Deciphering intercellular signaling complexes by interaction-guided chemical proteomics

Cell–cell communication through secreted ligands and cell surface receptors orchestrates key processes including proliferation, migration, and differentiation. Dysregulated ligand–receptor interactions in the tumor microenvironment promote cancer progression and metastasis, motivating systematic methods to map these intercellular signaling complexes in relevant biological contexts. The central research objective is to develop a robust, unbiased, and sensitive strategy to identify ligand–receptor interactions in situ on living cells, including low-abundance ligands and receptors on primary or limited cell numbers. The study introduces an interaction-guided cross-linking (IGC) chemical proteomics platform to address the current gaps in profiling extracellular interactions with high specificity and sensitivity under physiological conditions.

Existing approaches for mapping human cell surface interactomes include ex situ techniques (e.g., yeast two-hybrid, affinity purification–mass spectrometry) that may capture biophysical interactions not occurring in native microenvironments and often under-represent extracellular interactions due to membrane protein hydrophobicity, glycosylation, detergent solubilization, and the transient/weak nature of many ligand–receptor contacts. In situ crosslinking MS can capture low-affinity interactions but suffers from low inter-protein crosslinking efficiency and high sample complexity, limiting sensitivity. Ligand-based receptor capture methods (TRICEPS, HATRIC) enable selective crosslinking on living cells but typically require tens of micrograms of purified ligand and large cell numbers, constraining use with primary cells. Proximity labeling strategies (PUP-IT, µMap, LUX-MS, PhoTag) map surface interactomes of a bait protein but are hypothesis-driven. These limitations underscore the need for a hypothesis-free, sensitive, and scalable method for unbiased discovery of extracellular ligand–receptor pairs in situ.

The authors developed interaction-guided crosslinking (IGC), comprising two complementary implementations: Photo-IGC and Click-IGC, using trifunctional probes (Probes 1–3) each bearing: (i) a ligand coupling moiety [NHS ester (Probe 1) or aminooxy group for oxime ligation (Probes 2–3)], (ii) a crosslinking group [diazirine (Probe 1–2) or alkyne (Probe 3)], and (iii) a biotin handle for enrichment. The spacer arm (~60 Å) is suitable for inter-protein crosslinking. Overall workflow: secreted ligands in conditioned media (CM) or purified ligands are conjugated to probes; conjugates bind receptors on living cells at 4 °C; crosslinking is induced (UV for diazirine or Cu(I)-catalyzed azide–alkyne cycloaddition, CuAAC, for clickable glycans); crosslinked complexes are enriched via streptavidin, digested, and analyzed by LC-MS/MS; quantitative analysis (LFQ/iBAQ) reveals enriched receptors.

