Multi-locus genome-wide association study reveals the genetic architecture of kernel fat content in maize

Kernel fat content (KFC) is a key determinant of oil quality and nutritional value in maize, yet the genetic architecture underlying its natural variation remains incompletely understood. Here, we performed a genome-wide association study (GWAS) using 239 maize inbred lines genotyped with 1.25 million single nucleotide polymorphisms (SNPs), applying six complementary statistical models: MLM, MLMM, FarmCPU, BLINK, SUPER, and 3VmrMLM. KFC exhibited substantial phenotypic variation across three environments, ranging from 5.48% to 8.72%, with a broad-sense heritability of 0.81. In total, 200 significant quantitative trait nucleotides (QTNs) were detected, each explaining 0.53% to 23.49% of the phenotypic variance. Twenty QTNs were consistently identified across multiple models and/or environments. Within the co-localized QTL intervals, 58 candidate genes were annotated. Gene Ontology and KEGG enrichment analyses revealed significant enrichment in catalytic activities and metabolic pathways, notably α-linolenic acid metabolism—a core pathway in fatty acid biosynthesis. Based on multi-model co-localization and pathway evidence, Zm00001d025166 , encoding a plastid-localized putative quinone-oxidoreductase, was prioritized as the primary candidate gene. Haplotype analysis identified two major haplotypes, with Hap1 conferring modestly but significantly higher KFC than Hap2 (7.09% vs. 6.46%, p =  0.003). Selective sweep analysis detected moderate selection pressure on this locus in U.S. breeding germplasm compared with Chinese and CIMMYT accessions, while weak differentiation between tropical and temperate ecotypes suggested a conserved role in fatty acid metabolism. These findings provide a foundation for marker-assisted selection in high-oil maize breeding and demonstrate the value of multi-model GWAS strategies for dissecting complex quantitative traits.