Two-module architecture of ribosome-associated quality control in human brain: divergent co-expression networks of PELO, LTN1, and NEMF

Background Ribosome-associated quality control (RQC) resolves stalled translational complexes through sequential ribosome rescue (PELO), nascent chain ubiquitination (LTN1), and C-terminal alanine–threonine tail modification (NEMF). While the biochemical sequence of these events is well characterized, the broader transcriptional coordination of individual RQC components with other cellular systems in the human brain remains unexplored. Methods We performed genome-wide Pearson co-expression analysis for PELO, LTN1, and NEMF across 16,225 expressed genes in the GTEx v8 dataset, using Brain–Frontal Cortex (BA9; n = 209) as the primary region. The top 5% of co-expressed genes for each target were compared using Jaccard similarity, Fisher’s exact test, and Spearman rank correlation, with permutation-based null models (10,000 random gene pairs) and bootstrap confidence intervals (1,000 resamples) providing formal statistical assessment. Gene Ontology enrichment was performed via gProfiler. Multi-region replication was assessed across four additional brain regions, with cell-type deconvolution, covariate adjustment, and agonal stress sensitivity analysis as robustness checks. Results LTN1 and NEMF shared near-identical co-expression networks (Jaccard = 0.461, 95% CI [0.357, 0.546]; Spearman ρ = 0.986; permutation p = 0.006), while PELO diverged substantially (PELO–LTN1 J = 0.165, 95% CI [0.095, 0.242]; PELO–NEMF J = 0.156, 95% CI [0.082, 0.227]; permutation p > 0.10 for both). Bootstrap confidence intervals confirmed non-overlapping separation between the LTN1–NEMF module and PELO pairs. The two-module partition was robust across network thresholds from 1% to 20% (partition ratio 1.64×–8.78×). Genes unique to PELO (528 genes) were enriched for protein metabolic process (p = 9.35 × 10 ^− 22 ) and RNA metabolism (Reactome p = 1.58 × 10 ^− 20 ). Genes unique to NEMF (246 genes) were enriched for histone modifying activity (p = 2.16 × 10 ^− 5 ). Multi-region replication confirmed the two-module architecture (median cross-region ρ = 0.89–0.90), with PELO exhibiting uniformly high cross-region consistency (10/10 pairs > 0.80) and the LTN1–NEMF module showing hippocampus-specific divergence (all six sub-0.80 pairs involved hippocampus). All patterns were robust to agonal stress adjustment (rank preservation ρ ≥ 0.998). Conclusions RQC in human brain operates as a two-module system: a surveillance–mRNA decay module (PELO) is transcriptionally distinct from a nascent chain processing module (LTN1–NEMF). The hippocampus-specific divergence of the processing module suggests region-dependent transcriptional tuning of nascent chain quality control.