Preanalytical Impact of Sample Handling on Proteome Profiling Experiments with Matrix-Assisted Laser Desorption/Ionization Time-of-Flight Mass Spectrometry

Mass spectrometry (MS) offers a wide range of possibilities for analyzing biological samples containing complex mixtures of proteins and peptides, by generating proteome profiles. The aim of most profiling experiments is identification of changes in protein patterns that are related to a certain disease or clinical status and might therefore be used to improve diagnosis, staging, and monitoring (1). Reproducibility of spectra generated by matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) MS is crucial (2), however, and early enthusiastic reports about the diagnostic power of proteome-profiling approaches for early disease classification have recently been critically assessed and reevaluated (3). The reproducibility of profiling experiments depends on sample quality, which in turn is greatly influenced by preanalytical storage conditions and the choice of anticoagulants (4)(5)(6). In this study, we examined the impact of preanalytical sample handling and storage times on MALDI-TOF MS data for serum and plasma samples and identified a candidate marker to allow objective classification of given samples with respect to age and, hence, appropriateness for MS protein profiling and interpretation of data. We collected 10-mL venous blood samples into serum tubes, EDTA-coated tubes, and ammonium-heparin–coated tubes (Sarstedt). After collection, all samples were initially kept at room temperature for 30 min, to allow clot formation, before centrifugation at 3000g for 10 min in a precooled (4 °C) centrifuge (Rotina 48R; Hettich). After centrifugation, 1 aliquot of each sample was immediately frozen at −80 °C (time 0); other aliquots were incubated for distinct time periods at room temperature as follows: Specimens from 1 healthy volunteer were stored at room temperature until 1, 2, and 4 h after collection and then stored at −80 °C before further use. For classification analysis, serum samples from 6 healthy volunteers and from 20 patients consecutively entering the emergency unit of the university hospital at Mannheim were collected and processed as described above, with the exception that aliquots were prepared only at time 0 and from samples stored at room temperature for 4 h after collection. The study was conducted in accordance with the regulations of the local ethics committee. All samples were deidentified. Serum and plasma samples were processed with the following magnetic bead (MB)–based affinity chromatography resins, used according to optimized protocols recommended by the manufacturer (Bruker Daltonics): hydrophobic interaction chromatography C8 (MB-HIC-C8), weak cation-exchange chromatography (MB-WCX), and immobilized metal-ion affinity chromatography (MB-IMAC-Cu). Specifically, 5 μL of the sample was mixed with the appropriate MB suspension, washed 3 times with 100 μL of wash buffer, and eluted with 5 μL of appropriate elution solution for the MB-HIC-C8 and MB-WCX reagent sets and 10 μL of appropriate elution solution for the MB-IMAC-Cu reagent set. A portion of the eluted sample was diluted 1:10 in matrix solution, and 1 μL of the resulting mixture was spotted on a SCOUT 600-μm AnchorChip target (Bruker Daltonics), a sample support coated with a strongly water-repellent layer of poly(tetrafluoroethylene) (Teflon) harboring an array of small spots acting as hydrophilic sample anchors and holding the sample droplets in place during solvent evaporation. AnchorChip targets produce more homogenous matrix crystallization with better reproducibility of MS spectra than crystallization on steel targets (7). MALDI-TOF MS measurements were performed with an Autoflex II (Bruker Daltonics) operating in positive linear mode. Ions formed by a pulsed ultraviolet laser beam (nitrogen laser; wavelength, 337 nm; 25 Hz) were accelerated to 20 kV. Other instrumental settings were as follows: ion source 2 potential, 18.40 kV; focusing lens voltage, 7.8 kV; and pulsed-ion extraction, 250 ns. Matrix solution was prepared by dissolving 0.3 g/L α-cyano-4-hydroxycinnamic acid solution (Bruker Daltonics) in ethanol–acetone (2:1 by volume), and external mass calibration was performed. The MS spectra for peaks of 1–10 kDa were generated by summarizing 350 laser shots (50 laser shots at 7 different spot positions). To increase detection sensitivity, excess matrix was removed with 10 shots at a laser power of 90% before acquisition of spectra with 50 shots at a fixed laser power of 50%. For identification of the [M+H+] 1467-Da candidate mass peak, we performed postsource decay analysis in reflector mode with an accelerating voltage of 20 kV and a delayed extraction of 110 ns. Interpretation of the MS/MS spectra was performed with BioTools software (Bruker Daltonics) with integrated online linking to the Mascot database (www.matrixscience.com). For each spectrum, a peak-picking algorithm (centroid, signal-to-noise ratio >1.5) was applied by use of flexAnalysis software (Bruker Daltonics), and files containing mass lists and peak areas were exported to Excel file format for further analysis. Using an in-house–written Excel macro function, we sorted the m/z ratios and corresponding peak areas of 2 individual experiments for duplicate masses, accepting a tolerance of 0.1% mass deviation. The corresponding peak areas of 2 spectra were displayed graphically (see, for an example, Fig. 1 in the Data Supplement that accompanies the online version of this Technical Brief at http://www.clinchem.org/content/vol51/issue12) in a scatter plot, and we applied linear regression (add trend line) of Excel software to calculate the coefficient of determination (R2) (8). Because the correlation was related to the value of peak areas, the R2 values were calculated separately for spectra containing all peaks or only