Transcriptional regulation of maize (Zea mays) ear- length heterosis


 Background: Maize (Zea mays) ear length, which is an important yield component, exhibits strong heterosis. Understanding the potential molecular mechanisms of ear-length heterosis is critical for efficient yield-related breeding.Results: Here, a joint netted pattern, including six parent-hybrid triplets, was designed on the basis of two maize lines having long (T121 line) and short (T126 line) ears. Global transcriptional profiling of the young ears (containing meristem) was performed. Multiple comparative analyses revealed that 874 differentially expressed genes are mainly responsible for the ear-length variation between the T121 and T126 lines. Among them, four key genes, Zm00001d049958, Zm00001d027359, Zm00001d048502 and Zm00001d052138, were identified as being related to meristem development, which corroborated their roles in the superior genetic effects on ear length in the T121 line. Non-additive expression patterns were used to identify candidate genes related to ear-length heterosis. A non-additively expressed gene (Zm00001d050649) was associated with the timing of meristematic phase transition and was determined to be the homolog of tomato SELF PRUNING, which assists SINGLE FLOWER TRUSS in driving yield-related heterosis, indicating that Zm00001d050649 is a potential contributor to drive heterotic effect on ear length. Conclusion: Our results suggest that inbred parents provide genetic and heterotic effects on the ear lengths of their F1 hybrids through two independent pathways. These findings provide comprehensive insights into the transcriptional regulation of ear length and improve the understanding of ear-length heterosis in maize.

differentially expressed genes are mainly responsible for the ear-length variation between the T121 and T126 lines. Among them, four key genes, Zm00001d049958, Zm00001d027359, Zm00001d048502 and Zm00001d052138, were identi ed as being related to meristem development, which corroborated their roles in the superior genetic effects on ear length in the T121 line. Non-additive expression patterns were used to identify candidate genes related to ear-length heterosis. A non-additively expressed gene (Zm00001d050649) was associated with the timing of meristematic phase transition and was determined to be the homolog of tomato SELF PRUNING, which assists SINGLE FLOWER TRUSS in driving yield-related heterosis, indicating that Zm00001d050649 is a potential contributor to drive heterotic effect on ear length.
Conclusion: Our results suggest that inbred parents provide genetic and heterotic effects on the ear lengths of their F1 hybrids through two independent pathways. These ndings provide comprehensive insights into the transcriptional regulation of ear length and improve the understanding of ear-length heterosis in maize.

Background
Heterosis is a phenomenon in which F1 hybrid progeny exhibit superior performances compared with those of their parents [1][2][3]. It is used for hybrid crop breeding, which has greatly increased the productivity of many crops worldwide [4,5]. The successful exploitation of heterosis has also led researchers to determine its essential features. From the formation of two hypotheses (dominance and over-dominance) [6,7] to the identi cation of genetic components [8][9][10], as well as comprehensive analyses of genomes, transcriptomes and metabolomes [11][12][13][14][15][16], tremendous efforts have been made to elucidate the mechanisms responsible for heterosis. However, the underlying molecular mechanisms remain poorly understood.
Transcriptional regulation plays roles in various aspects of plant growth and development. Variation in transcriptional regulation promotes phenotypic diversity in all species [17] and, thus, is a potential source of heterosis that could explain the differences between F1 hybrids and their parental lines. Many transcriptome analyses between hybrids and inbred lines have been carried out in both maize (Zea mays) and rice (Oryza sativa), and a great number of differentially expressed genes (DEGs) were found in the F1 hybrids compared with their parents [4,[18][19][20]. Thus, heterosis was assumed to result from the global variation in gene expression between hybrids and inbred lines. However, several kinds of gene expression modes were observed in F1 hybrids: mid-parent (MP) (additivity), high and low parent (high and low parent dominance, respectively), above the high parent (over-dominance) and below the low parent (under-dominance) [21]. These data revealed that in hybrids, some genes exhibit non-additive expression patterns (not the expected MP level), which suggested a potential association with heterosis [22][23][24]. These expression differences may be caused by allele-speci c expression (ASE), which refers to the characteristic of preferentially expressing one parental allele in the hybrid owing to variations in regulatory sequences from the parental genome [25,26]. Trans-and cis-regulation frequently exist in different parental lines, and they might be responsible for inducing ASE in hybrids [17,27]. Consequently, analyzing transcriptional regulation is a valuable strategy for untangling the molecular basis of heterosis.
The measurement of heterosis involves a speci c trait. Moreover, the heterotic level is highly variable depending on the species, the cross parents and the trait(s) of interest [18]. Maize is an important food crop worldwide, and it exhibits superior heterosis for a wide range of traits. In addition, its inbred lines have been classi ed into several "heterotic groups" on the basis of their heterotic level [28,29]. Generally, crosses of parents within heterotic groups produce less heterosis than crosses of parents in different groups. This suggested that the inbred lines in each group may have speci c exclusive properties that contribute to heterosis. Thus, between-group crosses are more likely to produce greater heterosis. Maize ear length is a representative trait with a superior heterotic level, and it contributes greatly to grain yield [30]. A thorough knowledge of the transcriptional regulation of ear-length heterosis will aid in understanding the molecular basis of heterosis.
Maize ear length is predetermined, to some extent, by the activity of the ear primordium. As the ear primordium (meristem) differentiates, the visible young ear gradually elongates, revealing heterosis, and it is positively correlated with the nal ear length [30]. In this study, two speci c maize lines, T121 and T126, with long and short ears, respectively, were identi ed. When crossed with the other lines, the T121 line, compared with the T126 line, produced a series of hybrids with longer ears. However, equal earlength heterosis was observed in their corresponding hybrids. With these two speci c maize lines, we performed comprehensive transcriptional pro ling of the young ears using a joint netted pattern to determine the underlying cause of long ear in the T121 line and to gain insights into ear-length heterosis.