• Ligand coupling: (a) NHS ester chemistry (Probe 1) for general protein labeling; optimal ligand:probe mass ratio was empirically set at 1:2 for both large (HGF) and small (EGF) ligands. (b) Glycan-based oxime ligation (Probes 2–3) after mild periodate oxidation of glycans to generate aldehydes; aniline catalysis (50 mM) at 37 °C for 1.5 h; removal of small molecules by ultrafiltration. Glycan-based labeling minimizes perturbation at protein–protein interfaces and favors low-abundance glycoproteins.
• Crosslinking on living cells: Following 10 min incubation with ligand–probe at 4 °C, Photo-IGC employs 365 nm UV irradiation for 5 min to generate short-lived carbene intermediates from diazirines, enabling broad insertion into proximal residues. Click-IGC uses metabolic azido-sugar labeling of cell surface glycans (best-performing Ac4ManNAz among Ac4ManNAz, AcGalNAz, AcGlcNAz) and CuAAC with Probe 3’s alkyne: CuSO4/THPTA/aminoguanidine/sodium ascorbate at final 50 µM/250 µM/1 mM/2.5 mM for 15 min at 4 °C. Flow cytometry on K562 validated efficient surface labeling at 50 µM Cu(I).
• Enrichment and sample prep: Lysis in Triton X-100 buffer; streptavidin pull-down (optimized at 1 µL beads per 10^6 cells); stringent washes; on-bead reduction/alkylation; trypsin/Lys-C digestion; peptide cleanup.
• LC-MS/MS: For purified-ligand IGC, EASY-nLC 1000 with Q Exactive; defined gradients and DDA settings (full scan 70k resolution, top 10 HCD at 17,500). More sensitive instruments used for secretome-based IGC (per Supplementary).
• Data analysis: MaxQuant (UniProt human, mouse, Vero databases; 1% FDR peptide/protein; carbamidomethyl C fixed; variable: Met oxidation, N-terminal acetylation, Asn/Gln deamidation); LFQ and iBAQ enabled; match-between-runs active. R-based processing: filtering contaminants/reverse/only-by-site, requiring ≥2 razor+unique peptides, defined handling for match-only with low intensities; normalization to endogenous biotinylated carboxylases for pull-downs; missing value imputation; differential analysis with limma (empirical Bayes, BH-adjusted p-values). Significance: adjusted p<0.05 and |log2FC|>1. Interaction scores computed as log2(riBAQ/riBAQref) using surfaceome/secretome references.
• Biological systems: HeLa, NIH 3T3, Vero E6 for benchmarking known ligand–receptor pairs; pancreatic cancer cells (MIA PaCa-2, PANC-1) and pancreatic stellate cells (PSCs; CAFs) for paracrine mapping; typical inputs ranged from 0.1–6 million cells and ligands at 10–1600 ng depending on assay.
• Validation assays: Western blotting for pathway activation; CCK-8 proliferation; siRNA knockdown (PLAU, NRP1); co-culture and CM transfer; solid-phase binding ELISA to determine PLAU–NRP1 affinity (Kd).

• Specificity and sensitivity of Photo-IGC: Using purified ligands on living cells, Photo-IGC correctly identified known receptors with high selectivity: EGF→EGFR on HeLa (from as little as 0.2 million cells using ~10–50 ng EGF), HGF→MET, insulin→INSR/IGFIR on HeLa, PDGF-B→PDGFRA/PDGFRB on NIH 3T3, and SARS-CoV-2 RBD→ACE2 on Vero E6. Minimal irrelevant proteins co-purified, demonstrating high specificity.
• Optimization parameters: Ligand:probe mass ratio of 1:2 (Probe 1) maximized receptor MS/MS counts for both HGF and EGF; streptavidin beads at 1 µL per 10^6 cells optimized enrichment.
• Limitation of NHS ester for scarce ligands: In Photo-IGC, spiking HGF with excess BSA showed MET identification diminished and was lost at HGF:BSA 1:250.
• Click-IGC overcomes low-abundance constraints: Glycan-based oxime ligation and CuAAC on Ac4ManNAz-labeled cells enabled identification of MET receptors even with HGF diluted 1:250–1:500 in BSA and with extremely low inputs: confident MET identification from 0.1 million HeLa cells using only 10 ng HGF with 1000-fold BSA background, highlighting exceptional sensitivity.
• Secretome-to-surfaceome mapping: Integrating Click-IGC with quantitative secretome and surfaceome profiling enabled hypothesis-free discovery of paracrine interactions. In PCC-to-PSC mapping, 36 secreted proteins from MIA PaCa-2 and PANC-1 CM were captured on PSCs, and 6 putative receptors were commonly identified on PSCs, including CAF markers PDGFRA and PDGFRB. In the PSC-to-PCC direction, 40 putative ligands and 6 putative receptors were identified.
• Discovery of a novel ligand–receptor pair: PLAU (urokinase), strongly enriched and highest-scoring PCC-derived ligand interacting with PSCs, was identified. Click-/Photo-IGC and quantitative analysis implicated NRP1 and LRP1 as receptors; LRP1 is a known interactor for PLAU and PLAU–SERPINE1 complexes. Photo-IGC with PLAU on PSCs highlighted NRP1 (and after PNGase F treatment as control). Solid-phase binding assays measured PLAU–NRP1 affinity with Kd ≈ 9.5 nM (R^2=0.993).
• Functional validation: Recombinant PLAU (50 ng/mL) promoted PSC proliferation (CCK-8) and activated AKT and ERK signaling in a dose-dependent manner (10–100 ng/mL). Silencing PLAU in PCCs reduced PSC proliferation in co-culture and with CM. NRP1 knockdown in PSCs attenuated PLAU-induced ERK activation and proliferation, supporting a PLAU–NRP1 signaling axis in PCC-to-PSC paracrine communication.
• General performance: The IGC platform enables unbiased, all-to-all capture of secretome–surfaceome interactions in situ on living cells with high selectivity, using low ligand amounts and small cell numbers, suitable for limited primary cells.

