A novel genetic framework reveals transcriptional “butterfly effect” underlying heterosis in maize

Graphical abstract

Highlights

• •A single-segment substitution line (SSSL)-based framework effectively identifies heterosis genes.
• •Most heterosis-related DEGs function through trans-regulation beyond substitution regions.
• •Small genomic variations in heterozygous regions trigger genome-wide transcriptional shifts.
• •ZmWRKY67 is a potential transcriptional regulator of Incw1, linking sucrose metabolism to heterosis.
• •Findings support a “butterfly effect” in transcription and provide insight into heterosis mechanisms.

Abstract

Introduction

Maize is one of the first crops to benefit from heterosis, significantly enhancing commercial breeding. Despite extensive research, the molecular mechanisms of heterosis remain elusive.

Objectives

This study integrates a novel genetic framework with transcriptomic and phenotypic analyses to identify heterosis-related genes and uncover their regulatory mechanisms.

Methods

Single-segment substitution lines (SSSLs) were combined with a novel double single-factor differential subtraction (DSDS) strategy to identify heterosis-related differentially expressed genes (DEGs). Unlike traditional parent-hybrid comparisons, DSDS subtracts background expression noise and identifies heterosis-related genes within a controlled genomic context. Five hybrid groups, including mini-F₁ and sub-F₁ hybrids, were generated and analyzed. RNA sequencing was conducted at 15 days after pollination (DAP) to profile gene expression. Motif enrichment and dual-luciferase assays were used for functional validation.

Results

Heterosis exhibited a nonlinear trajectory with parental genetic distance, displaying a rapid increase at low genomic heterogeneity levels before reaching a saturation point at moderate divergence. DSDS identified 359 and 80 heterosis-related DEGs in two mini-F₁ hybrids, with most DEGs located outside the substitution regions, supporting a trans-regulatory mechanism. GO (gene ontology) analysis revealed enrichment in transport and cell expansion pathways, aligning with key processes in early kernel development. Notably, ZmWRKY67 was the only DEG in the substitution segment of mini-F₁ CL137 × Ye478. Initial functional assays confirmed its regulation of Incw1, a key sucrose metabolism gene. The findings suggest a transcriptional “butterfly effect”, where minor genomic variations in heterozygous regions trigger genome-wide expression changes to drive heterosis.

Conclusion

This study demonstrates the effectiveness of SSSL-based DSDS in identifying heterosis-related genes and provides a framework for dissecting transcriptional regulation in heterosis. The observed transcriptional “butterfly effect” offers a novel perspective on how small genomic variations contribute to heterosis. Future research should explore additional substitution-region DEGs to further unravel the molecular basis of heterosis in maize.

Introduction

Heterosis refers to the superior performance of F₁-hybrids compared to their parental inbred lines. This phenomenon, first systematically described by Charles Darwin in 1876, was later demonstrated in maize, where hybrids exhibited enhanced biomass, root growth, plant height, yield, and resilience to biotic and abiotic stresses [1,2]. The widespread use of heterosis in crop breeding has significantly contributed to agricultural productivity and economic development [3]. However, despite its importance, the precise molecular mechanisms underlying heterosis remain unclear, particularly the role of transcriptional regulation and gene network interactions in shaping hybrid performance.

Multiple classical genetic hypotheses have been proposed to explain heterosis. The dominance hypothesis suggests that deleterious recessive alleles in homozygous parental inbred lines are masked by dominant beneficial alleles in hybrids [[4], [5], [6], [7]]. The overdominance hypothesis suggests that heterozygous alleles at certain loci confer a stronger advantage than either homozygous state [8]. The epistasis hypothesis attributes heterosis to synergistic interactions between non-allelic genes across multiple loci [9,10]. Although each model explains aspects of heterosis, the regulatory mechanisms governing transcriptional changes and genetic interactions remain unresolved.

Maize (Zea mays L.) is one of the most extensively studied crops for heterosis research and serves as an ideal model system for investigating gene expression changes in hybrids [11]. The extent of heterosis is commonly measured using mid-parent heterosis (MPH) and better-parent heterosis (BPH). MPH indicates that the hybrid outperforms the mean value of its two parental lines (MPV), while BPH denotes cases where the hybrid surpasses the superior parent [12]. Studies have shown that heterosis is particularly strong during early developmental stages, with MPH values of 51 % in root traits during germination, and significant heterosis in embryo and endosperm size (40–50 %) within six days after pollination (DAP) [13,14]. At later developmental stages (25–35 days post-pollination), hybrid embryos significantly outgrow the mid-parental value, indicating sustained heterotic effects of gene expression [15].

Gene expression profiling has provided insights into the molecular basis of heterosis that heterosis is driven by non-additive gene expression patterns. Hoecker et al. found that hybrids exhibit conserved non-additive transcriptional patterns in primary root development, suggesting an intrinsic role of regulatory networks in heterosis [16]. Similarly, transcriptomic analyses revealed that at six days post-pollination, most DEGs in maize hybrid embryos and endosperms exhibit non-additive expression, with enrichment in biological regulation, stress responses, and transcriptional activity [14,17]. These findings suggest that heterosis is orchestrated by complex transcriptional reprogramming rather than simple allele complementation, but the precise regulatory factors driving these changes remain unclear.

The advancement of high-throughput sequencing technologies, particularly RNA-seq, has enabled global transcriptomic analysis with unprecedented resolution to facilitate the identification of genes and pathways involved in heterosis [18]. Studies have shown that maize hybrids express more genes than their parental inbred lines, suggesting extensive transcriptional reprogramming in hybrids [19,20]. Additionally, differential gene expression is more pronounced between parental inbred lines than between parents and hybrids or between reciprocal hybrids, indicating that hybridization alters gene regulatory networks in a systematic manner [18,21]. Despite the wealth of data generated by RNA-seq, identifying key heterosis-related genes remains challenging due to the complexity of gene regulatory networks underlying heterosis. While differential expression analysis, gene co-expression analysis, and functional enrichment studies have revealed genes and pathways involved in heterosis, pinpointing the causal regulatory factors has remained elusive.

To address these challenges, single-segment substitution lines (SSSLs) offer a powerful approach for dissecting heterosis-related gene regulation [22]. SSSLs are developed by introgressing a single chromosome segment from a donor parent into a receptor parent’s genetic background, allowing precise analysis of heterotic loci with minimal genetic background noise. Compared to traditional quantitative genetic approaches, SSSLs provide unique advantages in that they allow fine mapping of heterotic loci with reduced genomic variability. SSSL-based mapping has been employed in maize and rice to identify heterosis-related genes [[23], [24], [25]], revealing their potential in studying genetic and transcriptional determinants of heterosis. However, their integration with high-throughput transcriptomics for heterosis gene discovery remains underexplored.

