Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Nov 8.
Published in final edited form as: Anal Chem. 2022 Oct 28;94(44):15198–15206. doi: 10.1021/acs.analchem.2c01711

Coisolation of peptide pairs for peptide identification and MS/MS-based quantification

Ian R Smith 1, Jimmy K Eng 1, Anthony S Barente 1, Alexander Hogrebe 1, Ariadna Llovet 1, Ricard A Rodriguez-Mias 1, Judit Villén 1,*
PMCID: PMC9851627  NIHMSID: NIHMS1863326  PMID: 36306373

Abstract

SILAC-based metabolic labeling is a widely adopted proteomics approach that enables quantitative comparisons among a variety of experimental conditions. Despite its quantitative capacity, SILAC experiments analyzed with data dependent acquisition (DDA) do not fully leverage peptide pair information for identification and suffer from undersampling compared to label-free proteomic experiments. Herein, we developed a data dependent acquisition strategy that coisolates and fragments SILAC peptide pairs and uses y-ions for their relative quantification. To facilitate the analysis of this type of data, we adapted the Comet sequence database search engine to make use of SILAC peptide paired fragments and developed a tool to annotate and quantify MS/MS spectra of coisolated SILAC pairs. This peptide pair coisolation approach generally improved expectation scores compared to the traditional DDA approach. Fragment ion quantification performed similarly well to precursor quantification in the MS1 and achieved more quantifications. Lastly, our method enables reliable MS/MS quantification of SILAC proteome mixtures with overlapping isotopic distributions. This study shows the feasibility of the coisolation approach. Coupling this approach with intelligent acquisition strategies has the potential to improve SILAC peptide sampling and quantification.

Graphical Abstract

graphic file with name nihms-1863326-f0001.jpg

INTRODUCTION

Stable-isotope labeling with amino acids in cell culture (SILAC) is a powerful tool in quantitative proteomics1. In a SILAC experiment, proteome mixtures are typically analyzed by LC-MS/MS using data dependent acquisition (DDA) and the relative quantification of peptide pairs is obtained from the peptide precursor signals in MS1 scans1. Comparatively to label-free quantification, SILAC has better quantitative precision due to decreased technical variation, however SILAC suffers from increased MS1 spectral complexity and redundant sampling of peptides. As a result, fewer peptides are quantified2.

More recently, data independent acquisition (DIA) strategies with quantification on the MS/MS have also been applied to SILAC and other metabolically labeled samples, improving sampling reproducibility3-7. However, the systematic isolation used in DIA experiments can compromise MS/MS quantification, given that asymmetric isolation of the peptide pairs can lead to distorted SILAC ratios8.

Data-dependent acquisitions that are informed by SILAC peptide pairs have also been developed, such as triggering selected ion monitoring scans for precursors that are poorly quantified in MS19 and targeted coisolation and fragmentation of the SILAC peptide pair in direct infusion MS experiments10.

Experiments using other stable isotope labeling strategies, such as trypsin-catalyzed 16O-to-18O-exchange and d0/d3 methyl esterification, have leveraged peptide pairs to assist in de novo peptide sequencing11-13 and automated peptide search validation14. Additionally, coisolated heavy and light 16O-18O peptide pairs have been quantified using y-ion fragments15. Other methods using chemical16 or metabolic17 isotope labeling have shown feasibility of using isotopically distinct fragment ions for relative quantification.

We reasoned that analogously, coisolation of SILAC peptide pairs could result in a boost of b-ion fragment MS/MS signal and paired y-ion SILAC MS/MS signal that can be leveraged for database search identification and MS/MS-based quantification. To evaluate this, we implement a MS acquisition method to coisolate SILAC peptide pairs for MS/MS. To analyze coisolated MS/MS, we adapt Comet to perform peptide-spectrum matching using theoretical spectra of SILAC peptide pairs and we develop a tool to quantify SILAC y-ion pairs from MS/MS spectra. We demonstrate that our method can successfully identify SILAC peptide pairs, while enabling both MS1 and MS/MS-based quantification. We further expand the capabilities of our method to accurately quantify SILAC peptide pairs with overlapping isotopic distributions. Collectively, this work expands our proteomic toolkit for quantitative analysis of SILAC samples.

METHODS

Yeast growth and harvest

Saccharomyces cerevisiae yeast strains DBY10144 and BY4742 were used to generate SILAC proteome mixtures. Yeast cells were grown for ~8 doublings in synthetic complete media containing all amino acids except lysine, which was added as either Lys0 (regular lysine), Lys6 (13C6-lysine) or Lys8 (13C6,15N2-lysine). Cells were harvested using either centrifugation at 4oC or a TCA precipitation protocol. Further details in Supplementary Methods.

Cell lysis, protein reduction and alkylation, and protein digestion

Cell pellets were resuspended on ice with 600 μL lysis buffer composed of 8 M urea, 50 mM Tris pH 8.2, and 75 mM NaCl. Cells were lysed by bead-beating, proteins were reduced, alkylated, and digested with LysC with standard methods as described in the Supplementary Methods.

Peptide desalting

Peptides (1.5-1.7mg) were desalted by solid-phase extraction using reversed-phase tC18 Sep-Pak columns (Waters) of 50 mg beads and the protocol described in the Supplementary Methods. SILAC mixtures of lysine light to heavy ratios (10:1, 4:1, 2:1, 1:1, 1:2, 1:4, 1:10) were generated for mass spectrometry analysis. All sample aliquots were dried by vacuum centrifugation and stored at −80 °C.

