Mapping differential interactomes by affinity purification coupled with data-independent mass spectrometry acquisition

Abstract

Characterizing changes in protein-protein interactions associated with sequence variants (e.g., disease-associated mutations or splice forms) or following exposure to drugs, growth factors or hormones is critical to understanding how protein complexes are built, localized and regulated. Affinity purification (AP) coupled with mass spectrometry permits the analysis of protein interactions under near-physiological conditions, yet monitoring interaction changes requires the development of a robust and sensitive quantitative approach, especially for large-scale studies in which cost and time are major considerations. We have coupled AP to data-independent mass spectrometric acquisition (sequential window acquisition of all theoretical spectra, SWATH) and implemented an automated data extraction and statistical analysis pipeline to score modulated interactions. We used AP-SWATH to characterize changes in protein-protein interactions imparted by the HSP90 inhibitor NVP-AUY922 or melanoma-associated mutations in the human kinase CDK4. We show that AP-SWATH is a robust label-free approach to characterize such changes and propose a scalable pipeline for systems biology studies.

Figure 1

AP-SWATH pipeline. (a) MS analysis pipeline: each sample is processed separately for DDA and SWATH, and the spectral library built from all DDA runs within an experimental set is used to retrieve quantitative information from each of the SWATH runs. A series of tools are used to automatically match the DDA and SWATH spectra, extract quantitative information, normalize the transitions, peptides and proteins, and determine the Fold Change differences between samples, and the confidence on the Fold Change. (b) Schematic of the parameters used for normalization. Intensity ratio histograms are generated between pairs of samples, and a number of metrics are derived. (c) Effects of the normalization steps on the area ratio histograms demonstrated for a dataset consisting of 9 samples derived from CDK4 WT, 9 from CDK4 R24C and nine from CDK4 R24H. The top panel shows the ratio histograms before normalization; the middle panel, after normalization based on experimental types (here biological replicates); and the bottom panel shows the final results after normalization of the experimental bias.

Figure 2

AP-SWATH for scoring protein interactions. (a) Reproducibility metrics for the SWATH extraction at 1% FDR in relation to the binned % CV values; the upper boundary is indicated. (b) Reproducibility metrics for the common peptides identified in all DDA experiments and extracted from SWATH data with 1% FDR. The numbers of peptides and proteins identified/quantified within the %CV indicated are listed, and the overall percentages of peptides/protein within the 20% CV interval are indicated. (c) Fold Change calculation results for the FLAG-EIF4A2 bait in relation to a negative control, FLAG-GFP (left), and for FLAG-MEPCE in relation to FLAG-GFP (right). The proteins that changed ≥ 2-fold with a confidence ≥ 0.75 are displayed: increased proteins (yellow scale) are specific to the bait in relation to the control while the decreased proteins (blue scale) are more abundant in the negative control samples, and likely contaminants (these tend to be enriched in the Contaminant Repository for Affinity Purification, CRAPome.org). Several components of well-characterized protein complexes that were enriched with EIF4A2 and MEPCE are indicated by the colored dots to the right of the heatmaps (See Supplementary Figs 9-11 for an expanded view). In this and all other heatmaps, values exceeding the Fold Change (log₁₀) indicated in the color-coding bars are depicted as the maximal intensity values.

Figure 3

Selected biological samples. (a) Schematic representation of the effects of mutations, splice variants and chemical perturbations on the modulation of specific protein-protein interactions. In the reference interactome, interaction of the central protein with three binding partners is represented. If this protein is absent from the cells, all three interactions are lost (blue interactome; node removal). Interactions can also be selectively lost (green interactome; edge lost) or gained (red interactome, edge gain). In these cases, the loss/gain can be absolute (represented here by the presence or absence of an edge), or partial (depicted by changes in edge width; the magnitude of these changes can be measured by quantitative proteomics). (b) AP-western validation of a test case for monitoring interactome changes. FLAG-tagged CDK4 WT and mutant proteins are expressed at similar amounts (to each other and to the endogenous CDK4 protein) in Flp-In T-REx 293 cells and purified on an anti-FLAG resin. Association of the endogenous p18INK (CDKN2C) protein was detected by immunoblotting.

Figure 4

Identification of differential interactomes for CDK4 cancer-associated mutants. (a) Principal component analysis showing the clear separation of two control (FLAG alone) samples and two CDK4 WT samples in comparison to the two CDK4 R24 samples that show little separation. (b) Heatmap representation of the proteins passing the confidence threshold in one of the CDK4 baits relative to the negative controls. The grey cells indicate that the thresholds for confidence, Fold Change or signal-to-noise were not met in this particular pairwise comparison (See Supplementary Fig 12 for a global view of all the data without these missing values and Supplementary Figs 13-14 for expanded views). (c) Schematic of the scoring process for differential interactome mapping: in the first step, the potential interactions for a set of baits are collectively scored against a negative control and proteins confidently up-regulated with ≥ 1 baits are considered further. In the second step, systematic pairwise comparisons between all baits, or comparison to a bait used as a reference point, are performed. (d) Left: Heatmap depicting the high confidence proteins differentially detected in the R24C and R24H mutants in relation to the WT sample. Only proteins changing with a confidence ≥ 0.75 and that passed filtering criteria defined in Methods are depicted. See Supplementary Fig. 15-17 for all pairwise comparisons. Right: Heatmap showing the iTRAQ ratios of the high confidence SWATH proteins (see Supplementary Fig. 18 for iTRAQ ratio standard deviation). (e) Validation of selected regulated interactions by AP-western.

Figure 5

Use of AP-SWATH to probe drug-modulated interactions. (a) Increased association of kinase mutants with HSP90 as determined by LUMIER. (b) Heatmap depicting the high confidence differentially recovered proteins as a consequence of NVP-AUY922 treatment (all comparisons are pairwise, for the same bait treated with NVP in comparison to the mock treated sample). See Supplementary Fig. 24 for the heatmap of the first filtering step (normalization to the negative control) and Fig. 25 for an expanded view). (c) Fold Change and Median Absolute Variance (error bars) for all proteins from panel. b. See Supplementary Fig. 26-27 for an expanded view of protein and peptide level changes. (d) Fold change of selected proteins in the mutants as compared to the WT following NVP-AUY922 treatment. (e) Validation of selected regulated interactions by AP-western: 500nM NVP-AUY922 was used for 1 hour. (f, g) AP-western analysis of time course of CDK4 WT (f) and R24C mutant (g) dissociation from CDC37-HSP90 in the presence of 100nM NVP-AUY922. In e, f and g, * indicates the position of the FLAG-tagged bait protein; • indicates endogenous CDK4.