This study integrates a novel genetic framework integrating carefully-designed SSSL-derived hybrids with phenotypic and transcriptomic analyses to dissect the molecular basis of heterosis in maize. The objectives are to: (1) develop a genetic framework for efficient RNA-seq-based discovery of heterosis-related genes; (2) identify key transcriptional regulators and their roles in maize grain heterosis; (3) functionally validate key candidate heterosis-related genes.

Materials and methods

Development of single-segment substitution lines and hybrid construction

In our previous study [26], a large population of segment substitution lines were developed using Ye478 as the receptor parent and Qi319 as the donor parent. Through our previous molecular marker-assisted selection, CL86 and CL137, these two lines were found to be single segment substitution lines (SSSLs). Specifically, CL86 harbors a 174 Mb segment from Qi319 on chromosome 5 (chr5: 22,850,628–197,098,506 bp), while CL137 contains a 130 Mb segment from Qi319 on chromosome 10 (chr10: 9,256,114–139,016,967 bp). In this study, the two SSSLs and their two original parents were selected as the core inbred lines for hybrid construction.

Based on genome correlation between the parental lines, five core hybrids were categorized into three distinct groups, Qi319 × Ye478 as the original F₁ hybrid, CL86 × Ye478 and CL137 × Ye478 as mini-F₁ hybrids, CL86 × Qi319 and CL137 × Qi319 as sub-F₁ hybrids. The classification and genetic relationships of these inbred lines and hybrids are illustrated in Fig. 1A.

Fig. 1.

A genetic framework integrating core genetic materials and the double single-factor differential subtraction (DSDS) system used in this study for identifying heterosis-related differentially expressed genes (DEGs) in maize hybrids. (A) The four core inbred lines (Ye478, CL86, CL137, and Qi319) and five core hybrids: the original F₁ hybrid (Qi319 × Ye478), two mini-F₁ hybrids (CL86 × Ye478, CL137 × Ye478), and two sub-F₁ hybrids (CL86 × Qi319, CL137 × Qi319). Ye478 serves as the receptor parent (Pr), while Qi319 is the donor parent (Pd). The single segment substitution lines (SSSLs, CL86 and CL137) contain specific genomic fragments from Qi319 in the Ye478 background. Yellow represents the genetic composition from the donor parent Qi319, and blue represents the genetic composition from the receptor parent Ye478. (B) Schematic representation of the DSDS approach used to identify heterosis-related genes. The mini-F₁ vs. Pr contrast (union of sets A and B) identifies DEGs influenced by both the heterozygous substitution fragment and heterosis, while the SSSL vs. Pr contrast (union of sets B and C) isolates DEGs caused solely by genetic background differences. By removing overlapping DEGs (set B), the approach isolates heterosis-related genes (set A), which result from allele interactions within the heterozygous substitution fragment.

The genomic difference between Qi319 and Ye478 (defined as 100 % in the original F₁ hybrid) was used as a reference. Accordingly, the genomic differences in the two mini-F₁ hybrids (CL86 × Ye478 and CL137 × Ye478) were 8.3 % and 6.2 %, respectively, while the two sub-F₁ hybrids (CL86 × Qi319 and CL137 × Qi319) exhibited genomic differences of 91.7 % and 93.8 %, respectively.

To investigate the degree of heterosis and transcription pattern, two classic inbred lines B73 and Mo17 were selected as paternal parents and crossed with the core inbred lines (CL86, CL137, Ye478 and Qi319) following North Carolina II (NCII) mating design for a comprehensive assessment of genetic contributions to hybrid performance. As a result, eight additional hybrids were generated, increasing the total number of hybrids to thirteen, all of which were developed during the winter of 2018 in Hainan province.

Field phenotyping and statistical analysis for heterosis traits

Following hybrid construction, field trials were conducted across multiple locations to evaluate phenotypic variation and heterosis effects under different environmental conditions. The 13 hybrids and their 6 parental inbred lines were grown in 2019 at four locations in China: Shangzhuang, Beijing (116°11′E,40°8′N); Wuqiao, Hebei Province (116°26′E, 37°37′N); Quzhou, Hebei Province (115°0′E, 36°51′N); and Hebi, Henan province (114°19′E, 35°42′N). Field trials were conducted using a randomized complete block design (RCBD) with two replicates per site. Each plot consisted of four rows, with 17 plants per row, a row length of 4.0 m, and a row spacing of 0.6 m. To ensure accurate yield assessment, only the middle two rows were harvested. Field management practices followed standard local agronomic recommendations.

A total of eleven phenotypic traits were measured across all experimental environments, including plant height (PH, cm), ear height (EH, cm), days to heading (DTH, day), days to silking (DTS, day), days to anthesis (DTA, day), ear length (EL, cm), ear diameter (ED, cm), row number per ear (RNPE, count), kernel number per row (KNPR, count), hundred grain weight (HGW, g) and grain yield (GY, ton/ha) were measured. Detailed descriptions of these traits, along with their corresponding measurement protocols, are provided in Table S1.

To account for environmental effects and ensure robust genotype comparisons, we estimated variance components and computed Best Linear Unbiased Estimation (BLUE) values using statistical models aligned with the experimental design. Based on the experimental design, variance component profiling and BLUE value estimation were performed using the following model:

$$y_{\mathit{ijk}}=\mu +G_i+E_j+G_i\times E_j+R_k(E_j)+{\epsilon }_{\mathit{ijk}}$$ (1)

where $y_{\mathit{ijk}}$ is the phenotypic value of genotype i in Environment j and replication k; $\mu$ is the overall mean; $G_i$ is the genotypic effect of genotype i; $E_j$ is the effect of environment j; $G_i\times E_j$ is effect of the interaction between genotype i and environment j; $R_k(E_j)$ is the effect of replication k in environment j; ${\epsilon }_{\mathit{ijk}}$ is the random error.

All factors in model (1) were treated as random effects for variance component estimation, while the genotypic effect was considered a fixed effect for calculating BLUE values.

Mid-parent heterosis (MPH) was used as a standard metric to assess heterosis and was calculated based on BLUE values using the following formula:

$$\mathit{MPH}=G_{F_1}-\frac{1}{2}\left(G_{P_1}+G_{P_2}\right)$$ (2)

where $G_{F_1}$ is the genotypic value of an F₁ hybrid; $G_{P_1}$ and $G_{P_2}$ are the genotypic values of its two homozygous parental lines, respectively.

