Proteomic Methodological Recommendations for ... - ACS Publications

Proteomic Methodological Recommendations for Studies Involving Human Plasma, Platelets, and Peripheral Blood Mononuclear Cells Baukje de Roos,*,† Susan J. Duthie,† Abigael C. J. Polley,‡ Francis Mulholland,‡ Freek G. Bouwman,§ Carolin Heim,| Garry J. Rucklidge,† Ian T. Johnson,‡ Edwin C. Mariman,§ Hannelore Daniel,| and Ruan M. Elliott‡ Rowett Research Institute, Aberdeen, United Kingdom, Institute of Food Research, Norwich, United Kingdom, University of Maastricht, Maastricht, The Netherlands, and Technical University of Munich, Freising-Weihenstephan, Germany Received November 7, 2007

This study was designed to develop, optimize and validate protocols for blood processing prior to proteomic analysis of plasma, platelets and peripheral blood mononuclear cells (PBMC) and to determine analytical variation of a single sample of depleted plasma, platelet and PBMC proteins within and between four laboratories each using their own standard operating protocols for 2D gel electrophoresis. Plasma depleted either using the Beckman Coulter IgY-12 proteome partitioning kit or the Amersham albumin and IgG depletion columns gave good quality gels, but reproducibility appeared better with the single-use immuno-affinity column. The use of the Millipore Filter Device for protein concentration gave a 16% (p < 0.005) higher recovery of protein in flow-through sample compared with acetone precipitation. The use of OptiPrep gave the lowest level of platelet contamination (1:0.8) during the isolation of PBMC from blood. Several proteins (among which are R-tropomyosin, fibrinogen and coagulation factor XIII A) were identified that may be used as biomarkers of platelet contamination in future studies. When identifying preselected spots, at least three out of the four centers found similar identities for 10 out of the 10 plasma proteins, 8 out of the 10 platelet proteins and 8 out of the 10 PBMC proteins. The discrepancy in spot identifications has been described before and may be explained by the mis-selection of spots due to laboratory-to-laboratory variation in gel formats, low scores on the peptide analysis leading to no or only tentative identifications, or incomplete resolution of different proteins in what appears as a single abundant spot. The average within-laboratory coefficient of variation (CV) for each of the matched spots after automatic matching using either PDQuest or ProteomWeaver software ranged between 18 and 69% for depleted plasma proteins, between 21 and 55% for platelet proteins, and between 22 and 38% for PBMC proteins. Subsequent manual matching improved the CV with on average between 1 and 16%. The average between laboratory CV for each of the matched spots after automatic matching ranged between 4 and 54% for depleted plasma proteins, between 5 and 60% for platelet proteins, and between 18 and 70% for PBMC proteins. This variation must be considered when designing sufficiently powered studies that use proteomics tools for biomarker discovery. The use of tricine in the running buffer for the second dimension appears to enhance the resolution of proteins especially in the high molecular weight range. Keywords: plasma proteomics • platelet proteomics • PBMC proteomics • human nutrition intervention studies • technical variability

1. Introduction The potential value of proteomics for clinical and nutritional sciences has been recognized for some years.1–6 Especially in the area of nutritional sciences, a large number of reviews have * Corresponding author: Dr. Baukje de Roos, Rowett Research Institute, Division of Vascular Health, Greenburn Road, Bucksburn, Aberdeen AB21 9SB, United Kingdom; Tel, +44 (0)1224 716636; fax, +44 (0)1224 716629; e-mail, [email protected]. † Rowett Research Institute. ‡ Institute of Food Research. § University of Maastricht. | Technical University of Munich.

2280 Journal of Proteome Research 2008, 7, 2280–2290 Published on Web 05/20/2008

been published on proteomics, methods and concepts for research,1,7–11 but relatively few studies have actually described the use of proteome analysis as a tool, and most of these have involved the use of rodent models,12–15 or of human cells in culture.16–22 However, an advantage of the proteomics platform is the quantitative analysis of accessible human body fluids. Plasma, serum, or circulating cells, such as platelets and peripheral blood mononuclear cells (PBMC), can be obtained easily from donated blood.7 Proteomics technology has the potential to deliver specific and relevant biomarkers for health and reveal early indications for disease disposition. For nutri10.1021/pr700714x CCC: $40.75

 2008 American Chemical Society

Methodological Recommendations and Analytical Variability tion research, proteomics can assist in distinguishing and identifying dietary responders from nonresponders, and discover novel and beneficial food components.9 The plasma, platelet and PBMC proteomes are subject to rapid changes in response to external signals, for example, diurnal rhythm, postprandial metabolism and inflammatory state, but they are also hugely influenced by methods of blood sampling and sample preparation, giving rise to potentially large within- and between subject background variation. Another important source of uncertainty is the technical variation inherent in the use of 2D gel electrophoresis technology, which so far has not been assessed systematically in the context of intervention studies. These sources of variation may easily obscure the biological changes under investigation, and this has led some to argue that proteomic profiling of blood is a relatively immature technology in need of further refinement.23 This study was designed to determine analytical variation in 2D gel electrophoresis within and between laboratories, and to facilitate development, optimization and validation of protocols for blood processing for proteomic analysis of depleted plasma, platelets and PBMC. Such information is critical for the design of intervention studies that are powered to detect subtle changes in plasma, platelet and/or PBMC proteomes.

research articles

quently analyzed by ProteomWeaver software (BioRad) using the automatic spot detection procedure.

2. Materials and Methods

2.2.2. Comparison of Methods To Concentrate Protein Homogenates: Acetone Precipitation versus Centrifugal Concentration Device. The blood sample that was obtained in EDTA as described above was centrifuged at 1500g for 15 min at 4 °C to obtain plasma. This plasma sample was depleted using the IgY-12 high capacity proteome partitioning kit (ProteomeLab, Beckman Coulter), and five samples of 500 µL of flow-through were obtained, containing 0.35 µg/µL protein. The five aliquots were each mixed with 2 mL of ice-cold acetone. After vortexing, samples were incubated for 120 min at -20 °C. Samples were centrifuged for 10 min at 10 000g at 0 °C. The supernatant was removed and 50 µL of TBS dilution buffer (10 mM Tris-HCl, pH7.4, 150 mM NaCl) was added and vortexed thoroughly to dissolve the protein pellet. Protein concentration was measured using the BCA assay (Pierce). To concentrate protein using the centrifugal filter device, five samples of 500 µL of flow-through sample obtained from the IgY-12 high capacity proteome partitioning kit (ProteomeLab, Beckman Coulter), containing 0.35 µg/µL protein, were concentrated to ∼25 µL per sample using the Millipore Ultrafree 0.5 Centrifugal Filter Device (NMWL 5 KDa) as per manufacturer’s instructions. Proteins were measured in the original samples and in the concentrated samples using the BCA assay (Pierce).