Liquid chromatography mass spectrometry data acquisition strategies

SILAC peptide mixtures (500 ng) were subjected to nLC-MS/MS on an EASY-nLC 1200 (Thermo Fisher) coupled to a Orbitrap Eclipse Tribrid mass spectrometer (Thermo Fisher). Peptides were loaded on a trap and analytical column with C18 beads (Dr. Maisch) and separated using a gradient of 80% acetonitrile. Details of MS settings are in the Supplementary Methods. DDA scans had 1.6 m/z isolation windows (Narrow). For offset left and right wide window scans (Wide) applied to Lys0/Lys8 SILAC mixtures, the same MS acquisition parameters were used as the Narrow MS/MS scan however each triggered precursor was isolated with a 6.5 m/z isolation window, offset −4 Da and 4 Da (left and right respectively). For comparing Wide scans and Narrow scans, the same precursor was subjected to left and right Wide window scans and the Narrow scan. For Lys6/Lys8 SILAC mixtures, the isolation offset was set to −1 Da and 1 Da (left and right) with a 5.0 m/z isolation window.

SILAC peptide pair database search approach and new parameters

The Comet18,19 search engine was extended to perform peptide spectral matching on SILAC peptide coisolation theoretical fragments, controlled by the search parameter “silac_pair_fragments". In addition to specifying whether a standard database search or a coisolation search is performed, this parameter also controls whether to apply the coisolation fragment peaks on both the b- and y-ion series (silac_pair_fragments=1) or only on the y-ion series (silac_pair_fragments=2). Of note, we demonstrated that only paired y-ions should be considered (excluding possible paired b-ions) in peptide-spectrum matching (Supplementary Discussion, Figure S1).

To perform a SILAC coisolation search, the mass of the light lysine is defined by the amino acid mass (Lys0) or a static modification (Lys6) and the mass difference between the light and heavy SILAC reagents is set as a variable modification on lysine residues (e.g., 8.014199 for Lys0/Lys8 or 1.99407 for Lys6/Lys8). All matching candidate peptides are scored by the fast cross-correlation (Xcorr) algorithm20.

Standard E-value and Coisolation E-value for SILAC peptide pair prioritization

Comet calculates an expectation value (E-value) for each reported PSM18. Details on E-value calculation are in the Supplementary Methods. For coisolation search, Comet reports a second E-value (Coiso E-value or e.value_paired) based on Xcorr scores calculated on just the non-triggered fragment ion peaks (e.g. only the light fragment ions if the MS/MS spectrum was triggered on a heavy precursor or only the heavy fragment ions if the MS/MS spectrum was triggered on a light precursor). Coiso E-value is calculated similarly as Comet E-value with the Xcorr from non-targeted precursors. The Coiso E-value score is used to determine when a spectrum has correctly coisolated both light and heavy precursors or incorrectly coisolated only one of the two precursors. Details on Coiso E-value calculation are in the Supplementary Methods.

Database searching S. cerevisiae data with Comet

Prior to database searching, MS raw files were converted to .mzML format using msconvert21. The .mzML files from SILAC proteome mixtures were searched with Comet, which can be located at https://sourceforge.net/p/comet-ms/code/1622/tree/branches/release_2019015_silacpair/. The following database search parameters were used: a SGD S. cerevisiae protein sequence database (July 2014, decoys generated for sequence are pseudo-reversed22), searching for b-ions, y-ions, and H2O/NH3 neutral loss fragments, LysC endopeptidase specificity (C-terminal to lysine; max 2 missed cleavages), fixed modification of cysteine carbamidomethyl and [+6.020129] on lysines only when 13C6-lysine was the light SILAC label, variable modifications of oxidation on methionines, acetylation on protein N-terminus, and heavy lysine delta mass (based on mass difference between light to heavy lysine labels), mass tolerance of 20 ppm for precursor m/z, and mass tolerance of 0.02 Da for fragment ions. The parameter isotope_error was set to 3 for all Lys0/Lys8 SILAC mixtures and isotope_error was set to 1 for all Lys6/Lys8 SILAC mixtures. Coiso SILAC searches were designated silac_fragment_pairs = 1 or 2 (including and excluding paired b-ion fragments respectively, while traditional DDA searches had silac_fragment_pairs=0). Comet generated a pep.xml, pin, and tab delimited text output files for downstream analysis.

Peptide spectral match (PSM) FDR filtering and MS/MS spectral quantification

We developed a computational Python suite that integrates a variety of publicly available MS software (Dinosaur23, Pyteomics24,25, and mokapot26) with custom code to generate MS1 and MS/MS quantifications of PSMs from Comet database search results and .mzML27 spectral files. Details for the Python suite are in the Supplementary Methods. The source code and coiso_silac Python package can be accessed on GitLab at https://gitlab.com/public_villenlab/coiso_silac.

Lysine6/Lysine8 mass spectra deconvolution and quantification

For 13C6-lysine (Lys6) and 13C6,15N2-lysine (Lys8) SILAC proteome mixtures, we deconvoluted the overlapping light and heavy MS1 and MS/MS signals by extracting peak intensities for an extended isotopic distribution profile (monoisotopic peak and the first through fourth isotopic peaks) of the light precursor (MS1) or light peptide fragment (MS/MS) based on the calculated theoretical m/z values. In parallel, we determined the theoretical isotopic distribution for each precursor (MS1) or y-ion fragment (MS/MS) as previously described28. We determined the optimal SILAC ratio (Ropt) that minimizes the ratio error between the observed isotopic profile and the theoretical isotopic profile, using published methods29. Further details are in the Supplementary Methods.

Data analysis

Spectra for MS1 and MS/MS scans pertaining to figures were directly extracted from .mzML using custom code or Pyteomics24,25 and MS/MS spectra were annotated with our custom Coiso annotation code or spectrum_utils30. Resultant spectra were plotted in R (version 3.6.1) and RStudio (version 1.4.1103), and all data figures were generated in Adobe Illustrator CS5 (version 15.0.0) and R. All code and data analysis can be accessed via GitLab at https://gitlab.com/public_villenlab/coiso_silac_analysis. All mass spectrometry data and analysis files generated for this manuscript are deposited to the ProteomeXchange Consortium by the PRIDE partner. The PRIDE project identification number is PXD033016, and the reviewer username is reviewer_pxd033016@ebi.ac.uk; password oQ7OiEl7.