Results
The performance of ear length in different inbred-hybrid triplets Ear-length heterosis in maize is a very striking phenomenon resulting from a cross of two distinct inbred lines. To explore ear-length heterosis, we selected two speci c inbred lines (T121 and T126) with long and short ears, respectively. Additionally, two other inbred lines (PH4CV and PH6WC) were used to form a joint netted pattern (Fig. S1) that included six parent-hybrid triplets to adequately analyze ear-length heterosis.
During maize ear differentiation, the young ear gradually elongates and becomes visible. Moreover, the elongation capability of the growth cone determines the nal ear length to some extent. Here, we compared the morphologies of young ears of hybrids and their inbred parents at the 13-leaf stage when the young ears were initially apparent. The young ears of the T121 line were longer than those of the T126 line. Moreover, the F1 hybrids generated by T121 crosses (T121 × PH4CV and T121 × PH6WC) had   longer young ears than those generated by corresponding T126 crosses (T126 × PH4CV and T126 ×  PH6WC) (Table 1, Fig. 1a). In addition, the lengths of young ears from each inbred parent were less than those of their F1 hybrids (Table 1, Fig. 1a), indicating that ear-length heterosis had already emerged.
At the maturation stage, we measured the nal ear lengths of all the lines (inbred and hybrid). The T121 line, a long-ear inbred line, had an ear length that reached 19.52 ± 1.18 cm and was much longer than that of the T126 (13.52 ± 0.83 cm) line (Table 1, Fig. 1b). The six F1 hybrids, T121 × PH4CV, T121 × PH6WC, T126 × PH4CV, T126 × PH6WC, T121 × T126 and PH6WC × PH4CV, exhibited MP heterosis for ear length (Table 1). Interestingly, the hybrids produced by the T121 line (long ear), T121 × PH4CV and T121 × PH6WC, had longer ears than those produced by the short ear line T126 (Fig. 1b). However, for the MP heterosis, there was no signi cant difference between the other corresponding hybrids, such as T121 × PH4CV vs T126 × PH4CV (Table 1). These data indicate that the T121 line makes a superior contribution to ear length than the T126 line, but this is not a result of ear-length heterosis.
Transcriptome pro les of maize young ears among four inbred parents and six F1 hybrids To understand the comprehensive transcriptional regulation of maize ear-length heterosis, young ears of four inbred parents and six F1 hybrids were used to perform an RNA-sequencing (RNA-seq) analysis at the 13-leaf stage. In total, 20 libraries (10 × 2) were constructed for deep Illumina NovaSeq 150-bp pairedend sequencing. When the sequencing was completed, 332,543,189 raw reads were generated, ranging from 13.87 million to 21.76 million per library ( Table 2). After ltering, 320,828,384 clean reads, accounting for 96.48% of the total, were maintained ( Table 2). Based on the B73 maize reference genome (Version 4), the average unique mapping rate was 85.47%, with a range from 79.03% to 87.93% (Table 2). Moreover, the two biological replicates were in close agreement (Fig. S2). Finally, 25,199 unique genes were identi ed that had speci c expression levels in all the lines (Table S1). The RNA-seq data is available for further analyses of transcriptional regulation.
Global transcriptome changes from inbred parents to their hybrids Variation in gene expression is closely associated with phenotypic diversity. Thus, a series of transcriptional changes should occur from two inbred parents to one hybrid. For the T121-T126-T121 × T126 triplet, 64.97% of the genes in T121 × T126 hybrid kept their expression levels within the parental range, whereas the expression levels of the remaining genes (35.03%) were out of this rang (Table 3). This data indicated that the hybrid had the su cient potential to surpass the two parents. Using a differential expression analysis, 5,027 DEGs were identi ed between T121 and T126, 2,547 DEGs were identi ed between T121 × T126 and T121, and 2,431 DEGs were identi ed between T121 × T126 and T126 ( Fig. 2a; Table S2). Thus, the number of DEGs between a hybrid and one parent (T121 or T126) was less than that between the two parents. Moreover, similar scenarios, including the ranges of the gene expression levels and the numbers of DEGs, were found in other parent-hybrid triplets (Table 3; Fig. 2a; Table S2). Thus, some transcriptional regulatory mechanisms appeared to be universal and common in the production of hybrids from inbred parents.
