A total of 160 maize inbred line material populations were selected for the experiment. All materials were provided by the Maize Breeding Research Group of the College of Life Sciences at Shanxi Agricultural University (Biotechnology Research Center of Shanxi Academy of Agricultural Sciences). The names/sources of the materials were shown in Supplementary Table S1. 160 maize inbred lines were planted in Jinzhong City (111°25′∼114°05′E, 36°40′∼38°06′N) and Xinzhou City (110°53'∼113°58′E, 38°6′∼39°40′N). The field experiment was designed as a completely randomized block experiment with three replicates. Each material was sown in a single row with a row length of 5 m, a width of 2 m, and a plant spacing of 0.3 × 0.5 m. The plants were left in the holes, and the field management was conducted using conventional management methods. The main agronomic traits of the plant type were investigated.

Ten maize plants with normal growth 20 days after pollination were selected for plant type-related trait measurements in each cultivation experiment plot. The height from the ground surface to the top of the plant’s tassel, measured with a ruler (Unit: cm), was defined as the plant height. While measuring the plant height, the height from the ground surface to the uppermost ear bearing node was considered as the ear height. We also calculated the ratio of ear height to plant height. The length from the lowest branch of the main axis of the tassel to the top of the main axis of the tassel was measured and defined as the spindle length of tassel. The first-order branch number of tassel was determined as the count of first-order lateral branches present during the peak period of pollen shedding in the maize plant.

We extracted genomic DNA from 160 maize inbred lines using the CTAB method (Saghai-Maroof et al., 1984; Devos and Gale, 1992). The chip DNA library was conducted according to the Affymetrix Axiom data requirements protocol (Ganal et al., 2011; Ganal et al., 2012). Denaturing hybridization was performed using the Affymetrix GeneTitan system (Ganal et al., 2011; Ganal et al., 2012). After denaturation, the sample and Hyb Tray were placed on a 48°C metal block for operation. The denatured sample was added to the Hyb Tray and hybridized for 23.5–24 h at a designated location according to the requirements of the Affymetrix GeneTitan system. Finally, staining, cleaning, and scanning were performed in the study. The original data CELL file was classified based on several indicators: the percentage of successful loci in the sample Call relative to the total number of loci, Heterozygous Strength Offset, Fisher’s Linear Discriminant, Homozygote Ratio Offset, the number of minor alleles, and the minimum distance between the cluster centers of homozygous and heterozygous loci in HomFLD. We conducted a thorough analysis of the raw data quality, encompassing the loci deletion rate, minor allele frequency (MAF), sample deletion rate, and verification of hardy-weinberg equilibrium (HWE). Samples and loci with a high rate of missing typing and loci with a low MAF were filtered out in subsequent studies, as well as typing results that deviated from the HWE locus in normal controls (Tzeng and Sullivan, 2010).

The quality control of the SNPs obtained from 160 maize inbred lines was performed, and the loci with a typing success rate of less than 90% and a MAF of less than 0.05 were filtered out. The distribution of SNPs on chromosomes was obtained using high-quality genomes of maize (NCBI RefSeq assembly: GCF_902167145.1). Population structure analysis of 160 maize inbred lines was conducted using Admixture software, which employs a likelihood nested model (Alexander et al., 2009). The optimal number of groups was determined as the number of clusters for samples with the minimum cross-validation error value (Alexander et al., 2009). The genetic relationships among 160 maize inbred lines were calculated using SPAGeDi software (Hardy and Vekemans, 2002; Kim and Beavis, 2017), and the sample genetic relationship heatmap was plotted using R language’s pheatmap. The phylogenetic relationships of 160 maize inbred lines were constructed using the maximum likelihood method with the RAxML (v.8) software (Stamatakis, 2014).