2.1. Blood Sampling for the Isolation of Plasma, Platelets and PBMC. A single blood sample was taken from one healthy volunteer that had been fasting overnight, and after 10-15 min of rest. Tourniquet pressure was released before blood was withdrawn. Blood was collected into vacutainers containing potassium EDTA anticoagulant for the isolation of plasma and PBMC. These samples were kept on ice until the isolation of plasma. Blood was collected into a closed monovette system containing trisodium citrate anticoagulant for the isolation of platelets. The monovette system tubes were kept at 37 °C for a maximum of 30 min until the isolation of platelets. 2.2. Sample Preparation for Depleted Plasma Proteomics. 2.2.1. Comparison of Plasma Protein Depletion Kits: Amersham Albumin and IgG Removal Kit versus ProteomeLab/ Beckman IgY-12 High Capacity Proteome Partitioning Kit. The blood sample that was obtained in EDTA as described above was centrifuged at 1500g for 15 min at 4 °C to obtain plasma. This plasma sample was depleted of (i) the 12 most abundant proteins (albumin, IgG, fibrinogen, transferrin, IgA, IgM, HDL apoAI, HDL apoAII, haptoglobin, R1-antitrypsin, R1acid glycoprotein and R2-macroglobulin) using the IgY-12 high capacity proteome partitioning kit (ProteomeLab, Beckman Coulter) or (ii) albumin and IgG using the Amersham Albumin and IgG Removal Kit (GE Healthcare) as per manufacturer’s protocol (n ) 5 per depletion method). Flow-through samples were concentrated using Millipore Biomax 5K Dalton nominal molecular weight limit (NMWL) Membrane 0.5 mL Ultrafree Filters as per manufacturer’s instructions at 10 °C to a volume of ∼200 µL. A total of 400 µL of a modified rehydration buffer (7 M Urea, 2 M thiourea, 2% CHAPS, 0.5% IPG buffer, pH 4-7) was added to the sample in the column and the concentration/ washing steps were repeated twice more. Protein concentration was assessed using the Ettan sample preparation 2-D Quant Kit (GE Healthcare). A volume equivalent to 225 µg of depleted plasma protein was loaded per 2D gel and 5 2D gels were run per depletion method. Gels were stained with Flamingo (BioRad) according to the manufacturer’s instruction and subse-

2.2.3. Sample Preparation for Depleted Plasma Proteomics. The blood sample that was obtained in EDTA as described above was centrifuged at 1500g for 15 min at 4 °C to obtain plasma. This plasma sample was depleted by running 20 µL aliquots of plasma with a protein concentration of 65 µg/µL through an IgY-12 high capacity proteome partitioning kit according to the manufacturer’s instructions (ProteomeLab). The flow-through (500 µL per aliquot) was pooled and protein concentration measured using the BCA method. Forty-six runs of the IgY-12 high capacity proteome partitioning kit yielded 23 mL of flow-through containing 0.26 µg/µL protein. Therefore, 90% of protein was removed by the proteome partitioning kit. In four laboratories, 4.8 mL of flow-through was concentrated using the Millipore Ultrafree 0.5 Centrifugal Filter Device as per manufacturer’s instructions. Samples were concentrated to a volume of ∼25 µL, and concentrated, depleted plasma was removed. Modified rehydration buffer (210 µL), containing 7 M Urea, 2 M thiourea, 2% CHAPS, and 0.5% IPG buffer, pH4-7, was added to the Millipore Ultrafree-0.5 Centrifugal spin columns, aspirated up and down twice, and transferred to a tube containing the concentrated, depleted plasma. Another 210 µL of rehydration buffer was added to the Millipore Ultrafree-0.5 centrifugal spin columns, removed, and added onto the other fractions. For the 24 cm gels, 20 µL of 30% DTT was added to the tube containing extraction buffer and concentrated, depleted plasma. The sample was mixed, spun and loaded. Assuming a protein recovery of 76% using the Millipore Ultrafree-0.5 Centrifugal spin columns (Table 1), each laboratory loaded 4.8 mL of flow-through × 0.26 mg protein/ mL ) 1.25 mg protein × 0.76 ) 950 µg onto 5 gels. Therefore, approximately 190 µg of depleted plasma protein was loaded per 2D gel for 5 replicate gels in each of the four laboratories. 2.3. Sample Preparation for Platelet Proteomics. The blood sample that was obtained in eight monovette system tubes (with trisodium citrate as the anticoagulant), as described above, was centrifuged at 200g for 10 min at room temperature. Only the top one-third of platelet rich plasma (PRP) was Journal of Proteome Research • Vol. 7, No. 6, 2008 2281

research articles

de Roos et al.

Table 1. Mean Recovery of Flow-Through Proteins by Acetone Precipitation or Using a Millipore Ultrafree 0.5 Centrifugal Filter Device (cutoff 5000 Da) total amount of protein (µg)a method

Acetone precipitation (n ) 5) Centrifugal Filter Device (n ) 5)

flow-through sample

concentrated sample

recovery (%)

177

105 ( 11

60 ( 6

177

134 ( 5

76 ( 3

a Protein concentration was measured using the BCA protein assay. Values represent mean of five samples ( SD.

removed to prevent contamination with red and white blood cells. The PRP was placed in a warmed tube, adding ice-cold prostaglandin I2 (PGI2) dissolved in calcium-free Tyrodes buffer containing 140 mM NaCl, 3 mM KCl, 12 mM NaHCO3, 0.4 mM NaH2P04, 2 mM MgCl2, 0.35% (w/v) bovine serum albumin (BSA), 0.1% (w/v) glucose, pH 7.33, with Hepes (final concentration 50 ng/mL). For the isolation of platelets from the PRP, approximately 0.5 mL of 50% BSA (fatty acid free, pH 7.4) was added to fresh tubes and warmed to 37 °C with approximately 4 mL of PRP layered onto each albumin cushion. The tubes were centrifuged at 1600g for 15 min at room temperature. The platelet poor plasma (PPP) from each tube was removed and the platelet layer carefully resuspended in warmed Tyrodes buffer containing PGI2 (50 ng/mL). The platelet suspension was layered onto another 0.5 mL 50% BSA cushion in a fresh tube, warmed to 37 °C and centrifuged at 1600g for 15 min at room temperature. The supernatant was removed and the platelet layer resuspended in 1 mL of warmed Tyrodes buffer (containing PGI2 at 50 ng/mL). This sample was centrifuged at 10 000g for 5 min at 4 °C, the supernatant was removed, and the pellet was washed with cold Tyrodes buffer. The samples were then centrifuged at 10 000g for 5 min at 4 °C. The Tyrodes buffer was removed and platelet proteins were extracted by adding 100 µL of extraction buffer containing 7 M Urea, 2 M Thiourea, 2% CHAPS, and 0.06% proteinase inhibitor cocktail (Roche) to the platelet pellets. From 80 mL of blood, we extracted 5.0 × 109 platelets, yielding 6.15 mg of protein as assessed by the RC/DC assay (BioRad). A total of 200 µg of platelet proteins was loaded per 2D gel with 5 replicate gels at each of the four laboratories. 2.4. Sample Preparation for PBMC Proteomics. 2.4.1. Comparison of Methods for PBMC Isolation and Determination of Optimum Cell Number for Protein Extraction. Three methods for the isolation of PBMC were compared to establish degree of platelet contamination and protein yield. For this specific purpose, 60 mL of blood was collected, and for each of the three methods, two vacutainers (approximately 20 mL of whole blood) were used. The first method used Lymphoprep to isolate the PBMC from blood. For this, vacutainers were spun at 2400g at 4 °C for 15 min. The plasma layer was removed taking care not to disturb the buffy coat and red blood cell layer. The buffy coat (approximately 2 mL from each vacutainer) was removed carefully and added to an equal volume of RPMI 1640 medium (Lonza, Belgium) at room temperature (RT). PBMC from duplicate vacutainers were pooled at this point. The buffy coat was carefully layered on top of an equal volume of Lymphoprep at RT and centrifuged at 700g at RT for 30 min with no brake. The buffy coat was removed, washed with RPMI medium and centrifuged at 700g at RT for 15 min. The supernatant was removed and the cells were resuspended in 2282

