Quantitative Analysis of Human Cerebrospinal Fluid Proteins Using a

Oct 25, 2010 - Priscille Giron, Loıc Dayon, Natacha Turck, Christine Hoogland, and Jean-Charles Sanchez*. Biomedical Proteomics Group, Department of ...
0 downloads 0 Views 854KB Size
Download PDF
Quantitative Analysis of Human Cerebrospinal Fluid Proteins Using a Combination of Cysteine Tagging and Amine-Reactive Isobaric Labeling Priscille Giron, Loı¨c Dayon, Natacha Turck, Christine Hoogland, and Jean-Charles Sanchez* Biomedical Proteomics Group, Department of Structural Biology and Bioinformatics, University of Geneva, CH-1211 Geneva 4, Switzerland Received May 29, 2010

Highly complex and dynamic protein mixtures are hardly comprehensively resolved by direct shotgun proteomic analysis. As many proteins of biological interest are of low abundance, numerous analytical methodologies have been developed to reduce sample complexity and go deeper into proteomes. The present work describes an analytical strategy to perform cysteinyl-peptide subset enrichment and relative quantification through successive cysteine and amine-isobaric tagging. A cysteine-reactive covalent capture tag (C3T) allowed derivatization of cysteines and specific isolation on a covalent capture (CC) resin. The 6-plex amine-reactive tandem mass tags (TMT) served for relative quantification of the targeted peptides. The strategy was first evaluated on a model protein mixture with increasing concentrations to assess the specificity of the enrichment and the quantitative performances of the workflow. It was then applied to human cerebrospinal fluid (CSF) from post-mortem and ante-mortem samples. These studies confirmed the specificity of the C3T and the CC technique to cysteine-containing peptides. The model protein mixture analysis showed high precision and accuracy of the quantification with coefficients of variation and mean absolute errors of less than 10% on average. The CSF experiments demonstrated the potential of the strategy to study complex biological samples and identify differential brain-related proteins. In addition, the quantification data were highly correlated with a classical TMT experiment (i.e., without C3T cysteine-tagging and enrichment steps). Altogether, these results legitimate the use of this quantitative C3T strategy to enrich and relatively quantify cysteinecontaining peptides in complex mixtures. Keywords: Cysteine tagging • mass spectrometry • isobaric tags • covalent capture • cerebrospinal fluid

1. Introduction In classical shotgun proteomics, proteins are digested into peptides with trypsin and then analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS). Considering the heterogeneity of proteins, their post-translational modifications (PTMs), and wide dynamic range in concentration and expression, analyzing them at once is often difficult.1,2 As a result, strategies to reduce sample complexity and enrich low abundance proteins are required. For the reduction of sample complexity, several strategies have been developed, such as subcellular or multiple-round fractionation of proteins and/or peptides.3,4 Alternatively, several groups proposed chemical reagents that target specific features in proteins and peptides and permit the enrichment of a subproteome.5 Specific PTMs, functional groups, or rare amino acids were targeted this way.6-9 Cysteine is a rare and highly chemically reactive amino acid. It is therefore a target of choice for chemical tags for * To whom correspondence should be addressed. Biomedical Proteomics Group, Department of Structural Biology and Bioinformatics, Rue MichelServet 1, CH-1211 Geneva 4, Switzerland. Phone: +41 22 379 54 86. Fax: +41 22 379 55 05. E-mail: [email protected]. 10.1021/pr100535f

 2011 American Chemical Society

enrichment and/or reduction of sample complexity purposes. When considering the enrichment of the cysteinyl-peptide subset in a proteomic analysis, it was calculated from the annotated human database that such strategy reduces the number of peptides of about 90%, while still maintaining a sufficient coverage of the proteome (>95%).10 In addition, cysteines are implicated in a variety of biological processes such as cell recognition or apoptotic signaling.11 They can also undergo many PTMs of biological importance12,13 that can be therefore studied with these cysteine-tagging approaches. With the advent of mass spectrometry (MS) and the possibility to distinguish between isotopes, many of the cysteine-enrichment tags have been designed with stable isotopes to perform relative quantification through MS. Indeed, if different in mass, stable isotopes share the same physicochemical properties and thus behave similarly in the mass spectrometer. Basically, the tags are synthesized in a light and heavy form and react with two different sample populations. After combination of both populations and tryptic digestion, pairs of peptides are generated with the same ionization properties but a detectable mass difference. Relative quantification is obtained from the MS peak Journal of Proteome Research 2011, 10, 249–258 249 Published on Web 10/25/2010