To uncover the associations between maize plant type-related traits and molecular markers, a genome-wide scan of high-density molecular markers was performed on a population of maize inbred lines, resulting in the identification of significant SNPs associated with these traits. To analyze the population structure of the samples, we utilized the admixture software, assuming a range of 1–10 clusters (K values) for the samples and conducted clustering accordingly. We then performed cross-validation on the clustering outcomes to identify the optimal number of clusters, which was determined by the trough of the cross-validation error rate. We performed association analysis using both the Q + K and P + K models using the population structure data matrix, kinship matrix, and principal component (PC) scores as covariates, along with SNP markers and phenotypic data. We used the GAPIT3 software to perform genome-wide association analysis on five phenotypes, including first-order branch number of tassel, spindle length of tassel, ear height/plant height ratio, plant height and ear height (Wang and Zhang, 2021). To reduce false positives in association analysis results by combining population structure information and kinship matrix of samples, seven models including GLM, MLM, MLMLM, CMLM, ECMLM, SUPER, and Blink were used to perform association analysis for each trait. The model was then evaluated and selected based on the Quartile-Quartile plot (Q-Q plot) (Dehghan, 2018). We used the “qqman” package of R language to draw Q-Q and manhattan plots from GWAS results (Turner, 2014).

We selected the candidate regions 100 kb upstream and downstream of the significant SNPs, and simultaneously extracted all genes within the candidate regions as significant associated genes. The TBtool software was used to compare and extract significantly associated genes between Jinzhong and Xinzhou (Chen et al., 2020). Functional annotation of associated gene expression proteins was performed using Non-Redundant Protein Sequence Database (https://www.biostars.org/). Additionally, proteins were annotated in a hierarchical manner using the UniProt database. Initially, annotations were derived from the TrEMBL computer-predicted database, which were then refined through the manually curated Swiss-Prot database of protein sequences (UniProt-Consortium, 2023). The Pfam database was used to identify the structural domains and functional information of the associated gene expression proteins (Mistry et al., 2021). We conducted functional annotation and pathway enrichment of significantly associated genes with plant type traits in maize based on the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases (Gene-Ontology-Consortium, 2019; Jin et al., 2023).

We selected 160 maize inbred lines from Jinzhong and Xinzhou regions for testing their plant height, ear height, spindle length of tassel and first-order branch number of tassel characteristics. The maximum first-order branch number of tassel among the 160 maize inbred lines in Jinzhong and Xinzhou was 24 and 23, respectively, with an average of 6.29 and 6.33 per line. Notably, there were also inbred lines identified in both regions that lacked any first-order branch number of tassel (Figure 1A; Supplementary Table S2). The average ear height of 160 maize samples from Jinzhong and Xinzhou regions was 79 and 80 cm, respectively. The frequency distribution of ear height traits in maize inbred lines in Xinzhou region showed a large fluctuation (Figure 1B; Supplementary Table S2). The phenotypic analysis of 160 maize inbred lines from Jinzhong and Xinzhou showed that the maximum and minimum values of the spindle length of tassel in Jinzhong were 44.47 and 19.87 cm, respectively, while the maximum and minimum values of the spindle length of tassel in Xinzhou were 44.08 and 19.65 cm, respectively. There was no significant difference in the average length of the spindle length of tassel between the two regions (Figure 1C; Supplementary Table S2). The maximum plant heights of 160 maize inbred lines in Jinzhong and Xinzhou reached 279 and 272 cm, respectively, while the minimum heights were 93 and 101 cm, respectively. Similarly to the changes in ear height, the frequency distribution of plant height in maize inbred lines showed comparable variations, with a more pronounced fluctuation observed in Xinzhou (Figure 1D; Supplementary Table S2). The maximum ear height/plant height ratios of maize inbred lines in Jinzhong and Xinzhou regions peaked at 0.52 and 0.53, respectively, while the minimum ratios stood at 0.24 and 0.25, respectively. The average ear height/plant height ratio in Jinzhong was slightly lower than that in Xinzhou. Interestingly, both regions exhibited a gap in the frequency distribution at 0.43, and there was a notable fluctuation in the frequency distribution within the range of 0.34–0.35 (Figure 1E).

The average sample typing missing rate of the 160 sample raw data was 1.28% of the total number of missing loci, and the sample typing missing rate used was controlled within 4% (Supplementary Figure S1). The Affymetrix Axiom chip data had an average dish QC (DQC) value of 0.88, ranging from a maximum of 0.94 to a minimum of 0.83. Notably, over 95% of the high-quality probes met the QC detection criteria, with an average QC heterozygosity rate of 1.55%. Furthermore, the average detection rate across all loci in the samples was 99.29%, ensuring a sample quality control pass rate of 98.95%. The sample heterozygosity rate stood at 1.95% (Supplementary Table S3).

