A differentially methylated region of the ZmCCT10 promoter affects flowering time in hybrid maize

Zhiqiang Zhou,Xin Lu,Chaoshu Zhang,Mingshun Li,Zhuanfang Hao,Degui Zhang,Hongjun Yong,Jienan Han,Xinhai Li,Jianfeng Weng

Institute of Crop Sciences,Chinese Academy of Agricultural Sciences,Beijing 100081,China

Keywords: Maize Flowering time Hybrid performance QTL Epigenetic

ABSTRACT Flowering time(FT)is a key maize domestication trait,variation in which allows maize to grow in a wide range of latitudes.Although previous studies have investigated the genetic control of FT-related traits per se,few studies of FT hybrid performance have been published.We characterized the genomic architecture associated with hybrid performance for FT in a hybrid panel by testcrossing Chang 7-2 with 328 Ye478 × Qi319 recombinant inbred lines (RILs).We identified 11 quantitative trait loci (QTL) for hybrid performance in FT-related traits,including a major QTL qFH10 that controls hybrid performance and heterosis in a summer maize-growing region.However,this locus acts in regulating FT traits per se only in a spring maize-growing region.We validated ZmCCT10 as a candidate gene for qFH10 and found that differences between hybrids and their parental lines in DNA methylation in the differentially methylated region (DMR,-700 to -1520)of the ZmCCT10 promoter affected gene expression pattern and thereby FT in the summer maize-growing region.

1.Introduction

Human evolution has been closely connected with the domestication and improvement of crop plant species [1].Crop breeders currently face the daunting task of increasing crop yields by 100%-110% from 2005 to 2050 to meet future demands for food and feed [2].To achieve these yield increases,breeders are developing crops such as hybrid rice and maize that take advantage of heterosis,a phenomenon in which hybrid offspring exhibit performance superior to that of their parents[3].For this purpose,breeders must assess inbred lines not only by trait performanceper sebut also by their potential for creating hybrids with superior performance.However,because the performance of inbred lines poorly predicts the performance of hybrids,crosses must still be phenotypically evaluated [4,5].Improving our understanding of the genetic basis of hybrid performance is likely to lead to more effective exploitation of heterosis in crop improvement programs[6-8].

Three genetic models of heterosis,namely dominance[9],overdominance[10],and epistasis[11],have been proposed to explain heterosis,but our understanding of the molecular mechanisms underlying crop heterosis is still largely theoretical.The improved mapping of quantitative trait loci (QTL) in resequenced genomes and genome-wide association studies (GWAS) have enabled the identification of loci responsible for heterosis in crops such as maize and rice [10,12-15].A few heterotic QTL have been characterized systematically for traits such as grain yield and FT in hybrids [7,16,17].However,exactly how these QTL have improved hybrid performance remains unclear.High-throughput transcriptome profiling has revealed allelic variation leading to allele-specific expression of transcripts in hybrids that affects crop performance[18,19].Genes involved in the control of FT and circadian rhythm appear to be metabolically and physiologically linked to heterosis and hybrid performance inArabidopsis[20-22].

Epigenetic modifications such as DNA methylation in three sequence contexts(CG,CHG,and CHH(where H=A,C,or T))influence gene expression and phenotypic variation in rice,maize,tomato,and other plants [17,23,24].DNA methylation silences transposable elements (TEs) and other repetitive elements,and thus may result in altered chromatin states in hybrids [25].Insertion of a MITE transposon in theVgt1(vegetative to generative transition 1) locus was associated with a QTL for FT in maize[26,27].MITE insertion may disrupt a regulatory element ofVgt1and interrupt the stable secondary chromatin structure and/or change the chromatin state [28].DNA methylation may affect the genetic interactions of alleles such that nonadditive interactions result from methylated epialleles in hybrid progeny[29,30].In rice andArabidopsis,this notion has been supported by studies [29,30]that identified differentially methylated regions (DMRs) between parental lines that conferred nonadditive effects.Nonadditive genetic interactions resulting from differential methylation might also confer nonadditive effects on transcription and epigenetically controlled processes in hybrid progeny compared with their parental lines [30,31].But how DNA methylation and DMRs in the promoters of genes could affect hybrid performance remains unclear.