In hybrids, ASE frequently exists, increasing the plasticity of gene expression governed by diverse alleles from the two parents, and this may be the reason that non-additive patterns appear in F1 hybrids. Thus, we analyzed genes having ASE in all the parent-hybrid triplets and then compared them with nonadditively expressed genes identi ed in the same triplet. Quite a number of genes having ASE were detected in the hybrids (Table S4). However, few of them were non-additively expressed ( Fig. 2c-h). For instance, in the T121 × PH4CV hybrid, 1,702 genes having ASE were identi ed, but only 71 exhibited nonadditive expression patterns (Fig. 2c). These results indicated that in F1 hybrids, ASE might have a limited contribution to the production of non-additive expression-related variation.
The major genes responsible for the ear-length variation between T121 and T126 lines The T121 line produces longer ears than the T126 line at the 13-leaf stage. A transcriptional level analysis of young ears revealed a large number of DEGs (5,027) between the T121 and T126 lines. However, it was di cult to determine the major genes responsible for the ear-length variation.
Nevertheless, compared with the T126 line, the T121 had a longer ear and might pass this advantage to its F1 hybrid. When T121 and T126 were hybridized with the other parents (PH4CV and PH6WC), the former produced F1 hybrids with longer ears compared with the latter. Consequently, we performed a differential expression analysis between the corresponding F1 hybrids, T121 × PH4CV vs T126 × PH4CV and T121 × PH6WC vs T126 × PH6WC (Table S2). In total, 890 DEGs were found to overlap between the two groups (Fig. 3a). We compared these overlapped genes with the DEGs identi ed between lines T121 and T126. As expected, they shared many common genes (874) (Fig. 3b), which suggested that these genes take part in the regulation of ear elongation and are mainly responsible for the ear-length variation between the T121 and T126 lines.
A gene ontology (GO) enrichment analysis was performed to identify some major terms related to ear length, as well as the key genes implicated in ear-length heterosis. A total of 1,672 GO terms were enriched for these genes in biological process (Table S5). Furthermore, the top 10 GO terms were investigated, and they revealed several terms related to development, such as GO:0048582 (regulation of post-embryonic development) and GO:0048831 (regulation of shoot system development) (Fig. 3c; Table  S5). Among these terms, four genes (Fig. 3d), Zm00001d027359 (FUSCA homolog, FUS6), Zm00001d048502 (COP9 signalosome complex subunit 1, CNS1), Zm00001d052138 (E3 ubiquitinprotein ligase, COP1) and Zm00001d049958 (WD40 repeat domain family protein, CYP71), were found to also belong to GO:0048507 (meristem development), and they may make major contributions to earlength variation.
Non-additively expressed genes contributing to ear-length heterosis Non-additively expressed genes may be potential sources of heterosis [31]. To identify promising potential genes that contribute to ear-length heterosis derived from the T121 (or T126) line, we made multiple comparisons of non-additively expressed genes in these parent-hybrid triplets. For the T121 line, 47 non-additively expressed genes overlapped among hybrids produced by T121 × PH4CV, T121 × PH6WC and T121 × T126 (Fig. 4a). Whereas, for the T126 line, 50 common non-additively expressed genes were identi ed among hybrids produced by T126 × PH4CV, T126 × PH6WC and T121 × T126 (Fig.  4b). These genes should be involved in ear-length heterosis, because the ear lengths of all these F1 hybrids surpassed the MP values. Moreover, 19 genes were shared (Fig. 4c, d), and these genes displayed non-additive expression patterns in all the hybrids, suggesting that they had the potential to contribute to ear-length heterosis. The GO enrichment analysis revealed that the top 10 GO terms for the T121 and T126 lines were highly similar (Fig. 4e, f; Table S6, 7), suggesting that there are some common components of the mechanism underlying ear-length heterosis. Among the common GO terms, GO:0048506 (regulation of timing of meristematic phase transition) and GO:0048510 (regulation of timing of transition from vegetative to reproductive phase) were associated with meristem, and a shared gene, Zm00001d050649 (ZCN2), may be responsible for ear-length heterosis.

