|Home | About | Journals | Submit | Contact Us | Français|
Although two-dimensional gel electrophoresis (2DE) has long been used to study differential proteomics, its reproducibility has always been a major concern. In recent years, different methodological improvements have contributed to more robust 2DE workflows: use of immobilized isoelectric focusing (IEF) strips, fluorescence-based difference gel electrophoresis (DIGE), new software tools, etc. To assess the reproducibility of 2DE experiments across laboratories, we set up a multi-laboratory study, performed at 12 laboratories of the ProteoRed network (Spanish network of proteomics facilities). All participating laboratories received two protein extracts, prepared from cultured human adenocarcinoma MDA-MB-468 cells, treated or not with 50 ng/ml epidermal growth factor (EGF) for 24 h. Differential analysis was performed by a four-gel 2D-DIGE experiment, using four technical replicates of each sample, with Cy dye-swapping. Strictly defined 2DE conditions were followed by all laboratories. Each laboratory selected the 30 spots presenting the highest fold variations (with P<0.05) and attempted mass spectrometry (MS) protein identification.
The image data sets from the eight laboratories were combined into one experiment and analyzed using Progenesis SameSpots. A principal component analysis (PCA) biplot for the complete data set (all 1179 spots shown in gray) demonstrates a distinct clustering of the control gels (green circle) from the EGF-treated gels (red circle) (Fig. 3). The corresponding colors indicate the location of gels from each laboratory within the clustered data.
Using the same combined data analysis as for Figure 2, within groups, coefficients of variation [CV(%)] were calculated for each spot to assess the technical reproducibility within each laboratory and across all laboratories. In all cases, the CV(%) for control and treated were calculated, and the maximum value was selected for each spot. The bar chart shows the percentage of spots that achieved a CV(%) below the selected cut-off values of 5%, 10%, and 20% CV, respectively, within each laboratory (Fig. 4a). The graph shows the CV(%) calculated in the same manner across the laboratory; here, the data are plotted in ranked order as a percentile of the total number of spot features (Fig. 4b).
The 30 spots showing the highest fold variation (P<0.05) between control and EGF-treated samples were identified by each laboratory following independent SameSpots analysis. These data were then used to construct a heat map, overlaid on an image from Lab A, where red indicates the highest number of matches (Fig. 5). Table 1 shows the number of laboratories highlighting the same spot at each loaction. Of the five spots highlighted by all laboratories, six of the laboratories were able to obtain positive MS identification (Fig. 6) for the circled spot on the heat map.
Figure 6 shows the identical gel location of a differentially expressed spot selected for identification by six different laboratories. An identical protein identification (leukocyte elastase inhibitor) was obtained by all laboratories. The rank of the spot and the measured fold change (treated vs. control) were very similar.
Table 2 provides a summary of the MS identification results for those proteins identified by most laboratories. Cells in green correspond to laboratories successfully identifying the noted protein. The number is the rank position given to the spot by each laboratory. The total number of laboratories identifying the protein, the average fold change, and its CV(%) are shown for each spot. Cells in yellow mean the spot could not be identified. Orange denotes that a different identification was obtained. A white cell means that the spot was not picked for identification.
The results demonstrate a good within-lab and across-lab reproducibility. Within laboratories, 70–93% detected spots present with CVs <20%, 38–71% with CVs <10%, and 7–37% with CVs <5%. Across all laboratories, 72% and 32% of spots show CVs <20% and <10%, respectively. Selection of differentially expressed spots shows good reproducibility across laboratories, although there is a certain degree of subjectivity in the selection, as each laboratory applied its own filtering criteria. Overall, 19 spots were ranked among the top 30 by at least three laboratories and 17 by at least four. MS protein identification was, on average, 60% successful, and 13 spots were identified by at least three different laboratories. In those cases, identical gel locations corresponded to the same protein identification. In conclusion, the results of the study show the robustness of the methodology used and demonstrate the feasibility of across-laboratory validation schemes, pointing toward development of inter-laboratory quality control strategies for proteomic research.