research articles intensities/areas of peptide ions and as a consequence provides relative quantification of the parent protein itself.14 Among the various isotopic cysteine tags, the isotope-coded affinity tag (ICAT15) has played a pioneer role. The original tag consisted of a thiol-reactive group (iodoacetyl), an isotopically heavy or light linker, and an affinity group (biotin) for tagged-peptide enrichment with (strept)avidin. In addition to sample complexity reduction and relative quantification, the approach was later shown to be useful for the study of cysteine oxidation in proteins (i.e., redox ICAT,16 or the recent oxICAT17). Alternatively to biotin/avidin, other enrichments were used, such as immobilized metal chelate affinity chromatography (IMAC) with the polyhistidine group of the cysteine-reactive HysTag,18-21 or direct disulfide exchange on thiol resins with cysteinyl peptides.22 Instead of using isotopes in cysteine tags, some laboratories simply combined their cysteine-enrichment techniques with isobaric, metabolic, or enzymatic labeling. For example, the quantitative-peptide enrichment technology (QCET) used trypsin-catalyzed 18O labeling in combination with a thiolspecific covalent selection.23,24 Nowadays, isobaric tags are more and more used for quantitative proteomics because they perform quantification at the MS/MS level and allow multiplexing. These tags are specifically designed with 4 parts: a reactive group, a balancer group, a cleavable linker, and a reporter group. Upon MS/MS fragmentation, the tags are cleaved through the cleavable linker and release reporter-ions that appear in the low mass range of the tandem mass spectra. Several versions of the tag are synthesized, and because complementary isotopes are introduced in the balancer and the reporter groups, they all present the same mass in MS but release different reporter-ions in MS/ MS. Several versions of the tags allow multiplexed comparisons. These are clear advantages compared to the “ICAT-like” strategies that enable comparison of only 2 samples at once. In addition, the tags provide intense signals in a rather freemass region of the tandem mass spectra. The reporter-ions are therefore easily detected with no other interfering signals. Several isobaric tags can be found, and the most well-known are the iTRAQ (isobaric tag for relative and absolute quantification25), the TMT (tandem mass tag26), and the ExacTag,27 which allows comparing up to 10 samples. Several authors demonstrated the applicability of isobaric tags to relatively quantify and discover potential biomarkers from biological fluids such as plasma28 or cerebrospinal fluid (CSF).29 In addition, few groups used amine-reactive isobaric tagging in combination of an enrichment strategy to relatively quantify only peptide subsets. For instance, a biotin-hydrazide tag was used for the enrichment of carbonylated proteins followed by an iTRAQ labeling.30 We recently developed specific enrichment tags for the isolation of cysteine-containing peptides on an aldehyde resin through the covalent capture (CC) technique. These so-called cysteine-reactive covalent capture tags (C3T) react with cysteines at one end and with the CC resin at the other.31,32 To extend the method to quantitative proteomics, a quantitative C3T strategy was developed. It combines cysteine tagging with N-{2-((2-acryloyl)amino)ethyl}-1,3 thiazolidine-4-carboxamide) (ATC) and amine-reactive isobaric labeling with 6-plex TMT. It was first tested on a model protein mixture at increasing concentrations to assess the specificity, precision, and accuracy of the relative quantification. It was then applied to postmortem (PM) and ante-mortem (AM) human CSF samples. The efficiency and validity of the quantitative C3T strategy to enrich 250

Journal of Proteome Research • Vol. 10, No. 1, 2011

Giron et al. and relatively quantify cysteine-containing peptides in complex mixtures is demonstrated herein. The potential use of the strategy for redox proteomics is also highlighted.