FT has strongly influenced the adaptation of crops to specific growing regions and thereby crop domestication and dissemination [32].Maize is a typical short-day (SD) plant that was domesticated from teosinte (Zea maysssp.parviglumis),in southwestern Mexico～9000 years ago [33].Natural and artificial selection for reduced photoperiod sensitivity has allowed maize to adapt to higher latitude long-day (LD) environments [34,35].FT is a highly heritable trait that is controlled by a well-described gene regulatory network [36].FT in cereals is controlled by florigen,which may also influence grain yield [14,16,37].Genes involved in the regulation of maize FT,such asZCN8[38],dlf1[39],zfl1[40],andconz1[41],have been cloned by mutational strategies.But although many QTL affecting FT have been mapped in diverse maize populations[34,42,43],relatively few of these QTL,includingVgt1[26],ZmCCT9[44],andZmCCT10[45],have been systematically characterized and shown to affect photoperiod response.TE insertions affect the regulation of downstream genes by all three of these loci,suggesting that TEs might have developed distinct distribution patterns in the maize genome during the expansion of maize cultivation from low-to high-latitude regions [14].The CCT family of genes that encode transcription factors (TFs) such as CONSTANS,CO-like,and TOC1 that contain a conserved CCT domain,affect FT and grain yield in rice and sorghum [46].But despite the achievements made in elucidating in inbred lines the molecular details of photoperiod response in maize flowering,how these FT-associated genes,especially CCT-family genes,might regulate hybrid performance for FT in maize hybrids is still unclear.

To investigate the mechanisms underlying hybrid performance(heterosis)for FT in maize,we developed a testcross population by crossing a set of recombinant inbred lines(RILs)with a female parent.We evaluated the progeny in at least four environments to characterize the variation among hybrids and their parents for FT-associated traits.The objectives of this study were to(1)assess the correlation between the traitsper seof RILs and their hybrid performance,(2) identify the genetic basis of hybrid performance for FT in maize,(3) testZmCCT10as a candidate gene forqFH10,and(4)test whether a DMRs in the promoter of the candidate gene affected gene expression and flowering time in hybrid maize.

2.Materials and methods

2.1.Population development

A testcross (TC) progeny was developed from a cross between each of 328 recombinant inbred lines (RILs) and an elite inbred.The RIL population was derived from a set of 365 RILs produced by 11 generations of single-seed descent from a cross between Ye478,which carries a CACTA-like TE insertion at theZmCCT10promoter,and Qi319,which lacks this TE.This population has been described [47,48].Each RIL and the two parents were crossed as male parents to the elite inbred line Chang 7-2,which carries a TE inZmCCT10,to generate the TC progeny.Chang 7-2 is a line with high general combining ability derived from Huangzaosi.The three parental lines Ye478,Qi319,and Chang 7-2 were derived from the PA,PB and SPT heterotic groups,respectively [5],and all of them have been widely grown in the Yellow and Huai River Valley growing regions of China.To further define the function ofqFH10,a near-isogenic line(NILQi319)that covered the homozygousqFH10region was developed by DNA marker-assisted selection(MAS) using Ye478 as recurrent parent and Qi319 as donor.

2.2.Field design and phenotyping

The RIL population and its parents were grown in five locations across two years under distinct environmental conditions and latitudes spanning from 18.2°N to 43.4°N.Locations in the northeastern maize-growing region were Gongzhuling (GZL,43.4°N,124.8°E) and Beijing (BJ,39.4°N,116.2°E) in 2013 and 2014.Two summer maize-growing regions: Shijiazhuang (SJZ,37.2°N,113.3°E) in Hebei province and Xinxiang (XX,35.1°N,113.5°E) in Henan province,were chosen based on optimum growth conditions for the parental lines in 2015 and 2016 (Fig.S1).The other location was Sanya in Hainan province (SY,18.2°N,109.3°E) in 2016 and 2017.Meteorological information for these locations was retrieved from the China Meteorological Administration at https://www.cma.gov.cn/.Although Sanya is not defined as a summer maize-growing region,it was an excellent environment for identifying flowering phenotypes.The cluster analysis for locations was performed using the Vegan package inRas the Bray-Curtis distance based on growing-degree days(GDD)for maize in the tested years.GDD=∑[(Tmax+Tmin)/2-10].The mean of the daily maximum temperatures (Tmax) and minimum temperatures (Tmin) was used as daily mean temperature and 10 °C was set as the base developmental temperature [45].In a cluster analysis of GDD,the two high-latitude locations GZL and BJ grouped into one clade and SJZ and XX grouped with SY into another clade (Fig.S2).All plants in these locations were grown in randomized blocks with two biological replicates under standard practices for maize cultivation,except that in SY only one biological replicate was planted.

