To the Editor:

Wiśniewski et al. recently reported a sample preparation method for proteome analysis using spin filter microcentrifugation devices1. The procedure described is almost identical to a method we reported in 2005 (ref. 2). In our paper, we described the use of spin filters to remove sodium dodecyl sulfate (SDS) and other contaminants, followed by the reduction, alkylation and tryptic digestion of proteins on the filter and finally the isolation of peptides by centrifugation. We described the application of the spin filter preparation method to purified proteins, protein mixtures, cell lysates and subcellular fractions, which are the major elements of the method described by Wiśniewski et al.1. Our spin filter method already has seen considerable use: we are aware of at least 18 publications in which it was applied (Supplementary Note).

These publications show that this approach is useful in some applications, but is not necessarily “universal” as Wiśniewski et al.1 suggest. We and others have found that the use of spin filters has considerable limitations because of poor peptide recoveries when relatively small (<50 μg) protein samples are analyzed. Even at higher sample loads, digestion efficiencies and peptide recoveries are variable3. In our previous work with detergent-solubilized membrane vesicles, the spin filter preparation did not yield protein identifications, apparently owing to the difficulty of removing detergent (1% CHAPS) that interfered with protein digestion. In that work, we used a 'short' SDS–polyacrylamide gel electrophoresis (SDS-PAGE) separation ∼1–2 cm into the gel, followed by in-gel tryptic digestion and multidimensional liquid chromatography–tandem mass spectrometry (LC-MS/MS) to identify several dozen vesicle-associated proteins4. To analyze cell and tissue proteomes, we also have used the in-solution digestion method of Wang et al.5, which uses tri-fluoroethanol (TFE) instead of detergent to solubilize hydrophobic and membrane proteins.

We compared the performance of the spin filter method (performed as described by Wiśniewski et al.1) with that of the short SDS-PAGE and TFE methods for analysis of proteins from RKO colon carcinoma cells (Table 1 and Supplementary Data). We analyzed samples corresponding to a high protein load (50 μg) and a low protein load (150 ng). Then we analyzed all digests under identical conditions by reverse phase LC-MS/MS (Supplementary Methods). At the high protein load, the spin filter preparation yielded 8% more protein identifications than the gel and TFE methods. However, at the low protein load, the spin filter method yielded just 44% of the protein identifications found with the gel method and only 31% of the identifications found with the TFE method.

Table 1 Comparison of spin filter, short SDS-PAGE and TFE methods
Full size table

Thus we conclude that spin filter-based approaches are subject to substantial losses of identifications at low sample loads, probably owing to binding of proteins and peptides to the spin filters. We note that Wiśniewski et al.1 only analyzed complex cell proteomes with their spin filter method. However, nonspecific binding and protein or peptide losses would make this method a poor choice for the analysis of less complex samples (for example, multiprotein complex pull-downs), which often represent very small protein loads. The short SDS-PAGE approach is much better suited to such samples.

Note: Supplementary information is available on the Nature Methods website.