Investigators queried five complementary resources to assemble a comprehensive target list for Formononetin. TCMSP, CTD, HERB, SwissTarget and PharmMapper collectively yielded 413 unique predicted targets after deduplication. Of these, 335 overlapped with prostate cancer–associated genes curated from GeneCards and OMIM. The authors reported that three genes were predicted by all five platforms, signaling particularly robust candidacy, though the paper did not enumerate those three individually in the main text. Differential expression analysis of TCGA_PRAD (483 tumors versus 51 normals) identified thousands of DEGs, which supplied a disease-relevant filter for the compound-target universe. The combined approach produced a prioritized list of Formononetin-related genes for downstream network and functional analysis.
The study team applied weighted gene co-expression network analysis (WGCNA) to DKFZ_2018 transcriptomes, selecting a soft-threshold power that balanced scale-free topology and connectivity, and partitioning 11,647 filtered genes into co-expression modules. Modules with module–BCR correlation coefficients |r| > 0.3 were extracted, yielding 1,324 module-derived genes that intersected with the Formononetin target set. Through differential expression and module membership filtering, investigators narrowed candidates to 22 genes. Three orthogonal machine-learning strategies—LASSO Cox, random survival forest, and SVM-RFE—were then applied. Across these methods and subsequent survival analysis, GSTM2 emerged as the most important candidate. Data indicated that higher GSTM2 expression associated with increased immune infiltration metrics (by estimate and ssGSEA methods) and with predicted sensitivity to Bicalutamide using oncoPredict and GDSC2 training data.
To probe direct binding, the authors used the GSTM2 crystal structure (PDB ID: 3GUR) for docking with Formononetin. AutoDock Vina produced a lowest-energy pose which was then subjected to molecular dynamics using Amber99SB-ILDN (protein) and GAFF2 (ligand) force fields. The simulation protocol included system neutralization in 0.15 M NaCl, 1 ns NVT and 1 ns NPT equilibration, and a 100-ns production run. Post-simulation analyses reported by the team included RMSD, RMSF, radius of gyration (Rg), solvent-accessible surface area (SASA) and free energy landscape (FEL) metrics; investigators concluded the complex formed stably under the simulated conditions. The article describes stable complex formation but did not report experimental biophysical binding constants (for example, KD) from orthogonal assays.
Functional validation used human prostate cancer cell lines LNCaP and VCaP. Investigators treated cells with Formononetin and measured proliferation (CCK-8), colony formation, migration and invasion (Transwell with and without Matrigel). They also silenced GSTM2 using two siRNA duplexes to test dependency. The study team observed that Formononetin increased GSTM2 expression and that the anti-proliferative, anti-clonogenic and anti-migratory/invasive effects were lost or attenuated when GSTM2 was knocked down, indicating a GSTM2-required mechanism in these in vitro assays. All in vitro results are presented as mean ± SD and were analyzed using one-way ANOVA with Bonferroni-corrected comparisons according to the methods described.
Mechanistic assays linked GSTM2 activity to redox and STAT signaling changes. Intracellular reactive oxygen species were measured by flow cytometry using DCFH-DA; investigators reported that Formononetin reduced ROS levels in a GSTM2-dependent manner. Western blotting probed total and phosphorylated STAT proteins, and the team reported suppression of STAT3 phosphorylation (Tyr705) following Formononetin treatment that depended on GSTM2 expression. Downstream chemokine readouts included qRT-PCR and ELISA measurements: conditioned media assayed for CXCL10 and CCL22 showed that Formononetin treatment led to increased CXCL10 and decreased CCL22 secretion, consistent with a shift toward chemokines that recruit cytotoxic lymphocytes rather than regulatory T cells. Single-cell transcriptomic analysis (GSE157703) further supported these findings: GSTM2-positive malignant cells displayed enrichment of glutathione metabolism pathways and suppression of respiratory-burst signatures, consistent with enhanced antioxidant capacity that could blunt ROS-driven STAT3 activation.
The investigators did not report any in vivo efficacy or pharmacokinetic data for Formononetin in prostate cancer models in this article. The structural docking and MD support stable binding but lack experimentally measured binding affinities. Immune microenvironment modulation is inferred from chemokine changes, immune-infiltration correlations and single-cell program shifts; direct demonstration of altered immune cell recruitment in vivo was not provided. The authors noted that in vivo validation is necessary. Additional unanswered questions include dose–response relationships in vivo, safety and tolerability profiles, and whether the observed GSTM2-dependent chemokine changes translate to durable anti-tumor immune responses in animal models or clinical samples.