For the TC population,all hybrids were evaluated in SJZ and XX in 2015 and 2016 because the line Chang 7-2 is widely grown in the Yellow and Huai River Valley growing regions.Field evaluations of the hybrids were performed using an alpha lattice design with two replicates.Each plot contained two rows sown at a density of 60,000 plants ha-1.The field was managed according to standard agricultural practice.The cultivars CY and CQ (Ye478 and Qi319,crossed separately to Chang 7-2)were used as controls in each replication.The FT-related traits days to silking (DS),days to tasseling (DT),and days to pollen shed (DP) were estimated in all tested lines.DS was measured as the number of days between planting and date when the silks on half of the plants in a plot were visible on the ear.DT was scored as the date when half of the plants in a plot displayed male tassels.DP was assessed as the number of days between planting and the date when the central spikes of the tassels on half of the plants were shedding pollen.Mid-parent heterosis (MPH) was calculated as the phenotypic difference between the hybrid and midparent phenotypic values which defined as the mean of Chang7-2 and RILs[7].Best linear unbiased prediction(BLUP)values for each trait under spring maize-growing region,summer maize-growing region and SY conditions across two years were used for phenotype and QTL analysis.

2.3.Generation of F1 genotypes and neighbor-joining tree construction

All three parental lines were sequenced at effective sequencing depths of about 30× on the Illumina Infinium platform (Illumina,Inc).The RILs were then genotyped using genotyping by sequencing (GBS) technology with each line represented by～0.07× geno me coverage [47].Sequence data for the parental lines were filtered to remove all sites in the genome with incomplete information,triallelic sites,and any sites that did not exhibit polymorphism.After filtering,328 F1hybrid genotypes were generated by combining their corresponding parental genotypes[49],leaving 17,446 SNPs with MAF (minor allele frequency) ≥0.05 for further analysis.The IBD (identity by descent) analysis of Plink [14] was used to separately calculate the pairwise parental genetic distances between Chang 7-2 and each of the 328 RILs using with the parameters -maf 0.05 -map3 -noweb -cluster -distance matrix -out.MEGA6[50] was then used to generate and visualize the phylogenetic trees.

2.4.QTL mapping

An ultra-high density linkage map [47] consisting of 4602 bin markers was constructed previously using GBS technology to genotype the RIL population.A bin marker [47] was designated when consecutive 100-Kb intervals lacked a recombination event in the entire population.The map spanned 1533.72 cM with a mean distance of 0.33 cM between adjacent bin markers.QTL analyses were performed separately for each FT-related traitper se,hybrid performance,and heterosis.QTL analysis was performed using the composite-interval mapping (CIM) method with theR/qtlpackage[51].The thresholds of logarithm of odds (LOD) were set to 3,and confidence intervals for each QTL were set as 1.5-LOD drops on either side of the peak LOD scores.Thefitqtlfunction was used to calculate the proportion of phenotypic variation explained by each QTL.Any QTL effects identified in the testcross populations indicated differences between Chang 7-2/Qi319 and Chang 7-2/Ye478 heterozygotes.Hypothetically,QTL detected in both the hybrids and their corresponding RILs would exhibit dominant genetic effects,while QTL identified only in the hybrids without corresponding QTL in the corresponding RILs would exhibit overdominant genetic effects (Table S1) [12].The mapped QTL identified for traitsper sewere designated asqDSx(for DS),qDTx(for DT),andqDPx(for DP),wherexrepresents the maize chromosome number.Correspondingly,the QTL detected in the testcross population for hybrid performance and MPH were thus namedfDSx,fDTx,fDPx,hDSx,hDTx,andhDPxfor DS,DT,and DP,respectively.The termfrepresents the F1hybrids andhrepresents heterosis.Because we later identified one major QTL located on chromosome 10 with pleiotropic functions affecting hybrid performances for FT and MPH but not for FT-related traitsper sein the summer maizegrowing region,we tentatively named this QTLqFH10.