2. Experimental Section 2.1. Materials. Bovine serum albumin (BSA), horse heart myoglobin (MYG), bovine milk β-lactoglobulin (LAC), iodoacetamide (IAA), and tris(2-carboxyethyl) phosphine hydrochloride (TCEP) 0.5 M were purchased from Sigma-Aldrich (St. Louis, MO, USA). Hen egg white lysozyme (LYS), trifluoroacetic acid (TFA, Biograde), triethyl ammonium hydrogen bicarbonate buffer (TEAB) 1 M, pH 8.5, formic acid (FA), hydrochlorid acid, and sodium dodecyl sulfate (SDS) were from Fluka Biochemika (Buchs, Switzerland). Human serum albumin (HSA) was a gift from GeneProt (Geneva, Switzerland) and sequencing grade trypsin was from Promega (Madison, WI, USA). Millex-HV filters (0.45 µm, PVDF, 4 mm nonsterile) were from Millipore (Billerica, MA, USA). Hydroxylamine solution 50 wt % in H2O was from Aldrich (Milwaukee, WI, USA). HPLC gradient grade acetonitrile (CHROMASOLV) was from Sigma and analytical grade water (LICHROSOLV) from Merck (Darmstadt, Germany). The 6-plex tandem mass tags were provided by Proteome Sciences (Frankfurt am Main, Germany). ATC (N-{2-((2-acryloyl)amino)ethyl}-1,3 thiazolidine-4-carboxamide) was chemically synthesized as previously described.31 Aliphatic aldehydemodified amino-PEGA resin (starting resin 0.4 mmol/g dry weight) was prepared as previously described.33 The starting amino-PEGA resin was from Varian Inc. (Shropshire, UK), and an aldehyde substitution of 0.035 mmol/g of wet resin was calculated. Oasis HLB sample extraction cartridges (1 cm3/30 mg) were from Waters (Mildford, MA, USA). C18 Micro SpinColumns were from Harvard Apparatus (Holliston, MA, USA). 2.2. CSF Preparation. The 3 PM and 3 AM CSF samples were collected as previously described.29 Each patient or patient’s relatives gave informed consent prior to enrollment. The local institutional ethical committee board approved the clinical protocol. Each CSF sample was first filtrated on Millex-HV filters to remove all membrane and cellular debris. To estimate the protein content of the 6 samples, a Bradford assay (Protein assay, Bio-Rad, Hercules, CA, USA) was performed. From the protein concentrations obtained, a volume corresponding to 35 µg of proteins was taken from each sample and spiked with 700 ng of LAC as internal standard. A second series of samples was also prepared for a control experiment (see part 2.4). All samples were evaporated under speed-vacuum and kept at -80 °C until analysis. 2.3. General Strategy. The general strategy is depicted in Figure 1. It consists of protein disulfide-bond reduction, cysteine tagging with ATC, tryptic digestion, TMT labeling, pooling of labeled samples, isolation of the tagged peptides through the CC procedure, and LC-MS/MS analysis. 2.4. Cysteine Tagging (ATC). Model Protein Mixture. An amount of 250 µg of MYG, LYS, LAC, and HSA was dissolved in 1 mL of alkylation buffer (TEAB, 0.1 M, pH 8/TCEP, 1 mM/ SDS, 0.01%). From this sample, two series of 6 aliquots were prepared to have a total protein amount of 10, 20, 30, 30, 50, and 100 µg. One series followed the general workflow (Figure 1). The second one did not undergo the CC. Each tube was completed with alkylation buffer to 100 µL final volume and incubated for 1 h at 60 °C. Meanwhile, a 1 M stock solution of ATC in alkylation buffer was prepared, and the pH was carefully adjusted to 7.5 with diluted NaOH (1 N). ATC (10 µL) was added

Analysis of Human Cerebrospinal Fluid Protein

research articles

Figure 1. Quantitative C3T strategy.

