Mapping Extracellular Protein-Protein Interactions Using Extracellular Proximity Labeling (ePL)

Abstract

Proximity labeling (PL) has given researchers the tools to explore protein-protein interactions (PPIs) in living systems; however, most PL studies are performed on intracellular targets. We have adapted the original PL method to investigate PPIs within the extracellular compartment, which we term extracellular PL (ePL). To demonstrate the utility of this modified technique, we investigated the interactome of the matrisome protein TIMP2. TIMPs are a family of multifunctional proteins that were initially defined by their ability to inhibit metalloproteinases, the major mediators of extracellular matrix (ECM) turnover. TIMP2 exhibits broad expression and is often abundant in both normal and diseased tissues. Understanding the functional transformation of matrisome regulators, such as TIMP2, during disease progression is essential for the development of ECM-targeted therapeutics. Using dual orientation fusion proteins of TIMP2 with BioID2/TurboID, we describe the TIMP2 proximal interactome (MassIVE MSV000095637). We also illustrate how the TIMP2 interactome changes in the presence of different stimuli, in different cell types, in unique culture conditions (2D vs 3D), and with different reaction kinetics, demonstrating the power of this technique versus classical PPI methods. We propose that screening of matrisome targets in disease models using ePL will reveal new therapeutic targets for further comprehensive studies.

Keywords: interactomics; matrisome; proximity labeling.

Figure 1

Expression, secretion, and function of biotin ligase fusion proteins. (A) Structure of E. coli BirA (parental TurboID protein), A. aeolicus BirA (parental BioID2 protein), and the AlphaFold predicted structure of TIMP2. (B) Basic construct design and proposed structure of one fusion protein. (C) Immunoblots assessing the expression and secretion of fusion proteins in HT1080 cells. (D) Reverse zymography (10% acrylamide) of conditioned media from fusion protein expressing cells revealing retention of MP-inhibitory activity in the fusion orientation that maintains a free TIMP2 N-terminus, which is required for MP inhibition. Endogenous/wild-type TIMP2 (22 kDa) is indicated by the green arrow. (E) Streptavidin pulldown and blotting of SDS-PAGE nitrocellulose blots with streptavidin-HRP reveals unique patterns of biotinylated proteins in fusion protein expressing samples.

Figure 2

Extracellular proximity (ePL) labeling to identify the proximal interactome of TIMP2 in HT1080 cells. (A) Schematic describing the basic workflow of ePL reactions in HT1080 cells. (B) Data acquired through LC-MS/MS was scored and proximal interactors identified using a defined scoring system. (C) Proximal interactors were uploaded into the Contaminant Repository for Affinity Purification-mass spectrometry (CRAPome) and persistent contaminants were removed based on a defined threshold (identified in >99 control experiments). Duplicate experiment identified proximal interactors are illustrated in the quadrant based on the biotin ligase utilized and the fusion protein orientation. (D) Crystal structure of the TIMP2:proMMP2 complex (Protein Data Bank #1GXD) reveals a strong C-terminal interaction (metalloproteinase inhibition independent) that is (E) significantly perturbed in the presence of a C-terminal fused biotin ligase. (A+C) Created with BioRender.com.

Figure 3

Cellular treatments can reveal proximal interactome dynamics. (A) Ingenuity Pathway Analysis canonical pathways and upstream regulators of transcriptome data from HT1080 cells treated with 40nM PMA or 40 μg/mL ConA for 24 h. (B) Venn diagram comparing the transcriptome changes in PMA versus ConA treated HT1080 cells. (C) Transcript abundance changes in defined TIMP2 interactors or previously identified proximal interactors from PMA and ConA treated HT1080 cells. (D) Quantified ePL data using TurboID fusions with TIMP2 acquired through LC-MS/MS that was processed, scored, and proximal interactors identified using a defined system. (E) Tables illustrating the identified TIMP2 proximal interactors (TurboID) in HT1080 cells treated with PMA or ConA. Gained or lost proximal interactors (versus unstimulated cells in analogous TurboID experiments) are highlighted in magenta (lost) or cyan (gained), created with BioRender.com. (F) Predicted structure of TIMP2 interacting with full length active MMP14, created by superimposing the AlphaFold MMP14 structural prediction over the TIMP2-MMP14 (catalytic domain) crystal structure (Protein Data Bank # 1BQQ). The structural domains of MMP14 are color-coded, with disordered linker regions colored gray. (G) Comparison of the normalized abundance of MMP14 across HT1080 TurboID ePL assays that were untreated or treated with ConA or PMA.

Figure 4

ePL can be performed across multiple live in vitro systems. (A) Basic workflow of HT1080 3D spheroid ePL experiments, created with BioRender.com. (B) Key illustrating the color-coded nodes in subsequent figure panels. (C) Summary of TIMP2 proximal interactors (BioID2) identified in 3D spheroid cultures, compared with previous BioID2 ePL experiments. (D) Overall summary of the TIMP2 proximal interactome in HT1080 cells, a total of 5 experiments each performed in duplicate with both fusion orientations using both TurboID and BioID2. (E) STRING analysis of the HT1080 proximal interactome for TIMP2 reveals potential interactome neighborhoods using previously reported direct protein–protein interactions and by incorporating additional potential interactions through STRING’s text mining feature. (F) TIMP2 TurboID ePL experiments in HS-5 bone marrow stromal cells reveals duplicate identified novel proximal interactors and shared interactors compared with the TIMP2 proximal interactome in HT1080 cells.

Figure 5

Proximity ligation assay (PLA) corroborates the identification of CCN1/CCN2 as candidate proximal interactors for TIMP2. (A) Normalized quantification, (B) example control images, and (C) PLA images indicating a high abundance of PLA foci between TIMP2 coimmunostained with MMP14, CCN1, and CCN2. Co-staining between TIMP2 and LOXL2 required a high exposure to observe a surplus of PLA foci versus controls (p = 0.057). Arrows indicate regions of interest (not quantification). * = p < 0.05, Two-tailed Mann–Whitney test. Scale bar = 10 μm.

Figure 6

CRAPome filtering can cause data loss through false-negative reporting.