Association analyses were conducted using the FaSTLMM algorithm in the easyGWAS framework [52] with 17,446 combined SNPs in 328 hybrids.For the testcross population,association analyses were performed using overdominance genotype encoding for the SNP matrix,wherein the codes for both of the homozygous genotypes were 0 and the code for heterozygous genotypes was 1.A kinship matrix was then calculated for the encoded data.Association mapping was performed using an overdominance model for both hybrid performances and MPH for each of the FT-related traits.To compensate for Type I error associated with multiple comparisons,a conservative 5% Bonferroni correction [49] of 0.05/[number of tested SNPs (17,446)]=2.8 × 10-6was applied to identify modestly significant SNPs associated with each trait.

2.5.2.5 Genotyping of ZmCCT10

To develop DNA markers to help determine the function ofZmCCT10in FT-related traits,the promoter and coding regions ofZmCCT10were amplified from the genomic DNA of the three parental lines using TransStart FastPfu DNA polymerase (TransGen Biotech,China) for sequencing.A TE-associated marker was designed based on theCACTA-like TE and its flanking sequences to detect the presence or absence of this TE in theZmCCT10promoter in these three parenal lines.TheZmCCT10element was detected in Qi319 genomic DNA,but no polymerase chain reaction(PCR) product was amplified from Ye478 and Chang 7-2 genomic DNA.All primers used for PCR amplification and DNA sequencing in the study are listed in Table S3.

2.6.Measurement of ZmCCT10 expression

Seeds of the parental lines Ye478,Qi319,NILQi319,and Chang 7-2 and the F1hybrids CY and CQ were planted in a chamber under short-day (SD,9 h light/15 h dark) or long-day (LD,15 h light/9 h dark) conditions.Temperatures in the chamber were maintained at 22 °C in the dark and 26 °C in the light.Leaf samples were collected randomly from four plants that had reached the V4 growth stage.Total RNA was isolated as described above,and cDNA was synthesized using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems,Foster City,CA,USA).Measurement ofZmCCT10transcript expression was performed with real-time(RT)-PCR on a 7500 Fast Real-Time PCR System (Applied Biosystems),using the housekeeping gene GADPH as an internal control.All PCR amplifications were performed with three biological and technical replications under the following conditions: 95 °C for 5 min,followed by 40 cycles of 95 °C for 10 s and 60 °C for 20 s[45].Relative expression ofZmCCT10was calculated using the 2-ΔΔCTmethod [53].ZmCCT10and GAPDH sequences were retrieved from the maize database in Gramene (https://ensembl.-gramene.org/Zea_mays/Location).Primers for amplifying these two genes are shown in Table S3.

2.7.Methylation sequencing

Genomic DNA was extracted by a standard CTAB method as described previously [47].For each sample,about 1 μg genomic DNA isolated from V4-stage leaves of Ye478,Qi319,Chang 7-2,CY,and CQ grown under SD conditions was fragmented to 200 bp using a Biorupter Pico Sonicator (Diagenode,Belgium).The fragmented DNA was then end-repaired,dA-tailed,and ligated to Illumina adapters with all cytosines methylated using the Meth-Cap Fast Library Preparation Kit (iGeneTech,Beijing,China).The 2.4-kb promoter ofZmCCT10was enriched using four degenerate probes with the TargetSeq Enrichment Kit (Thermo Fisher,USA),treated with bisulfite using the EpiTect Fast DNA Bisulfite Kit(Qiagen),and then amplified using AceQ U+Probe Master Mix(Vazyme,Nanjing,China)to generate a bisulfite sequencing library of the target region.Four degenerate probes that covered 95.54%of theZmCCT10promoter region(-36 to-487,-505 to-825,-847 to-1193,and-1225 to-2428 relative to-1 at the transcription start site(TSS))were used to amplify bisulfite-treated DNA before deep sequencing using the HiSeq X Ten system (Illumina,Inc).An Illumina NovaSeq 6000 sequencer was used to obtain paired-end 150-bp reads from the target region bisulfite sequencing library.The proportion of cytosine methylation in each context (CG,CHH,or CHG) was calculated [30] for each genotype and differences in methylated regions between these genotypes were tested with Student’st-test.