RESULTS

Rationale for coisolation SILAC acquisition and database searching

To coisolate SILAC pairs, we use a DDA approach where the detected precursor m/z’s are isolated in a 6.5 m/z-wide isolation window that is offset to the center of the SILAC pair. For a light precursor the offset would be set to the right and for heavy precursors to the left. Our main goal here is to provide proof of principle of the coisolation approach. To avoid relying on the instrument designating precursors as light or heavy, we acquired MS/MS for both offset isolations, and compared to standard DDA MS/MS (i.e., with 1.6 m/z-wide isolation window) for the same precursor (Figure 1a). We adapted Comet to be able to assign peptides from MS/MS spectra of coisolated SILAC pairs. In the manuscript, we refer to the combination of the wide window MS/MS acquisition and the coisolation Comet search as “Wide Coiso”, and the standard DDA MS/MS and Comet search as “Narrow DDA”.

Figure 1: Coiso SILAC computation workflow and MS acquisitions.

Figure 1:

Open in a new tab

a) MS1 scan indicating isolation windows for Coiso (6.5-m/z) and narrow DDA (1.6-m/z) for a heavy peptide from Leu1. b) MS/MS of left and right Coiso scans and narrow window DDA scan for the triggered Leu1 peptide in (a). Theoretical fragments for Coiso search are in the bottom panel at theoretical m/z values. Peptide spectral matched b-ions (purple), light y-ions (green), and heavy y-ions (blue) are highlighted, and other MS/MS peaks are grey. c) Adapted Comet database searching performs peptide spectral matching to SILAC peptide paired fragments. PSMs are matched to Pyteomics-parsed MS/MS spectral data, FDR-filtered with mokapot, mapped to Dinosaur MS1 features, and PSM y-ion paired fragments are annotated and quantified.

We expect the coisolation and analysis of peptide pairs (Wide Coiso) will yield three improvements. First, fragmentation spectra of coisolated SILAC peptide pairs feature merged b-ions and split y-ions (Figure 1b top panel). In a Comet search, the increased ion representation should increase Xcorr values (Figure 1b). Second, to overcome the increased complexity in a wide isolation window, we apply a narrow mass tolerance to increase the specificity of the precursor. We expect these two features will prioritize the true pair over other candidates and paired decoys, thus improving Comet E-values and overall identifications. Third, quantification can be performed in the MS/MS using the y-ions, which are expected to have higher signal-to-noise ratios than precursor signals31.

Here, we assessed the performance of the method by comparing peptide-spectrum matching metrics and identifications between SILAC pair MS/MS and DDA MS/MS for the same precursors. Also, we evaluate quantification precision and accuracy for MS1 precursor and MS/MS y-ion pair features.

Comet database search and quantification tool for SILAC peptide pairs

To analyze SILAC pair data, a two-step analysis pipeline was generated. First, Comet was adapted to perform peptide spectral matching (PSM) for SILAC pairs using heavy and light fragments. For all candidate targets and decoys within the precursor mass tolerance, theoretical paired spectra are generated and matched to calculate cross correlation (Xcorr) scores20. With calculated Xcorr values, PSMs can be assigned an E-value, or an expectation score, as a metric to standardize the confidence of the reported PSMs and calibrate Xcorr values across scans. E-values serve as an effective single metric to differentiate target and decoy PSMs18, thus enabling FDR-based PSM filtering to establish a high confidence set of identifications for downstream analysis.

The second component of the analysis pipeline is an open-source Python package, coiso_silac, that performs SILAC quantification with coisolation data. In the package, we integrate the publicly available software mokapot26 to FDR filter Comet PSMs, Pyteomics24,25 to extract MS spectra m/z and intensity, and Dinosaur23 to identify MS1 precursor features. With this information, we annotate peptide fragments in MS/MS spectra, calculate MS/MS-based SILAC ratios, and map MS1 features to PSMs (Figure 1c). MS/MS SILAC ratios are generated for a variety of different summation strategies and variable number of most abundant topN paired y-ions. This quantification pipeline generates a result file containing 1% FDR filtered PSMs with scoring metrics and MS1 and MS/MS quantifications for downstream analysis.

Coisolating SILAC pairs results in similar peptide identification metrics than narrow window DDA scans

We applied the MS isolation schema from Figure 1a to analyze seven S. cerevisiae proteome samples mixed at different SILAC ratios (10:1, 4:1, 2:1, 1:1, 1:2, 1:4, and 1:10; light:heavy; Lys0:Lys8). For simplicity, initial coisolation MS acquisitions were designed based on SILAC peptides with a single lysine. Thus, peptide identifications that did not have exactly one lysine were removed from the analysis. Since both wide offset MS/MS scans are analyzed for each precursor, only the wide offset scan containing the SILAC pair was compared to the matching DDA MS/MS scan.

To identify which wide offset scan contained the SILAC pair, we developed a “Coiso E-value". This score is built from the Xcorr contribution of the y-ions of the non-targeted SILAC precursor (Figure S2), which should only be present with coisolation of peptide pairs. Compared to the standard Comet E-value (Figure S3), the Coiso E-value improved our capacity to identify coisolated wide offset scans from the non-coisolated scans (Figure S4). Thus, we choose the wide offset MS/MS with greater −log10(Coiso E-value) for our analysis.

