Improved Precursor Characterization for Data-Dependent Mass Spectrometry

Abstract

Modern ion trap mass spectrometers are capable of collecting up to 60 tandem MS (MS/MS) scans per second, in theory providing acquisition speeds that can sample every eluting peptide precursor presented to the MS system. In practice, however, the precursor sampling capacity enabled by these ultrafast acquisition rates is often underutilized due to a host of reasons (e.g., long injection times and wide analyzer mass ranges). One often overlooked reason for this underutilization is that the instrument exhausts all the peptide features it identifies as suitable for MS/MS fragmentation. Highly abundant features can prevent annotation of lower abundance precursor ions that occupy similar mass-to-charge (m/z) space, which ultimately inhibits the acquisition of an MS/MS event. Here, we present an advanced peak determination (APD) algorithm that uses an iterative approach to annotate densely populated m/z regions to increase the number of peptides sampled during data-dependent LC-MS/MS analyses. The APD algorithm enables nearly full utilization of the sampling capacity of a quadrupole-Orbitrap-linear ion trap MS system, which yields up to a 40% increase in unique peptide identifications from whole cell HeLa lysates (approximately 53 000 in a 90 min LC-MS/MS analysis). The APD algorithm maintains improved peptide and protein identifications across several modes of proteomic data acquisition, including varying gradient lengths, different degrees of prefractionation, peptides derived from multiple proteases, and phosphoproteomic analyses. Additionally, the use of APD increases the number of peptides characterized per protein, providing improved protein quantification. In all, the APD algorithm increases the number of detectable peptide features, which maximizes utilization of the high MS/MS capacities and significantly improves sampling depth and identifications in proteomic experiments.

Figure 1

Basic workflow of the advanced peak determination algorithm. The three algorithm steps build upon each other to efficiently analyze charge states in local isotope clusters (step II) and to correlate isotope distributions originating from multiply charged analytes (step III). Lists of isotope distributions and deconvolved masses forming charge envelopes in the spectrum are obtained as the result of the analysis.

Figure 2. Annotation of MS¹ spectra of three overlapping isotope clusters with and without APD

a) With standard peak determination (top), only the green cluster (z = 3) and a subset of the blue cluster are correctly annotated. The missed or incorrect annotations are highlighted in grey and the charge states can be easily calculated with the noted m/z values. The APD algorithm (bottom) correctly annotates all 3 clusters observed in the example spectra. b) The same clusters are observed 1–1.4 seconds later in the gradient now with the blue cluster being the most abundant. Without using the APD algorithm, only the most abundant cluster is annotated while APD again correctly annotates all three species.

Figure 3. Maximizing precursor sampling capacity with the APD algorithm

When acquiring data with an MS/MS scan rate of approximately 35 Hz, the standard peak determination algorithm utilizes only 72% of the available sampling capacity (a) because the instrument runs out of annotated precursors to select. APD (b), on the other hand, nearly maximizing sampling capacity by providing more annotated precursor ions to select for MS/MS fragmentation.

Figure 4. The APD algorithm enables sampling and identification of more peptide features

a) Standard peak determination (top) allows selection of only a fraction of the available peptide features for MS/MS analysis. The APD algorithm (bottom) significantly increases the number of features sampled and identified via MS/MS, especially in the most densely populated regions of the MS¹ spectra (i.e., 400–900 m/z). b) The SPD algorithm inadvertently limits the MS/MS sampling of impure peptides. Peptide selection is less affected by purity when using APD. The success rate (identified/sampled) for each bin from the APD data is plotted in red c) Because of the crude filtering function in SPD, when spectral acquisition speeds are low enough that precursors are never exhausted, precursors selected for MS/MS will have greater dynamic range. The combination of finding more precursors, with APD, and faster spectral acquisition leads to more comprehensive sampling.

Figure 5. The APD algorithm improves peptide and protein identifications over the standard algorithm

a) Improved precursor annotation with APD results in more MS/MS scans, as more precursor ions meet the charge state criterion for selection. With APD, higher MS¹ resolutions correlate with increased MS/MS events. b) The APD algorithm shows significant benefits in peptide identification at 120K and 240K MS1 resolutions because the instrument can perform ion trap MS/MS in parallel while taking advantage of the improved spectral MS¹ quality from the longer transients. c) Even with longer gradients (i.e., higher peak capacity), the APD algorithm enables more unique peptide identifications. d) The benefits of APD translate to protein identifications, as well, including the analysis of pre-fractionated samples. Note, the y-axes do not start at zero and error bars represent minimum and maximum values for two replicates.

Figure 6. Performance of the APD algorithm for multiple proteases and phosphopeptides

a) More unique peptides from mouse brain tissue are identified using the APD algorithm for each of the three alternative proteases investigated. Error bars represent minimum and maximum values of two replicates. Run times were 60 minutes for each analysis. b) When acquiring MS/MS scans in the ion trap (i.e., when the instrument can parallelize MS¹ and MS/MS acquisitions), the APD algorithm enables identification of ~15% more unique phosphopeptide sequences. When acquiring MS/MS scans in the Orbitrap, however, (i.e., when the instrument is operating at lower scan speeds) the standard method outperforms APD. Run times were 3 hrs for each analysis.

Figure 7. APD gains depend upon spectral acquisition rate

a) Average (n=2) FT MS/MS spectra collected and % improvement ((APD-SPD)/SPD) as a function of FT MS/MS resolving power using APD or SPD. b) Average (n=2) unique peptides identified (1% FDR) and % improvement ((APD-SPD)/SPD) as a function of FT MS/MS resolving power using APD or SPD. APD produces the greatest increases in both MS/MS scans and unique peptides at fastest acquisition rates (lowest resolving power).