Statistical analysis of relative labeled mass spectrometry data from complex samples using ANOVA

Statistical tools enable unified analysis of data from multiple global proteomic experiments, producing unbiased estimates of normalization terms despite the missing data problem inherent in these studies. The modeling approach, implementation, and useful visualization tools are demonstrated via a case study of complex biological samples assessed using the iTRAQ relative labeling protocol.     

Application of the random forest classification method to peaks detected from mass spectrometric proteomic profiles of cancer patients and controls

The random forest classification method was applied to classify samples from 76 breast cancer patients and 77 controls whose proteomic profile had been obtained using mass spectrometry. The analysis consisted of two stages, the detection of peaks from the profiles and the construction of a classification rule using random forests. Using a peak detection method based on finding common local maxima in the smoothed sample spectra, 444 peaks were detected, reducing to 365 robust peaks found in at least 7 out of 10 random subsets of samples. Subjects were classified as cases or controls using the random forest algorithm applied to the 365 peaks. Based on the prediction of the status of out-of-bag samples, the total error rate was 16.3%, with a sensitivity of 81.6% and a specificity of 85.7%. Measures of importance of each of the peaks were calculated to identify regions of the spectrum influencing the classification, and the four most important peaks were identified as mz3863_13, mz2943_12, mz3193_44 and mz8925_94. Combining initial peak detection with the random forest algorithm provides a high-performance classification system for proteomic data, with unbiased estimates of future performance.     

Classification and identification of bacteria by mass spectrometry and computational analysis

Background

In general, the definite determination of bacterial species is a tedious process and requires extensive manual labour. Novel technologies for bacterial detection and analysis can therefore help microbiologists in minimising their efforts in developing a number of microbiological applications.

Methodology

We present a robust, standardized procedure for automated bacterial analysis that is based on the detection of patterns of protein masses by MALDI mass spectrometry. We particularly applied the approach for classifying and identifying strains in species of the genus Erwinia. Many species of this genus are associated with disastrous plant diseases such as fire blight. Using our experimental procedure, we created a general bacterial mass spectra database that currently contains 2800 entries of bacteria of different genera. This database will be steadily expanded. To support users with a feasible analytical method, we developed and tested comprehensive software tools that are demonstrated herein. Furthermore, to gain additional analytical accuracy and reliability in the analysis we used genotyping of single nucleotide polymorphisms by mass spectrometry to unambiguously determine closely related strains that are difficult to distinguish by only relying on protein mass pattern detection.

Conclusions

With the method for bacterial analysis, we could identify fire blight pathogens from a variety of biological sources. The method can be used for a number of additional bacterial genera. Moreover, the mass spectrometry approach presented allows the integration of data from different biological levels such as the genome and the proteome.     

Proteomic analysis of ubiquitinated proteins from human MCF-7 breast cancer cells by immunoaffinity purification and mass spectrometry

Post-translational modification of proteins via the covalent attachment of Ubiquitin (Ub) plays an important role in the regulation of protein stability and function in eukaryotic cells. In the present study, we describe a novel method for identifying ubiquitinated proteins from a complex biological sample, such as a whole cell lysate, using a combination of immunoaffinity purification and liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. We have demonstrated the applicability of this approach by identifying 70 ubiquitinated proteins from the human MCF-7 breast cancer cell line after treatment with the proteasome inhibitor MG132. This method will aid the study of protein ubiquitination and may be used as a tool for the discovery of novel biomarkers that are associated with disease progression.     

