Multi-omics convergence screening reveals dual-cargo antitumor mechanisms of orally administered Houttuynia cordata-derived exosome-like nanoparticles

Abstract

Plant-derived exosome-like nanoparticles (PELNs) transport multi-component molecular cargos with cross-kingdom regulatory potential and have emerged as a promising class of cancer nanomedicines. A persistent bottleneck, however, is causally linking individual cargo species to in vivo antitumor mechanisms, because the simultaneous presence of proteins, miRNAs, and metabolites obscures which cargo drives which downstream phenotype. We isolated Houttuynia cordata aerial-herbage exosome-like nanoparticles (aHELNs) and underground-root nanoparticles (uHELNs) by differential and ultracentrifugation. We evaluated aHELN antitumor activity in A549, H1975, and H520 non-small cell lung cancer (NSCLC) cell lines and in a subcutaneous A549 xenograft model by oral gavage (10, 20, or 30 mg protein/kg every other day for 18 days). Cargo was profiled by label-free proteomics, small RNA sequencing, and headspace SPME-GC-MS metabolomics, and tumor tissue was profiled by 4D-DIA proteomics. We developed Multi-omics Convergence Screening (MOCS), a transparent, rule-based prioritization strategy that anchors predicted miRNA- and metabolite-target relationships to in vivo tumor differentially expressed proteins from the same treatment cohort. Selected MOCS nominations were evaluated by dual-luciferase reporter assays, an 8-gene RT-qPCR panel including a MOCS-negative control (PRKAR1B), molecular docking, molecular dynamics (MD), a drug affinity-responsive target stability (DARTS) assay, and Western blotting, and were anchored to The Cancer Genome Atlas lung adenocarcinoma cohort (TCGA-LUAD; n = 517 tumor, 58 normal, 57 matched pairs) and MSigDB Hallmark gene-set enrichment analysis (GSEA). aHELNs produced dose-dependent tumor suppression without overt organ toxicity. MOCS resolved two complementary candidate modules: a miRNA-dominant module comprising TRPC7, ABL2, SENP1, and CXCL14 and a metabolite-associated module comprising SHBG, ADH1C, and FABP3. Dual-luciferase reporters confirmed seed-dependent 3′UTR suppression for hco-miR166e-3p targeting TRPC7 and for hco-miR447a-3p targeting ABL2 (WT+mimic vs. WT + NC fold = 0.73, t = -16.51, P < .001, Cohen’s d = -10.44). An 8-gene RT-qPCR panel supported MOCS-predicted transcriptional directionality; SENP1, CXCL14, ADH1C, and FABP3 remained expression-supported exploratory candidates, while the MOCS-negative control PRKAR1B remained unchanged ( F = 0.83, P = .512). Computational modeling placed β-caryophyllene within the SHBG ligand pocket (docking energy = −8.526 kcal/mol; stable MD RMSD), while DARTS showed concentration-dependent protection of SHBG from proteinase K digestion and Western blotting confirmed concentration-dependent SHBG accumulation. In TCGA-LUAD, TRPC7 and SHBG were both reduced in tumor versus normal (Mann-Whitney P < .001 for both); a combined logistic classifier achieved AUC = 0.774 (95% CI 0.718 to 0.826; DeLong bootstrap n = 2000). GSEA linked TRPC7 to proliferative Hallmark programs (G2M, E2F, Myc) and SHBG to metabolism-related programs (oxidative phosphorylation, bile-acid, fatty-acid metabolism). MOCS rankings were stable across a 27-combination parameter-perturbation grid (top-5 Jaccard median = 0.750; SHBG recovered in 27 of 27 combinations). aHELNs are orally deliverable, multi-component anti-NSCLC nanoparticles for which MOCS prioritizes a dual miRNA- and metabolite-associated candidate architecture. MOCS is a transparent, reproducibility-first framework for prioritizing cargo–target relationships in medicinal nanoparticles.