Relative mid-parent heterosis (MPHr) was calculated using the following formula:

$$\mathit{MPHr}=\frac{\mathit{MPH}}{\frac{1}{2}\left(G_{P_1}+G_{P_2}\right)}\times 100\%$$ (3)

Broad-sense heritability ($H^2$) was calculated as:

$$H^2={\sigma }_G^2/({\sigma }_G^2+{\sigma }_{\mathit{GE}}^2/n+{\sigma }_{\epsilon }^2/nr)$$ (4)

where ${\sigma }_G^2$ is the genetic variance, ${\sigma }_{\mathit{GE}}^2$ is the variance of interaction between genotype and environment, ${\sigma }_{\epsilon }^2$ is the random error variance, n is the number of environments and r is the number of replications. All the analyses above were conducted with the R package sommer (v4.1.2) [27,28].

Multiple comparisons were performed using LSD.test() in the R package agricolae with model (1). Means with the same letter did not significantly differ at p < 0.05, according to LSD test.

RNA isolation and sequencing for maize kernel transcriptomes

We selected 15 DAP for RNA-seq analysis because it represents a critical stage of maize kernel development, marked by peak synthesis of starch and storage proteins [29,30]. This stage has been widely used to evaluate transcriptional regulation associated with domestication [31], transcription factor function [[32], [33], [34]], and grain filling [35,36]. Its central role in determining sink strength and yield potential makes it an ideal developmental window for studying the molecular basis of heterosis.

All the 6 parental inbred lines and 13 F₁ hybrids were planted in 2019 at the Shangzhuang Experimental Station of China Agricultural University, Beijing, under standard field management conditions. For each line, three biological replicates of ear were harvested at 15 DAP. Kernels were immediately separated, flash-frozen in liquid nitrogen, and stored at −80 °C for subsequent RNA extraction. Total RNA was extracted using the RNAprep Pure Plant Kit (QIAGEN, Cat#DP432) per the manufacturer’s instructions. RNA concentration was quantified using a Qubit 3.0 Fluorometer (Life Technologies, CA, USA), while RNA integrity was assessed using an Agilent 2100 Bioanalyzer with the RNA Nano 6000 Assay Kit (Agilent Technologies, CA, USA).

The mRNA sequencing libraries of maize kernels were constructed using the NEBNext® Ultra™ RNA Library Prep Kit for Illumina® (#E7530L, NEB, USA) following the manufacturer’s protocol. Briefly, mRNA was purified from total RNA using poly-T oligo-attached magnetic beads. Fragmentation was carried out using divalent cations under elevated temperature in NEBNext First Strand Synthesis Reaction Buffer (5X). First strand cDNA was synthesized using random hexamer primer and RNase H. Second strand cDNA synthesis was subsequently performed using buffer, dNTPs, DNA polymerase I and RNase H. The library fragments were purified with QiaQuick PCR kits and elution with EB buffer, then terminal repair, A-tailing and adapter added were implemented. The aimed products were retrieved and PCR was performed, then the library was completed. Unique index codes were added to assign sequences to each sample. Library RNA concentration was preliminarily quantified using the Qubit® RNA Assay Kit on a Qubit® 3.0 Fluorometer and diluted to 1 ng/μL. Insert size was evaluated using an Agilent Bioanalyzer 2100, while accurate library quantification was performed using a StepOnePlus™ Real-Time PCR System, ensuring a valid concentration of >10 nM. Clustering of the index-coded samples was conducted on a cBot Cluster Generation System using the HiSeq PE Cluster Kit v4-cBot-HS (Illumina) following the manufacturer’s instructions. Finally, sequencing was performed on an Illumina platform, generating 150 bp paired-end reads.

Processing and differential expression analysis of RNA-seq data

After sequencing, raw reads were processed to remove low-quality bases and mapped to the maize reference genome for downstream differential expression analysis. Raw RNA sequencing data were preprocessed to obtain clean reads by removing low-quality reads, i.e., those containing more than 5 % ambiguous bases (N), and adapter sequences. After quality control, an average of 45.8 million clean reads were retained per sample, with an average sequencing depth of approximately 3.5 × . On average, 93.1 % of reads were successfully mapped to the B73 reference genome (B73_RefGen_v4.38, https://plants.ensembl.org/index.html), indicate high data quality suitable for transcriptome analysis (Table S2). The cleaned reads were aligned using HISAT2 (v2.1.0) [37], and uniquely mapped reads for each gene in each sample were counted using HTSeq (v0.6.0) [38]. Gene expression levels were estimated using FPKM (fragments per kilobase of transcript per million fragments mapped) and TPM (transcripts per million). Genes with FPKM > 1 were considered expressed.

Three biological replicates per line were independently sequenced and analyzed using DESeq2 (v1.30.1) [39]. Rather than averaging replicate values, differential expression analysis was performed using a negative binomial generalized linear model to account for replicate variation. P-values were then adjusted using the Benjamini-Hochberg (BH) method to obtain adjusted p-values (p_adj). To identify DEGs (Table S3), we applied a threshold of |log₂FC| ≥ 1.7 and adjusted p-value (padj) ≤ 0.05. The log₂FC cutoff was derived based on a Student’s t-test, corresponding to the minimum fold change with a statistically significant deviation from zero at α = 0.05 [40], thereby reducing false positives and enhancing specificity in expression comparisons across samples. Genes with FPKM < 1 in all samples were removed, resulting in 26,380 retained genes for principal component analysis (PCA) using prcomp() in R.

Double single-factor differential subtraction for identifying heterosis-related genes

To differentiate heterosis-related DEGs from those arising due to genetic background differences, we applied a double single-factor differential subtraction (DSDS) approach, allowing precise identification of heterosis-related genes. This method isolates expression changes attributable to heterosis while minimizing the influence of the genetic background variation. In this system, an SSSL (CL86 or CL137) was crossed with the receptor parent (Pr, Ye478) to generate an SSSL × Pr hybrid (Fig. 1B), which is heterozygous only in the substituted segment. Two single-factor contrasts were performed: i) the mini-F₁ (SSSL × Pr) vs. Pr comparison (Fig. 1B, union of set A and B) isolates differentially expressed genes (DEGs) caused by both the substitution segment and heterosis effects; ii) the SSSL vs. Pr comparison (Fig. 1B, union of sets B and C) identifies DEGs due to genomic differences in the substitution segment alone. By excluding overlapping DEGs (Fig. 1B, set B) shared between the two comparisons, we isolated DEGs specific to the hybridization event (Fig. 1B, set A), which were considered heterosis-related genes resulting from the interaction of alleles within the heterozygous substitution fragment. Given this design, in which hybrids and parents differ only in the substituted segment, background effects were effectively controlled, and a separate statistical test was not performed.

GO enrichment analysis

GO enrichment analysis was performed using the R package clusterProfiler (v3.18.1) [41]. The database was obtained online using the R package biomart (v2.46.3) [42,43]. Only the genes with GO annotations were retained as background, and GO enrichment analysis was conducted for heterosis-related genes. Items with P-values < 0.05 are considered to be significantly enriched.