Journal of Proteome Research • Vol. 7, No. 6, 2008

RPMI medium and counted. The second method combined Lymphoprep and OptiPrep for the isolation of PBMC from blood. For this, PBMC were isolated as described in Method 1, resuspended in 1 mL of RPMI medium and layered on top of 3 mL of OptiPrep density gradient (1 mL of OptiPrep + 4.4 mL of RPMI medium). The sample was spun at 350g for 15 min at RT and the pellet washed in RPMI medium. The sample was then centrifuged at 700g for 15 min at RT and the cells were counted. The third method used OptiPrep only for the isolation of PBMC from blood. For this, a working solution of OptiPrep was prepared by mixing 4 vol of OptiPrep with 2 vol of RPMI medium. Working solution (2.7 mL) was added to each vacutainer of blood to adjust the plasma density of the whole blood to approximately 1.096 g/mL. The density barrier of 1.078 g/mL was prepared by mixing 5 vol of working solution with 9.6 vol of RPMI medium. Diluted blood (5 mL) was carefully placed under 5 mL of Density Barrier and 0.5 mL of RPMI medium placed on top of the Density Barrier. The tubes were centrifuged at 700g for 20 min at RT with no brake. The PBMC layer, lying on top of the density barrier, was removed into fresh RPMI medium. PBMC from the two tubes were combined at this point and centrifuged at 400g for 10 min at RT. The supernatant was removed and the cell pellets were combined and resuspended in RPMI medium and counted. PBMCs isolated by each method from 20 mL of whole blood were counted using a Sysmex KX-21N, which also measured platelet contamination. Protein concentration in aliquots of 5 × 103 PBMC (2-3 per preparation) was determined in quadruplicate using the BioRad RC-DC protein assay. 2.4.2. Sample Preparation for PBMC Proteomics. The blood sample that was obtained in EDTA as described under Blood Sampling for the Isolation of Plasma, Platelets and PBMC was used to isolate PBMC by Method 3 (OptiPrep). Protein was extracted by adding 10 vol of extraction buffer (7 M Urea, 2 M Thiourea, 2% CHAPS, 0.06% proteinase inhibitor cocktail (Roche)) to 1 vol of pellet. The extracted protein fractions were pooled and protein concentrations were determined using the RC/DC assay (BioRad). A total of 250 µg of PBMC proteins was loaded per 2D gel in 5-fold in each of the four laboratories. 2.4.3. Proteomics of PBMC Spiked with Platelets. To assess the impact of contamination of PBMC with platelets, 150 mL of blood from a single volunteer was collected for this specific purpose into a closed monovette system containing trisodium citrate anticoagulant and pure fractions of both PBMC and platelets were isolated as described above. PBMC and platelet fractions were counted using a Sysmex KX-21N cell counter and combined in the ratio of 1:3 (actual control value postisolation) as a starting reference sample, and 1:25, 1:50 and 1:100 as “contaminated” samples. Protein (250 µg per ‘ratio’) was loaded per 2D gel in triplicate.24–25 The 12 gels were matched using PDQuest (BioRad) software and spots that were considered unique for low or high platelet contamination (i.e., either showing a significantly higher spot density in the starting reference sample or showing a significantly higher spot density in the highest contaminated sample) were cut, trypsinized and identified using MALDI-TOF methods.24–25 2.5. Two-Dimensional Gel Electrophoresis. 2.5.1. Comparison of Nondifferential Staining Techniques. Three hundred micrograms of protein from a single mouse liver protein homogenate was run in triplicate on a 17 cm 2D electrophoresis gel.12–13 BioRad immobilized pH gradient strips (pH 5-8) were used for separation of proteins in the first dimension. One gel was stained with Sypro Ruby (BioRad) and one gel was stained

research articles

Methodological Recommendations and Analytical Variability with Flamingo (BioRad) according to the manufacturer’s instruction. The third gel was stained with Coomassie Blue. For this, the gel was placed into a fixation solution of 50% ethanol, 2% ortho-phosphoric acid and 48% H2O for a minimum of 3 h. The gel was then washed for at least 1 h with two changes of deionized water and shaken in a staining solution of 34% methanol, 2% ortho-phosphoric acid and 64% H2O containing 17% (NH4)2SO4 before 1 mg/mL Coomassie blue was sprinkled on top of the staining solution. After scanning, gel images were uploaded into PDQuest software and an automatic spot count was performed using the same sensitivity parameters for all three gels. 2.5.2. 2D-Gel Electrophoresis of Depleted Plasma, Platelet and PBMC Proteins. Each of the four laboratories ran five replicate gels per sample type, that is, depleted plasma, platelet and PBMC, according to their own standard operating protocols. Laboratory 1 ran large (25 × 21 cm) Duracryl gradient (8-16%) gels, 1 mm thick.12–13 Laboratory 2 ran large Duracryl isocratic (10%) gels, 1 mm thick.26–27 Laboratory 3 ran large acrylamide isocratic (12.5%) gels, 1.5 mm thick.24–25 Laboratory 4 ran large acrylamide isocratic (12.5%) gels, 1 mm thick.19–20 All gels were stained with Flamingo (BioRad) according to the manufacturer’s instructions. After scanning of gels, individual match sets were created from the 5 depleted plasma, platelet, and PBMC gels. These were subsequently analyzed by PDQuest software (BioRad) at laboratories 1 and 3, and by ProteomWeaver software (BioRad) at laboratories 2 and 4. Gels were cropped and prepared for automatic spot detection, setting the sensitivity parameters in such a way that approximately 1500 spots would initially be detected per gel. After the automatic spot detection, the total number of spots, matching rates, correlation coefficients and overall mean coefficient of variation of the 3 match sets (depleted plasma, platelets and PBMC) were recorded. All remaining valid spots not initially matched, or wrongly matched, during the automatic matching were then manually matched. 2.6. Matching and Identification of Specific Proteins across Gels and across Laboratories. From each match set (depleted plasma, platelets and PBMC), 10 random relatively abundant spots from different regions of the gel were selected for spot cutting, trypsinization and identification by either LC-MS/MS (laboratory 1) or MALDI-TOF (laboratories 2-4). Spots were excised from the SDS-PAGE gels, trypsinized and identified using standard operating protocols at each of the four laboratories.12,13,19,20,24–27 Briefly, in laboratory 1, the identity of all spots were analyzed by an ‘Ultimate’ nano LC system (LC Packings, Camberly, Surrey, U.K.) using a C18 PepMap 100 nanocolumn, (LC Packings). The mass spectrometry was performed using a Q-Trap triple quadrupole mass spectrometer fitted with a nanospray ion source (Applied Biosystems/MDS Sciex, Warrington, U.K.). The total ion current (TIC) data was submitted for database searching using the MASCOT search engine (Matrix Science) using the MSDB database with the following search criteria: allowance of 0 or 1 missed cleavages; peptide mass tolerance of (1 Da; fragment mass tolerance of (0.8 Da, trypsin as digestion enzyme; carbamidomethyl modification of cysteine; methionine oxidation as partial modification; and charged state as MH+. In laboratory 2, the identity of all spots was determined using an Ultraflex MALDI-ToF/ ToF mass spectrometer (Bruker Daltonics Ltd.). A 200 Hz nitrogen laser was used to desorb/ionize the matrix/analyte material, and ions were detected in positive ion reflectron mode. All spectra were acquired automatically using the Bruker