Validation of candidate gene expression by quantitative real-time PCR
The application of RNA-seq technology has greatly enhanced the global understanding of transcriptional regulatory networks. To verify the accuracy of the RNA-seq analysis, we performed a quantitative realtime PCR (qRT-PCR) analysis of ve candidate genes, including four DEGs having genetic effects on ear length, Zm00001d027359, Zm00001d048502, Zm00001d052138 and Zm00001d049958, and one nonadditively expressed gene having heterotic effects on ear length, Zm00001d050649. Primers were designed to speci cally amplify each of the ve genes (Table S8). These primers were used to conduct qRT-PCR on three biological replications of RNA from re-prepared samples. All the assayed genes showed expression patterns similar to those determined by RNA-seq (Fig. 5), verifying the reliability of our RNAseq analysis.

Discussion
Two independent pathways of genetic and heterotic effects on ear length in maize Three major hypotheses, dominance [6,32], over-dominance [7,33] and epistasis [34,35], have served as the foundation for exploring the genetic and molecular causes of heterosis. The core notions are complementation within alleles, interactions within alleles and interactions between alleles, respectively [2]. All of them highlight the potential contributions of alleles (preferably considered as genetic loci) from two parental inbred lines to the F1 hybrid. Indeed, conventional genetic loci refer to quantitative trait loci (QTL) that control the genetic effects for trait performance, while heterotic effects are determined by special genetic loci, de ned as heterotic loci (HL), based on MP heterosis [36]. Using recombinant inbred lines and immortalized F 2 populations, some researchers have identi ed large numbers of QTL and HL, respectively. Interestingly, very few of the QTL and HL overlap, indicating that two independent pathways are responsible for their respective contributions to trait performance [10,37]. However, quite a few overlapped loci were revealed in another study [9]. More information is needed to determine the relationships between genetic and heterotic effects.
Gene expression is a complex process involving a series of transcriptional regulations that affect an individual's phenotype [17]. Transcriptional regulation plays a role in explaining the molecular mechanisms of heterosis that bene t F1 hybrid individuals with a superior trait performance compared with the parental inbred individuals [19,38]. Transcriptome pro les of two inbred lines and their hybrids have been determined to investigate variations in global gene expression [39][40][41][42]. More genes are differentially expressed between the parents than between one parent and the hybrid. Moreover, the majority of DEGs between two parents are presented at the MP level (additive pattern) in their hybrids. Similarly, in this study, there were more DEGs greatly between the two parents than between each parent and the hybrids in all the triplets, and only a few genes (< 6%) displayed non-additive patterns (Fig. 2b). Thus, the prevailing additive pattern appears to limit the difference between one parent and the hybrids and makes a limited contribution to heterosis in the hybrids. Additionally, the T121 line produced longer ears than the T126 line, which indicated that the T121 line had a superior genetic effect on ear length. However, their corresponding hybrids, such as T121 × PH4CV vs T126 × PH4CV, produced nearly equal levels of heterosis for ear length (Table 1). Thus, the genetic effect (parental variation) and heterotic effect appeared to be uncorrelated. Similarly, the genetic distance between two parents is a limited predictor of heterosis in their F1 hybrids [18].
Indeed, the ear lengths of T121 × T126 hybrids easily surpassed that of the T121 line (extremely long ear). This was attributed to ear-length heterosis. Thus, the DEGs between T121 × T126 and T121 should be involved in ear-length heterosis, which increased the ear length compared with that of the T121 line. However, some of these genes were also identi ed as being responsible for the ear-length variation between the T121 and T126 lines, but their variation trends were opposite (up-/down-regulation) (Fig.   S3a, b). This indicated that these shared genes may not play roles in the ear-length heterosis. In this context, the two inbred parents would provide the genetic and heterotic effects on the ear lengths of their F1 hybrids through two independent pathways. The superior performance of the hybrid over that of the better inbred parent bene ts from the altered regulation of speci c genes having non-additive expression patterns.
A possible pathway resulting in the non-additive expression pattern from inbred parents to their hybrids From two inbred parents to one hybrid, it is tempting to infer that some new transcriptional regulations (non-additive expression patterns) contribute to heterosis. Determining the pathway responsible for the non-additive expression pattern may help to elucidate the mechanisms of heterosis.
Most transcriptional variation may be caused by sequence variation in regulatory regions of genes (cisregulation) or by functional variations in a regulator (trans-regulation) [43]. In hybrids, all genes consist of a pair of alleles derived from two parental inbred lines, respectively. Usually, the biased expression of alleles (ASE) takes place in some genes. If cis-regulation is present, then the allelic expression is expected to be additive in the hybrids, resulting in an additive expression pattern. In this study, we found that few genes having ASE present the non-additive expression pattern in all parent-hybrid triplets (Fig. 2c-h). Thus, this indicated that ASE is mainly caused by cis-regulation and makes a limited contribution to the non-additive expression pattern. This indicated that ASE is mainly caused by cis-regulation and makes a limited contribution to the non-additive expression pattern. Indeed, numerous studies [44][45][46][47] have revealed that cis-regulation plays major roles in the regulation of allelic expression in the hybrids of many species, and this may result in the production of large numbers of additively expressed genes.