to each tube and reacted for 2 h at 37 °C. Each protein sample was then digested overnight at 37 °C by adding 10 µL of freshly prepared trypsin at 0.2 µg/µL in TEAB 0.1 M, pH 8. CSF Samples. Two series of samples were prepared; one was used for a “control” experiment using the classical TMT workflow,29 and the other followed the quantitative C3T strategy (Figure 1). The 12 evaporated CSF samples were dissolved in 100 µL of alkylation buffer to be reduced for 1 h at 60 °C. The series CSF samples for the quantitative C3T strategy was reacted 2 h at 37 °C with 10 µL of ATC 1 M, pH 7.5 (in alkylation buffer). In parallel, the series for the classical TMT quantitation was reacted for 30 min, in the dark, at room temperature (rt), with 10 µL of IAA (400 mM in water). Both series were then digested overnight at 37 °C by adding 10 µL of freshly prepared trypsin at 0.2 µg/µL in TEAB 0.1 M, pH 8. 2.5. Isobaric TMT Labeling. TMT labeling was performed as previously described.29 A volume of 40.3 µL of TMT solution (i.e., 0.83 mg in CH3CN) was added to each tube. After 60 min of reaction at rt, 8 µL of hydroxylamine 5% (w:v) was added to stop the TMT labeling and reverse potential undesired labeling of alcohol moieties (15 min under agitation). For the protein model mixture experiments, each aliquot (10, 20, 30, 30, 50, and 100 µg) of the two series was differentially labeled with TMT with respective reporter-ion m/z from 126.1 to 131.1 Thomsons (Th). For the CSF experiments, the PM samples were labeled with TMT with reporter-ions at m/z ) 126.1, 128.1, and 130.1 Th, and the AM samples were labeled with TMT with reporter-ions at m/z ) 127.1, 129.1, and 131.1 Th. After labeling, the six samples of each series were combined in a new tube and evaporated under vacuum (i.e., 4 new tubes). The samples tagged with ATC needed an additional step of tag deprotection. Therefore, samples tagged with ATC were dissolved in 200 µL of O-methyl hydroxylamine 500 mM, TCEP 10 mM, pH 3.5, and reacted for 2 h at 37 °C. All samples were desalted on Oasis HLB extraction cartridges following the manufacturer’s protocol. They were dried under vacuum. The two tubes that did not follow the CC isolation step were then desalted with C18 Micro SpinColumns and dried under vacuum for following LC-MS/MS analysis, while the two others were submitted to the CC procedure (one sample of the model protein mixture experiment and one of the CSF experiment). 2.6. Covalent Capture (CC). CC was performed as previously described.31 Dried samples were resuspended in 400 µL of capture buffer, consisting of 100 mM of acetate buffer and 10 mM TCEP in H2O/CH3CN (50:50) at pH 6-6.5. Each solution was then mixed with 50 mg of aldehyde resin (i.e., large excess per theoretical cysteinyl peptide) and incubated overnight at rt under gentle agitation. After filtering and washing-off the