2.8.Data availability

The authors state that all data necessary for confirming the conclusions presented in the article are represented fully within the article.All original genotypic datasets of the RIL population have been submitted to the Sequence Read Archive(SRA)at the National Center for Biotechnology Information (NCBI) under accession ID PRJNA627044.

3.Results

3.1.Comparison of flowering time between RIL and testcross populations

The phenotypic values for DS,DT,and DP in RILs and TC hybrids were compared(Fig.1).The values of DS,DT,and DP varied widely and showed approximately normal distributions in both populations.The means of the three traits in the TC populations were lower than the corresponding values in the RILs,suggesting that the performance of each heterozygous F1was superior to that of its corresponding parental lines (Fig.1A).All hybrids showed negative MPH with large variation for all three traits (Fig.1B).Significant correlations for each trait mean were detected (Fig.1C,|r| ＞0.50;P＜ 1.90E-05) among individual TCs,parental lines and MPH,suggesting that genetic variation among paternal lines was associated with flowering-time differences among hybrids.Because heterosis observed in hybrids has been hypothesized to be proportional to the genetic distance between parental lines,if multiple paternal lines are testcrossed with a common maternal line,the heterosis of their hybrids should correlate positively with the genetic distance between parents.To test this hypothesis,17,446 high-quality SNPs were used to estimate genetic distances between Chang 7-2 and each of the 328 RILs.Although the paternal lines were more genetically diverse than Chang 7-2,no correlation between the genetic distance between parental lines and MPH for FT in hybrids was found(Fig.1D,E),suggesting that interparental genetic distance does not necessarily determine or predict hybrid performance for FT in maize.

Fig.1.Extensive phenotypic variation and heterosis for flowering time(FT)-related traits in 328 maize hybrids.(A)Means and ranges of three FT-related traits measured in the RILs and their F1 progenies.(B)MPH for FT in 328 maize hybrids.(C)Significant correlations for each trait mean among individual TCs,their parental lines and MPH.(D)The phylogenetic positions of 328 parental lines and 328 F1 in the neighbor-joining tree,with colors for clades indicating genotypes.(E)No correlation was detected between MPH for FT and parental genetic distance.Parental genetic distance between Chang 7-2 and each paternal line was calculated with Plink software with 17,446 SNPs(MAF ＞0.05).DS,days to silking;DT,days to tasseling;DP,days to pollen shed.

3.2.Flowering time-associated QTL in the RIL and testcross population.

A set of 11 QTL explaining from 2.28% to 34.30% of variation in FT-related traits was identified (Fig.2).The major QTL clusterfDS10/fDT10/fDP10explained ＞20%of the variance and alleles from the Chang 7-2/Ye478 heterozygote delayed FT compared with alleles from the Chang 7-2/Qi319 heterozygote (Table 1).The confidence interval of the major QTL spanned a physical distance from 91.2 to 100.4 Mb on chromosome 10.In order to further reveal loci that might influence hybrid performance,an association analysis was also performed to identify significant SNPs for FT hybrid performance by an overdominant model with 17,446 SNPs of 328 F1genotypes(MAF ＞0.05).Significant SNPs(P＜2.8×10-6)collapsed into six regions that were significant for FT-related traits (Fig.S1).One of these regions in which the most significant SNPs collapsed into chromosome 10 for both three FT-related traits overlapped withfDS10/fDT10/fDP10confidence intervals.Of seven QTL for MPH,three were detected simultaneously for hybrid performance including the major QTLhDS10/hDT10(R2＞14%) located on chromosome 10 for DS and DT (Table 1).The major QTL for MPH also could be detected by the overdominant model(Fig.S3).This cluster of multiple major QTL was renamedqFH10.