Identification of regulatory motifs in heterosis-related genes

To further explore the regulatory mechanisms underlying heterosis-related DEGs, we performed motif enrichment analysis to identify potential transcription factor binding sites associated with gene expression changes. Motif enrichment analysis was conducted using the MEME Suite (v5.5.5) in Linux, applying simple enrichment analysis (SEA) [44]. The upstream 2000 bp sequence and the initial 100 bp sequence of 79 genes were extracted as input sequences. To generate a background control, randomized sequences were created using the “Shuffled input sequences” option. For motif identification, we used the JASPAR (NON-REDUNDANT) DNA and JASPAR CORE (2022) plant databases. Results were filtered based on the significance threshold (Q_value ≤ 0.05) to retain only significantly enriched motifs.

To complement the motif enrichment analysis, we further evaluated the enrichment of WRKY transcription factor (TF) targets among the 79 heterosis-related DEGs. WRKY target gene sets, publicly available from Tu et al. [45] and identified by using a probabilistic model applied to ChIP-seq data from maize leaves, were used to assess enrichment among the 79 DEGs. To test for enrichment, we performed a one-sided proportion test in R using the function prop.test(x, n, p, alternative = “greater”), where x represents the number of overlapping genes between the WRKY TF targets and the 79 DEGs, n is the total number of DEGs (79), and p is the proportion of WRKY TF targets among all genes in the ChIP-seq dataset.

Functional validation of ZmWRKY67 as a transcriptional regulator of heterosis-related genes

Based on ChIP-seq enrichment and functional annotations, Incw1 (cell wall invertase 1) was prioritized for downstream validation as a candidate WRKY67 target. Incw1 encodes an apoplast-localized invertase that hydrolyzes sucrose into glucose and fructose, facilitating carbohydrate unloading to sink tissues such as developing kernels [46,47]. Although its paralog Incw2 (also known as Mn1) is better characterized, studies show that Incw1 can compensate for Incw2 loss, contributing to seed development in mn1 mutants [48,49]. The rice ortholog OsCIN1 similarly supports grain filling [50], and Incw1 has been co-mapped with QTLs for kernel weight and size in maize [51,52]. In Arabidopsis, its ortholog AtcwINV2 regulates seed weight, and Incw1 complementation rescues the mutant phenotype [51]. Together with its predicted regulation by multiple WRKYs (including ZmWRKY67), these findings provided a rationale for further investigation of Incw1.

To functionally validate the regulatory role of ZmWRKY67, we conducted a dual-luciferase reporter assay to test its direct activation of the target gene Incw1. The full-length coding sequence of ZmWRKY67 was cloned into the pGreenII 62-SK vector to generate effector. The promoter region of the target gene (Incw1), with a length determined by the binding motif position, was inserted into the pGreenII 0800 to generate reporter. The renilla LUC (REN) gene, driven by Cauliflower mosaic virus (CaMV) 35S promoter (p35S) in the pGreenII 0800-LUC vector, served as the internal control.

Protoplasts were extracted from the yellowing leaves of 12-day-old B73 maize seedlings, following the protocol described [53]. The experiment included three biological replicates. LUC and REN activities were measured using the Dual-Luciferase Reporter Assay System (Promega) with a Spark multimode reader (TECAN, Zurich, Switzerland). The relative LUC activity was determined by calculating the LUC to REN (LUC/REN) activity ratio.

Results

Heterosis in hybrids and its nonlinear relationship with genetic distance

To quantify the degree of heterosis in hybrids, we conducted statistical analyses across multiple environments. We observed substantial phenotypic variation across parental lines and their hybrids for all eleven traits. The full raw phenotypic measurements are provided in Supplementary Table S4. This allowed us to assess the significance and stability of key agronomic traits across different genotypes. ANOVA of the four-environmental trial showed that, all the genotypes were highly significant for all 11 traits in both inbred lines and hybrids (Table S5). Broad-sense heritability estimates ranged from 0.70 (DTA in inbred) to 0.98 (RNPE in hybrid), except for DTS, which exhibited lower heritability in both inbred lines (0.45) and hybrids (0.58). These results confirm the stability and reliability of the field experiment.

Based on the BLUE values (Table S6), MPH and their heritabilities were estimated for all traits (Table S7). Results showed that, except for HGW in CL86 × Ye478, almost all traits showed apparent MPH and displayed clear heterotic effects when CL86 and CL137 were crossed with any inbred parent. The negative MPH observed for HGW in this hybrid indicates relatively small kernel size, while its overall yield heterosis was compensated by increases in row number per ear and kernel number per row (Table S7 and Table S4), suggesting a trade-off among yield components. Notably, hybrid lines derived from the two SSSL lines (CL86 and CL137) and their parental line (Ye478) crossed with two classic inbred lines B73 and Mo17 exhibited no statistical significance in GY or other growth-related traits (p-value > 0.05 for multiple comparisons), which (CL86 × B73, CL137 × B73, and Ye478 × B73; CL86 × Mo17, CL137 × Mo17, and Ye478 × Mo17) indicated that, the derives lines retained heterotic potential similar to Ye478 (Table S8). This suggests that the SSSL-derived lines CL86 and CL137 exhibit no disadvantage due to their single-segment substitutional fragments on chromosome 5 or chromosome 10, preserving their heterotic potential.

To further quantify the magnitude of heterosis, relative mid-parent heterosis (MPHr) was estimated (Table S9) and visualized in Fig. 2 and Fig. S1. Specifically, the GY-MPHr values for CL86 × Ye478, CL137 × Ye478, CL86 × Qi319, CL137 × Qi319, Qi319 × Ye478 were 65.0 %, 114.3 %, 188.5 %, 177.0 %, and 178.5 % respectively. Similarly, the KNPR-MPHr values were 29.9 %, 43.7 %, 60.3 %, 78.0 %, and 51.6 %, respectively. These results, along with the MPHr expression patterns observed in other traits (Fig. S1), indicate that both mini-F₁s exhibited significantly high MPH, whereas sub-F₁s displayed MPH levels comparable to the original F₁.

Fig. 2.

Mid-parent heterosis (MPHr) and its nonlinear relationship with genomic heterogeneity in core hybrids. (A) MPHr in grain yield across five hybrid combinations: CL86 × Ye478, CL137 × Ye478, CL86 × Qi319, CL137 × Qi319, and Qi319 × Ye478. Bars represent female parent (F), male parent (M), and hybrid (H), with MPHr percentages labeled above hybrids. (B) MPHr in kernel number per row across the same hybrid combinations, following the format in (A). (C) The nonlinear relationship between genomic heterogeneity (%; defining Qi319 × Ye478 as 100 %) and MPHr of grain yield. Hybrid combinations are plotted with a curving trend illustrating the saturation trend in heterosis gain. MPHr increases rapidly at low-to-moderate genomic heterogeneity (∼6.2–8.3 %) but plateaus at higher levels (∼91.7–93.8 %), supporting a nonlinear heterosis trajectory. (D) The nonlinear relationship between genomic heterogeneity (%) and MPHr of kernel number per row, mirroring the pattern in (C). In (C) and (D), genomic heterogeneity between any two materials is defined as the percentage of the heterozygous genomic regions in their hybrid relative to the total genome length. The error bars represent standard errors (SE) estimated from four environmental replicates.