We assess method feasibility by comparing identification metrics between our Wide Coiso approach and traditional Narrow DDA for matching precursors. Herein, the metrics for PSM spectral representation (Xcorr) and confidence in PSM assignment (Standard Comet E-value) were compared. In the 1:1 SILAC labeled sample, the Wide Coiso approach showed higher Xcorr scores for 96% of the PSMs compared to Narrow DDA (Figure 2a), likely due to the additional, paired y-ions and intensity increase in b-ion signals. To benefit from this increase in Xcorr with Wide Coiso in the PSM filtering step, we should observe a greater increase for true matches than for false matches. Indeed, we observed the standardized E-value metric showed a 42% improvement for Wide Coiso compared to Narrow DDA for matching PSMs at 1% FDR (Figure 2b).

Figure 2: Coiso SILAC Comet identification performance is comparable to traditional DDA.

Figure 2:

Open in a new tab

a) Density-binned scatter plot of Xcorr from correctly coisolated Wide Coiso scan and Narrow DDA for matched precursors of 1:1 S. cerevisiae SILAC mixture. b) Same as (a) for Comet E-values. c) Number of PSMs at specified FDR using E-value target-decoy competition Wide Coiso and Narrow DDA on the same precursors from (a). Inlet zooms over a 0.01 FDR cut-off. d) PSM identifications at 0.01 FDR based on Comet E-values for PSMs with one lysine across the scan acquisition for seven S. cerevisiae SILAC ratio proteome mixtures for Wide Coiso and Narrow DDA back-to-back scans on the same precursors.

Across the run, Wide Coiso and Narrow DDA showed similar E-value distributions for both decoy hits and target hits (Figure S5a). However, we observed a shifted distribution toward higher −log10(E-values) in the Wide Coiso approach (Figure S5b,c).

Comet E-values are often used to filter PSMs to a defined FDR and generate a high-confidence set of PSMs for downstream analysis. We assessed the filtering capabilities of the E-value metric for Wide Coiso and Narrow DDA by estimating PSM false discovery rate (FDR) using target-decoy competition. In the 1:1 SILAC sample, we observed that the target/decoy discriminatory capability of Wide Coiso E-values is similar to that of Narrow DDA E-values, however Narrow DDA is slightly more sensitive (Figure 2c). When comparing total PSM identifications at 1% FDR across SILAC mixtures, Wide Coiso scans capture 92% for 1:10 and 10:1, 94% for 1:4 and 4:1, 97% for 1:2 and 2:1, and 97% for the 1:1 SILAC ratio samples compared to Narrow DDA scans for the same precursors (Figure 2d). Thus, we conclude that our coisolation pair method provides a similar capability to identify SILAC peptide pairs as narrow DDA scans for the same precursors.

MS/MS quantification of SILAC peptide pairs

To assess quantification performance, we analyzed seven SILAC S. cerevisiae proteome mixtures (labeled with Lys0 and Lys8) acquiring Wide MS/MS with both offsets but without the Narrow MS/MS. PSMs were filtered for the offset with the higher −log10(Coiso E-value) and peptides with one lysine (See Methods). For each PSM, we calculated SILAC ratios at the MS1 level using precursor signals and at the MS/MS level using the most abundant y-ion signals. The number of quantified PSMs at the MS/MS level and its overlap with quantified MS1 precursors vary based on the number of top y-ions used for quantification (Figure S6-7). Here, we highlight MS/MS quantification using the top3 and top4 most abundant y-ions, since they showed the best balance between the number of PSMs quantified and quantification accuracy and precision.

In a 1:1 Lys0:Lys8 S. cerevisiae proteome mixture, we observed a 14% increase in quantified PSMs by top3 y-ion pair MS/MS compared to MS1 (Figure 3a). This increase was further amplified as the SILAC ratio deviated from 1:1. We observed a 15% improvement with the 1:2 and 2:1 mixtures, 23% improvement with 1:4 and 4:1, and 40-51% improvement with 1:10 and 10:1 (Figure 3a). MS1 and MS/MS quantifications using the top3 and top4 y-ion pairs largely overlapped (Figure 3b-c).

Figure 3: In Coiso SILAC MS/MS quantification is more precise than precursor quantification.

Figure 3:

Open in a new tab

a) Barplot of the number of quantified PSMs for MS1 (green) and the two best Coiso MS/MS quantification methods (purple). Plot considers the same set of PSMs for each of the SILAC S. cerevisiae proteome mixtures (Lys0:Lys8 ratio respectively). b-c) Number of quantified PSMs as in (a). Stacked barplot of MS1 quantifications and MS/MS quantifications considering top3 (b) or top4 (c) y-ion fragments. d) Box plot representing the distribution of all quantifiable PSMs from (a). The horizontal line represents the median, box designates the IQR, and the whiskers indicate 1.5 x IQR from the box ends. e) Barplot of percentage PSMs that are outliers from the samples in (a). Outliers are PSM log2(ratios) > Q3 + (1.5 x IQR) or log2(ratios) < Q1 - (1.5 x IQR).

We explored multiple ways to calculate SILAC ratios at the MS/MS level using the sum, median, or linear regression across the topN y-ion pairs (Figure S8). For top3 and top4 y-ion pairs, median-based MS/MS quantifications showed narrower distributions than MS1 quantifications (Figure 3d), suggesting improved precision. For 1:1, 1:2, and 2:1 Lys0:Lys8 SILAC mixtures, the distributions of heavy-to-light ratios for MS/MS quantifications accurately center at the expected values. However, SILAC ratios further from 1:1 showed mild (1:4 and 4:1) to moderate (1:10 and 10:1) ratio compression. The extent of ratio compression did not correlate with MS/MS fragment ion signal, and the relationship between SILAC ratios and MS intensities were similar between MS1 and MS/MS quantifications (Figure S9).