fuzzy logic algorithm (FlexControl 3.0, Bruker) and a Biotools 3.0 search routine. The resulting mass data was interrogated using an offline version of the MASCOT search engine (Matrix Science) using the SPtrEMBL database with the following search criteria: allowance of 0 or 1 missed cleavages; peptide mass tolerance of (50 ppm; trypsin as digestion enzyme; carbamidomethyl modification of cysteine; methionine oxidation as partial modification; and charged state as MH+. In laboratory 3, the identity of all spots were analyzed by a MALDI-LR mass spectrometer (Waters) operating in positive reflector mode. A peptide mass list was generated for the subsequent database search with the Mascot search engine (Matrix Science) against the Swiss-Prot database (http://expasy.ch/sprot) with the following search criteria: allowance of 0 or 1 missed cleavages; peptide mass tolerance of 100 ppm; trypsin as digestion enzyme; carbamidomethyl modification of cysteine; methionine oxidation as partial modification; and charged state as MH+. In laboratory 4, the identity of all spots were analyzed with the Autoflex mass spectrometer (Bruker Daltonics) and identified with the MASCOT database (Matrix science) using the following parameters: allowance of 0 or 1 missed cleavages; peptide mass tolerance of (1 Da; fragment mass tolerance of (0.8 Da, trypsin as digestion enzyme; carbamidomethyl modification of cysteine; methionine oxidation as partial modification; and charged state as MH+. 2.7. Calculation of Between Laboratory Variation. Calculation of between laboratory coefficient of variation (CV) was based on the 10 spots from each match set (depleted plasma, platelets and PBMC) selected for spotcutting, trypsinization and identification by either LC-MS/MS or MALDI-TOF as described above. Different gel analysis programs were used to match the gels in the four laboratories. Spot intensity measurements were, therefore, not directly comparable between laboratories, but the relative intensities of different spots would be expected to be maintained. To compare these, spot intensities were first normalized by the mean intensity of the 10 spots studied on each gel. The between laboratory variability per protein spot and per gel in these normalized intensities was obtained from an analysis of variance: the component of variance was estimated as SDL ) [(MSL - MSw)/m]1/2, where SDL is the standard deviation between laboratories, and MSL and MSW are the mean square terms for between- and within-laboratory variance in the ANOVA, and m ) 5 is the number of replicate gels per laboratory. When MSW > MSL, SDL was estimated as zero. An F-test was used to examine whether the between laboratory variance was significantly greater than zero per protein spot and per gel. In some cases, the established protein identity was inconclusive, and so the above analysis was repeated with these spots removed. In this case, the normalization was based only on those spots which were consistently identified by all laboratories, and the CV was based only on those laboratories where the spot identity agreed. Normalizing the spot intensities on the basis of all 10 spots, or only those where there was agreement between laboratories, rather than a normalization based on the whole gel will increase the between laboratory CV, so that our estimates of this will tend to be overestimates.

3. Results and Discussion 3.1. Comparison of Nondifferential Staining Techniques. Staining of proteins with colloidal Coomassie brilliant blue is popular as it allows almost a background-free detection of Journal of Proteome Research • Vol. 7, No. 6, 2008 2283

research articles

de Roos et al.

Figure 1. Stain comparison. 2D gels were stained with Sypro Ruby (BioRad) stain, Flamingo (BioRad) stain or Coomassie Blue stain and matched using PDQuest software as described in Materials and Methods.

proteins, and a good quantitative linearity and compatibility with mass spectrometry. Moreover, it is an easy-to-use and lowcost stain.7 Coomassie brilliant blue and silver stain have a low dynamic range (approximately 10-fold), whereas fluorescent stains like SYPRO Ruby represent a much higher dynamic range (1000-fold). Fluorescent stains are more expensive and more difficult to handle, but they offer a higher sensitivity, with a detection limit of about 5 ng of protein. They also provide a higher specificity than silver stains, which also detect nucleic acids, lipids, lipopolysaccharides and glycolipids, and produce a higher reproducibility, in particular in the quantification of low-intensity protein spots.7 A recent advance is the introduction of difference gel electrophoresis (DIGE) technology, allowing for direct quantitative measurements between differentially labeled samples using cyanine fluorescent dyes prior to gel electrophoresis. When absolute protein variation between two or three samples is the primary target, this method is more reproducible and accurate and not limited by distortion from gel-to-gel variation.28 To choose the optimal (nondifferential) stain for our study, we ran a single mouse liver protein sample using 2D gel electrophoresis and stained the gel with either Sypro Ruby fluorescent stain, Flamingo fluorescent stain or Coomassie Blue (Figure 1). Staining with Sypro Ruby (BioRad) yielded 56% more spots, and staining with Flamingo (BioRad) yielded 108% more spots compared with Coomassie Blue staining after automatic spotmatching in PDQuest (BioRad) software. The Flamingo stain also gave a clearer background. Therefore, Flamingo stain was selected for use in this study. Although Flamingo staining was more sensitive in detecting the spots on scanners in the four laboratories, detection of these spots was more difficult when using the spot cutters. Flamingo does not have a strong UV range excitation spectrum which is the type of light available on the spot pickers which use a transilluminator. Laser scanners use white light range wavelengths to excite which is the effective area for Flamingo. More appropriate filters for Flamingo staining are now available, and therefore, this stain, because of its higher dynamic range, appears a better choice compared with the Sypro Ruby stain. 3.2. Depleted Plasma Proteomics. 3.2.1. Comparison of (A) IgY-12 Proteome Partitioning Kit (Beckman) versus (B) Amersham Plasma Protein Depletion Columns. Sample load capacity of 2D gels is severely limited by the presence of highly abundant proteins. The protein complement of a body fluid or cell represents thousands of different proteins and the variation in abundance between these proteinssthe dynamic rangesdiffers in plasma by a factor of 100 000. Furthermore, nine proteins (e.g., albumin, immunoglobulins, transferrin, fibrinogen and haptoglobulin) make up 90% of plasma pro2284


teins.29–30 Depletion of the major abundant proteins enables 10- to 20-fold higher amounts of depleted serum or plasma samples to be applied to 2D gels.31 However, the number of new spots detected on 2D gels after depletion of major abundant proteins still may be relatively modest, and most newly visualized spots are minor forms of relatively abundant proteins.31 Individual antibody methods have proven to be more specific in depleting targeted proteins and give more complete removal of abundant proteins. Several commercial products are available (mainly immuno-affinity columns) to obtain “depleted plasma”, resulting in a diluted product or “flow-through”. Figure 2 shows a representative plasma 2D gel image where plasma was depleted of 12 abundant proteins (A; Beckman Coulter IgY-12 proteome partitioning kit) or of 2 abundant proteins (B; Amersham Albumin and IgG removal kit). Automated spot detection of all 10 gels revealed that the average ((SD) spot count for the 5 Beckman Coulter IgY-12 proteome partitioning kit gels (986 ( 66) was 5% higher (p < 0.09) than the average ((SD) spot count for the 5 Amersham Albumin and IgG removal kit gels (937 ( 31). This suggests that by removing 12 instead of only 2 of the most abundant proteins from plasma, additional less abundant proteins may have been detected because of a reduction in the dynamic range. Although both methods produced good quality gels, there were significant differences in the ease of use of each of the columns. The Beckman Coulter IgY-12 proteome partitioning kit is a single immuno-affinity column used for all samples with regeneration steps throughout the process. Although the column should deplete up to 100 samples, we found that after 25 samples the protein content of the flow-through increased, suggesting that efficiency for retaining high-abundance proteins had decreased, and consistent depletion conditions could not be guaranteed for each individual sample. The Amersham Albumin and IgG depletion columns are single-use, immuno-affinity columns with one column employed per sample, with multiple samples run concurrently. Use of this column is approximately 4 times more expensive per sample compared with the Beckman Coulter IgY-12 proteome partitioning kit (based on depletion of 100 plasma samples). However, sample preparation time is significantly less, and consistent depletion can be assumed for each of the plasma samples. Therefore, the use of single-use, immuno-affinity columns may be preferred, as the reproducibility of this approach appears better and, as a consequence, the between-sample variation will be lower. 3.2.2. Comparison of Protein Concentration Methods. Table 1 compares the protein yields after concentration of proteins using acetone precipitation or the Millipore Ultrafree 0.5 Centrifugal Filter Device (cutoff 5000 Da). Use of the Millipore Filter Device resulted in a 16% (p < 0.005) higher

Methodological Recommendations and Analytical Variability

research articles

Figure 2. Representative plasma 2D gel images of plasma depleted of 12 abundant proteins (A; Beckman Coulter Yg-12 proteome partitioning kit) or 2 abundant proteins (B; Amersham Albumin and IgG removal kit). A total of 225 µg of protein was loaded onto each gel and five 2D gels were run per depletion method. 2D electrophoresis was carried out as described in Materials and Methods.