Fig. 2C and D illustrate the nonlinear relationship between genomic difference and MPHr for GY and KNPR, respectively. Both traits exhibit a saturation effect, where MPHr increasing rapidly at low to moderate genomic difference (6.2 %–8.3 %) but plateaus at higher heterogeneity levels (91.7 %–93.8 %). Mini-F₁ hybrids (CL86 × Ye478 and CL137 × Ye478) displayed greater heterosis than expected, while sub-F₁ hybrids showed heterosis levels comparable to the original F₁ hybrid.

Heterosis-related heterozygous loci influence genome-wide gene expression through trans-regulation

The period between 10 and 15 DAP is a critical phase for kernel formation [54,55] and a key stage for the accumulation of yield heterosis. Given its role as a transcriptional turning point in kernel development, 15 DAP was selected for transcriptome analysis to capture expression changes underlying heterosis (see Methods). In total, expression of 38,272 genes was detected. The stability of transcription data is crucial for downstream analysis, and the principal component analysis (PCA) of transcriptional data (Fig. S2) confirmed that biological replicates clustered closely, indicating data reliability.

Using DSDS approach (Fig. 1B), we identified 359 DEGs in CL86 × Ye478 and 80 DEGs in CL137 × Ye478 (Fig. 3A & C), suggesting a strong association with grain yield heterosis. Notably, only 22 and 1 of these DEGs were located within the substituted genomic region of CL86 × Ye478 and CL137 × Ye478, respectively (Fig. 3B & D). In contrast, the majority of heterosis-related DEGs were found in the homozygous genomic regions, suggesting that differential gene expression in the heterozygous genomic regions can influence gene expression beyond their own loci. This implies a potential trans-regulatory mechanism, where heterozygous loci influence global gene expression dynamics, contributing to heterosis at the transcriptional level. To further support this hypothesis, we examined the publicly available maize 15 DAP kernel regulatory network (MaizeGRN: n13e dataset) [56] and found that ZmWRKY67, the only DEG within the substituted region of CL137 × Ye478, is predicted to regulate a broad set of downstream genes across multiple chromosomes (Table S10). Several predicted targets, including cipk20, dof2, and ereb97, were located on chromosomes 7 and 2, consistent with the observed trans-distribution of DEGs.

Fig. 3.

Identification of heterosis-related differentially expressed genes (DEGs) using the double single-factor differential subtraction (DSDS) approach and the gene ontology (GO) enrichment analysis. (A, C) Venn diagrams showing the distribution of DEGs in CL86 × Ye478 and CL137 × Ye478. Heterosis-related DEGs are highlighted in orange and indicated by arrows. (B, D) Genomic distribution of heterosis-related DEGs across chromosomes in CL86 × Ye478 (B) and CL137 × Ye478 (D). Yellow marks the substituted genomic segments from the donor parent Qi319 (Chr5 in B and Chr10 in D), while blue represents the receptor parent Ye478. The number of heterosis-related DEGs (Het. DEGs No.) per chromosome is indicated. The three numbers on Chr5 (6, 22, 8) and Chr10 (0, 1, 2) correspond to the substitution segment and its two flanking regions; a total of 22 heterosis-related DEGs in CL86 × Ye478 and 1 heterosis-related DEG in CL137 × Ye478 are located within the substituted segments. (E, F) GO enrichment analysis of heterosis-related DEGs in CL86×Ye478 (E) and CL137×Ye478 (F). The top eight significantly enriched GO terms are presented.

Heterosis-related DEGs enrich transport and cell expansion pathways

To better understand the biological roles of heterosis-related genes, we performed Gene Ontology (GO) enrichment analysis, and the top 8 significantly enriched GO terms were visualized in Fig. 3E and F for CL86 × Ye478 and CL137 × Ye478, respectively. The most significant and biologically relevant GO terms could be categorized into two major functional groups. The first category encompasses two distinct processes: chemical transfer (catalyzed by transferase enzymes) and physical translocation of molecules. Chemical transfer involves terms like: “transferase activity, transferring acyl groups other than amino-acyl groups” (p = 0.001 in CL86 × Ye478, p = 0.007 in CL137 × Ye478), “UDP-glycosyltransferase activity” (p = 0.003 in CL137 × Ye478), “transferase activity, transferring hexosyl groups” (p = 0.005 in CL137 × Ye478). Physical transport refers to terms such as “xenobiotic transport” (p = 0.019 in CL86 × Ye478, p = 0.005 in CL137 × Ye478). The second category consists of cell reproduction and expansion-related processes, including “cell wall organization” (p = 0.008 in CL86 × Ye478), “cell wall” (p = 0.009 in CL137 × Ye478), and “cell wall modification” (p = 0.003 in CL137 × Ye478). The genes in second GO category entries are highly correlated and directly act on the process of plant cell expansion. They regulate the mechanical properties (strength, rigidity, plasticity, and ductility) of the cell wall by synthesizing the cell wall skeleton (cellulose, hemicellulose, pectin), reshaping the cell wall network (xyloglucan endotransglucosylase/hydrolase, expansin), and precisely regulating the properties of the cell wall matrix (pectinesterase). Enable cells to expand in a controllable and usually anisotropic manner under the drive of bentonite pressure. This is the core mechanism of plant cell growth and organ morphogenesis. Although they do not directly drive cell division (the core meaning of “reproduction”), they are crucial for the growth of daughter cells after division.

Identification and functional validation of ZmWRKY67 in heterosis

Among the 80 DEGs identified in mini-F₁ CL137 × Ye478 (Fig. 3C), only Zm00001d024376 (ZmWRKY67) was located within the substitution fragment region on chromosome 10, while the remaining 79 DEGs were distributed across the genome (Fig. 3D). This unique localization suggests that ZmWRKY67, in its heterozygous state, is a key regulatory gene contributing to the heterosis phenotype of CL137 × Ye478. ZmWRKY67 is annotated as a WRKY-transcription factor 67 in MaizeGDB (https://www.maizegdb.org/gene_center/gene/Zm00001d024376, accessed 13 March 2025) and shows evolutionary conservation with orthologs in sorghum [20,57].