Due to MS1 complexity, many precursor signals fall close to the noise level, resulting in poor quantifications. We expect that our coisolation method’s gas phase enrichment would improve signal-to-noise resulting in fewer outlier quantifications. To compare, we define an outlier as a PSM quantification that lies 1.5 times the interquartile range (IQR) away from the first and third quartile. The percentage of outlier PSMs varies by MS/MS quantification approach and top y-ion pair filters (Figure S10). Remarkably, for all SILAC ratios, top3 and top4 median-based MS/MS quantifications had fewer outliers compared to MS1 quantifications (Figure 3e).

Quantifying SILAC peptide pairs with overlapping isotopic distributions

An avenue to improve the SILAC coisolation method is to increase its multiplexing capacity. This can be achieved by combining isotopically labeled amino acids with smaller delta masses, while maintaining an equivalent size isolation window. MS1-based quantification of SILAC mixtures with overlapping isotopic distributions are challenging to deconvolute using traditional DDA. The coisolation method could improve SILAC quantification due to reduced isotopic overlap of peptide fragment ions in the MS/MS. Thus, we set out to test the coisolation method for proteomes with SILAC labels separated by 2 Da, potentially enabling a 5-plex over an 8 Da range.

We generated 13C6-lysine (Lys6) and 13C6,15N2-lysine (Lys8) labeled S. cerevisiae proteome mixtures spanning seven ratios (10:1, 4:1, 2:1, 1:1, 1:2, 1:4, and 1:10 Lys6:Lys8). We applied our coisolation method using 1 Da left and right offsets with 5 m/z wide isolations. PSMs were assigned using the version of Comet we adapted for SILAC pairs and filtered for the offset MS/MS with a higher −log10(Coiso E-value) and one lysine (See Methods).

In Lys6:Lys8 SILAC mixtures, the second isotope of the lighter (Lys6) precursor of the SILAC pair has the exact same mass as the monoisotopic peak of the heavier (Lys8) precursor, aggregating their signals. This overlap continues for heavier isotopes of the distribution. We deconvolved signals for precursors and y-type fragments using a mathematical approach that fits the theoretical isotopic distribution28 to the observed isotopic spectra to calculate an optimal SILAC ratio (Ropt)29 (Figure 4a).

Figure 4: Quantification of Lys6:Lys8 mixtures using Coiso SILAC.

Figure 4:

Open in a new tab

a) Coiso SILAC MS/MS spectra (bottom panel) of PSA1 ETFPILVEEK light peptide with peptide fragments (black) and unannotated peaks (grey). Top3 peptide paired fragments (y7++, y5+, y7+) observed isotope distribution (black) and the inferred light (green) and heavy (blue) intensity contributions based on the model-defined optimal SILAC ratio (Ropt = Log2(heavy/light)). b) Barplot of number of quantified PSMs for topN and top3 fragments. MS1 quantifications from precursor pair signals in apex or three most intense successive MS1 scans. c) Box plot for the distributions of all quantified PSMs from (b). The horizontal line represents the median, box designates the IQR, and the whiskers indicate 1.5 x IQR from the box ends.

For MS/MS-based quantification, we applied this deconvolution approach to calculate Ropt ratios for each y-ion fragment and then calculated a SILAC ratio for each PSM as the median Ropt among the top3 and topN most abundant y-ion pairs. For MS1-based quantification, we applied the deconvolution method for overlapping precursor signals, generating a Ropt for the chromatographic peak apex and a median Ropt for the three most intense successive MS1 scans (See Methods).

The deconvolution approach resulted in similar numbers of quantified MS1 and MS/MS features in Lys6:Lys8 samples (Figure 4b). In mixtures 4:1 through 1:4, we observed a similar number of quantified PSMs based on y-ion pairs for Lys6:Lys8 and Lys0:Lys8 (Figure 4b, Figure S11a). However, at the most extreme ratios, more PSMs were quantified by MS/MS in Lys6:Lys8 samples compared to Lys0:Lys8 samples (Figure S11a). Surprisingly, the Lys6:Lys8 mixture with the most quantified PSMs was the 1:2 Lys6:Lys8 ratio, possibly due to the asymmetrical light isotopic contribution boosting the heavy precursor signal.

We observed Lys6:Lys8 MS/MS-based quantifications accurately mapped to the expected SILAC ratios (Figure 4c) and had similar precision to Lys0:Lys8 mixtures (Figure S11b). In 1:4 and 1:10 Lys6:Lys8 SILAC mixtures, we observe mild SILAC ratio compression, while the 4:1 and 10:1 SILAC mixtures deviate towards more extreme ratios. This is likely due to the limitations to the isotopic deconvolution, which is supported by our observation of ratios being more extreme for shorter peptides (i.e., less overlap between isotopic envelopes) in the 10:1 Lys6:Lys8 sample (Figure S12).

Collectively, we demonstrate the coisolation method can reliably and precisely quantify proteome mixtures containing SILAC labels with overlapping isotopic distributions.

DISCUSSION

Here we developed a SILAC coisolation analysis platform equipped with novel MS/MS acquisition schema, database search strategy, and quantification pipeline. We use an offset, wide window isolation to coisolate SILAC peptide pairs for joint MS/MS analysis. The adapted Comet search for SILAC pairs leverages light and heavy fragments for peptide-spectrum matching. When considering back-to-back scans on the same precursors, the coisolation method improves Comet identification scores and achieves a similar number of PSMs compared to traditional DDA. We generated a Coiso E-value score that prioritizes MS/MS with successfully coisolated peptide pairs and enables FDR control for SILAC quantification. Furthermore, our coisolation method offers MS/MS-based quantification with little to no interference, using y-ion pairs. Quantification of coisolated SILAC pairs from MS/MS scans outperformed MS1 quantification in the number of quantified PSMs, quantification precision, and percentage of outlier quantifications. However, the method suffered from ratio compression and reduced accuracy at extreme SILAC ratios. Of note, the coisolation method provides both, standard MS1 quantifications and the additional MS/MS quantifications.