A total of 55,229 SNPs were retrieved from the Affy_SNPs database for the 160 sample loci probes, with an average typing success rate of 99.11%. The lowest typing success rate was 41.05% for probe locus: AX-91594479. Among all the probes, the average number of minor alleles was 150.7, the number of clusters was 2.7, the number of genotypes AA was 157.7, the number of genotypes AB was 12.5, and the number of genotypes BB was 206.3. The 36,300 probes previously exhibited the PolyHighResolution type, 2,305 probes exhibited the NoMinorHom type, 4,667 probes exhibited the OTV type, 3,635 probes exhibited the MonoHighResolution type, and only 858 probes exhibited the CallRateBelowThreshold type. Additionally, 13.5% of all probes belonged to other types (Figure 2A). After rigorous filtering, a total of 42,240 high-quality SNPs were identified. As depicted in Figure 2B, these SNPs were unevenly distributed across the 10 chromosomes of maize. Chromosome 1 hosts the highest number of SNPs, constituting 14.31% of the total, followed closely by chromosome 3 with 10.47%. Chromosomes 4 and 5 also contributed significantly, each accounting for over 10% of the total SNPs. Conversely, chromosome 10 has the lowest distribution of SNPs, accounting for just 6.51% of the total (Figure 2B; Supplementary Figure S2).

The optimal clustering number K = 6 was determined based on the minimal cross-validation error value achieved at this point, with subsequent variations remaining stable (Supplementary Figure S3). Consequently, 160 maize inbred lines were classified into six distinct substructures: Substructure 1, Substructure 2, Substructure 3, Substructure 4, Substructure 5, and Substructure 6 (Figure 3). The improved Substructure 4 group, represented by YE478 and ZHENG58, boasts the highest number of maize inbred lines among the six subgroups, totaling 55 inbred line materials. The Substructure 5 group comprised 28 maize inbred lines, notable examples of which were PH6WC and XY686M. The Substructure 6 group consisted of 23 maize inbred lines, featuring D1M, D3M, and 38P05M. The Substructure 3 group encompassed 15 maize inbred lines, with AMD5 and AMD293 among its past representatives (Figure 3; Supplementary Table S4).

In order to study the nucleic acid diversity of maize subgroups, the π value of nucleic acid diversity was calculated for each subgroup. The Substructure 5 (π = 1.23 × 10⁻⁴), Substructure 1 (π = 1.25 × 10⁻⁴), and Substructure 2 (π = 1.50 × 10⁻⁴) displayed notably lower average π values, suggesting a reduced nucleic acid diversity potentially due to stronger selection pressure from natural selection or other factors. The Substructure 4 exhibited a higher π value of 2.44 × 10⁻⁴, indicating a reduced likelihood of it being subjected to significant selection pressure from natural selection or other external factors (Supplementary Figure S4). A total of 79,946 allelic variation loci were identified across 160 maize inbred lines, utilizing 42,240 high-quality SNPs. The comparison of genetic allele frequencies among taxa revealed the largest divergence between the Substructure 6 and the Substructure 5, while the Substructure 3 and Substructure 2 exhibited the smallest difference in allele frequencies (Table 1; Supplementary Figure S5). The six subgroups exhibited an average gene diversity of 0.356 and a mean polymorphism information content of 0.245. The genetic distances between these subgroups ranged from 0.06 to 0.38, while the genetic differentiation index varied from 0.83 to 0.85, averaging at 0.59.

An analysis of the genetic relationships and phylogenetic development among 160 maize inbred lines, utilizing SNP markers, revealed that maize within the same subpopulation exhibited closer genetic ties. The phylogenetic analysis categorized the 160 maize inbred lines into six distinct branches, mirroring closely the subgroup division obtained from the population structure analysis (Supplementary Figure S5; Supplementary Material S1). The Substructure 1 and Substructure 5 exhibited a close phylogenetic proximity, indicating a relatively frequent exchange of genetic material. The Substructure 6 exhibited the greatest phylogenetic distance from the other subgroups, suggesting limited gene flow between them. Only a few Substructure 6 shared close relationships with the Substructure 1 and Substructure 5. The gene exchange between various subgroups provided important basic materials for the improvement of maize inbred lines.

Among the seven models selected for genome-wide association analysis, the GLM model of GWAS was the most suitable for maize first-order branch number of tassel and plant height traits, while the SUPER model was the most suitable for spindle length of tassel, ear height/plant height ratio and ear height traits (Figure 4; Supplementary Figures S9–S13). The GWAS between the first-order branch number of tassel and SNPs in 160 maize samples from Jinzhong and Xinzhou revealed no discernible systematic bias (Figure 4; Supplementary Figure S10). The GWAS of the first-order branch number of tassel in the Jinzhong region revealed four significant loci, primarily located on chromosomes 1, 7, and 8 (Figure 5; Supplementary Table S5). The GWAS analysis for the first-order branch number of tassel in Xinzhou region yielded five significant loci, predominantly located on chromosomes 2, 7, and 8 (Supplementary Table S5). Maize from Jinzhong and Xinzhou shared three loci significantly associated with the first-order branch number of tassel. The GWAS in Jinzhong identified 15 significant genes associated with 4 loci, while the analysis in Xinzhou revealed 14 significant genes linked to 5 loci in maize inbred lines. A total of 12 genes related to the first-order branch number of tassel were commonly associated in both Jinzhong and Xinzhou inbred lines, comprising 70.6% of the relevant genomic region (Figure 5). The molecular functional annotation and enrichment results showed that the UGT75L6 protein had transferase activity, the IPK2 protein had kinase activity, the AIP1-2 protein had protein binding activity, and the ABCB28 protein had ATP binding, TP enzyme activity, and ATPase activity, respectively. At the same time, ABCB28 was involved in transmembrane transport and IPK2 was involved in the biological process of phosphoinositide biosynthesis (Supplementary Table S6).

The GWAS of the spindle length of tassel and SNPs across 160 maize inbred lines from Jinzhong and Xinzhou did not exhibit any discernible systematic deviation (Supplementary Figures S12, S14). The GWAS analysis of the spindle length of tassel in Jinzhong region identified 14 significant associated loci, mainly distributed on chromosomes 2, 4, 5, 7, and 10, with 42.9% of the significant associated loci distributed on chromosome 2 (Supplementary Table S5). GWAS analysis of the spindle length of tassel in Xinzhou region revealed nine significant association loci, mainly distributed on chromosomes 2, 5, 7, and 10. All significant association loci in Xinzhou region were included in the significant association loci in Jinzhong region (Supplementary Table S5). The GWAS in Jinzhong region revealed 55 significant genes associated with 14 loci, while the analysis in Xinzhou yielded 35 significant genes linked to 9 loci in maize inbred lines. A total of 35 genes were commonly associated with the spindle length of tassel in both Jinzhong and Xinzhou inbred lines, comprising 63.6% of the relevant genomic region (Supplementary Figure S14; Supplementary Table S6). Among the shared genes between Jinzhong and Xinzhou, 88.6% of the expressed proteins had annotations in the Swiss-Prot database, while 82.9% of the gene names had been confirmed. At the same time, these genes were functionally annotated and enriched, such as ERD7 gene involved in endocytosis pathway, CML32 gene involved in plant-pathogen interaction, GLU3 gene involved in starch and sucrose metabolism, and ATG10 gene involved in autophagy regulation (Supplementary Table S6).

The GWAS association analysis of ear height trait in maize inbred lines in Jinzhong region obtained 13 significant loci, mainly distributed on chromosomes 1, 3, 4, 5, 7, 8 and 10 (Supplementary Figures S12, S14; Supplementary Table S5). The GWAS analysis for ear height in maize inbred lines in Xinzhou region revealed eight significant loci, predominantly found on chromosomes 1, 3, and 4 (Supplementary Table S5). Based on the candidate regions 100 kb upstream and downstream of the significant loci, 57 significantly associated genes were obtained for 13 loci significantly associated with maize inbred lines in Jinzhong region, and 52 significantly associated genes were obtained for 8 loci significantly associated with maize inbred lines in Xinzhou region. A total of 27 genes are commonly associated with the ear height trait in maize inbred lines from both Jinzhong and Xinzhou regions (Supplementary Figure S5; Supplementary Table S6). The functions of genes related to the ear height trait were annotated and enriched. The results showed that the PBS1 gene was enriched in the plant-pathogen interaction pathway, and the IAA4 gene was enriched in the auxin signaling pathway of the plant hormone signaling pathway (Supplementary Table S6).

The GWAS analysis of plant height traits in maize inbred lines in Jinzhong region identified five significant loci, mainly distributed on chromosomes 2, 3, 7, and 9 (Supplementary Figures S12, S14; Supplementary Table S5). The GWAS analysis of plant height traits in maize inbred lines in Xinzhou region identified three significant loci, mainly distributed on chromosomes 3 and 7 (Supplementary Table S5). Significant association loci were detected at 169,704,259 bp on chromosome 3 and 95581049 bp on chromosome 7 in the maize inbred lines from Jinzhong and Xinzhou, respectively (Supplementary Table S5). The GWAS in Jinzhong identified 26 significant genes associated with 5 loci, while the analysis in Xinzhou revealed 3 significant genes linked to 3 loci in maize inbred lines. Notably, all significant genes in Xinzhou were encompassed within those identified in Jinzhong (Supplementary Figure S15). We used multiple databases to compare and annotate the genes associated with plant height traits in maize inbred lines from Jinzhong and Xinzhou, which were annotated as BT2 (NM_001136739.1) and (ICMT NM_001365697.1 and XM_008674752.3) genes, respectively (Supplementary Table S6).

Based on the phenotypic data of ear height and plant height, we further conducted a genome-wide association analysis of the ratio of ear height to plant height. Compared with ear height and plant height traits, GWAS of ear height/plant height ratio obtained more SNPs. Among them, 23 significant loci were obtained by GWAS analysis of ear height/plant height ratio of maize inbred lines in Jinzhong region, mainly distributed on chromosomes 1, 2, 3, 4, 5, 8, 9 and 10 (Supplementary Figures S12, S14; Supplementary Table S5). The GWAS analysis of the ear height/plant height ratio trait of maize inbred lines in Xinzhou region identified 17 significant loci, primarily located on chromosomes 1, 3, 5, 6, 8, and 10 (Supplementary Figures S12, S14; Supplementary Table S5). At the same time, high-density loci were identified on chromosome 5 within the range of 24.50 Mb–24.72 Mb in the maize inbred lines from Jinzhong and Xinzhou. The GWAS in Jinzhong identified 136 significant genes associated with 23 loci, whereas the analysis in Xinzhou revealed 106 significant genes linked to 17 loci in maize inbred lines. Between the regions of Jinzhong and Xinzhou, 89 genes were commonly associated with the ratio of ear height to plant height in maize inbred lines, comprising 58.2% of the total identified genes (Supplementary Figure S15; Supplementary Table S6). The 89 shared genes between Jinzhong and Xinzhou regions were enriched in seven distinct pathways, including glutathione metabolism, metabolism of xenobiotics by cytochrome P450, drug metabolism - cytochrome P450, carotenoid biosynthesis, protein processing in endoplasmic reticulum, ribosome biogenesis in eukaryotes, plant hormone signal transduction (Supplementary Table S6).

We identified 12, 35, 27, 3, and 89 genes that were significantly associated with first-order branch number of tassel, spindle length of tassel, ear height, plant height, and ear height/plant height ratio, respectively. There were unannotated genes and alternative splicing of genes in these significantly associated genes of maize plant type (Figure 6; Supplementary Table S6). We screened and eliminated genes that were unannotated genes and alternative splicing. Finally, 6, 22, 14, 2, and 37 genes were identified as significantly associated with first-order branch number of tassel, spindle length of tassel, ear height, plant height, and ear height/plant height ratio, respectively (Supplementary Table S6). The identification of these genes not only sheds light on the genetic underpinnings of maize plant type traits but also serves as a valuable genetic resource for future biological breeding efforts. By understanding the specific functions and interactions of these genes, researchers can develop targeted breeding strategies to enhance desirable plant type traits in maize.