A Nine-Category Hepatic Transcriptomic Classification Framework Identifies Coordinated Cholesterol Biosynthesis Suppression and Fructose-Specific Regulatory Variants in Dietary NAFLD

Dietary fructose synergizes with high-fat feeding to accelerate hepatic steatosis and injury, yet the transcriptomic and genomic mechanisms that distinguish fructose-specific from fat-driven regulation remain incompletely defined. Here we present a three-layer integrative analysis of hepatic gene expression and expressed sequence variants (ESVs) in male C57BL/6N mice fed a chow diet (CD), high-fat Western diet (HFD), or HFD supplemented with liquid fructose/sucrose (HFSD) for 12 weeks, using RNA-seq data from GEO accession GSE89296. To systematically dissect the contribution of each dietary component, we developed and applied a nine-category classification framework that partitions differentially expressed genes (DEGs) across three pairwise comparisons (HFD vs. CD, HFSD vs. CD, HFSD vs. HFD) into biologically interpretable regulatory programs. Analysis of 16,626 expressed genes using Salmon/DESeq2 (GRCm39/GENCODE M38) yielded 716 classified DEGs; 209 (29.2%) were fructose-specific, absent in HFD alone. The most striking finding was a coordinated, pathway-wide suppression of the entire cholesterol biosynthesis cascade across all 16 mevalonate-pathway enzymes-a novel result for this model not reported in the original study-with KEGG fold enrichment of 35.6x (padj = 7.0x10⁻ ²³ ). Fructose further amplified glutathione S-transferase induction above fat-driven levels (Gsta1 Δlog₂FC = + 1.26; Gsta13 Δlog₂FC = + 1.34), and specifically suppressed the glycolytic checkpoint gene Pfkfb3 (HFSD log₂FC = − 1.67, padj = 1.3x10⁻⁹ vs. HFD log₂FC = − 0.48, ns). In parallel, joint variant calling with STAR/FreeBayes across all 14 samples identified 21,125 quality-filtered variants, 29 of which exhibited a strict HFSD-specific allele frequency gradient. In silico functional prediction via AlphaGenome (Google DeepMind, 2025), cross-validated against ENCODE4 liver epigenomic tracks, demonstrated 78% concordance (7/9 loci) including a confirmed non-regulatory negative control. Three high-confidence regulatory candidates were mechanistically characterized: an insertion near Rpl3l/Msrb1 predicted to disrupt a liver CTCF architectural domain (RNA log₂FC = + 0.942; CTCF signal = 150.8); a splicing-disrupting insertion within the Ddx19b promoter CpG island (ΔΨ = 0.734) coinciding with the sole HFSD-specific DESeq2-significant variant locus (log₂FC = − 0.39, padj = 0.035); and a chromatin-remodeling insertion near Trp53i11 at an ENCODE4-confirmed cis-regulatory element (ATAC = 59.6). Together, these results demonstrate that the nine-category framework resolves biologically distinct regulatory programs obscured by conventional DEG lists, and that ESV analysis of existing RNA-seq data can pinpoint diet-responsive regulatory loci with high precision. The analytical pipeline described here provides a replicable template for extracting transcriptomic and regulatory variant landscapes from legacy datasets.