Another advantage of the coisolation method is the capability to deal with labeling schemes that produce overlapping isotopic envelopes. Here we show the performance of our method with Lys6:Lys8 labeling. But the method can be applied to labeling schemes that combine Lys0:Lys2:Lys4:Lys6:Lys8 to increase sample multiplexing. Use of stable carbon (13C) and nitrogen (15N) isotopes are preferred to deuterium (2H), given that deuterium can alter peptide elution32.

The two main limitations of this study are related to the current coisolation MS/MS acquisition implementation. First, we performed wide window MS/MS for both offsets to ensure coisolation in at least one of the offsets. This duplicates the number of scans per SILAC precursor pair, slows down acquisition, and thus reduces the number of fragmented precursors. Second, like in data independent acquisition, wide window isolations result in complex MS/MS spectra, which can be difficult to match to a peptide sequence. These two limitations result in fewer peptide identifications than using traditional DDA.

The main goal of our study was to demonstrate the feasibility of the coisolation method when the SILAC peptide pair is successfully coisolated. In future work, we will explore strategies to decrease spectral complexity and improve proteome sampling in combination with the coisolation method. For example, we can apply offline HPLC fractionation to reduce sample complexity or online ion mobility separations (IMS) to separate SILAC peptide pairs by their collisional-cross section (CCS) as in DIA-SIFT33.

With programmatic access to the mass spectrometer34, we envision greater improvements to the coisolation method. We propose implementing an acquisition workflow in which MS/MS scans are triggered on SILAC pair MS1 features detected in real-time, with dynamic adjustment of the center and width of the isolation window. This adjustment will accelerate acquisition by eliminating the need to acquire both offset MS/MS scans for each precursor. Additionally, to increase sampling and reach proteome depths similar to label-free approaches, dynamic exclusion could be programmed at the level of the SILAC pair rather than individual precursors. Finally, sequential isolation of paired light and heavy precursors with an MSX approach35 can minimize the MS/MS spectral complexity while increasing SILAC pair signals. Collectively, these developments can improve the coisolation method maximizing the number of quantified peptides achieved per run.

In future, we envision the coisolation SILAC method to offer unique advantages to MS analysis of phosphoproteomes. During peptide-spectrum matching, site diagnostic fragment ions that distinguish between phosphate localizations must be observed to precisely localize the phosphosite. We expect that the paired y-ions collected in the MS/MS of the coisolation method will offer additional site diagnostic fragments and increase the confidence in site localization assignments. Additionally, site diagnostic ions could be used in the coisolation method to separately quantify phosphopeptide positional isomers, which is not possible if the quantification is done at the MS1 level.

This study presents a coisolation MS method that can successfully identify and quantify SILAC labeled proteomes by coisolating and fragmenting peptide pairs. This approach expands our proteomics toolkit for analyzing SILAC samples and offers exciting new opportunities for future development.

CONCLUSIONS

We provide proof of principle implementation of a MS method that performs MS/MS on coisolated SILAC peptide pairs for peptide identification and MS/MS-based quantification. In a comparison with narrow window isolation DDA scans collected back-to-back on the same precursors, we show the benefits of the paired y-ions and boost in redundant b-ions for increased spectral representation and sensitive MS/MS quantification. Future developments in MS intelligent acquisitions for peptide pair feature detection and dynamic exclusion control can help maximize the potential of this method by increasing proteome sampling and make this method a viable alternative to current DDA approaches to sample SILAC proteomes.

Supplementary Material

Supporting Information

ACKNOWLEDGEMENTS

We would like to thank the members of the Villén Lab for scientific discussions pertaining to the project, particularly Bianca Ruiz, Kyle Hess, and Mario Leutert. Also, we would like to thank Will Fondrie, Andy Keller, Bill Noble, and Devin Schweppe for valuable discussions pertaining to the method’s development. I.R.S. was supported by the NIH training grant T32HG000035. A.S.B. was supported by the NIH training grant T32LM012419. A.H. was supported by DFF International Postdoctoral Grant (0131-00031B) and EMBO non-stipendiary Postdoctoral Fellowship (ALTF 481-2020). This work was supported by the NIH grant R35GM119536 to J.V. The Villén lab is additionally supported by NIH grants R01AG056359, R01NS098329, and RM1HG010461, Human Frontiers Science Program grant RGP0034/2018, and a Research Program grant from the W.M. Keck Foundation. This work is also supported in part by the University of Washington Proteome Resource UWPR95794.

Footnotes

COMPETING INTERESTS

The authors declare no competing interests.

SUPPLEMENTARY INFORMATION

Additional details in Supplementary Methods, Supplementary Discussion, and Supplementary Figures.