Figure 3. Representative images of depleted plasma, platelet and PBMC protein 2D electrophoresis gels ran in each of the four laboratories. 2D gel electrophoresis was carried out as described in Materials and Methods.

recovery of protein from flow-through sample compared with acetone precipitation, and is therefore the preferred method for protein concentration. 3.2.3. Depleted Plasma Proteomics. Proteins circulating in human plasma serve as valid markers of health and disease and are interesting targets for proteome analysis in nutritional studies. Under the umbrella of the Human Proteome Organization (HuPO), a Human Plasma Proteome Project (HPPP) was launched and the first results were published in 2005.31–34 Following HPPP recommendations, we collected plasma using EDTA as anticoagulant, stored plasma samples at -80 °C and used antibody-based depletion of the most abundant plasma proteins. Figure 3 (first row) shows representative images of depleted plasma 2D electrophoresis gels from each of the four laboratories. The gels from laboratory 2 showed a higher resolution of proteins especially in the high molecular weight range. This might be due to consistent use of tricine in the running buffer for the second dimension. Automatic matching of the gels using

PDQuest (BioRad) software in laboratories 1 and 3, and using Proteom Weaver (BioRad) software in laboratories 2 and 4, revealed coefficient of variations in this match set between 18% and 69% (Table 2). Manual matching of the gels improved the overall mean coefficient of variation per match set by 1-16% (Table 2). The CV was lowest in the match set from laboratory 2 where the best resolution of spots was observed (Table 2). The between laboratory CV was originating from the use of different standard operating procedures for 2D gel electrophoresis in the four laboratories, as well as from the different operators using these proteomics platforms, although the current study design does not allow us to separate the technology-derived variation from the human-derived variation. The average between laboratory CV for depleted plasma proteomics was 23% (range 4-54%) (Figure 4A), and 17% (range 0-37%) when accounting for the missing values (Figure 4B). For 6 of the 10 spots, a test of between laboratory variance was significant. The relatively large range in between laboratory CV values for depleted plasma proteomics was due to a high value Journal of Proteome Research • Vol. 7, No. 6, 2008 2285

2286

30%

34%

52%

55%

22%

21%

26%

0.76 [0.63-0.79] 0.81 [0.77-0.88] 0.98 [0.98-0.99] 0.98 [0.98-0.99] 0.86% [0.83-0.87] 0.92 [0.90-0.93] 0.7 [0.5-0.8] 1.0 [0.9-1.0] 66% [60-71%] 68% [62-75%] 71% [65-77%] 80.1 [72.9-84.1] 47% [41-51%] 78% [74-82%] 83% [74-84%] 79% [76-80%] 28%

44%

68%

69%

16%

18%

35%

706 [652-1057] 361 [332-528] 1554 [1448-1722] 1560 [1486-1713] 2271 [1633-2772] 1242 [1162-1566] 789 [713-1109] 628 [470-672] 37%

0.77 [0.75-0.77] 0.78 [0.77-0.79] 0.99 [0.98-0.99] 0.99 [0.99-0.99] 0.85 [0.70-0.87] 0.89 [0.75-0.89] 0.9 [0.9-0.9] 0.9 [0.9-0.9] 57% [54-58%] 61% [58-62%] 69% [67-97%] 76.6 [70.2-80.1] 69% [50-71%] 82% [71-83%] 71% [68-80%] 84% 80-88%] Manual

Automatic 4

Manual

Automatic 3

Manual

Automatic 2

Manual

953 [922-997] 490 [458-781] 970 [691-1087] 948 [906-1034 947 [868-1550] 783 [711-1056] 881 [703-918] 357 [338-536] Automatic 1


a Number of valid spots on each of the 5 gels per match set. b Percentage of matched spots relative to the total number of spots on the reference gel. c Correlation coefficient of valid spots with matched spots on the reference gel. d Overal mean coefficient of variation of all valid spots in each of the 5 gels per match set. Values represent median and [range].

26%

38%

25%

26%

26%

22%

21%

22%

0.83 [0.79-0.88] 0.89 [0.86-0.91] 0.98 [0.97-0.99] 0.99 [0.99-1.00] 0.93 [0.91-0.94] 0.95 [0.93-0.96] 0.7 [0.5-0.8] 1.0 [0.9-1.0] 714 [429-1675] 496 [472-519] 1221 [1160-1248] 1294 [1240-1296] 1428 [1409-1620] 1179 [1100-1231] 1668 [1635-1815] 747 [684-758] 31%

89% [87-91%] 94% [91-96%] 67% [66-71%] 74% [74-77%] 72% [68-76%] 89% [86-89%] 72% [71-72%] 77% [75-80%]

CVd R2c PBMC gels

match rateb valid spotsa CVd R2c match rateb valid spotsa

platelet gels

de Roos et al.

CVd R2c depleted plasma gels

match rateb valid spotsa matching laboratory

Table 2. Automatic and Manual Matching of Five 2D Electrophoresis Gels of Depleted Plasma, Platelet and PBMC Proteins Using PDQuest (laboratory 1 and 3) or ProteomWeaver (laboratory 2 and 4) Software

research articles

for spot 7. This high value may be explained by a relatively low spot intensity for spot 7, or the fact that one of the laboratories was not able to provide an identification for this spot, and therefore, it was treated as a ‘missing value’ (Figure 4B). 3.3. Platelet Proteomics. Platelets play a significant role in common diseases, especially in atherothrombosis and coronary artery disease. They are involved in maintaining vascular integrity by sensing and responding to endothelial damage, in wound healing as well as in activation of inflammatory and immune responses.35 Two-dimensional gel electrophoresis has been used for many years to study platelet biology.36–37 The absence of a nucleus prevents platelets from being studied using a classic molecular approach, and proteomics offers a powerful way to approach their biology. The description of the platelet proteome has used two general approaches, involving either the global cataloguing of proteins present in resting platelets or the characterization of changes in response to stimulation or intervention.35 These combined studies highlight the abundance of signaling (24%) and cytoskeletal proteins (15%) present in the platelet.38 Platelets can be obtained in large quantities from relatively small amounts of blood. In this study, we obtained 5 × 109 platelets from 80 mL of blood, yielding 6.15 mg of protein. In previous studies, a slightly higher platelet yield was observed: in general 100 mL of blood yielded on average 2 × 1010 platelets, and from that number of platelets, between 16 and 24 mg of protein was obtained.39 Our approach aimed toward isolation of a very pure platelet preparation, minimizing platelet activation. Platelets were isolated immediately after blood donation to avoid changes in their physiology and viability. In addition, only the upper third of the platelet rich plasma was isolated to avoid contamination from other blood cells such as erythrocytes, leukocytes and plasma proteins.40–41 Platelets also underwent additional centrifugation steps to minimize potential contamination with plasma proteins normally present in platelet-rich plasma which could influence the outcome of the experiment.39 The method used for platelet isolation is critical. The platelet proteome is subject to rapid changes in response to external signals, giving rise to potentially large within- and between subject variation.40 Another important source of variation is the technical variation within and between laboratories, which so far has not been systematically investigated. In this study, we established the within- and between laboratory variation in proteomics of the resting platelet using 2D gel electrophoresis technology. Figure 3 (middle row) shows representative images of platelet 2D electrophoresis gels from each of the four laboratories. Automatic matching of the gels using PDQuest (BioRad) software in laboratories 1 and 3, and using ProteomWeaver (BioRad) software in laboratories 2 and 4, revealed CVs between 21% and 55% in this match set (Table 2). Manual matching of the gels improved the overall mean CV per match set by only 3-5% in three out of four laboratories, but increased the overall mean CV for this match set by 1% in laboratory 2 (Table 2). The average between laboratory CV for platelet proteomics was 22% (range 5-60%), and 22% (range 5-60%) after accounting for missing values (Figure 4). For all 10 spots, a test of between laboratory variance was significant. The relatively large range in between laboratory CV values for depleted plasma proteomics was due to a high value for spot 2, probably because of the relatively low spot intensity for spot 2 (Figure 4).