Table 1Information for QTL for FT-related traits detected in TC and RIL populations in the summer maize-growing region.

Fig.2.Distribution of QTL in RIL and TC populations for three FT-related traits.The heatmap boxes illustrates the density of FT-associated QTL across the genome in the summer maize-growing region.The cutoff area is defined by -log10 (P) ＞3.DS,days to silking;DT,days to tasseling;DP,days to pollen shed.

In order to assess the genetic correlation between theper seof RILs and their hybrid performance,the genetic basis of flowering time traitsper sealso be detected.Using the high-density bin map of the RIL population,a total of 12 QTL for all FT-related traits were mapped and explained 2.76%-8.83% of phenotypic variance in FT (Fig.2;Table 1).In the RILs and in the TC populations for hybrid performance under the same conditions,onlyfDT3andfDT7could both be simultaneously detected (Table 1).In order to determine the effect of QTL below the threshold,we also performed phenotypic (for RILs,F1,and MPH) analysis using the bin marker at the peak position of each QTL,there was little difference in FT phenotype among diverse genotypes,a finding consistent with the minor effect of QTL below the threshold (Fig.S4).The major QTLqFH10,which had the largest phenotypic variation for DS and DT,was stably detected on chromosome 10 only in the TC population.

To further define the function ofqFH10,we appraised the FT phenotype using a near-isogenic line (NILQi319) that covered the homozygousqFH10locus.In comparison with Ye478,NILQi319delayed DS by 2-4 days in the spring maize growing region,but DS was delayed much less in the summer maize growing region(Fig.S5),suggesting that photoperiodic response genes are likely to be located within theqFH10locus.Although the FT of parent inbred lines Ye478,Qi319 and test line Chang 7-2 were similar in the summer maize-growing region,the hybrid offspring of their testcrosses to Chang 7-2 differed significantly (Fig.S6).Thus,the effect of the photoperiodic response genes differed in Ye478 versus Qi319 and CY versus CQ in the summer maize-growing region.

3.3.ZmCCT10 as a candidate gene for qFH10

To identify gene(s) underlying theqFH10cluster on chromosome 10 that explained ＞20%of the variance for FT hybrid performance and MPH,the candidate gene(s) were validated with DS.The peak bin(-log10(P)=34.01)forqFH10spanned a 260-kb interval containing three annotated genes: GRMZM2G174773,GRMZM2G174671,and GRMZM2G381691 (ZmCCT10) based on the B73 RefGen_v3 genome (Fig.2).ZmCCT10,a homolog of riceGhd7,has previously been identified as a pleiotropic major QTL responsible for FT-related traits in maize[54].We isolated genomic DNA from three parental lines and sequenced the promoter and CDS ofZmCCT10to confirm the insertion of a 2.4 kbCACTA-like TE in the promoter region of the Ye478 and Chang 7-2 alleles as the genetic determinant of allelic variation with respect to Qi319(Fig.3A).Because photoperiod strongly influences the flowering phenotype of maize in spring and summer maize-growing regions,we measured the expression ofZmCCT10gene under LD and SD conditions.qRT-PCR analysis ofZmCCT10expression patterns in Ye478 and NILQi319revealed significant differences of numbers of transcripts(P＜0.001)only under the LD condition(Fig.3B),consistent with the NILQi319delayed flowering time and theqDP10-1/qDT10-1/qDS10-2only could be identified under spring maizegrowing region (Table S2).ZmCCT10was accordingly selected as a candidate gene forqFH10.

Fig.3. ZmCCT10 effects.(A) PCR sequencing revealed the DNA sequence polymorphism of ZmCCT10 between parents.Black boxes represent exons.Pink box indicates the CACTA-like transposon in the promoter.Gray background indicates the InDel site.Letters below exons indicate amino acid changes caused by nonsynonymous mutations.TE,transposable element.(B)Expression of ZmCCT10 under short-day(9 h light/15 h dark)or long-day(15 h light/9h dark)conditions by qRT-PCR.***,P ＜0.001;LD,long-day;SD,short-day.