REFERENCES

  • (1).Ong S-E; Blagoev B; Kratchmarova I; Kristensen DB; Steen H; Pandey A; Mann M Stable Isotope Labeling by Amino Acids in Cell Culture, SILAC, as a Simple and Accurate Approach to Expression Proteomics*. Mol. Cell. Proteomics 2002, 1 (5), 376–386. 10.1074/mcp.M200025-MCP200. [DOI] [PubMed] [Google Scholar]
  • (2).Stepath M; Zülch B; Maghnouj A; Schork K; Turewicz M; Eisenacher M; Hahn S; Sitek B; Bracht T Systematic Comparison of Label-Free, SILAC, and TMT Techniques to Study Early Adaption toward Inhibition of EGFR Signaling in the Colorectal Cancer Cell Line DiFi. J. Proteome Res 2020, 19 (2), 926–937. 10.1021/acs.jproteome.9b00701. [DOI] [PubMed] [Google Scholar]
  • (3).Meyer JG; D’Souza AK; Sorensen DJ; Rardin MJ; Wolfe AJ; Gibson BW; Schilling B Quantification of Lysine Acetylation and Succinylation Stoichiometry in Proteins Using Mass Spectrometric Data-Independent Acquisitions (SWATH). J. Am. Soc. Mass Spectrom 2016, 27 (11), 1758–1771. 10.1007/s13361-016-1476-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (4).Salovska B; Zhu H; Gandhi T; Frank M; Li W; Rosenberger G; Wu C; Germain P-L; Zhou H; Hodny Z; Reiter L; Liu Y Isoform-Resolved Correlation Analysis between MRNA Abundance Regulation and Protein Level Degradation. Mol. Syst. Biol 2020, 16 (3), e9170. 10.15252/msb.20199170. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (5).Pino LK; Baeza J; Lauman R; Schilling B; Garcia BA Improved SILAC Quantification with Data-Independent Acquisition to Investigate Bortezomib-Induced Protein Degradation. J. Proteome Res 2021, 20 (4), 1918–1927. 10.1021/acs.jproteome.0c00938. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (6).Wu C; Ba Q; Lu D; Li W; Salovska B; Hou P; Mueller T; Rosenberger G; Gao E; Di Y; Zhou H; Fornasiero EF; Liu Y Global and Site-Specific Effect of Phosphorylation on Protein Turnover. Dev. Cell 2021, 56 (1), 111–124.e6. 10.1016/j.devcel.2020.10.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (7).Salovska B; Li W; Di Y; Liu Y BoxCarmax: A High-Selectivity Data-Independent Acquisition Mass Spectrometry Method for the Analysis of Protein Turnover and Complex Samples. Anal. Chem 2021, 93 (6), 3103–3111. 10.1021/acs.analchem.0c04293. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (8).Ludwig C; Gillet L; Rosenberger G; Amon S; Collins BC; Aebersold R Data-Independent Acquisition-Based SWATH-MS for Quantitative Proteomics: A Tutorial. Mol. Syst. Biol 2018, 14 (8), e8126. 10.15252/msb.20178126. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (9).Graumann J; Scheltema RA; Zhang Y; Cox J; Mann M A Framework for Intelligent Data Acquisition and Real-Time Database Searching for Shotgun Proteomics. Mol. Cell. Proteomics MCP 2012, 11 (3), M111.013185. 10.1074/mcp.M111.013185. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (10).Meyer JG; Niemi NM; Pagliarini DJ; Coon JJ Quantitative Shotgun Proteome Analysis by Direct Infusion. Nat. Methods 2020, 17 (12), 1222–1228. 10.1038/s41592-020-00999-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (11).Shevchenko A; Chernushevich I; Ens W; Standing KG; Thomson B; Wilm M; Mann M Rapid ‘de Novo’ Peptide Sequencing by a Combination of Nanoelectrospray, Isotopic Labeling and a Quadrupole/Time-of-Flight Mass Spectrometer. Rapid Commun. Mass Spectrom 1997, 11 (9), 1015–1024. . [DOI] [PubMed] [Google Scholar]
  • (12).Qin J; Herring CJ; Zhang X De Novo Peptide Sequencing in an Ion Trap Mass Spectrometer with 18O Labeling. Rapid Commun. Mass Spectrom 1998, 12 (5), 209–216. . [DOI] [PubMed] [Google Scholar]
  • (13).Goodlett DR; Keller A; Watts JD; Newitt R; Yi EC; Purvine S; Eng JK; von Haller P; Aebersold R; Kolker E Differential Stable Isotope Labeling of Peptides for Quantitation and de Novo Sequence Derivation. Rapid Commun. Mass Spectrom 2001, 15 (14), 1214–1221. 10.1002/rcm.362. [DOI] [PubMed] [Google Scholar]
  • (14).Volchenboum SL; Kristjansdottir K; Wolfgeher D; Kron SJ Rapid Validation of Mascot Search Results via Stable Isotope Labeling, Pair Picking, and Deconvolution of Fragmentation Patterns*. Mol. Cell. Proteomics 2009, 8 (8), 2011–2022. 10.1074/mcp.M800472-MCP200. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (15).Heller M; Mattou H; Menzel C; Yao X Trypsin Catalyzed 16O-to-18O Exchange for Comparative Proteomics: Tandem Mass Spectrometry Comparison Using MALDI-TOF, ESI-QTOF, and ESI-Ion Trap Mass Spectrometers. J. Am. Soc. Mass Spectrom 2003, 14 (7), 704–718. 10.1016/S1044-0305(03)00207-1. [DOI] [PubMed] [Google Scholar]
  • (16).Bamberger C; Pankow S; Park SKR; Yates JR Interference-Free Proteome Quantification with MS/MS-Based Isobaric Isotopologue Detection. J. Proteome Res 2014, 13 (3), 1494–1501. 10.1021/pr401035z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (17).Merrill AE; Hebert AS; MacGilvray ME; Rose CM; Bailey DJ; Bradley JC; Wood WW; El Masri M; Westphall MS; Gasch AP; Coon JJ NeuCode Labels for Relative Protein Quantification*. Mol. Cell. Proteomics 2014, 13 (9), 2503–2512. 10.1074/mcp.M114.040287. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (18).Eng JK; Jahan TA; Hoopmann MR Comet: An Open-Source MS/MS Sequence Database Search Tool. Proteomics 2013, 13 (1), 22–24. 10.1002/pmic.201200439. [DOI] [PubMed] [Google Scholar]