Trans-regulation also occurs widely among regulatory networks, and it regulates the expression of many genes [17]. A de ciency in trans-acting factors in one parent leads to the differential expression of their target genes between parents. However, because F1 hybrids have the same genetic background, the sharing of trans-acting factors facilitates the balanced expression of allelic genes, resulting in nonadditive expression patterns. Here, we compared the two blocks: non-additively expressed genes in the F1 hybrids and the DEGs between the two parents. Several common genes were discovered in each triplet ( Fig. S4a-f). For example, in the T121-PH4CV-T121 × PH4CV triplet, 395 genes overlapped between the two blocks (Fig. S4a). This revealed that trans-regulation may cause the variation in gene expression between the two parents and produce the non-additive expression pattern in the F1 hybrids. However, such genes would be responsible for the variation in ear length between the two parents (genetic effect) but would not contribute to ear-length heterosis.
Indeed, there are numerous isolated non-additively expressed genes ( Fig. S4a-f), which were not differentially expressed between the two parents and presented non-ASE patterns in the F1 hybrids. One possible scenario is that cis-and trans-interactions occur. If the cis-regulated alleles also harbor variations in functions (trans-regulation), then non-additive expression patterns would be produced. An excellent example has been reported in tomato hybrids, in which yield was improved by ne-tuning the expression of a transcription factor (MADS-box) and its trans-effects on the target alleles [48]. Alternatively, two or more trans-acting factors may combine to activate or suppress the expression of the target alleles in the F1 hybrids, leading to a non-additive expression pattern. For example, the two maize transcription factors, B and Pl, interact to up-regulate the expression of genes A1, A2 and Bz1, which control anthocyanin production [49]. An inbred line with a nonfunctional b or pl allele displays a green phenotype owing to the low, or absent, expression of genes A1, A2 and Bz1, but a hybrid with B/b Pl/pl alleles has high expression levels for genes A1, A2 and Bz1 and a red phenotype [46]. Thus, once the two parents are crossed, some speci c interactions, rather than either cis-or trans-regulation, play major roles in generating the non-additive expression patterns found in hybrids, contributing to heterosis.
Key genes having superior genetic effects on ear length in the T121 line One inbred line often transfers its excellent characteristic to its progeny, including F1 hybrids, and this can be attributed to its superior genetic effect. In this study, the T121 line produced an extremely long ear, and its ear length was far greater than that of the T126 line. Likewise, its F1 hybrids exhibited longer ears compared with those of the T126 line, indicating that the T121 line had a superior genetic effect on ear length.
Additionally, Arabidopsis CYP71 is a unique immunophilin with a WD40 domain, and it interacts with histone H3 to regulate gene expression patterns that determine plant organogenesis [58,59]. The CYP71 gene is preferentially expressed in meristem and other actively dividing tissues, and a loss of CYP71 function causes the arrest of apical meristem development [58,59]. Similarly, higher expression levels of Zm00001d049958 (CYP71) in the T121 line and its progeny (Fig. 3d; Fig. 5) was conducive to ear growth. In this context, a possible scenario is that the expression of Zm00001d049958 (CYP71) in maize axillary meristem suppresses the expression of Zm00001d027359 (FUS6/CNS1), Zm00001d048502 (CNS1) and Zm00001d052138 (COP1) genes and then promotes ear elongation. Overall, these four genes may be responsible for the genetic effects on ear length in the T121 line.
The potential contributors to the ear length of heterosis The performance of F1 hybrids mainly bene ts from two aspects: the genetic and heterotic effects derived from the two parents. Heterosis is speci c to different traits and may be attributed to speci c loci for a particular trait [46]. Independent of the loci, drastic transcriptional variations in key genes must take place in hybrids. Genes having non-additive expression patterns have been studied owing to their huge contributing potential to heterosis [31,60,61].
Maize ear length is an important agronomic trait that often exhibits super heterosis [46]. In this study, we analyzed global transcriptomes of young maize ears from six parent-hybrid triplets derived from four inbred lines. The Zm00001d050649 (ZCN2) gene displayed a non-additive expression pattern in all the triplets and belonged to GO:0048506 (regulation of timing of meristematic phase transition), implying its contribution to ear-length heterosis. Maize ZCN2 is a member of the TERMINAL FLOWER1 (TFL1)-like gene family, which is highly conserved in plants and is thought to function in the maintenance of meristem indeterminacy [62]. In Arabidopsis, TFL1 and FLOWERING LOCUS T (FT) are two antagonistic integrators of the oral transition pathways that function in repressing and promoting owering, respectively [63][64][65]. The tomato SINGLE FLOWER TRUSS gene, an ortholog of Arabidopsis FT, drives the heterosis for yield in an over-dominant pattern [66]. These heterotic effects depend on the genetic background having a mutation in SELF PRUNING (SP), an ortholog of Arabidopsis TFL1. If plants carry a functional SP gene, then heterosis is eliminated [66,67]. This suggested that the sp gene is a required contributor that drives heterosis for yield in tomato. Like the tomato sp gene, the lack of expression of the homologous maize ZCN2 gene in hybrids (Fig. 5) contributes to ear-length heterosis.