research articles


Figure 4. Individual values and the mean (bold line) for between laboratory CV (in percentages) of the 10 spots from each of the three match sets (depleted plasma, platelets and PBMC) selected for spotcutting, trypsinization and identification by either LC-MS/MS or MALDI-TOF as described in Material and Methods. (A) CV based on all data; (B) CV based only on those laboratories where the spot identity was agreed (see also Figure 6). Table 3. Efficiency of PBMC Isolation and Degree of Platelet Contamination Using Three Different Methods for PBMC Isolation: (1) Lymphoprep, (2) Lymphoprep and Optiprep, or (3) Optiprepa method

PBMC cell number

platelet cell number

contamination ratio

pellet weight (mg)

protein weight (mg)

1 2 3

1.47 × 107 1.00 × 107 1.52 × 107

34.35 × 107 3.20 × 107 1.16 × 107

1:23.3 1:3.2 1:0.8

13.70 ( 0.53 6.65 ( 0.35 4.93 ( 0.61

0.88 ( 0.04 0.40 ( 0.06 0.24 ( 0.05

a Blood for PBMC isolation was obtained from 2 vacutainers (18 mLs of blood) per method. Counts were performed on a Sysmex KX-21N cell counter after PBMC/platelet isolation. PBMC and platelet number reflect total cell number recovered per original 18 mL of blood. Protein concentration was assessed in quadruplicate on 2-3 aliquots of 5 × 106 PBMCs per treatment by the RC/DC assay (BioRad) as described in Materials and Methods. Values represent mean ( SD.

Figure 5. Representative 2D gel of a PBMC protein homogenate contaminated with platelets in a 1:3 ratio (A) or a 1:100 ratio (B). A total of 250 µg of protein per ‘ratio’ was loaded per 2D gel in triplicate. Gels were matched using PDQuest (BioRad) software and spots that were considered unique for platelet contamination were cut, trypsinized and identified using MALDI-TOF methods as described in Materials and Methods.

3.4. PBMC Proteomics. 3.4.1. Comparison of Methods for PBMC Isolation and Determination of Optimum Cell Number for Protein Extraction. Various methods exist for isolation of PBMC from whole blood. However, a significant problem with each of these methods is contamination of PBMC with platelets, which obviously impairs the usefulness and validity of PBMC proteomics. Table 3 shows the total PBMC and platelet counts obtained after isolation of PBMC from 20 mL of whole blood with three different methods. Using Lymphoprep and OptiPrep together (Method 2) gave the lowest PBMC yield, but significantly reduced platelet contamination compared with Lymphoprep alone (Method 1). Using OptiPrep alone (Method 3) provided as good a PBMC recovery as Method 1 and the lowest level of platelet contamination (1:0.8) and is therefore optimal for PBMC isolation. Because of the lower platelet contamination, protein yield was correspondingly lower

with this method (Table 3). Total protein isolated with Method 3 was linear with the number of PBMC recovered and subsequent pellet weight (data not shown). 3.4.2. PBMC Proteomics. Figure 3 (bottom row) shows representative images of PBMC 2D electrophoresis gels from each of the four laboratories. Automatic matching of the gels using PDQuest (BioRad) software in laboratories 1 and 3, and using Proteom Weaver (BioRad) software in laboratories 2 and 4, revealed CVs per match set between 22% and 38% (Table 2). Manual matching of the gels improved the overall mean CV per match set by 1-12% in 3 out of 4 laboratories, but increased the overall mean CV for this match set by 4% in laboratory 2 (Table 2). The average between laboratory CV for PBMC proteomics was 34% (range 18-70%), and 29% (range 13-49%) when taking the missing values into account (Figure 4). For 9 out of the 10 spots, a test of between laboratory Journal of Proteome Research • Vol. 7, No. 6, 2008 2287

2288


P33241 P33241 P12004 P33241 P08670 P08865 P09382 P52566 P60709 O15144 P60709 P04083 P60709 P14868 P60709 P06702 P06702 P32119 P35232 Q13347 P21281 P06702 P30084 Q8WU71 P50213 P00558 P40121 P00558 P67936 P67936 P09493 P07951 P67936 P09493 P09493 P24844 P02679 P18206 P02679 P78417 Q86UX7 P00488 O00151

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43

Lymphocyte-specific protein 1 Lymphocyte-specific protein 1 Proliferating cell nuclear antigen Lymphocyte-specific protein 1 Vimentin 40S ribosomal protein SA Galectin-1 Rho GDP-dissociation inhibitor 2 Beta-Actin Actin-related protein 2/3 Beta-Actin Annexin A1 Beta-Actin Aspartyl-tRNA synthetase Beta-Actin Calgranulin-B Calgranulin-B Peroxiredoxin-2 Prohibitin Eukaryotic translation initiation factor 3 Vacuolar ATP synthase subunit B Calgranulin-B Enoyl-CoA hydratase Alpha-enolase Isocitrate dehydrogenase Phosphoglycerate kinase 1 Macrophage capping protein Phosphoglycerate kinase 1 Alpha-tropomyosin 4 Alpha-tropomyosin 4 Alpha-tropomyosin 1 Beta-Tropomyosin 2 Alpha-tropomyosin 4 Alpha-tropomyosin-1 Alpha-tropomyosin 1 Myosin Fibrinogen Vinculin Fibrinogen gamma Glutathione transferase omega-1 Unc-112-related protein 2 Coagulation factor XIII A PDZ and LIM domain protein 1

Protein

584 ( 89 945 ( 103 300 ( 3 767 ( 142 276 ( 35 1229 ( 95 1783 ( 329 11726 ( 544 1228 ( 115 1388 ( 141 5327 ( 649 559 ( 28 3067 ( 367 222 ( 19 2776 ( 399 3078 ( 467 2557 ( 290 2880 ( 246 783 ( 50 650 ( 99 331 ( 36 8904 ( 989 357 ( 39 743 ( 40 505 ( 31 860 ( 67 713 ( 50 553 ( 79 2189 ( 314 272 ( 78 411 ( 64 193 ( 23 182 ( 86 270 ( 59 126 ( 23 430 ( 52 2942 ( 363 4(1 4792 ( 568 4(1 125 ( 55 4(1 403 ( 46

94 ( 18 108 ( 20 62 ( 4 94 ( 2 52 ( 24 265 ( 13 275 ( 43 5248 ( 811 191 ( 7 256 ( 57 704 ( 122 37 ( 48 303 ( 86 54 ( 21 475 ( 136 508 ( 144 424 ( 105 582 ( 123 179 ( 27 134 ( 37 32 ( 10 1901 ( 306 99 ( 20 86 ( 14 70 ( 18 151 ( 32 166 ( 48 137 ( 25 9850 ( 959 1768 ( 254 2018 ( 156 866 ( 40 661 ( 41 1282 ( 5 602 ( 106 2093 ( 3 9334 ( 1702 591 ( 97 22288 ( 2620 580 ( 129 718 ( 125 2240 ( 78 1845 ( 391

Low platelet High Platelet contamination contamination (mean spot density ( SD)

0.013 0.0063 0.00022 0.014 0.0012 0.0028 0.015 0.0030 0.0042 0.0045 0.0061 0.0026 0.0053 0.0097 0.0092 0.0083 0.0041 0.0044 0.0024 0.012 0.0067 0.0041 0.0053 0.00081 0.0017 0.0018 0.0028 0.0090 0.00077 0.0018 0.00079 0.000038 0.017 0.0010 0.0040 0.00034 0.011 0.0044 0.0057 0.0082 0.0033 0.00021 0.0086

p-value

67 76 68 73 173 74 80 86 46 97 70 75 70 223 67 67 51 89 81 103 166 95 91 72 109 129 72 73 137 131 69 88 67 68 116 106 94 74 180 71 130 80 165

Mascot score

23 28 24 30 52 35 55 29 17 28 18 24 16 43 25 63 5 33 31 33 56 70 31 19 32 38 16 26 36 37 15 40 32 16 41 57 26 11 49 31 21 30 47

Sequence coverage %

5 6 6 6 32 7 6 5 17 9 6 7 4 21 7 6 5 7 8 12 26 8 9 7 20 14 5 9 15 14 7 16 12 6 18 10 11 11 17 8 15 26 17

# peptides matched

a Lymphocyte and platelet fractions were counted using a Sysmex KX-21N cell counter and combined in the ratio of 1:3 (lowest contamination level) and 1:100 (highest contamination level). Proteomics using 2D gel electrophoresis was performed on protein homogenates in triplicate as described in Material and Methods. Spots with densities differentially affected by platelet contamination were cut, trypsinised and identified using MALDI-TOF as described in Materials and Methods.

