Haplotype-rich cis-regulation underlies transcriptomic diversity across the breeding history of maize (Zea mays)

Abstract

The architecture of gene-expression variation and its relationship to population differentiation remain incompletely characterized in crops. Here, we integrated leaf RNA-seq and genome-wide variation from 631 maize inbred lines to characterize transcriptomic diversity and fine-map cis-regulatory signals. Subpopulation identity explained a mean of 5.3\% of expression variation across genes, indicating that most expression diversity segregates within rather than among breeding populations. We identified 17,468 genes with cis-eQTLs and resolved putatively causal eQTLs for 16,270 of them. Most fine-mapped eGenes (77.7%) were haplotype-rich (at least two credible sets), and approximately 39% of credible sets contained a single variant. Modeling all credible sets increased median R2 from 0.16 to 0.27. Individual effects were generally modest, and genes with more signals tended to have higher expression heritability. Fine-mapped variants were enriched in transcribed regions, accessible chromatin, and transcription-factor-binding regions, with the latter enrichment persisting after accessibility matching. Genes with high gene-body sequence constraint had modestly smaller regulatory effects and fewer credible sets. Across five population contrasts, including Stiff Stalk--Non-Stiff Stalk comparisons within two breeding eras, genes with stronger statistical evidence of differential expression were associated with higher Fst at fine-mapped variants. These results indicate that single-variant summaries provide an incomplete picture of cis-regulatory loci, and offer a mechanism for why transcriptome-wide association studies identify genes missed by genome-wide association studies in the same populations.