  • (19).Eng JK; Hoopmann MR; Jahan TA; Egertson JD; Noble WS; MacCoss MJ A Deeper Look into Comet—Implementation and Features. J. Am. Soc. Mass Spectrom 2015, 26 (11), 1865–1874. 10.1007/s13361-015-1179-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (20).Eng JK; Fischer B; Grossmann J; MacCoss MJ A Fast SEQUEST Cross Correlation Algorithm. J. Proteome Res 2008, 7 (10), 4598–4602. 10.1021/pr800420s. [DOI] [PubMed] [Google Scholar]
  • (21).Chambers MC; Maclean B; Burke R; Amodei D; Ruderman DL; Neumann S; Gatto L; Fischer B; Pratt B; Egertson J; Hoff K; Kessner D; Tasman N; Shulman N; Frewen B; Baker TA; Brusniak M-Y; Paulse C; Creasy D; Flashner L; Kani K; Moulding C; Seymour SL; Nuwaysir LM; Lefebvre B; Kuhlmann F; Roark J; Rainer P; Detlev S; Hemenway T; Huhmer A; Langridge J; Connolly B; Chadick T; Holly K; Eckels J; Deutsch EW; Moritz RL; Katz JE; Agus DB; MacCoss M; Tabb DL; Mallick P A Cross-Platform Toolkit for Mass Spectrometry and Proteomics. Nat. Biotechnol 2012, 30 (10), 918–920. 10.1038/nbt.2377. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (22).Elias JE; Gygi SP Target-Decoy Search Strategy for Increased Confidence in Large-Scale Protein Identifications by Mass Spectrometry. Nat. Methods 2007, 4 (3), 207–214. 10.1038/nmeth1019. [DOI] [PubMed] [Google Scholar]
  • (23).Teleman J; Chawade A; Sandin M; Levander F; Malmström J Dinosaur: A Refined Open-Source Peptide MS Feature Detector. J. Proteome Res 2016, 15 (7), 2143–2151. 10.1021/acs.jproteome.6b00016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (24).Goloborodko AA; Levitsky LI; Ivanov MV; Gorshkov MV Pyteomics—a Python Framework for Exploratory Data Analysis and Rapid Software Prototyping in Proteomics. J. Am. Soc. Mass Spectrom 2013, 24 (2), 301–304. 10.1007/s13361-012-0516-6. [DOI] [PubMed] [Google Scholar]
  • (25).Levitsky LI; Klein JA; Ivanov MV; Gorshkov MV Pyteomics 4.0: Five Years of Development of a Python Proteomics Framework. J. Proteome Res 2019, 18 (2), 709–714. 10.1021/acs.jproteome.8b00717. [DOI] [PubMed] [Google Scholar]
  • (26).Fondrie WE; Noble WS Mokapot: Fast and Flexible Semisupervised Learning for Peptide Detection. J. Proteome Res 2021, 20 (4), 1966–1971. 10.1021/acs.jproteome.0c01010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (27).Bakalarski CE; Elias JE; Villén J; Haas W; Gerber SA; Everley PA; Gygi SP The Impact of Peptide Abundance and Dynamic Range on Stable-Isotope-Based Quantitative Proteomic Analyses. J. Proteome Res 2008, 7 (11), 4756–4765. 10.1021/pr800333e. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (28).Yergey J; Heller D; Hansen G; Cotter RJ; Fenselau C Isotopic Distributions in Mass Spectra of Large Molecules. Anal. Chem 1983, 55 (2), 353–356. 10.1021/ac00253a037. [DOI] [Google Scholar]
  • (29).Chavez JD; Keller A; Mohr JP; Bruce JE Isobaric Quantitative Protein Interaction Reporter Technology for Comparative Interactome Studies. Anal. Chem 2020, 92 (20), 14094–14102. 10.1021/acs.analchem.0c03128. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (30).Bittremieux W Spectrum_utils: A Python Package for Mass Spectrometry Data Processing and Visualization. Anal. Chem 2020, 92 (1), 659–661. 10.1021/acs.analchem.9b04884. [DOI] [PubMed] [Google Scholar]
  • (31).Venable JD; Dong M-Q; Wohlschlegel J; Dillin A; Yates JR Automated Approach for Quantitative Analysis of Complex Peptide Mixtures from Tandem Mass Spectra. Nat. Methods 2004, 1 (1), 39–45. 10.1038/nmeth705. [DOI] [PubMed] [Google Scholar]
  • (32).Ong S-E; Kratchmarova I; Mann M Properties of 13C-Substituted Arginine in Stable Isotope Labeling by Amino Acids in Cell Culture (SILAC). J. Proteome Res 2003, 2 (2), 173–181. 10.1021/pr0255708. [DOI] [PubMed] [Google Scholar]
  • (33).Haynes SE; Majmudar JD; Martin BR DIA-SIFT: A Precursor and Product Ion Filter for Accurate Stable Isotope Data-Independent Acquisition Proteomics. Anal. Chem 2018, 90 (15), 8722–8726. 10.1021/acs.analchem.8b01618. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (34).Schweppe DK; Eng JK; Yu Q; Bailey D; Rad R; Navarrete-Perea J; Huttlin EL; Erickson BK; Paulo JA; Gygi SP Full-Featured, Real-Time Database Searching Platform Enables Fast and Accurate Multiplexed Quantitative Proteomics. J. Proteome Res 2020, 19 (5), 2026–2034. 10.1021/acs.jproteome.9b00860. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (35).Egertson JD; Kuehn A; Merrihew GE; Bateman NW; MacLean BX; Ting YS; Canterbury JD; Marsh DM; Kellmann M; Zabrouskov V; Wu CC; MacCoss MJ Multiplexed MS/MS for Improved Data-Independent Acquisition. Nat. Methods 2013, 10 (8), 744–746. 10.1038/nmeth.2528. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supporting Information
NIHMS1863326-supplement-Supporting_Information.pdf (1.5MB, pdf)

RESOURCES