peaks <100 000, 50 000, and 25 000 arbitrary units. We calculated time-dependent changes within MS profiles by subtracting the R2 values of different timepoints (1, 2, and 4 h after sample collection) from the R2 value at the corresponding time 0. To determine the method reproducibility, we performed measurements of time 0 samples for 8 individually prepared aliquots and calculated the mean (SD) values of R2. Sample preparations at other times (1, 2, and 4 h) were performed by duplicate spotting of single sample preparations. Serum samples showed profound time-dependent changes in proteome profiles compared with plasma samples. Most changes within MS spectra occurred after a storage time of 4 h at room temperature in serum samples, regardless of the MBs (WCX, HIC-C8, and IMAC-Cu) used for sample processing (Table 11 ). Serum samples from 6 individual donors were collected, and aliquots were prepared only at time 0 and 4 h after blood withdrawal. Serum samples were processed with MB-WCX before MALDI-TOF MS. Samples from 3 randomly selected blood donors were chosen as a training set. In all cases, correct classification of the serum samples from the test set was possible with respect to the sample age (data not shown). Systematic analysis of the time-related changes within mass spectra identified a single [M+H+] peak of 1467 Da that displayed an obvious decrease in signal intensity corresponding to extended storage of serum samples (Fig. 1A1 ). The 1467-Da peak was prominent in the MS spectra of serum samples processed with MB-IMAC-Cu and MB-HIC-C8 and was visible as a minor peak in spectra of serum samples processed with MB-WCX but was absent in plasma samples (see Fig. 2 in the online Data Supplement). Postsource decay analysis allowed identification of this peak as part of the N-terminal fragment of fibrinogen α. For sera from 20 patients, mass spectra of serum samples processed with MB-HIC-C8 also showed a consistently prominent 1467-Da [M+H+] peak. The time-dependent decreases in the relative peak areas is shown in Fig. 1B1 . Because values for the 2 groups did not overlap, the time-dependent decrease in the area of the 1467-Da peak was sufficient to classify samples as aged. Initial signal intensity and time-dependent degradation of the 1467-Da mass peak were not related to acute-phase markers such as C-reactive peptide concentrations (data not shown). We have concluded that different preanalytical sample-handling conditions lead to a significant nonbiological experimental bias that invalidates bioinformatics analysis of proteomic profiling data. Standardized sample collection procedures are required for advantageous use of high-throughput technologies in the proteomics field (9). Our results indicate that plasma samples may be more suitable for MS protein profiling (Table 11 ). During coagulation, release of cellular components, mainly from thrombocytes and leukocytes, is induced (10)(11). Furthermore, proteases seem to be more active in serum samples (12). Through the course of thrombin activation, fibrinogen has been shown to release 3 major forms of fibrinopeptide A (FPA) (13): the unmodified form ADSGEGDFLAEGGGVR (FPA), which constitutes ∼70%, an NH2-terminal truncated FPA (des-Ala FPA; ∼10%), and a product in which Ser-3 is phosphorylated (P-FPA; ∼20%). Only the des-Ala FPA (1467 Da), which is one of the most prominent peaks in freshly obtained serum samples, is sufficiently ionized by the MALDI process (14)(15). As shown in this study, the decrease of the 1467-Da [M+H+] peak is strongly dependent on extended storage of serum samples and appears to be a suitable quality indicator for excluding inappropriate serum samples from protein-profiling experiments. Further work will concentrate on the identification of additional decay markers for assessment of sample quality and their validation in larger clinical studies. Summary of R2 difference calculations.1 R2 differences were calculated by subtracting the R2 of timepoint 0 [R2(t0)] from R2 values of different timepoints [R2(tx)]. Calculations were performed for different specimen types (serum, heparin-plasma, and EDTA-plasma) and peak-area categories (all peaks and peak areas smaller than 100 000, 50 000, and 25 000). R2 differences exceeding 0.2 are underlined. Note the clustering of underlined fields in the lower left corner (serum at 4 h). Bold font indicates R2 differences that can be seen in Fig. 3 of the online Data Supplement. Summary of R2 difference calculations.1 R2 differences were calculated by subtracting the R2 of timepoint 0 [R2(t0)] from R2 values of different timepoints [R2(tx)]. Calculations were performed for different specimen types (serum, heparin-plasma, and EDTA-plasma) and peak-area categories (all peaks and peak areas smaller than 100 000, 50 000, and 25 000). R2 differences exceeding 0.2 are underlined. Note the clustering of underlined fields in the lower left corner (serum at 4 h). Bold font indicates R2 differences that can be seen in Fig. 3 of the online Data Supplement. Time-dependent changes in the 1467-Da [M+H+] peak. (A), serum from 1 individual donor. Samples were processed with MB-IMAC-Cu and spotted on the target in duplicate. tp, timepoint. (B), values for sera from 20 patients, processed with MB-IMAC-Cu. For any sample, the relative (rel) peak area of the 1467-Da [M+H+] peak was calculated, and values for timepoint 0 and timepoint 4 h are displayed as box-and-whisker plots. At timepoint 0, the 1467-Da [M+H+] peak in any of the 20 serum samples was prominent, with a mean peak area of 8.9%; the peak area decreased to ∼1.7% after 4 h of sample storage at room temperature. The line inside each box is the mean; the limits of each box are the 25th and 75th percentiles, and the whiskers are the minimum and maximum values. We gratefully acknowledge that the costs of publication and reprints were supported by the LESSER-LOEWE Foundation e.V.