Differential transcriptomic modulation by histone deacetylase inhibitor SAHA in LUAD and LUSC.

BACKGROUND

Histone deacetylases (HDACs) are central epigenetic regulators in non-small cell lung cancer (NSCLC), yet responses to HDAC inhibitors (HDACi) vary markedly between lung adenocarcinoma (LUAD) and lung squamous carcinoma (LUSC). We asked how the pan-HDAC inhibitor suberoylanilide hydroxamic acid (SAHA, vorinostat) rewires lineage-specific transcriptional programs and whether SAHA-aligned modules of genes, rather than individual loci, capture clinically relevant vulnerabilities in each subtype.

METHODS

LUAD-like NCI-H1299 (TP53del, NRASQ61K) and LUSC-like NCI-H1703 (TP53WT, PDGFRAamp, PIK3CAE542K) cells were treated with SAHA (10 µM, 24 h) or DMSO. Bulk RNA-seq data were analysed with edgeR (FDR < 0.05, |log2FC|> 1), followed by GO/Reactome over-representation, Hallmark GSEA, and STRING-based protein-protein interaction mapping. We quantified apoptosis (Annexin V/PI) and motility (scratch assays under mitomycin C). SAHA "feature-sensing" modules were constructed by intersecting SAHA-responsive DEGs with overall-survival-associated genes from GEPIA2 and were scored in 592 LUAD and 551 LUSC tumours. Correlations between HDAC isoforms and module scores were used to define subtype-biased HDAC-module neighbourhoods.

RESULTS

SAHA reprogrammed the transcriptome in both lines (1,098 DEGs in H1299; 1,532 in H1703), enforcing a shared suppression of E2F/G2-M programs but diverging in non-cell-cycle outputs. In LUAD-like H1299, SAHA upregulated morphogenesis/adhesion and KRAS_SIGNALING_DN/EMT-adjacent signatures while dampening interferon/stress pathways, and significantly reduced migration at low dose. In LUSC-like H1703, SAHA triggered a dominant cell-cycle checkpoint shutdown coupled to complement/ECM and inflammatory induction, with stronger apoptosis but only modest short-term migration restraint. Survival-anchored analysis yielded four SAHA feature-sensing modules; the LUAD_RISK module was enriched for cell-cycle/mitotic genes and was attenuated by SAHA, whereas the LUSC_RISK module captured checkpoint, ECM, and stress-response programs preferentially down-regulated in H1703. Both risk modules stratified prognosis and were linked to distinct HDAC-centred neighbourhoods (HDAC7/9-LUAD_RISK and HDAC4/6-LUSC_RISK) in TCGA tumours.

CONCLUSIONS

SAHA imposes a common anti-proliferative core but engages distinct lineage-conditioned risk modules in LUAD and LUSC-cell-cycle/migration-linked in LUAD and checkpoint/stress-linked in LUSC. These SAHA feature-sensing modules provide a mechanistic and clinically anchored framework for subtype-tailored HDAC-directed combinations and for future development of HDACi-aligned biomarkers in NSCLC.