous scoring of their variability and bias in quantitative shotgun proteomics

The abundance ratio between the light and heavy isotopologues of an isotopically labeled peptide can be estimated from their selected ion chromatograms. However, quantitative shotgun proteomics measurements yield selected ion chromatograms at highly variable signal-to-noise ratios for tens of thousands of peptides. This challenge calls for algorithms that not only robustly estimate the abundance ratios of different peptides but also rigorously score each abundance ratio for the expected estimation bias and variability. Scoring of the abundance ratios, much like scoring of sequence assignment for tandem mass spectra by peptide identification algorithms, enables filtering of unreliable peptide quantification and use of formal statistical inference in the subsequent protein abundance ratio estimation. In this study, a parallel paired covariance algorithm is used for robust peak detection in selected ion chromatograms. A peak profile is generated for each peptide, which is a scatterplot of ion intensities measured for the two isotopologues within their chromatographic peaks. Principal component analysis of the peak profile is proposed to estimate the peptide abundance ratio and to score the estimation with the signal-to-noise ratio of the peak profile (profile signal-to-noise ratio). We demonstrate that the profile signal-to-noise ratio is inversely correlated with the variability and bias of peptide abundance ratio estimation.     

Method development for metaproteomic analyses of marine biofilms

The large-scale identification and quantitation of proteins via nanoliquid chromatography (LC)-tandem mass spectrometry (MS/MS) offers a unique opportunity to gain unprecedented insight into the microbial composition and biomolecular activity of true environmental samples. However, in order to realize this potential for marine biofilms, new methods of protein extraction must be developed as many compounds naturally present in biofilms are known to interfere with common proteomic manipulations and LC-MS/MS techniques. In this study, we used amino acid analyses (AAA) and LC-MS/MS to compare the efficacy of three sample preparation methods [6 M guanidine hydrochloride (GuHCl) protein extraction + in-solution digestion + 2D LC; sodium dodecyl sulfate (SDS) protein extraction + 1D gel LC; phenol protein extraction + 1D gel LC] for the metaproteomic analyses of an environmental marine biofilm. The AAA demonstrated that proteins constitute 1.24% of the biofilm wet weight and that the compared methods varied in their protein extraction efficiencies (0.85-15.15%). Subsequent LC-MS/MS analyses revealed that the GuHCl method resulted in the greatest number of proteins identified by one or more peptides whereas the phenol method provided the greatest sequence coverage of identified proteins. As expected, metagenomic sequencing of the same biofilm sample enabled the creation of a searchable database that increased the number of protein identifications by 48.7% (≥1 peptide) or 54.7% (≥2 peptides) when compared to SwissProt database identifications. Taken together, our results provide methods and evidence-based recommendations to consider for qualitative or quantitative biofilm metaproteome experimental design.     

Improving protein identification sensitivity by combining MS and MS/MS information for shotgun proteomics using LTQ-Orbitrap high mass accuracy data

We investigated and compared three approaches for shotgun protein identification by combining MS and MS/MS information using LTQ-Orbitrap high mass accuracy data. In the first approach, we employed a unique mass identifier method where MS peaks matched to peptides predicted from proteins identified from an MS/MS database search are first subtracted before using the MS peaks as unique mass identifiers for protein identification. In the second method, we used an accurate mass and time tag method by building a potential mass and retention time database from previous MudPIT analyses. For the third method, we used a peptide mass fingerprinting-like approach in combination with a randomized database for protein identification. We show that we can improve protein identification sensitivity for low-abundance proteins by combining MS and MS/MS information. Furthermore, "one-hit wonders" from MS/MS database searching can be further substantiated by MS information and the approach improves the identification of low-abundance proteins. The advantages and disadvantages for the three approaches are then discussed.     

N-terminal isotope tagging strategy for quantitative proteomics: results-driven analysis of protein abundance changes

Comparing the relative abundance of each protein present in two or more complex samples can be accomplished using isotope-coded tags incorporated at the peptide level. Here we describe a chemical labeling strategy for the incorporation of a single isotope label per peptide, which is completely sequence-independent so that it potentially labels every peptide from a protein including those containing posttranslational modifications. It is based on a gentle chemical labeling strategy that specifically labels the N-terminus of all peptides in a digested sample with either a d5- or d0-propionyl group. Lysine side chains are blocked by guanidination prior to N-terminal labeling to prevent the incorporation of multiple labels. In this paper, we describe the optimization of this N-terminal isotopic tagging strategy and validate its use for peptide-based protein abundance measurements with a 10-protein standard mixture. Using a results-driven strategy, which targets proteins for identification based on MALDI TOF-MS analysis of isotopically labeled peptide pairs, we also show that this labeling strategy can detect a small number of differentially expressed proteins in a mixture as complex as a yeast cell lysate. Only peptides that show a difference in relative abundance are targeted for identification by tandem MS. Despite the fact that many peptides are quantitated, only those few showing a difference in abundance are targeted for protein identification. Proteins are identified by either targeted LC-ES MS/MS or MALDI TOF/TOF. Identifications can be accomplished equally well by either technique on the basis of multiple peptides. This increases the confidence level for both identification and quantitation. The merits of ES MS/MS or MALDI MS/MS for protein identification in a results-driven strategy are discussed.     

An automated multidimensional protein identification technology for shotgun proteomics.

We describe an automated method for shotgun proteomics named multidimensional protein identification technology (MudPIT), which combines multidimensional liquid chromatography with electrospray ionization tandem mass spectrometry. The multidimensional liquid chromatography method integrates a strong cation-exchange (SCX) resin and reversed-phase resin in a biphasic column. We detail the improvements over a system described by Link et al. (Link, A. J.; Eng, J.; Schieltz, D. M.; Carmack, E.; Mize, G. J.; Morris, D. R.; Garvik, B. M.; Yates, J. R., III. Nat. Biotechnol. 1999, 17, 676-682) that separates and acquires tandem mass spectra for thousands of peptides. Peptides elute off the SCX phase by increasing pI, and elution off the SCX material is evenly distributed across an analysis. In addition, we describe the chromatographic benchmarks of MudPIT. MudPIT was reproducible within 0.5% between two analyses. Furthermore, a dynamic range of 10000 to 1 between the most abundant and least abundant proteins/peptides in a complex peptide mixture has been demonstrated. By improving sample preparation along with separations, the method improves the overall analysis of proteomes by identifying proteins of all functional and physical classes.