resin several times to remove noncaptured peptides,31 the captured peptides were eluted from the resin using O-methyl hydroxylamine 200 mM in H2O/CH3CN (50:50) for 4 h. After elution, samples were desalted with C18 Micro SpinColumns and dried under vacuum for following LC-MS/MS analysis. 2.7. LC-MS/MS. The 4 desalted and dried samples were dissolved in H2O/CH3CN/FA 94.9%/5%/0.1% to be subjected to LC-MS/MS. LC-MS/MS was performed on an linear trap quadrupole Orbitrap (LTQ-OT) XL from Thermo (San Jose, CA, USA) coupled with a NanoAcquity LC system from Waters as recently described.34 2.8. Data Processing. Protein Identification. Peak lists (.mgf) were generated from the raw data using an in-house written Perl script. In addition, the collision-induced dissociation and higher-energy C-trap dissociation spectra were merged using a custom-made program as recently described.34 These merged peak lists were submitted to the Phenyx software (Genebio, Geneva, Switzerland), operating on a local server. The protein identification was performed using the following parameters: Protein Model Mixture. Database: SwissProt/UniProt (UniProtKB/Swiss-Prot release 57.11, 24 November 2009). Taxonomy: NO_TAXONOMY. Accession Number (AC): P02768, P02754, P68082, P00698, and P00761 for, respectively, HSA, LAC, MYG, LYS, and trypsin. Modifications: TMT labeling of Nterminal and ε-amino group of lysines was set as fixed (+229.1629 Da). ATC tagging (+217.0885 Da) of cysteines was set as variable as well as oxidation of methionine and the replacement of a glycine by an aspartic acid corresponding to a genetic variant of LAC. Enzyme: trypsin, with one missedcleavage, in normal cleavage mode. Scores: The protein and peptide scores were 5.5. Parent tolerance: 5 ppm. Peptide thresholds: length g6, p value e0.001. Only one search round was used with selection of “turbo” scoring. The search was also performed in the reverse database using the same parameters, to assess the peptide false discovery rate (FDR). The peptides and proteins scores were chosen to have a FDR e 1%. CSF Samples. Database: SwissProt/UniProt (UniProtKB/ Swiss-Prot release 57.11, 24 November 2009). Taxonomy: Homo sapiens. Modifications: TMT labeling of N-terminal and ε-amino group of lysines was set as fixed. Oxidation of methionine was set as variable. For the sample that followed the classical TMT workflow, cysteine modification was “carbamidomethylation” while for the sample that followed the quantitative C3T strategy it was ATC tagging. In each case the cysteine modification was set as variable. Enzyme: trypsin, with one missedcleavage, in normal cleavage mode. Scores: The protein and peptide scores were 4.8 and 5.0 for respectively the classical TMT and the quantitative C3T workflow. Parent tolerance: 10 Journal of Proteome Research • Vol. 10, No. 1, 2011 251

research articles ppm. Peptide thresholds: length g6, p value e0.001. Only one search round was used with selection of “turbo” scoring. The search was also performed on the reverse database using the same parameters, to assess the FDR. The peptide and protein scores were chosen to have a FDR e 2%. An additional search was performed to identify the spiked LAC standard using the same parameters but with no taxonomy specified and P00698 as AC. For the classical TMT sample, proteins matching 2 different peptide sequences were selected while for the quantitative C3T sample all peptide hits were considered. Protein Quantification. Protein quantification was mainly based on a previously described procedure29 with slight modifications. The reporter-ion intensities were directly obtained from the dedicated Phenyx export tool. After an isotopic correction of the reporter intensities, peptides with one or more missing reporter-ion values were removed from the quantification. When isotopic correction generated negative values, the corresponding peptides were also removed. For the CSF experiment, the spiked LAC standard was used to correct experimental bias. A normalization factor was determined for each reporter according to the theoretical relative abundance (RA) of the reporters for LAC divided by the experimental RA for LAC. The RA of a reporter was calculated as the reporter intensity divided by the sum of all reporter intensities. The “Out?Lier” freeware35 was used to remove outlier peptides. This freeware applies 4 different statistical outlier tests: the Grubbs, the Dixon, the Inter-Quartile Range (IQR), and the Gauss g-test (see Web site for precisions). For each protein, the RAs of each reporter channel were tested. When one RA was considered as an outlier, the corresponding peptide was removed. For each protein, the average of RAs of each reporter was calculated as well as the standard deviation (SD). For the model protein mixture, the averages and the SDs of all proteins were also calculated. From these values, ratios with respect to reporter at m/z ) 126.1 Th were calculated as the mean RA multiplied by the sum of the theoretical ratios 1:2: 3:3:5:10 (i.e., 24). Calibration curves between experimental and theoretical ratios were constructed. A linear regression was assessed. For the CSF samples, to assess the basal variation within the same population (respectively PM and AM), an intermediate RA was calculated for each population with respective reporters at m/z ) 126.1, 128.1, and 130.1 Th and 127.1, 129.1, and 131.1 Th. Intermediate RAs were calculated for each reporter as the intensity of the reporter divided by the sum of the intensities of the 3 reporter-ions of the same population. The averages and the SDs were calculated. Additionally, individual ratios were calculated between each couple of patients within the same population (i.e., 126.1/128.1, 126.1/130.1, 128.1/131.1 and 127.1/129.1, 127.1/131.1, 129.1/131.1 for PM and AM respectively). From the average of these ratios a global ratio of all proteins was determined. The RAs were finally calculated for each sample as described above, the outliers were removed, and the averages and the SDs were determined. The final protein ratios PM/AM corresponded to the sum of averaged PM RAs divided by the sum of averaged AM RAs. In addition, from the RAs of each peptide, an individual peptide ratio was calculated. A correlation test was performed using the Prism software (GraphPad Software Inc., La Jolla, CA, USA): one for each condition (i.e., classic TMT quantification and TMT quantification combined with the C3T), to assess the correlation between experiments. As the protein 252