Conclusions
In this study, multiple comparative analyses of the transcriptional pro les of six parent-hybrid triplets revealed that the genetic and heterotic effects on ear length in maize contribute to the performance of F1 hybrids through two independent pathways. Four key genes, Zm00001d049958 (CYP71), Zm00001d027359 (FUS6/CNS1), Zm00001d048502 (CNS1) and Zm00001d052138 (COP1), were identi ed as being responsible for the superior genetic effects on ear length in the T121 line. Cis-and trans-regulatory interactions mainly caused the emergence of non-additive expression patterns in F1 hybrids, providing the potential to drive ear-length heterosis. The lack of expression of a non-additively expressed gene, Zm00001d050649 (ZCN2) was identi ed as potentially contributing to ear-length heterosis just as its homologous tomato SP gene contributes to yield heterosis. This will lead to investigations of the mechanism behind the silencing of Zm00001d050649 (ZCN2) in F1 hybrids, which will help further elucidate the mechanisms of heterosis. The present work provides insights into the transcriptional regulation of the maize ear-length characteristic from two parents to one hybrid. These ndings improve our understanding of ear-length heterosis in maize.

Materials And Methods
Experimental design and plant growth Four maize inbred lines were selected in this study. T121 and T126 are inbred lines having long and short ears, respectively, which were derived from our breeding lines. The other two inbred lines, PH4CV and PH6WC, are the two parents of the excellent hybrid 'Xianyu 335'. Then, each of them were crossed to six F1 hybrids (excluding reciprocal hybridization), resulting in a joint netted pattern (Fig. S1).
The four parental inbred lines and six F1 hybrids were planted in a specially designed plot.