Direct analysis of the kinetic profiles of organophosphate-acetylcholinesterase adducts by MALDI-TOF mass spectrometry

A sensitive matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry procedure has been established for the detection and quantitation of acetylcholinesterase (AChE) inhibition by organophosphate (OP) compounds. Tryptic digests of purified recombinant mouse AChE (mAChE) were fractionally inhibited by paraoxon to form diethyl phosphoryl enzyme. The tryptic peptide of mAChE that contains the active center serine residue resolves to a molecular mass of 4331.0 Da. Phosphorylation of the enzyme by paraoxon results in covalent modification of the active center serine and a corresponding increase in molecular mass of the tryptic peptide by 136 Da. The relative abundance of AChE peptides containing a modified active center serine strongly correlates with the fractional inhibition of the enzyme, achieving a detection range of phosphorylated to nonphosphorylated enzyme of 5-95%. Modifications of AChE by OP compounds resulting in dimethyl, diethyl, and diisopropyl phosphoryl adducts have been monitored with subpicomole amounts of enzyme. The individual phosphorylated adducts of AChE that result from loss of one alkyl group from the inhibited enzyme (the aging reaction) and the reappearance of unmodified AChE (spontaneous reactivation) have been resolved by the kinetic profiles and relative abundance of species. Further, the tryptic peptide containing the active center serine of AChE, isolated from mouse brain by anion-exchange and affinity chromatography, has been monitored by mass spectrometry. Native brain AChE, purified from mice treated with sublethal doses of metrifonate, has demonstrated that enzyme modifications resulting from OP exposure can be detected in a single mouse brain. For dimethyl phosphorylated AChE, OP exposure has been monitored by the ratio of tryptic peptide peaks that correspond to unmodified (uninhibited and/or reactivated), inhibited, and aged enzyme.     

Comparative expression analysis of four breast cancer subtypes versus matched normal tissue from the same patients

Gene expression studies have been widely used in an effort to identify signatures that can predict clinical progression of cancer. In this study we focused instead on identifying gene expression differences between breast tumors and adjacent normal tissue, and between different subtypes of tumor classified by clinical marker status. We have collected a set of 20 breast cancer tissues, matched with the adjacent pathologically normal tissue from the same patient. The cancer samples representing each subtype of breast cancer identified by estrogen receptor ER(+/-) and Her2(+/-) status and divided into four subgroups (ER+/Her2+, ER+/Her2-, ER-/Her2+, and ER-/Her2-) were hybridized on Affymetrix HG-133 Plus 2.0 microarrays. By comparing cancer samples with their matched normal controls we have identified 3537 overall differentially expressed genes using data analysis methods from Bioconductor. When we looked at the genes in common of the four subgroups, we found 151 regulated genes, some of them encoding known targets for breast cancer treatment. Unique genes in the four subgroups instead suggested gene regulation dependent on the ER/Her2 markers selection. In conclusion, the results indicate that microarray studies using robust analysis of matched tumor and normal samples from the same patients can be used to identify genes differentially expressed in breast cancer tumor subtypes even when small numbers of samples are considered and can further elucidate molecular features of breast cancer.     

Profiling of apoptotic changes in human breast cancer cells using SELDI-TOF mass spectrometry.

Apoptosis is a key process in the response of tumours to chemotherapeutic agents. Tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) induces apoptosis in many tumor cells, while sparing most normal cells. Several chemotherapeutic drugs synergize with TRAIL in reducing tumor growth and inducing apoptosis. Because some tumour cells respond poorly to these treatments, biomarkers that predict clinical responsiveness are needed. This study used surface-enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF MS) to identify novel apoptotic markers in TRAIL and etoposide (T+E)-treated MDA-MB-231 and ZR-75-1 breast cancer cells and MCF-10A non-transformed breast cells. T+E induced apoptosis, increasing caspase-3 activity at 4-8h, in all cell lines. Protein profiles revealed two prominent peaks, m/z 10090 and 8560, which decreased significantly during apoptosis. Mass spectrometry sequencing of tryptic peptides identified these proteins as S100A6 (confirmed immunologically) and ubiquitin (confirmed against a purified standard), respectively. Caspase inhibition prevented the decrease in both proteins during T+E-induced apoptosis whereas proteasome inhibition combined with T+E further decreased ubiquitin, possibly by preventing its recycling. Using SELDI-TOF MS we have identified S100A6 and ubiquitin as potential protein markers of apoptosis. Further validation using patient samples is required to confirm their potential utility in monitoring the effectiveness of anti-cancer drugs in inducing tumour cell apoptosis.     

