Mapping Dysfunctional Protein-Protein Interactions in Disease

Abstract

Protein-protein interaction (PPI) networks are dynamically remodeled in disease, yet most systems biology approaches focus on changes in protein abundance, overlooking critical interaction-level dysfunction. Here, we present a robust, chemoproteomic method-dysfunctional Protein-Protein Interactome (dfPPI)-that enables high-throughput, systematic, disease-contextual mapping of PPI network dysfunctions in cells and primary human tissue. This method integrates chemical biology probes that selectively capture epichaperome-based interactome assemblies with label-free liquid chromatography-tandem mass spectrometry (LC-MS/MS) and network-based computational analysis, to uncover the rewiring of protein networks not apparent from transcriptomic or proteomic data alone. The dfPPI platform can be applied across disease states, species, and tissues to identify actionable nodes of dysfunction and enable high-resolution, systems-level insights into disease progression. In this protocol, we demonstrate step-by-step procedures for sample preparation, chemical probe treatment, affinity enrichment, label-free LC-MS/MS analysis, and bioinformatics workflows used to generate and interpret dfPPI datasets. This article aims to promote reproducibility and accessibility of this approach, supporting its adoption by the broader systems biology and translational research communities.

Figure 1:. Overview of the dfPPI platform.

(A) The dfPPI (dysfunctional Protein-Protein Interactome) platform enables the identification of disease-specific PPI network dysfunctions through a chemoproteomic workflow. Section 1 involves sample preparation and interactome capture using epichaperome-targeting chemical probes immobilized on affinity matrices (e.g., PU-beads). Sections 2 and 3 outline two compatible digestion workflows-on-bead or in-gel-followed by label-free LC-MS/MS for interactome identification. Computational analyses, including statistical modeling and pathway enrichment, are performed using the dfPPI bioinformatics pipeline, which is not covered in this protocol but is freely available at https://epichaperomics.mskcc.org/. (B) Principle of dfPPI. In disease-relevant contexts, endogenous epichaperomes form supramolecular scaffolds that reorganize PPI networks. Epichaperome-directed chemical probes (e.g., PU-beads or equivalent) capture these scaffolds together with their bound interactors directly from intact cells or from native lysates, yielding a multi-bait pull-down in a single experiment. Matched controls (e.g., inert-bead control and/or competition with free probe) can be processed in parallel. Captured material is eluted and digested on-bead or in-gel, analyzed by LC-MS/MS, and proteins are quantified (intensity or MS/MS spectral counts). The resulting interactome is used to compute differential interactors across conditions and to perform pathway/network enrichment, revealing context-specific PPI dysfunction at the systems level. The workflow requires no genetic tagging or overexpression, is compatible with both cells and tissues, and scales to cohort studies. Conversely, traditional affinity-purification-MS centers on one tagged bait at a time, producing a local neighborhood per construct and typically requiring many constructs/runs to sample a network. In contrast, dfPPI uses endogenous epichaperomes as the bait, capturing many dysfunctional PPIs simultaneously from the native system, which accelerates the discovery of disease-specific network rewiring and provides pathway-level readouts from a single capture. Please click here to view a larger version of this figure.

Figure 2:. PU-bead lot validation and capture specificity.

(A) Biological specificity and lot-to-lot consistency. MDA-MB-468 (epichaperome-high) and ASPC1 (epichaperome-low) cells were lysed in native buffer (20 mM Tris, pH 7.4; 20 mM KCl; 5 mM MgCl₂; 0.01% NP-40; protease/phosphatase inhibitors). For each capture, 40 μL of PU-bead slurry was incubated with 250 μg of total protein (1 μg/μL) for 3 h at 4 °C with rotation. Beads were washed in native buffer; complexes were denatured/eluted in ~100 μL of SDS sample buffer. 5-10 μL of each eluate was resolved by SDS-PAGE and immunoblotted for HSP90, HSC70, and HOP. Input lysates are shown for reference. Two PU-bead lots (Batch 1, fresh; Batch 2, aged) yield strong enrichment in MDA-MB-468 and minimal signal in ASPC1, demonstrating preserved specificity and lot consistency. (B) Global cargo versus control beads. MDA-MB-468 lysates were processed as in (A) with either PU-beads or matched control beads. ~20 μL of each eluate was loaded, separated by SDS-PAGE, and Coomassie-stained. PU-beads recover a complex, high-MW cargo characteristic of epichaperome-bound assemblies, whereas control beads show minimal background. Molecular-weight markers (kDa) are indicated. Please click here to view a larger version of this figure.

Figure 3:. Representative results from dfPPI affinity capture, interactome identification, and bioinformatics analyses.

(A) Coomassie-stained SDS-PAGE gel of proteins eluted from PU-bead affinity captures of epichaperome-bound interactomes from four representative human brain samples (PT1-PT4). Bead-bound protein assemblies were denatured by boiling in SDS sample buffer, resolved by SDS-PAGE, and stained to assess interactome recovery across molecular weights. Each sample was loaded into a separate lane, with an empty lane between samples to minimize cross-contamination; a faint signal occasionally observed in these empty lanes reflects minor sample spillover during electrophoresis. These gels are representative of samples submitted to the MS facility, where each lane was sectioned into five slices for in-gel digestion and protein identification by MS. (B) Principal component analysis (PCA) of identified protein intensities from on-bead digested epichaperome interactomes from eight representative murine brain samples (run in duplicate) reveals tight clustering of technical replicates and clear biological separation along PC1 and PC2. This confirms high technical reproducibility and sufficient biological variance for robust dfPPI analysis. The data structure demonstrates that interaction-level differences-not technical artifacts-drive sample separation, providing a strong foundation for downstream differential interactome modeling without requiring artificial normalization. (C) Coefficient of variation (CV) distributions for precursor ion intensities across the eight sample groups show consistently low variance, with median CVs ranging from 9.7% to 11.9% (indicated by vertical dashed lines). These low CVs highlight the reproducibility and robustness of the MS detection and quantification pipeline. The consistency supports the use of raw intensity or MS/MS values as reliable inputs for dfPPI analysis, where variance in interaction strength-not global protein expression-is the biological signal of interest. (D) Hierarchical clustering heatmap of log₂-transformed protein intensity values across all samples (with technical duplicates shown). Samples cluster by biological origin rather than replicate, confirming robust capture of epichaperome-bound interactomes and preservation of biologically relevant variance. Broad dynamic range and consistent detection across proteins further support dataset quality for dfPPI analysis. Scale bar represents log₂-transformed protein intensity values, ranging from 2 (blue, low abundance) to 14 (yellow, high abundance). (E) Manhattan plot of pathway enrichment analysis performed on differentially interacting proteins identified by dfPPI. Enrichment was conducted using g:Profiler, and significant terms are plotted by −log₁₀(p-value), colored by annotation source: GO Molecular Function (GO:MF), GO Biological Process (GO:BP), GO Cellular Component (GO:CC), KEGG, Reactome, WikiPathways, and Human Phenotype Ontology (HP). This output represents the core utility of dfPPI: uncovering biologically meaningful functional pathways perturbed through epichaperome-driven interactome remodeling. The broad annotation coverage underscores the systems-level insight achievable through this approach. Please click here to view a larger version of this figure.