To explore whether ZmWRKY67 regulates other DEGs, motif enrichment analysis using MEME and ChIP-seq target analysis based on published maize ChIP-seq data [45] was conducted. The analysis identified two enriched WRKY motifs (Table S11) and five WRKY-related transcription factor binding sites (Table S12), suggesting that ZmWRKY67 may, via trans-regulation, affect a broader set of DEGs among the 79 genes distributed in other chromosomes. Through enrichment analysis of published maize ChIP-seq data, we screened genes likely regulated by WRKY transcription factors. Among the predicted targets, only a few dozen genes had functional annotations. Incw1 (annotated as cell wall invertase 1, CWI1) emerged as a compelling candidate, appearing among the genes targeted by ZmWRKY53, ZmWRKY25, ZmWRKY82, and ZmWRKY95. Incw1 encodes an apoplast-localized invertase that catalyzes the irreversible hydrolysis of sucrose into glucose and fructose, facilitating carbohydrate unloading from source to sink tissues such as developing kernels. Its previously characterized physiological functions (detailed in Methods) combined predicted WRKY regulation, supported Incw1 as a downstream target of ZmWRKY67 for further functional validation.

To investigate the molecular mechanism underlying the transcriptional relationship between ZmWRKY67 and Incw1, we examined the expression levels of their expression levels in the hybrid CL137 × Ye478 and its parents. RNA-seq results showed that ZmWRKY67 was significantly upregulated in the hybrid compared to its parental lines (Fig. 4A). Similarly, Incw1 also exhibited increased expression in the hybrid (Fig. 4B), suggesting a potential regulatory relationship between ZmWRKY67 and Incw1. Promoter sequence analysis of the Incw1 promoter identified two putative ZmWRKY67-binding sites (W-box motifs) within the 2000-bp upstream region (Fig. 4C). To test whether ZmWRKY67 directly activates Incw1, we performed dual-luciferase assays in maize protoplasts. A reporter construct containing the Incw1 promoter fused to the LUC gene was co-expressed with either a control vector or a ZmWRKY67 overexpression construct (Fig. 4D). The results showed a significant increase in LUC/REN ratio in the presence of ZmWRKY67, confirming its role in transactivating Incw1 (Fig. 4E). A negative control using the ZmWRKY67 promoter without Incw1 ruled out nonspecific activation (data not shown). These findings provide functional evidence that ZmWRKY67 directly regulates Incw1, suggesting that ZmWRKY67-mediated transcriptional control enhances sucrose hydrolysis and transport, which may contribute to kernel development in the grain filling stage through improved carbohydrate partitioning.

Fig. 4.

Functional link between ZmWRKY67 and Incw1. (A, B) Expression levels of ZmWRKY67 (A) and Incw1 (B), measured in TPM, in female (F), hybrid (H), and male (M) individuals from the CL137 × Ye478 cross. The black horizontal line represents the median parental expression value. Significant differences (P < 0.05, LSD test) are indicated by different letters. (C) Schematic representation of WRKY binding motifs (W-box) in the promoter region of Incw1. (D) Dual-luciferase assay design. The Incw1 promoter (Pro Incw1) was fused to a LUC reporter and co-expressed with either a control vector or 35S:ZmWRKY67 effector in transient expression assays. (E) Relative promoter activity of Incw1, represented as the LUC/REN ratio. Renilla (REN) was used as an internal control. White bar: pIncw1 reporter with a control vector; Grey: pIncw1 and 35S:ZmWRKY67 effector co-transformed. Data represent three independent experiments (mean ± SD, n = 3). P < 0.001 (***), Student’s t-test. A control luciferase assay using the ZmWRKY67 promoter alone (35S:ZmWRKY67 + reporter construct backbone without Incw1) showed no nonspecific activation (data not shown).

Discussion

Heterosis is particularly evident during the early stages of embryo and endosperm development in maize hybrids, where gene expression often follows a non-additive pattern [14]. Since kernel development plays a critical role in determining hybrid maize yield, most studies on grain heterosis have focused on the F₁ generation derived from cross-pollinated parental inbred lines. However, there is limited research on how the ears produced by F₁ hybrids, which contribute to yield heterosis, lead to enhanced grain production. To address this gap, we analyzed F₁ seeds following self-pollination to gain deeper insights into the molecular mechanisms underlying yield heterosis.

Nonlinear relationship between heterosis and genetic distance

Predicting heterosis or hybrid performance based on parental genetic distance has been a long-standing goal in plant genetics and breeding. A classic heterosis estimation formula, $H_{F_1}=\sum d{y}^2$ [58], describes the contribution of each dominance effect (d) and allele frequency difference (y) between parental populations. When applied to inbred lines, where y = 1, this equation simplifies to $H_{F_1}=\sum d$, suggesting a linear relationship between heterosis and genetic distance. However, increasing evidence indicates that the relationship is nonlinear [59,60]. A humped quadratic polynomial model has been proposed, suggesting that heterosis peaks within an optimal range of parental genetic divergence before plateauing at higher levels of genomic differentiation [59,60].

A key challenge in assessing heterosis is the genetic background effect, which varies across hybrids, making it difficult to compare heterotic effects under a consistent genomic context. In this study, we leveraged SSSLs, where CL86 and CL137 each contain a specific genomic fragment from Qi319, while maintaining the Ye478 background. This controlled genomic framework enabled precise heterosis comparisons across hybrids while minimizing confounding genetic effects.

Our findings support a saturation model, where heterosis increases rapidly at low-to-moderate genomic heterogeneity (6.2 %–8.3 %) but plateaus at higher levels (91.7 %–93.8 %). Mini-F₁ hybrids (CL86 × Ye478 and CL137 × Ye478) exhibited greater heterosis than expected, whereas sub-F₁ hybrids (CL86 × Qi319 and CL137 × Qi319) showed heterosis levels comparable to the original F₁ (Qi319 × Ye478).

The observed pattern resembles the rising phase of a humped quadratic curve [59,60], supporting the hypothesis that heterosis follows a nonlinear trajectory with respect to genetic distance. However, since our study lacks hybrids with extreme genetic divergence, the dataset does not capture a potential decline in heterosis beyond an optimal divergence threshold. A complete humped quadratic curve, which depicts an initial rise followed by a decline, is typically observed when genetic distance spans a broader range, capturing both beneficial and detrimental effects of increasing divergence. In this study, the SSSL-derived hybrids and controlled genomic substitutions do not include hybrids with high genetic divergence. The original F₁ hybrid (Qi319 × Ye478) exhibited the highest heterosis, meaning our dataset represents only the increasing phase of the curve. To determine whether heterosis eventually decreases at extreme divergence levels, additional hybrids with broader genetic distances would be required.

