Domestication and improvement have led to dramatic changes in the plant architecture and inflorescence in maize. These transformations were achieved by selecting desirable alleles for target traits to meet human needs and local environment adaptation, such as grain yield and flowering time (Doebley et al., 2006). Indeed, the selection of KRN2 and KRN4 genes contributed to the enlargement of ear size and the selection of ZCN8, ZmCCT9 and ZmCCT10 genes promoted the expansion of maize from tropical to temperate regions by accelerating flowering (Chen et al., 2022; Guo et al., 2018; Liu et al., 2015).

Our previous study revealed that the KERNEL NUMBER PER ROW6 (KNR6) gene regulated ear size variations (Jia et al., 2020). The alleles with insertions of two linked transposable elements (TEs) in the promoter and 5′ untranslated region (5′UTR) suppressed the expression of KNR6 and decreased ear size, indicating that the TE-insertion allele was undesirable for ear size. Phenotyping of two independent RNA-interference (RNAi) lines and two overexpression (OE) lines showed earlier flowering in the KNR6^RNAi lines (average 2.3–3.4 days earlier in days-to-tasseling (DTT) and days-to-silking (DTS) (Figure 1a,c,d) and later flowering in KNR6-OE lines (average 1.8–3.7 days later in DTT and DTS) than in the non-transgenic lines (Figure 1b,e,f). We next found that the near-isogenic line (NIL) with TE-insertion allele promoted flowering by 1.6 days in DTT and 1.5 days in DTS compared with the NIL without TE-insertion allele (Figure 1g,h). The marker–trait association showed the presence/absence variation (PAV) of TE in 5′UTR rather than the structural variation around the TE in the promoter was significantly associated with flowering time (Figure S1), and hereafter, the alleles with and without 5′UTR-TE were referred to the KNR6^TE+ and KNR6^TE- alleles, respectively. Subsequently, we tested the allelic effects in a Chinese widely grown maize hybrid, Zheng58 (ZH58)/Chang7-2 (C7-2). When the KNR6^TE+ allele was substituted by the KNR6^TE− allele, the improved hybrids with heterozygous alleles (ZH58/iC7-2, iZH58/C7-2) or homozygous KNR6^TE− alleles (iZH58/iC7-2) showed 1.9–3.8 days later in flowering time than the original ZH58/C7-2 hybrid (Figure 1i,j), but a 3.3%–5.6% increase in grain yield production (Figure 5a–g of Jia et al., 2020). These findings indicate that KNR6 affects both ear size and flowering time, and the long-ear KNR6^TE− allele shows a delayed flowering time. Thus, managing the KNR6 trade-off between flowering time and grain yield would be an efficient way to breed elite lines with high grain yield and appropriate flowering time.

We next genotyped the 5’UTR-TE PAV in 189 teosinte accessions, 275 tropical/subtropical and 357 temperate maize inbreds (Data S1). The KNR6^TE+ allele was not found in teosintes and its frequency in tropical/subtropical and temperate germplasms was 2.4% and 18.5% respectively (Figure 1k). Nucleotide diversity analysis in teosintes (n = 43) and maize inbreds (n = 275) showed a strong selection signal in the surrounding region of the 5′UTR-TE PAV (Figure 1l). Both KNR6^TE+ and KNR6^TE− alleles retained only 3.8% and 13.0% of the nucleotide diversity from teosinte to maize, respectively (Figure 1l), indicating both alleles were selected during maize evolution. Additionally, the KNR6^TE+ frequency in temperate inbreds was higher than that in tropical/subtropical inbreds, indicating that the KNR6^TE+ allele might contribute to maize adaptation to temperate regions. Similar results were observed by the geographical distribution of the KNR6^TE+/KNR6^TE− alleles in 470 landraces, showing that landraces with KNR6^TE+ allele were primarily located in the northern United States with high latitudes (Figure 1m), suggesting that the KNR6^TE+ allele might be positively selected to promote the adaptation of maize to high latitudes by accelerating flowering. However, the lower frequency of the KNR6^TE+ allele in the modern inbreds (Figure 1n) indicates that it was negatively selected during maize improvement due to its negative effect on grain yield.

The duration of the crop life cycle is often restricted by the local farming system. For example, the conventional double-cropped winter wheat–summer maize system in the Chinese Huanghuaihai region requires early flowering for maize in summer to facilitate wheat cultivation immediately after maize harvesting. Geographical distribution of the KNR6^TE+/KNR6^TE− alleles in maize inbreds showed that the KNR6^TE+ allele was more enriched in Chinese lines (18.5%) than in the lines from Latin America (3.6%) and America (10.9%) in modern maize (Figure 1n). Interestingly, 70.7% (29/41) of Chinese lines harbouring KNR6^TE+ allele distributed in the Huanghuaihai region (Figure S2), where short life cycle maize is in high demand. Notably, the KNR6^TE+ allele frequency was high in proprietary germplasms that derived from the American maize hybrid 3382 (52.5%) and the Chinese landrace TangSiPingTou (55.6%) (Figure 1o), which are two of the most widely used germplasms in the current breeding programmes in China. Thus, optimizing the flowering time to adapt to the local environment would be an effective way to maximize maize production (Parent et al., 2018).

Finally, we estimated the effects of KNR6 and four known flowering genes, ZmCCT9, ZmCCT10, VGT1 and ZCN8, on flowering time and ear length in a diverse maize population containing 508 maize inbreds. As expected, all genes had strong effects on flowering time, and only KNR6 had a trade-off effect on ear length (Figure S3). Notably, the early-flowering allele of VGT1 still promoted flowering by 2.1 days when the early flowering alleles of ZmCCT9 and ZmCCT10 were fixed in the lines with long-ear/late-flowering KNR6^TE− allele of KNR6 (Figure S4). Moreover, the early-flowering alleles of VGT1, ZmCCT9 and ZmCCT10 were not fixed in modern inbreds (Figure S5). Thus, we proposed an operational strategy, as shown in two aforementioned Chinese germplasms, to improve grain yield of the lines by replacing the short-ear/early-flowering KNR6^TE+ allele with the long-ear/late-flowering KNR6^TE− allele of KNR6 and balancing the late-flowering effect by pyramiding the early-flowering alleles of ZmCCT9, ZmCCT10, VGT1 and other flowering-time genes via marker-assisted selection (Figure 1o).

Collectively, we found that KNR6 underwent divergent selection during maize breeding to extend adaptation and increase grain yield, and offered a knowledge-driven strategy for engineering KNR6 to maximize maize grain production.