DEG analysis
The criteria (statistical signi cance) of p-value < 0.05 and abs (log 2 (fold-change) > 1 were used to identify DEGs between two lines with the DESeq package (http://www.bioconductor.org/packages/). Between each two lines (inbred parents and F1 hybrids), a differential expression analysis was performed. Moreover, the non-additive expression pattern was analyzed in each parent-hybrid triplet. Nonadditively expressed genes were de ned as having differential expression levels between those of the F1 hybrid and the MP value.
A GO enrichment analysis was conducted to determine the essential functions of the DEGs (https://www.omicshare.com/tools). The top 10 GO terms were investigated to determine the major candidate genes. The threshold p-value < 0.05 were used for the analysis.

ASE identi cation
A speci c lter was required for mapping reads. Using a customized Perl script, desired reads that were perfectly mapped to one parental sequence and had single nucleotide polymorphisms mapped to the other were retained. Then, the re ltered reads were assembled according to the previously reported criteria [25]. In each triplet, the re ltered reads from the F1 hybrid were divided into two sets: set 1, reads aligned against one parent, and set 2, reads aligned against the other parent, to distinguish parent-speci c reads in the single nucleotide polymorphism calling step. The normalization of these read numbers was performed using the function estimateSize Factors from the DESeq package [73]. For each gene, ASE was called if the reads of each set deviated signi cantly from 1:1 by simple random sampling, which was validated by 1,000 permutations at a false discovery rate < 0.05.

Quantitative real-time PCR analysis
The same samples were re-prepared for a quantitative real-time PCR (qRT-PCR) analysis in an attempt to validate the expression patterns of key genes. The qRT-PCR was performed in a Bio-Rad CFX96 Real-Time PCR System with SYBR Green PCR Master Mix (Takara Bio). Three technical replicates were included in each plate for qRT-PCR. The Zm00001d013873 (ACTIN-2) gene was used as an internal standard to normalize gene expression, and the relative gene expression levels were measured using the 2 -ΔΔCt method [74]. Primers were designed online (https://www.ncbi.nlm.nih.gov/) for the genes studied, and the primer information is provided in Table S8.

Consent for publication
Not applicable.

Availability of data and materials
The datasets generated and analysed during the current study are available in the NCBI Sequence Read Archive (SRA) database under Bioproject PRJNA682653 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA682653).

Supplementary Information
Additional le 1: Additional le 5: Table S1 The mapped genes and their expression levels.
Additional le 6: Table S2 The differential expression analysis between different varieties. Additional le 7: Table S3 The non-additive expression analysis for the six parent-hybrid triplets.
Additional le 8: Table S4 The ASE analysis for the six parent-hybrid triplets.

Page 21/23
Additional le 9: Table S5 The GO enrichment analysis of the major genes responsible for the ear length variation between T121 and T126 lines.
Additional le 10: Table S6 The GO enrichment analysis of the non-additive expression genes for T121 line.
Additional le 11: Table S7 The GO enrichment analysis of the non-additive expression genes for T126 line.