Journal of Proteome Research • Vol. 10, No. 1, 2011

Giron et al. RAs did not follow a Gaussian distribution, the nonparametric Spearman with a two tailed p-value was chosen as the correlation test. The test was performed on the log-transformed ratios. Finally, the results of the LC-MS/MS duplicate analyses of each condition were summed to calculate the mean RAs of all PM and AM samples (i.e., 6 measurements for each population). The RA of each protein came therefore from 6 measurements in each population, even for proteins quantified with a single peptide. In order to determine if the RAs of AM and PM populations were significantly different, a nonparametric Mann-Whitney test was performed using the Prism software. A p-value inferior to 0.05 was considered to traduce a significant change. The correlation between the PM/AM protein and peptide ratios found in each strategy was calculated using the Spearman test. 2.9. Immunoblot Verification. A 20 µg amount of each AM and PM CSF sample was separated with 1-dimensional (1-D) SDS-polyacrylamide gel electrophoresis (GE) on a 12% acrylamide gel. Proteins separated were further electroblotted onto a nitrocellulose membrane as described by Towbin et al.36 Membranes were stained with red ponceau and destained with water. Membranes were then incubated 1 h in 5% of milk with 0.05% of PBS-Tween for blocking. Immunodetection was performed using an anti-human peroxiredoxin 5 rabbit polyclonal antibody (Abcam, Cambridge, UK) diluted 1/2000 in 1% of milk with 0.05% of PBS-Tween. After several washing steps, the membrane was incubated with a goat anti-rabbit HRP polyclonal secondary antibody (Dako, Denmark A/S) for 1 h at 1/2000 dilution. ECL plus Western Blotting detection system kit (GE Healthcare, Uppsala, Sweden) was used for detection. The membrane was finally scanned with a Typhoon 9400 (GE Healthcare).

3. Results 3.1. Model Protein Mixture. The quantitative C3T strategy is depicted in Figure 1. Six model protein samples with increasing concentrations were tagged with ATC, digested with trypsin, labeled with TMT, pooled, enriched through CC, and analyzed with LC-MS/MS. Two experiments were conducted: one without and one with CC enrichment. The quantification reliability of peptides enriched or not on the CC resin was evaluated. The relative quantification results of HSA, LYS, LAC, and MYG proteins are summarized in table 1. The corresponding calibration curves are given in Figure 2 (see Supporting Information, SD1, for detailed results). Without CC, 154 peptides were identified with 62 unique peptides for the 5 selected proteins (i.e., HSA, LAC, LYS, MYG, and trypsin). From these unique peptides, 25% contained a cysteine. Incomplete tagging of cysteine residues was observed. About 90% of the identified peptides were quantified. The relative quantification gave ratios of 1.02:2.01:2.75:2.93:5.01: 10.26 with less than 10% of coefficient of variation (CV). The mean absolute error was less than 3% with respect to the theoretical ratios of 1:2:3:3:5:10. The calibration curve showed a high linearity between the experimental ratios found and the theoretical ones, with a coefficient of determination of 0.9987 (Figure 2). In addition, the quantification of trypsin gave ratios of 0.96:1.01:0.92:1.07:1.04:1.01 with CV less than 14% (data not shown). These ratios were consistent with the theoretical ratios expected of 1:1:1:1:1:1, since the same amount of trypsin was added in each tube for digestion. The mean absolute error was of about 4%.