A key question arising from these findings is why small genomic differences generate disproportionately high heterosis levels. One possibility is that minor genetic changes in key regulatory loci trigger transcriptional cascades and epistatic interactions, amplifying heterosis beyond expectations based on genetic distance alone. The DSDS system developed in this study enabled precise identification of heterosis-related genes, leading to the discovery of ZmWRKY67 as a central regulatory factor in heterosis.

The DSDS approach enhances heterosis-related gene identification

Understanding the genetic basis of heterosis remains a significant challenge due to the complex interplay of parental genetic divergence, allele-specific interactions, and gene expression regulation. Conventional transcriptomic analyses often yield large thousands of DEGs, making it difficult to pinpoint the key regulators in heterosis. Previous studies have attempted to refine DEG identification by reducing interference from parental genetic backgrounds. For instance, in rice, hybrids containing Oryza rufipogon introgressions exhibited enhanced yield as compared to control hybrids that lack the introgression, and the observed differences were attributed to introgressed hybrid loci [61]. Similarly, in maize, researchers removed 400 DEGs shared between hybrids and their inbred parents, revealing 2,531 uniquely expressed DEGs related with heterosis [62]. However, these approaches still capture a mix of background genetic variation and hybridization effects.

In contrast to these approaches, the DSDS approach developed in this study, integrates both parental genetic differences and the heterozygous background, allowing for a more refined identification of heterosis-related genes. This approach allowed to identify 359 and 80 heterosis-related DEGs in two mini-F₁s CL86 × Ye478 and CL137 × Ye478 (Fig. 3A & C), respectively, which significantly narrowing down candidate genes. Notably, only 22 DEGs in CL86 × Ye478 and a single DEG in CL137 × Ye478 were located within the substituted genomic fragment (Fig. 3B & D). The single DEG located within the substituted genomic fragment of CL137 × Ye478, identified as Zm00001d024376 (ZmWRKY67), was functionally validated in this study, which suggests its role as a potential key regulatory factor in heterosis. We also explored to address the relevance of the other 22 DEGs on the substituted region of CL86 × Ye478. Among them, Zm00001d016590 (ZmSWEET15b) belongs to the SWEET (sugars will eventually be exported transporter) family which play diverse roles in regulating plant development. The maize sweet15b knockout mutant has significantly reduced plant height and grain size [63], suggesting a potential role in heterosis-related pathways. The gene Zm00001d015064 is highly homologous to AT5G09790 (ATXR5) and AT5G24330 (ATXR6) in Arabidopsis, which encodes the histone-lysine N-methyltransferase and regulates H3K27me1 histone marks [64]. Our findings highlight the effectiveness of SSSL-based DSDS method in narrowing down heterosis-related genes, reducing false positives, and refining candidate selection compared to conventional transcriptomic approaches.

ZmWRKY67 as a potential transcriptional hub in heterosis

This study identified ZmWRKY67 as a candidate gene associated with heterosis, with experimental validation confirming its regulatory effect on Incw1, which encodes cell wall invertase 1 (CWI1) as a key enzyme for sucrose hydrolysis and carbohydrate partitioning [46]. Increased ZmWRKY67 expression in heterozygous regions correlated with elevated Incw1 expression in homozygous regions of mini-F₁ hybrids at 15 DAP, suggesting its role in enhancing carbohydrate transport during grain filling. Incw1 has been linked to yield traits: it is located within the qHKW5-2 locus associated with hundred-kernel weight [52], and has been identified in QTLs from multiple RIL populations showing significant associations with grain size and weight [51]. A conserved function is also supported by studies in rice (OsCIN1) and Arabidopsis (AtcwINV2) [50,51].

Notably, we observed this regulation in mini-F₁ CL137 × Ye478 but not in the other mini-F₁ CL86 × Ye478, indicating that multiple regulatory pathways may contribute to heterosis in maize. Incw2, the same family as Incw1, was reported to have a natural mutant mn1 at the Mn1 locus. Mn1 locus is essential for kernel development, and its loss results in the mn1 mutant phenotype, characterized by > 70 % seed weight reduction [48,49]. Notably, mn1 mutants is, however, non-lethal presumably because it retains a residual low level, ∼1% residual CWI activity through compensatory Incw1 expression, highlighting Incw1′s critical role in sucrose homeostasis [65]. Its regulation by ZmWRKY67 suggests a transcriptional mechanism contributing to carbohydrate metabolism in hybrids.

ZmWRKY67 was significantly upregulated in immature maize kernels following Aspergillus flavus inoculation, suggesting its role involved in resistance potentially via salicylic acid and ethylene signaling pathways [66]. In MaizeGDB, ZmWRKY67 is annotated as synonymous with ZmWRKY114 (https://www.maizegdb.org/gene_center/gene/Zm00001d024376, accessed 13 March 2025), which has been previously characterized in rice. ZmWRKY114 functions as a negative regulator in salt stress response via the ABA-mediated pathway [67]. It also binds to the W-box motif of the OsGA2ox4 promoter, regulating GA metabolism by activating GA2ox4, which reduces active gibberellin levels and inhibits plant height and growth [68]. These findings suggest that ZmWRKY67 may serve as a transcriptional regulator that integrates hormonal signaling, stress responses, and carbohydrate metabolism, all of which contribute to heterosis.

Transcription factors play a central role in heterosis by coordinating gene expression networks. A previous study demonstrated that many transcription factors exhibit non-additive expression patterns in hybrids, particularly during early ear development [69]. Hub regulators such as ZmEBP1 and ZmARF12 have been linked to tissue growth via cell proliferation and maize seed development via cell division activity in hybrids, respectively [69,70]. Additionally, transcription factors from the ARF, MYB, and bZIP families have been implicated in seed development and hybrid performance [70]. Given its role in transcriptional regulation of sucrose metabolism identified in the study, along with prior evidence of its regulatory functions, ZmWRKY67 may function as a similar transcriptional hub in heterosis, optimizing carbohydrate allocation, hormonal signaling, and stress adaptation to enhance hybrid growth and kernel development. Further studies are needed to elucidate its genome-wide targets and regulatory interactions to fully understand its contribution to heterosis.

The “Butterfly effect”: Small genomic variations trigger regulatory cascades to drive genome-wide expression changes in heterosis

While ZmWRKY67 was the only differentially expressed gene within the substitution fragment in CL137 × Ye478, its regulatory effects extended genome-wide, suggesting that heterozygous regions act as transcriptional hubs. This cascading effect, akin to a “butterfly effect”, demonstrates how small genomic changes can trigger large-scale transcriptomic shifts in hybrids. Our DSDS analysis supports this hypothesis, revealing that only a small number of DEGs were located in the substitution fragments, whereas most DEGs appeared in non-heterozygous regions. These findings indicate that heterozygous loci influence gene expression beyond their immediate genomic context, likely through epistatic interactions.