Accession

Spot

Table 4. Proteins Where Levels Were Significantly Higher with Low Platelet Contamination (spot number 1-28) or with High Platelet Contamination (spot number 29-43)a

research articles de Roos et al.


research articles

Figure 6. Identification of 10 preselected plasma protein spots, 10 preselected platelet protein spots and 10 preselected PBMC protein spots in four laboratories. Proteins were identified using MALDI-TOF (n ) 3 laboratories) or LC-MS/MS (1 laboratory) as described in Materials and Methods. Red arrows/numbers indicate agreement on protein identity in all four laboratories; blue arrows/numbers indicate agreement on protein identity in 3 out of 4 laboratories; black arrows/numbers indicate agreement on protein identity in less than 3 out of 4 laboratories.

variance was significant. The relatively large range in between laboratory CV values for depleted plasma proteomics was due to a high value for spot 2 without taking the missing values into account (Figure 4A). However, as none of the laboratories were able to identify this protein, the range in between laboratory CV values was lower when spot 2 was treated as a missing value (Figure 4B). 3.4.3. Proteomics of PBMC Contaminated with Platelets. To determine the impact that platelet contamination has on PBMC proteomics, we investigated how increasing relative platelet numbers would affect the PBMC proteome. Figure 5 shows representative 2D gel images representing PBMC contaminated with platelets at a ratio of 1:3 (A) or 1:100 (B). Matching of the gel set identified 28 spots with significantly higher spot densities (p < 0.05) in gel A compared with gel B, and 15 spots with significantly higher spot densities (p < 0.05) in gel B compared with gel A. Therefore, the first 28 spots represent proteins primarily expressed in PBMC, whereas the second 15 spots represent proteins primarily expressed in platelets. Table 4 lists the identity of each of the 43 spots of which spot densities were differentially affected by platelet contamination. Gel B (showing the highest platelet contamination) also showed a significant increase in levels of albumin protein on the 2D gel. This may not have come from the platelets per se, but may result from BSA being used during platelet isolation to prevent activation. By spiking a PBMC preparation with a platelet protein extract that had contained only low levels of platelet contamination, we were able to demonstrate changes in overall protein composition that were visible on 2D protein gels. The quantities of platelet protein spiked into the PBMC protein extract were within a range representative of the possible contamination of PBMC preparations with platelets. This confirms that platelet contamination can profoundly affect the 2D protein map of

PBMC preparations and that needs to be considered as a serious analytical problem. 3.5. Matching and Identification of Specific Proteins across Gels and across Centers. Mass spectrometry has increasingly become the method of choice for the analysis of complex protein samples and is currently the most important analytical tool in proteomics research. MS-based proteomics, which routinely achieves femtomole sensitivity, has an increasingly important role in biomedical research where limited sample material is available.42 Mass spectrometric technology has progressed enormously with regards to mass accuracy, resolution and peptide sequencing power.8 In four different laboratories, we used mass spectrometry tools to compare protein identities of 30 preselected spots from the 2D gels representing the plasma, platelet and PBMC samples (Appendix 1 in Supporting Information). The four centers obtained similar identities for a total of 19 of 30 spots (9 out of 10 plasma proteins, 7 out of 10 platelet proteins, and 3 out of 10 PBMC proteins). At least 3 out of 4 laboratories obtained similar identities for 26 of 30 spots (10 out of 10 plasma proteins, 8 out of 10 platelet proteins, and 8 out of 10 PBMC proteins; Figure 6). The discrepancy in spot identifications may be explained by the mis-selection of spots as a result of laboratory-to-laboratory variation in gel formats, low scores on the peptide analysis leading to no or only tentative identifications, or incomplete resolution of different proteins in what appears as a single abundant spot. Mass spectrometry can be sensitive to changes in the part of the proteome that it can measure accurately. As has been reported before, it can be difficult to obtain stable, reproducible mass spectrometry results over time and across laboratories,23 and differences in sample collection or sample handling affect the plasma proteome to a degree that can dominate biological changes.23 Our results provide further Journal of Proteome Research • Vol. 7, No. 6, 2008 2289

research articles insights into the origins of these problems, and some practical strategies for their resolution.

4. Conclusions Here we provide a systematic assessment of within- and between laboratory variation in depleted plasma, platelet and PBMC proteomics, using 2D gel electrophoresis as the main method for protein separation. Within laboratory CV (based on all visual spots on the 2D gel) ranged from 18 to 68%, whereas between laboratory CV (based on 30 preselected abundant protein spots) ranged from 22 to 34%. This variation must be considered when designing sufficiently powered human intervention studies that use proteomics tools to elucidate the (often subtle) effects of dietary compounds on potential novel biomarkers of health. Several steps in the proteomics procedure appear critical for and contribute to this variation, among which are the use methods that deplete plasma of its most abundant proteins in a reproducible way, the careful isolation of circulating cells from blood so that contamination with other cells is minimized, the use of tricine to improve the running conditions of the 2D gel electrophoresis procedure, the acquisition of high resolution images for comparative analysis, and the careful assessment of spot identifications obtained by MALDI-TOF or LC-MS/MS analysis.

Acknowledgment. We acknowledge NuGO (The European Nutrigenomics Organisation: linking genomics, nutrition and health research (NuGO, CT-2004-505944) for funding with work. NuGO is a Network of Excellence funded by the European Commission’s Research Directorate General under Priority Thematic Area 5 Food Quality and Safety Priority of the Sixth Framework Programme for Research and Technological Development. Emma Massie, Karen Ross and Lynn Pirie are acknowledged for the sample preparation work. Martin Reid, Gary Duncan, Lynn Olivier and Mike Naldrett are acknowledged for helping with the proteomics and LC-MS/MS analysis. Graham Horgan is acknowledged for his help with the statistical analysis. The Rowett Research Institute is funded by the Scottish Government Rural and Environment Research and Analysis Directorate (RERAD). The Institute of Food Research is funded by the Biotechnology & Biological Sciences Research Council (BBSRC), U.K. Supporting Information Available: Appendix 1, measures of confidence for protein identification and characterization by MALDI-TOF and MS/MS analysis. This material is available free of charge via the Internet at http://pubs.acs.org. References (1) Trayhurn, P. Br. J. Nutr. 2000, 83, 1–2. (2) Xiao, Z.; Prieto, D.; Conrads, T. P.; Veenstra, T. D.; Issaq, H. J. Mol. Cell. Endocrinol. 2005, 230, 95–106. (3) Veenstra, T. D.; Conrads, T. P.; Hood, B. L.; Avellino, A. M.; Ellenbogen, R. G.; Morrison, R. S. Mol. Cell. Proteomics 2005, 4, 409–418. (4) Jacobs, J. M.; Adkins, J. N.; Qian, W. J.; Liu, T.; Shen, Y.; Camp, D. G.; Smith, R. D. J. Proteome Res. 2005, 4, 1073–1085. (5) Fu, Q.; Van Eyk, J. E. Expert Rev. Proteomics 2006, 3, 237–249. (6) Hu, S.; Loo, J. A.; Wong, D. T. Proteomics 2006, 6, 6326–6353. (7) Fuchs, D.; Winkelmann, I.; Johnson, I. T.; Mariman, E.; Wenzel, U.; Daniel, H. Br. J. Nutr. 2005, 94, 302–314. (8) Kussmann, M.; Affolter, M. Curr. Opin. Clin. Nutr. Metab. Care 2006, 9, 575–583. (9) Kussmann, M.; Raymond, F.; Affolter, M. J. Biotechnol. 2006, 124, 758–787. (10) Schweigert, F. J. Ann. Nutr. Metab. 2007, 51, 99–107.

2290


de Roos et al. (11) Trujillo, E.; Davis, C.; Milner, J. J. Am. Diet. Assoc. 2006, 106, 403– 413. (12) de Roos, B.; Rucklidge, G.; Reid, M.; Ross, K.; Duncan, G.; Navarro, M. A.; Arbones-Mainar, J. M.; Guzman-Garcia, M. A.; Osada, J.; Browne, J.; Loscher, C. E.; Roche, H. M. FASEB J. 2005, 19, 1746– 1748. (13) de Roos, B.; Duivenvoorden, I.; Rucklidge, G.; Reid, M.; Ross, K.; Lamers, R. J.; Voshol, P. J.; Havekes, L. M.; Teusink, B. FASEB J. 2005, 19, 813–815. (14) Tom, D. H.; Doring, F.; Fuchs, D.; Roth, H. P.; Daniel, H. J. Nutr. 2005, 135, 199–205. (15) Arbones-Mainar, J. M.; Ross, K.; Rucklidge, G. J.; Reid, M.; Duncan, G.; Arthur, J. R.; Horgan, G. W.; Navarro, M. A.; Carnicer, R.; Arnal, C.; Osada, J.; Roos, B. D. J. Proteome Res. 2007, 6, 4041–4054. (16) Fuchs, D.; Erhard, P.; Rimbach, G.; Daniel, H.; Wenzel, U. Proteomics 2005, 5, 2808–2818. (17) Fuchs, D.; Erhard, P.; Turner, R.; Rimbach, G.; Daniel, H.; Wenzel, U. J. Proteome Res. 2005, 4, 369–376. (18) Fuchs, D.; Pascual-Teresa, S.; Rimbach, G.; Virgili, F.; Ambra, R.; Turner, R.; Daniel, H.; Wenzel, U. Eur. J. Nutr. 2005, 44, 95–104. (19) Fuchs, D.; Dirscherl, B.; Schroot, J. H.; Daniel, H.; Wenzel, U. Mol. Nutr. Food Res. 2006, 50, 58–69. (20) Fuchs, D.; Dirscherl, B.; Schroot, J. H.; Daniel, H.; Wenzel, U. J. Proteome Res. 2007, 6, 2132–2142. (21) Herzog, A.; Kindermann, B.; Doring, F.; Daniel, H.; Wenzel, U. Proteomics 2004, 4, 2455–2464. (22) Wenzel, U.; Herzog, A.; Kuntz, S.; Daniel, H. Proteomics 2004, 4, 2160–2174. (23) Coombes, K. R.; Morris, J. S.; Hu, J.; Edmonson, S. R.; Baggerly, K. A. Nat. Biotechnol. 2005, 23, 291–292. (24) Bouwman, F.; Renes, J.; Mariman, E. Proteomics 2004, 4, 3855– 3863. (25) Bouwman, F.; Suylen, D.; Renes, J.; Mariman, E. Rapid Commun. Mass Spectrom. 2005, 19, 2465–2468. (26) Holmes, K.; Mulholland, F.; Pearson, B. M.; Pin, C.; McNichollKennedy, J.; Ketley, J. M.; Wells, J. M. Microbiology 2005, 151, 243– 257. (27) Polley, A. C.; Mulholland, F.; Pin, C.; Williams, E. A.; Bradburn, D. M.; Mills, S. J.; Mathers, J. C.; Johnson, I. T. Cancer Res. 2006, 66, 6553–6562. (28) Issaq, H. J.; Veenstra, T. D. Electrophoresis 2007, 28, 1980–1988. (29) Anderson, N. L.; Anderson, N. G Electrophoresis 1998, 19, 1853– 1861. (30) Anderson, N. L.; Anderson, N. G. Mol. Cell. Proteomics 2002, 1, 845–867. (31) Echan, L. A.; Tang, H. Y.; Ali-Khan, N.; Lee, K.; Speicher, D. W. Proteomics 2005, 5, 3292–3303. (32) Huang, L.; Harvie, G.; Feitelson, J. S.; Gramatikoff, K.; Herold, D. A.; Allen, D. L.; Amunngama, R.; Hagler, R. A.; Pisano, M. R.; Zhang, W. W.; Fang, X. Proteomics 2005, 5, 3314–3328. (33) Rai, A. J.; Gelfand, C. A.; Haywood, B. C.; Warunek, D. J.; Yi, J.; Schuchard, M. D.; Mehigh, R. J.; Cockrill, S. L.; Scott, G. B.; Tammen, H.; Schulz-Knappe, P.; Speicher, D. W.; Vitzthum, F.; Haab, B. B.; Siest, G.; Chan, D. W. Proteomics 2005, 5, 3262–3277. (34) Omenn, G. S.; States, D. J.; Adamski, M.; Blackwell, T. W.; Menon, R.; Hermjakob, H.; Apweiler, R.; Haab, B. B.; Simpson, R. J.; Eddes, J. S.; Kapp, E. A.; Moritz, R. L.; Chan, D. W.; Rai, A. J.; Admon, A.; Aebersold, R.; Eng, J.; Hancock, W. S.; Hefta, S. A.; Meyer, H.; Paik, Y. K.; Yoo, J. S.; Ping, P.; Pounds, J.; Adkins, J.; Qian, X.; Wang, R.; Wasinger, V.; Wu, C. Y.; Zhao, X.; Zeng, R.; Archakov, A.; Tsugita, A.; Beer, I.; Pandey, A.; Pisano, M.; Andrews, P.; Tammen, H.; Speicher, D. W.; Hanash, S. M. Proteomics 2005, 5, 3226–3245. (35) Macaulay, I. C.; Carr, P.; Gusnanto, A.; Ouwehand, W. H.; Fitzgerald, D.; Watkins, N. A. J. Clin. Invest. 2005, 115, 3370–3377. (36) Jenkins, C. S.; Phillips, D. R.; Clemetson, K. J.; Meyer, D.; Larrieu, M. J.; Luscher, E. F. J. Clin. Invest. 1976, 57, 112–124. (37) Gravel, P.; Sanchez, J. C.; Walzer, C.; Golaz, O.; Hochstrasser, D. F.; Balant, L. P.; Hughes, G. J.; Garcia-Sevilla, J.; Guimon, J. Electrophoresis 1995, 16, 1152–1159. (38) Garcia, A.; Prabhakar, S.; Hughan, S.; Anderson, T. W.; Brock, C. J.; Pearce, A. C.; Dwek, R. A.; Watson, S. P.; Hebestreit, H. F.; Zitzmann, N. Blood 2004, 103, 2088–2095. (39) Garcia, A.; Watson, S. P.; Dwek, R. A.; Zitzmann, N. Mass Spectrom. Rev. 2005, 24, 918–930. (40) Gnatenko, D. V.; Perrotta, P. L.; Bahou, W. F. Blood 2006, 108, 3983– 3991. (41) O’Neill, E. E.; Brock, C. J.; von Kriegsheim, A. F.; Pearce, A. C.; Dwek, R. A.; Watson, S. P.; Hebestreit, H. F. Proteomics 2002, 2, 288–305. (42) Aebersold, R.; Mann, M. Nature 2003, 422, 198–207.

PR700714X

Proteomic Methodological Recommendations for ... - ACS Publications

Recommend Documents