research articles

Analysis of Human Cerebrospinal Fluid Protein

Table 1. Relative Quantification Obtained for a Model Protein Mixture (HSA, LAC, LYS, MYG) Subjected to the Quantitative C3T Strategy without or with the Covalent Capture Enrichment Stepa ID

relative abundance for TMT 126

relative abundance for TMT 127

relative abundance for TMT 128

relative abundance for TMT 129

relative abundance for TMT 130

relative abundance for TMT 131

ALBU_HUMAN LACBBOVIN LYSC_CHICK MYG_HORSE mean corresponding ratio

0.047 ( 0.014 0.042 ( 0.007 0.042 ( 0.006 0.037 ( 0.005 0.042 ( 0.004 1.02 ( 0.100

0.088 ( 0.013 0.080 ( 0.007 0.085 ( 0.007 0.081 ( 0.009 0.084 ( 0.004 2.01 ( 0.092

Without CC 0.119 ( 0.010 0.115 ( 0.008 0.113 ( 0.007 0.112 ( 0.008 0.115 ( 0.003 2.75 ( 0.066

0.127 ( 0.015 0.122 ( 0.010 0.128 ( 0.009 0.112 ( 0.005 0.122 ( 0.007 2.93 ( 0.175

0.206 ( 0.014 0.209 ( 0.014 0.205 ( 0.008 0.216 ( 0.009 0.209 ( 0.005 5.01 ( 0.121

0.413 ( 0.041 0.429 ( 0.018 0.426 ( 0.009 0.442 ( 0.019 0.428 ( 0.012 10.26 ( 0.280

ALBU_HUMAN LACB_BOVIN LYSC_CHICK MYG_HORSEb mean corresponding ratio

0.034 ( 0.005 0.032 ( 0.001 0.046 ( 0.020 0.088 0.037 ( 0.008 0.89 ( 0.181

0.068 ( 0.010 0.059 ( 0.0004 0.067 ( 0.001 0.061 0.065 ( 0.005 1.56 ( 0.115

With CC 0.114 ( 0.008 0.118 ( 0.005 0.115 ( 0.008 0.102 0.116 ( 0.002 2.78 ( 0.051

0.130 ( 0.003 0.116 ( 0.007 0.105 ( 0.003 0.130 0.117 ( 0.013 2.81 ( 0.304

0.215 ( 0.006 0.243 ( 0.033 0.243 ( 0.018 0.199 0.234 ( 0.016 5.61 ( 0.392

0.439 ( 0.015 0.432 ( 0.037 0.424 ( 0.028 0.421 0.432 ( 0.007 10.36 ( 0.171

a The theoretical ratios are 1:2:3:3:5:10 for TMT reporter ions ranging from 126.1 to 131.1, respectively. b With the CC, MYG was not considered to calculate the mean of the RAs since the protein does not contain cysteines in its sequence and was identified with a single peptide.

Figure 2. Calibration curves for the model protein mixture with 6 increasing concentrations, subjected to the quantitative C3T strategy: without (0) or with (b) the CC enrichment.

With hyphenation of the CC, 26 peptides were identified, with 18 unique peptides for the 5 selected proteins. From these peptides, 50% were quantified. Only one peptide did not contain a cysteine, matching a MYG peptidic sequence. The others were all tagged with ATC. The experimental ratios obtained were 0.89:1.56:2.78:2.81:5.61:10.36 with CVs less than 21% for the less concentrated sample and a mean absolute error inferior to 11%. The calibration curve showed a high linearity, with a coefficient of determination of 0.9931 (Figure 2). MYG was not considered for the calculation, since it does not contain any cysteine. 3.2. CSF Samples. Three different PM and AM CSF samples were subjected to the classical TMT workflow29 or to the quantitative C3T strategy. In the classical TMT workflow, after a reducing step, the 6 CSF samples were alkylated with IAA, digested with trypsin, differentially labeled with TMTs, pooled, and analyzed with LC-MS/MS. For the quantitative C3T workflow, the 6 CSF samples were reduced, tagged with ATC, digested with trypsin, differentially TMT-labeled, pooled, enriched with the CC, and analyzed with LC-MS/MS (Figure 1). In both experimental designs, the 3 PM samples were labeled

Figure 3. Decomplexification and enrichment power of the C3T strategy. Number of unique peptides and proportion of cysteinecontaining peptides identified in CSF using the classical TMT (A) and the quantitative C3T strategies (B). With the C3T strategy, the number of peptide to analyze was dramatically reduced of about 80% and the enrichment of cysteinyl tagged-peptides was 100% specific.