While our findings suggest that heterozygosity at a small substituted segment can trigger genome-wide transcriptional changes, the current evidence remains correlative rather than causal. ZmWRKY67 stands out as a strong candidate regulator based on its unique localization, expression pattern, and activation of downstream targets like Incw1, but its broader regulatory role remains to be confirmed. We use the term “butterfly effect”, or more scientifically “transcriptional cascade”, to describe this widespread regulatory influence. This concept is consistent with the “omnigenic” model depicted by Jakobson and Jarosz [71], where a few core regulators trigger widespread transcriptomic changes through both direct and indirect interactions with peripheral genes. Unlike classical epistasis, which involves defined interactions between two loci [72], our results reflect a diffuse, network-driven propagation. A simplified schematic is included in the graphical abstract (panel C). Further studies using chromatin mapping, network analysis, or transgenic validation will be needed to establish causality.

Epistasis plays a well-documented role in plant genetics. A two-way epistasis analysis in maize identified 19 gene × gene interactions associated with flowering time control [73]. In sorghum, the Maturity locus 1 (Ma1) suppresses the expression of the flowering activator Early heading date 1 (Ehd1), which in turn triggers anther production at the flowering stage [74]. Further epistatic QTL analysis demonstrated a significant interaction between Ma1-associated QTLs and flowering-time-associated QTLs [75]. Genome-wide association studies (GWAS) have also highlighted the dominant role of epistasis in trait regulation, showing that cumulative epistatic effects often exceed those of individual loci. For instance, a candidate gene-based GWAS revealed that the Hap-7A-15 locus on chromosome 7A interacting with multiple major and minor loci on multiple other chromosomes, accounting for 8.9 %–11.6 % of phenotypic variance [76].

A natural functional defect of the tRNA-modifying enzyme gene PDD^OL in rice chloroplasts provides a compelling example of how a small variation can trigger multiple cascading molecular effects. In a near-isogenic background, the defective PDD^OL allele leads to a reduction in tRNA modification in chloroplasts, resulting in translation errors in chloroplasts and impaired PEP activity. This disruption causes abnormal chloroplast translation and transcription, triggering retrograde signal transduction from chloroplasts to the nucleus, enhancing nuclear-encoded RNA polymerase activity and altering nuclear genes’ expression [77]. It demonstrated the “butterfly effect” of natural variations of tRNA-modifying enzymes in rice chloroplasts leading to pleiotropic developmental defects [78]. Similarly, in our study, we found that a genomic difference as small as 6.2 % can double heterosis levels (Fig. 2C & D) and induce widespread transcriptional variation across the genome (Fig. 3). Notably, ZmWRKY67, a key gene uniquely differentially expressed within a heterozygous genomic segment, appears to function as a regulatory hub, that is, to initiate genome-wide expression alterations much like a butterfly’s wings setting off a storm [71]. This regulatory cascade directly affects key genes such as Incw1, alongside more complex secondary responses.

Collectively, these findings suggest that the “butterfly effect”, where differential expression in heterozygous fragments induces widespread transcriptional changes in non-heterozygous regions, may be a fundamental driver of heterosis.

Future directions and implications of this study

A preliminary dual-reporter assay confirmed ZmWRKY67 activation of Incw1, suggesting a role in sucrose metabolism and transport that may contribute to heterosis by improving source-to-sink carbohydrate partitioning. However, further in vivo validation is needed to confirm ZmWRKY67′s regulatory network in hybrid maize. Additionally, sucrose transport and kernel-filling assays in ZmWRKY67-edited hybrids will help define its contribution to yield heterosis. These approaches will establish a causal link between ZmWRKY67 function, transcriptional regulation, and heterosis, advancing our understanding of heterosis mechanisms in maize. Notably, for proof-of-concept functional validation, we only examined ZmWRKY67, the sole DEG within the substitution segment in mini-F₁ CL137 × Ye478. However, in mini-F₁ CL86 × Ye478, 22 DEGs were identified in the substitution fragment region. A more comprehensive validation of these genes could provide deeper insights into additional regulatory mechanisms underlying heterosis. Future work should expand functional validation efforts to these DEGs, assessing their contributions individually and collectively to transcriptional reprogramming and phenotypic enhancement in hybrids.

For hybrid breeding, this study highlights the potential importance of strategic parental selection beyond simply maximizing genetic divergence. Our results suggest that intermediate levels of genomic divergence may contribute more effectively to hybrid performance than extreme divergence, though this pattern requires further validation across diverse genetic backgrounds. Additionally, the observation of a transcriptional “butterfly effect”, in which small heterozygous differences in substitution fragments appear to trigger widespread gene expression changes, offers a novel perspective on regulatory complexity in heterosis. Applying such insights to breeding remains challenging due to the complexity of epistatic interactions and the limited understanding of large-scale regulatory networks. Further work is needed to explore these mechanisms and to assess how such insights can be incorporated in future breeding strategies.

Conclusion

This study highlights the effectiveness of the SSSL-based DSDS approach in dissecting the molecular mechanisms of heterosis while minimizing genetic background noise. Using the SSSL-derived hybrids, the observed relationship between heterosis and genetic distance indicates that hybrid performance follows a nonlinear trend, where rapid gains occur at low-to-moderate divergence levels before reaching a plateau. Transcriptomic analyses reveal that most heterosis-related DEGs function beyond substitution regions, supporting a trans-regulatory mechanism. Among the identified DEGs, ZmWRKY67 emerged as a potential transcriptional hub, with experimental evidence linking it to Incw1, a key regulator of sucrose metabolism. Our findings suggest a “butterfly effect”, in which small genomic variations in heterozygous regions trigger widespread transcriptional changes, potentially shaping hybrid performance. These insights provide a foundation for refining hybrid breeding strategies by optimizing genomic heterogeneity and leveraging regulatory interactions. Future research should expand functional validation to additional DEGs in substitution regions to uncover diverse regulatory pathways contributing to heterosis. Incorporating transcriptional regulatory networks into hybrid prediction models could further enhance breeding efficiency and improve crop productivity.

Data Availability Statement

The original phenotypic data can be obtained from the attachment file (Table S4).

The transcriptome sequencing data have been deposited in the Genome Sequence Archive (GSA) at the National Genomics Data Center (NGDC) under the accession number CRA026054.

Compliance with Ethics Requirements

This article does not contain any studies with human or animal subjects.

Funding

This research was supported by the National Key Research and Development Program of China (2022YFD1200802 & 2016YFD0101201), the Science and Technology Innovation Team of Maize Modern Seed Industry in Hebei (21326319D), the Key Research and Development Program of the Seed Industry Innovation of Yangling (Ylzy-ym-05) and the Agricultural Science and Technology Innovation Program (CAAS-ZDRW202503).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: