Great advances have been made in sensitivity and acquisition speed on the Orbitrap mass analyzer, enabling increasingly deep proteome coverage. However, these advances have been mainly limited to the MS2 level, whereas ion beam sampling for the MS1 scans remains extremely inefficient. Here we report a data-acquisition method, termed BoxCar, in which filling multiple narrow mass-to-charge segments increases the mean ion injection time more than tenfold as compared to that of a standard full scan. In 1-h analyses, the method provided MS1-level evidence for more than 90% of the proteome of a human cancer cell line that had previously been identified in 24 fractions, and it quantified more than 6,200 proteins in ten of ten replicates. In mouse brain tissue, we detected more than 10,000 proteins in only 100 min, and sensitivity extended into the low-attomolar range.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We acknowledge all members of the Department of Proteomics and Signal Transduction for help and fruitful discussions; in particular, we thank G. Borner, S. Doll, D. Hornburg, N. Kulak, M. Murgia, M. Raeschle and K. Sharma. We thank G. Sowa, I. Paron and K. Mayr for technical support and P. Treit and N. Skotte for comments on the manuscript. This research was partially supported by funding from the German Research Foundation (DFG–Gottfried Wilhelm Leibniz Prize) (F.M., M.M.), the European Union's Horizon 2020 research and innovation program under grant agreement 686547 (MSmed project) (F.M., P.E.G., S.V.W., J.C., M.M.) and the Max Planck Society for the Advancement of Sciences (F.M., P.E.G., S.V.W., J.C., M.M.).
Integrated supplementary information
Supplementary Figure 1 Investigation of key BoxCar acquisition parameters using a D-optimal Design of Experiment to model linear and quadratic effects.
The objective of the DoE was to maximize the number of detected peptide features (m/z 400–1,200, z > 1) in 45-min runs of 1 μg HeLa digest. a,b, The 4D response contour plot illustrates the effect of the following factors: the number of BoxCar scans per cycle (x axis), the number of boxes per scan (y axis) as well as the effect of the maximum ion injection time in percent of the transient time (Fill) for a resolution of 60,000 (a) and 120,000 (b) at m/z 200. The results indicate that the number of features increases with the maximum fill time, which is in accordance with the expected increase in dynamic range and improved signal-to-noise ratios. Furthermore, the benefits of increasing the resolving power overcompensate for the downside of lengthening the cycle time. The effects of varying the number of BoxCar scans and boxes were less prominent; however, the results imply that a combination of three scans with about 12 boxes each yields best performance. Stars indicate the settings used for data acquisition in the present study for 45-min (white) and 100-min (green) gradients.
a, Number of detected multiply charged features from 1 μg of HeLa digest over a 45-min gradient with the standard and BoxCar methods. b, Dynamic range of the detected features as a function of retention time.
a,b, Quantification of peptide (N = 24,487) and protein (N = 1,647) (b) ratios from a human cancer cell line in a two-channel SILAC experiment, acquired in triplicate single runs with the BoxCar method and applying the intensity correction as described in the main text. The heavy and light channels were mixed in a 1:3 ratio, which is accurately reflected in the density plots.
Supplementary Figure 4 Peptide label-free quantification in ten replicate 45-min runs with the BoxCar method and our standard shotgun method.
a, Pearson correlation analysis of the non-normalized peptide intensities. The median pairwise correlation coefficients were 0.96 and 0.97 for the BoxCar and shotgun replicates, and 0.95 for cross-correlated pairs (total N = 36,736 peptides). b, Pairwise peptide feature intensity ratios of a representative BoxCar–shotgun pair as a function of m/z (N = 30,924).
E. coli lysate was mixed with a human cancer cell line (HeLa) lysate in 1:2 and 1:12 ratios (peptide w/w, E. coli:HeLa). a,b, The scatterplot indicates median MaxLFQ ratios of human (red) and E. coli (blue) proteins that were fully quantified in triplicate single runs (N = 3) of each sample with the shotgun (N = 5,214 proteins) (a) and BoxCar (N = 5,699 proteins) (b) acquisition method. One-sided Student's t-tests return in total 962 significantly changing E. coli proteins at a permutation-based FDR below 0.05 for BoxCar, which is 35% more than with the standard method.
Comparison of isotope patterns (features) from human plasma samples that were commonly identified and quantified in triplicate 45-min shotgun and BoxCar runs (m/z 400–1,200, N = 11,918 for each). a, Detection time per feature as a measure of sensitivity. b, Number of detected isotope peaks as a measure of intra-scan dynamic range.
Comparison of the number and dynamic range of identified features by matching from a deep library into single shotgun (gray) and BoxCar (red) runs. a–c, Analysis of a human cancer cell line digest in a 45-min gradient. d–f, Analysis of a mouse cerebellum digest in a 100-min single run.
Supplementary Figure 8 Absolute abundance range of mouse cerebellum proteins quantified in 100-min BoxCar single runs.
a, Estimated protein amount on the analytical column per 1-μg peptide injection (N = 10,569). b, Estimated copy numbers per cell for proteins annotated with the GOCC term ‘transcription factor complexes’ (N = 185).
Supplementary Figures 1–8, Supplementary Tables 1–3 and Supplementary Note 1
BoxCar acquisition and data analysis
Peptide and protein identifications and quantitative information for all proteomics experiments