SNAIL2驱动杂合上皮-间质转录网络，揭示HNSCC中选择性、可成药的抗侵袭易损性

In nearly all solid tumors, metastatic progression remains the major barrier to long-term disease control and is the most common reason for death. Current anticancer agents largely target proliferation, causing toxicity in normal proliferating cells and limiting their use as long-term adjuvant therapy to suppress metastatic dissemination. Notably, no FDA-approved therapy selectively inhibits invasion without impairing normal cell viability, underscoring a major gap in metastasis-directed treatment. In head and neck squamous cell carcinoma (HNSCC), invasion and metastasis are driven by hybrid epithelial-mesenchymal (E/M) states, yet the upstream regulatory circuitry governing these states remains poorly defined and therapeutically unaddressed. Here, through comprehensive correlative and loss-of-function analyses of the six canonical EMT transcription factors (TFs), we identify SNAI2 (encoding Snail2) as the dominant EMT-TF regulating hybrid-E/M states in HNSCC. Functional assays across multiple models demonstrate that SNAI2 is both necessary and sufficient to drive invasion and metastasis. We mapped genome-wide binding sites of Snail2 (via Calling Cards) and integrated this with RNA-seq to uncover the first genome scale direct Snail2 regulon in hybrid-E/M cells. To determine which genes controlled by Snail2 act as effectors for its metastatic phenotypes, we performed a large-scale functional screen to identify invasion-specific vulnerabilities and nominate druggable nodes. Remarkably, of the 48 direct SNAI2 targets we analyzed, 21 (43 percent) function as bona fide invasion drivers in multiple lines establishing the first systematic delineation of the upstream transcriptional architecture governing the hybrid-E/M invasive phenotype in HNSCC. Importantly, we identify two key downstream effectors, NID1 (basement membrane glycoprotein) and BNC2 (zinc-finger transcription factor) as actionable therapeutic targets. Pharmacologic inhibition of NID1 and BNC2 markedly reduces invasion in vitro and suppresses metastasis in vivo at doses ~10-fold lower than those associated with toxicity. Notably, BNC2 activates SNAI2 expression, forming a reinforcing SNAI2-BNC2 positive feedback transcriptional circuit that sustains the hybrid-E/M invasive program. Together, these findings position SNAI2 at the apex of the hybrid-E/M transcriptional network in HNSCC and expose selective, drug-vulnerable anti-metastatic nodes with high translational potential. This work establishes a blueprint for developing the first generation of long-term low-toxicity anti-metastatic therapies deployable in the adjuvant setting to prevent metastatic progression.