Linkage analysis of quantitative traits: increased power by using selected samples.

Although a number of methods have been developed for linkage analysis of quantitative traits, power is relatively poor unless there is a single major locus of very large effect. Here it is demonstrated that the use of selected samples (i.e., ascertainment of a proband with an extreme score on the quantitative measure) can dramatically increase power, especially when proband selection is performed on the tail of a distribution with an infrequent recessive gene. Depending on gene action and allele frequency, selected samples permit detection of a major locus that accounts for as little as 10%-20% of the phenotypic variation. The judicious use of selected samples can make an appreciable difference in the feasibility of linkage studies for quantitative traits.     

Reproducibility of mass spectrometry based protein profiles for diagnosis of breast cancer across clinical studies: a systematic review

Serum protein profiling by mass spectrometry has achieved attention as a promising technology in oncoproteomics. We performed a systematic review of published reports on protein profiling as a diagnostic tool for breast cancer. The MEDLINE, EMBASE, and COCHRANE databases were searched for original studies reporting discriminatory protein peaks for breast cancer as either protein identity or as m/ z values in the period from January 1995 to October 2006. To address the important aspect of reproducibility of mass spectrometry data across different clinical studies, we compared the published lists of potential discriminatory peaks with those peaks detected in an original MALDI MS protein profiling study performed by our own research group. A total of 20 protein/peptide profiling studies were eligible for inclusion in the systematic review. Only 3 reports included information on protein identity. Although the studies revealed a considerable heterogeneity in relation to experimental design, biological variation, preanalytical conditions, methods of computational data analysis, and analytical reproducibility of profiles, we found that 45% of peaks previously reported to correlate with breast cancer were also detected in our experimental study. Furthermore, 25% of these redetected peaks also showed a significant difference between cases and controls in our study. Thus, despite known problems related to reproducibility, we were able to demonstrate overlap in peaks between clinical studies indicating some convergence toward a set of common discriminating, reproducible peaks for breast cancer. These peaks should be further characterized for identification of the protein identity and validated as biomarkers for breast cancer.     

Direct cocktail analysis of drug discovery compounds in pooled plasma samples using liquid chromatography-tandem mass spectrometry

Direct plasma injection technology coupled with a LC-MS/MS assay provides fast and straightforward method development and greatly reduces the time for the tedious sample preparation procedures. In this work, a simple and sensitive bioanalytical method based on direct plasma injection using a single column high-performance liquid chromatography (HPLC) and tandem mass spectrometry (MS/MS) was developed for direct cocktail analysis of double-pooled mouse plasma samples for the quantitative determination of small molecules. The overall goal was to improve the throughput of the rapid pharmacokinetic (PK) screening process for early drug discovery candidates. Each pooled plasma sample was diluted with working solution containing internal standard and then directly injected into a polymer-coated mixed-function column for sample clean-up, enrichment and chromatographic separation. The apparent on-column recovery of six drug candidates in mouse plasma samples was greater than 90%. The single HPLC column was linked to either an atmospheric pressure chemical ionization (APCI) or electrospray ionization (ESI) source as a part of MS/MS system. The total run cycle time using single column direct injection methods can be achieved within 4 min per sample. The analytical results obtained by the described direct injection methods were comparable with those obtained by semi-automated protein precipitation methods within +/- 15%. The advantages and challenges of using direct single column LC-MS/MS methods with two ionization sources in combination of sample pooling technique are discussed.     

Proteomic comparison of colorectal tumours and non-neoplastic mucosa from paired patient samples using iTRAQ mass spectrometry

Quantitative mass spectrometry using iTRAQ was used to identify differentially expressed proteins from 16 colorectal cancer (CRC) tumours compared to patient-paired adjacent normal mucosa. Over 1400 proteins were identified and quantitated, with 118 determined as differentially expressed by >1.3-fold, with false discovery rate < 0.05. Gene Ontology analysis indicated that proteins with increased expression levels in CRC tumours include those associated with glycolysis, calcium binding, and protease inhibition. Proteins with reduced levels in CRC tumours were associated with loss of ATP production through: (i) reduced β-oxidation of fatty acids, (ii) reduced NADH production by the tricarboxylic acid cycle and (iii) decreased oxidative phosphorylation activity. Additionally, biosyntheses of glycosaminoglycans and proteoglycans were significantly reduced in tumour samples. Validation experiments using immunoblotting and immunohistochemistry (IHC) showed strong concordance with iTRAQ data suggesting that this workflow is suitable for identifying biomarker candidates. We discuss the uses and challenges of this approach to generate biomarker leads for patient prognostication.     

Classification of some lactic acid bacteria from vacuum-packed meats by direct probe mass spectrometry

The technique of direct probe mass spectrometry (DPMS) has been applied to the classification of 40 strains of lactic acid bacteria from refrigerated vacuum-packed beef, pork, lamb and bacon. Relationships between strains were examined by multi-variate statistical techniques using sets of ions selected for reproducibility and sample discrimination. Five groups were distinguished which corresponded closely to those detected in a previous numerical taxonomic study. Two groups contained all 12 representatives of a cluster of unidentifiable non-aciduric streptobacteria whose sub-division is supported by other taxonomic evidence. All twenty-one strains from a cluster of aciduric streptobacteria provisionally identified with Lacto-bacillus sake were contained in two further groups. The sub-division of these acid-uric strains revealed by DPMS has not been verified by other techniques and requires further investigation. The fifth group contained Leuconostoc strains. The study demonstrates the value of DPMS in confirming and clarifying classification schemes obtained by conventional methods.     

Classification of some lactic acid bacteria from vacuum-packed meats by direct probe mass spectrometry

The technique of direct probe mass spectrometry (DPMS) has been applied to the classification of 40 strains of lactic acid bacteria from refrigerated vacuum-packed beef, pork, lamb and bacon. Relationships between strains were examined by multi-variate statistical techniques using sets of ions selected for reproducibility and sample discrimination. Five groups were distinguished which corresponded closely to those detected in a previous numerical taxonomic study. Two groups contained all 12 representatives of a cluster of unidentifiable non-aciduric streptobacteria whose sub-division is supported by other taxonomic evidence. All twenty-one strains from a cluster of aciduric streptobacteria provisionally identified with Lactobacillus sake were contained in two further groups. The sub-division of these aciduric strains revealed by DPMS has not been verified by other techniques and requires further investigation. The fifth group contained Leuconostoc strains. The study demonstrates the value of DPMS in confirming and clarifying classification schemes obtained by conventional methods.     

Decision tree classification of proteins identified by mass spectrometry of blood serum samples from people with and without lung cancer

A classification and regression tree (CART) model was trained to classify 41 clinical specimens as disease/nondisease based on 26 variables computed from the mass-to-charge ratio (m/z) and peak heights of proteins identified by mass spectroscopy. The CART model built on all of the specimens (no cross-validation) had an error rate of 4/41 = 10%. The CART model suggests that mass spectra peaks in the 8000-10,000, 20,000-30,000, 45,000-60, 000, and >125,000 m/z ranges may be valuable in distinguishing between the disease/nondisease specimens. The area under the receiver operating characteristics curve was 0.80 +/- 0.07 for leave-one-out cross-validation.     

Quantitative analysis of estrogens and estrogen metabolites in endogenous MCF-7 breast cancer cells by liquid chromatography-tandem mass spectrometry