The study addresses the challenge of mapping biologically relevant extracellular ligand–receptor interactions by establishing the IGC platform, which crosslinks interactions in situ on living cells and enriches the complexes for MS identification. Photo-IGC demonstrated high specificity in identifying canonical receptor–ligand pairs across multiple cell types, validating the approach. Click-IGC, leveraging glycan-based ligand conjugation and metabolic glycan labeling with CuAAC, markedly increased sensitivity and preserved interaction integrity, enabling discovery even for low-abundance ligands amidst high protein background and with minimal cell inputs. Integrating IGC with secretome/surfaceome quantification provided a systems-level view of paracrine signaling between pancreatic cancer cells and CAFs. This revealed a rich landscape of secreted ligands but comparatively fewer cell surface receptors captured, reflecting biological and technical factors in receptor detectability. Critically, the platform led to the identification and validation of a previously unrecognized ligand–receptor interaction between PLAU and NRP1 on PSCs. Biochemical binding (Kd ~9.5 nM), signaling activation (AKT/ERK), and genetic perturbations (siNRP1) collectively support functional relevance, illustrating how hypothesis-free IGC can uncover key intercellular signaling axes in the tumor microenvironment. Overall, IGC complements and extends existing ex situ and proximity labeling methods by offering unbiased, sensitive, native-context discovery of secreted ligand–surface receptor complexes.

This work introduces interaction-guided crosslinking (IGC)—comprising Photo-IGC and Click-IGC—as a hypothesis-free chemical proteomics strategy to systematically decipher extracellular signaling complexes in situ. The approach combines efficient ligand conjugation (including glycan-based oxime ligation), precise in situ crosslinking (UV or CuAAC), and sensitive MS-based identification to reveal ligand–receptor pairs with high specificity from small sample inputs. Benchmarking against known interactions validated performance, and application to pancreatic cancer–CAF communication uncovered and validated a novel PLAU–NRP1 axis promoting PSC proliferation via ERK/AKT signaling. IGC opens avenues for comprehensive mapping of intercellular signaling networks, with potential applications in primary cells, disease microenvironments, and drug target discovery. Future work could extend to broader primary tissue contexts, refine quantitative interaction scoring, and systematically validate additional inferred ligand–receptor pairs across diverse biological systems.

The study notes that NHS ester-based ligand labeling (Photo-IGC) is limited for low-abundance ligands, as receptor identification failed when HGF was diluted to a 1:250 ratio with BSA. Although Click-IGC improves sensitivity markedly, paracrine analyses still identified a limited number of receptors relative to secreted ligands. Experimental validation focused on selected interactions (notably PLAU–NRP1, with additional LIF signaling evidence), so many inferred ligand–receptor pairs remain to be biochemically validated. Moreover, some workflows (e.g., Click-IGC) rely on metabolic glycan labeling and copper-catalyzed click reactions, which may not be uniformly applicable across all cell types or conditions.