with TMTs with reporter-ions at m/z ) 126.1, 128.1, and 130.1 Th, while the 3 AM samples were labeled with TMTs with reporter-ions at m/z ) 127.1, 129.1, and 131.1 Th. Each sample was analyzed twice with LC-MS/MS using an LTQ-OT mass spectrometer. The complete lists of identified and quantified peptides and proteins, as well as tables summarizing the results of each experiment are provided in supplemental data 2 (SD2). Protein Identification in CSF. Phenyx searches provided 789 and 769 peptide hits with the classical TMT procedure from each LC-MS/MS duplicate, corresponding respectively to 430 and 412 unique peptides. These peptides were matched with a total of 86 and 82 proteins, which were all quantified. When merging results from both duplicates, 92 proteins were identified with a total of 465 unique peptides from which 30% contained at least one cysteine (Figure 3). Phenyx searches on the quantitative C3T experiment generated 100 and 120 peptide hits for each LC-MS/MS duplicate with respectively 84 and 96 unique peptides. These peptides were matched with 48 and 52 proteins from which 45 and 48 were quantified respectively. A total of 57 proteins with 107 unique peptides were identified from both duplicates. All identified tryptic peptides contained at least one cysteine (Figure 3). Except one, they were all tagged with ATC. Forty-six proteins and 67 peptides were common to Journal of Proteome Research • Vol. 10, No. 1, 2011 253

research articles

Giron et al.

Figure 4. Spearman correlations between the LC-MS/MS duplicates for the classical TMT quantification (A) and the quantitative C3T strategy (B). Each dot is a protein ratio represented with a logarithmic scale. The correlation coefficient (r) and the associated p value are also provided.

Figure 5. Spearman correlation between the two quantification strategies, namely the classical TMT (y axis) and the quantitative C3T (x axis) with in A the protein ratios and in B the peptide ratios. Each dot is a protein (A) or a peptide (B) ratio represented with a logarithmic scale. The correlation coefficient (r) and the associated p value are also provided. Dash lines (----) show 2-fold change boundaries; red dots are protein and peptide ratios outside the arbitrary boundaries.

the two experiments (classical TMT and quantitative C3T). These numbers represented about 50% of the proteins and 14% of the peptides from the classical TMT experiment while in the quantitative C3T, it represented about 80% of the proteins and 60% of the peptides. Application to CSF. To assess the basal variation within the same population (i.e., PM and AM), an initial comparison was done between the reporter-ion values of each population (i.e., at m/z ) 126.1, 128.1, 130.1 and 127.1, 129.1, 131.1 Th, respectively; see Experimental Section; data not shown). The RA of the PM population of both strategies provided an overall mean CV of 25% and a global protein ratio of 0.95. For the AM population, an overall mean CV of 18% was calculated and a global protein ratio of 1.1. The RAs representing the tryptic peptide abundances in the 6 compared samples were then calculated (see Supporting Information, SD2). From these values, the overall mean CV was less than 30%, for the PM population in each duplicate and experimental design, while for the AM population it was about 20%. For each LC-MS/MS analysis, the ratio PM/AM of each protein was calculated and for each duplicate the protein ratios were plotted in a logarithmic scale to perform a nonparametric Spearman test. Figure 4 summarizes these results. The correlation coefficients (r) were 0.9138 and 0.9323 with significant p-values of less than 0.0001 for the classical TMT and the quantitative C3T experiments, respectively. After assessment of 254

Journal of Proteome Research • Vol. 10, No. 1, 2011

the correlation between the LC-MS/MS duplicates, the results of the duplicates were summed to increase the number of quantified peptides per protein. The correlation between the ratios obtained with both experimental designs was also assessed and appears in Figure 5. The correlation coefficient was 0.7623 with a significant p-value of less than 0.0001 between the 46 proteins commonly quantified with both strategies. In the same manner, the 50 unique peptides that served for the quantification, showed a correlation coefficient of 0.8772 with a significant p-value (