To study the roles of estrogens and estrogen metabolites (EMs) in breast carcinogenesis, we reported a quantitative liquid chromatography-tandem mass spectrometry (LC-MS/MS) method utilizing selective reaction mode (SRM) to analyze estrogens and EMs in the extracellular and intracellular compartments of endogenous MCF-7 breast cancer cells through simple ethyl acetate (EA) extraction and dansyl chloride derivatization. Under a 35-min LC gradient elution on a reversed phase C18 column, the method was shown to simultaneously quantify 12 estrogens and EMs: estrone (E1) and its 2-, 4-, 16α-hydroxy derivatives (2-OHE1, 4-OHE1, 16α-OHE1), and 2-, 4-methoxy derivatives (2-MeOE1, 4-MeOE1); 17β-estradiol (E2) and its 2-, 4-hydroxy derivative (2-OHE2, 4-OHE2) and 2- and 4-methoxy derivatives (2-MeOE2 and 4-MeOE2); and estriol (E3), using ethinylestradiol (EE2) as the internal standard (IS). Using a calibration curve-standard addition hybrid method, we were able to determine the amount of estrogens and EMs in not only the treated cells but also the non-treated cells. The limits of quantification (LOQs) were determined to range from 0.05-80 pg on column with an inter-batch accuracy around 72-123% and precision around 1-10%. Results indicated that trace amounts (<0.9 fg/cell) of E1 and E2 were present in both the extra- and intra-cellular compartments under non-treated condition but DMSO could induce E1 and E2 as well as trace amounts (<2.25 fg/cell) of EMs in the cell. E2 treatment substantially increased not only E1 and E2 in the intra-cellular (60 fg/cell) and extra-cellular (3000 fg/cell) compartment but also substantially induced EMs primarily in the extracellular compartment (0.6-25 fg/cell). These data implied that EMs could be quickly generated and distributed to the extracellular compartment by E2 within 24h of treatment and DMSO solvent could potentially induce slight estrogen effects.     

Quantitative analysis of phospholipids in functionally important membrane domains from RBL-2H3 mast cells using tandem high-resolution mass spectrometry.

We recently showed that ligand-mediated cross-linking of FcepsilonRI, the high-affinity receptor for immunoglobulin E, on RBL-2H3 mast cells results in its co-isolation with detergent-resistant membranes (DRM) and its consequent tyrosine phosphorylation by the co-localized tyrosine kinase Lyn that is a critical early event in signaling by this receptor [Field et al. (1997) J. Biol. Chem. 272, 4276-4280]. As part of efforts to determine the structural bases for these interactions, we examined the phospholipid composition of DRM vesicles isolated from RBL-2H3 cells under conditions that preserve FcepsilonRI association. We used positive and negative mode electrospray Fourier transform ion cyclotron resonance mass spectrometry to compare quantitatively the phospholipid composition of isolated DRM to that of total cell lipids and to a plasma membrane preparation. From these analyses, over 90 different phospholipid species were spectrally resolved and unambiguously identified; more than two-thirds of these were determined with a precision of +/-0.5% (absolute) or less. Quantitative characterization of lipid profiles shows that isolated DRM are substantially enriched in sphingomyelin and in glycerophospholipids with a higher degree of saturation as compared to total cellular lipids. Plasma membrane vesicles isolated from RBL-2H3 cells by chemically induced blebbing exhibit a degree of phospholipid saturation that is intermediate between DRM and total cellular lipids, and significant differences in the headgroup distribution between DRM and plasma membranes vesicles are observed. DRM from cells with cross-linked FcepsilonRI exhibit a larger ratio of polyunsaturated to saturated and monounsaturated phospholipids than those from unstimulated cells. Our results support and strengthen results from previous studies suggesting that DRM have a lipid composition that promotes liquid-ordered structure. Furthermore, they demonstrate the potential of mass spectrometry for examining the role of membrane structure in receptor signaling and other cellular processes.     

Structural analysis of alpha-helical proteins from wool using cysteine labelling and mass spectrometry

A simple reduction/labelling/extraction protocol has been developed to fractionate cortical matrix proteins from filament proteins in wool. Through differential labelling of cysteine residues their relative accessibility in the wool fibre has been investigated. This has allowed the preliminary development of a map of the chemical functionality that is accessible within wool fibres under native conditions. Protein analyses of wool subjected to mechanical action, wet chemical permonosulphate/sulphite treatment and dry argon plasma treatment revealed that none of these detectably improved the accessibility of functional groups at the wool cortex. It is anticipated that this analytical method can be extended to improve the sensitivity and scope with which chemical functionality within native fibres can be mapped and lead to a better understanding of the potential limits/opportunities for fibre modification.