3.4.DNA methylation in the ZmCCT10 promoter affects gene expression patterns

The parental lines Ye478 and Chang 7-2 both carry the TE insertion in theZmCCT10promoter but exhibited differential expression of this gene(Fig.4A).For this reason,the expression of endogenousZmCCT10differed(P＜0.01)between CY and CQ;lower expression ofZmCCT10was detected in crosses with the inbred line Qi319.Changes in DNA methylation affected by TEs can sometimes explain the divergence of alleles and the degree of methylation levels in promoters can correlate with differences in gene expression.We used bisulfite sequencing to investigate CG,CHG,and CHH methylation in theZmCCT10promoter region.CG sites were usually more methylated than CHG or CHH sites,and the TSS region (-408 to -1008) was more methylated than were other regions,particularly～700 bp upstream of the TSS(Fig.4B).In comparison with theZmCCT10allele in Qi319,which does not carry a TE,the TE insertion in the Ye478 and Chang 7-2ZmCCT10allele was associated with dramatically greater methylation at CG,CHG,and CHH in the promoter region (Fig.4C).However,the parental alleles displayed non-additive DNA methylation levels in the hybrids.When the methylomes of each hybrid and its parents were compared,a DMR (-700 to -1520) in which parental methylation levels differed was identified among these lines (Fig.4B).At this DMR,Chang 7-2 exhibited greater methylation than either Ye478 or Qi319.Although Ye478 and Qi319 exhibited similarly low methylation in this DMR,after testcrossing to Chang 7-2,their hybrid offspring displayed different (P＜ 0.01) methylation(Fig.4D).Differences in methylation at CG in the DMR in the hybrids and parental lines were negatively correlated withZmCCT10expression (Fig.4A,D).These findings revealed a close relationship between DNA methylation and differences in patterns ofZmCCT10expression.

Fig.4.DNA methylation effects on ZmCCT10 expression.(A) ZmCCT10 expression in parental lines and their hybrids under short-day (SD) conditions.Gene expression is presented relative to the housekeeping gene GAPDH.Standard deviations are shown as bars.(B)Dot plot of CG,CHG,and CHH methylation changes in Qi319(a),Chang 7-2(b),Ye478(c),CY(d),and CQ(e)in the intact promoter region.Red,blue,and green heatmap circles indicate CG,CHG,and CHH methylation(closed)or no methylation(open).TSS,transcription start site;DMR,differentially methylated region;DMC,differentially methylated cytosines.(C)Percentages of CG,CHG,and CHH methylation changes in the intact promoter region between hybrids and parental lines.(D) Percentages of methylation changes in the DMR.**, P ＜0.01;NS,non-significant;MML,mean methylation level.

4.Discussion

Flowering time affects fitness and yield.Late-maturing hybrids that produce more biomass than early maturing hybrids can also produce higher yields[16].Because mechanized harvesting is more suitable for early-maturing varieties with low grain moisture content [55],QTL associated with maize FTper seand hybrid performance are of interest.Almost all of 39 QTL which we detected under ten environmental conditions at latitudes spanning 18°N to 43°N for FT traitsper seshowed small additive effects(Table S2).This result supports the previous study [34] showing that variation in flowering time is due to the combined small effects of many additive QTL.One consistent QTL cluster,qDS7/qDT7/qDP7,influenced variation in FT-related traits in inbred lines and hybrids even under several sources of experimental variation such as differences in location,latitude,temperature,and photoperiod.This QTL cluster may regulate photoperiod response and/or maturity [56].We also detected 13 QTL under single conditions,including the major QTLqDS10/qDT10/qDP10(R2＞15.12%) in the bin 10.04 region on chromosome 10 that overlaps with a largeeffect flowering time QTL and domestication traits detected in previous studies [45,56,57].This QTL detected exclusively in the spring maize-growing region may represent gene(s) encoding a photoperiod response element(Table S2).Although positive correlations were detected (Fig.1B,r＞0.50)between hybrids and their parental lines for FT-related traits,only two QTL(qDT3-2andqDS7)were detected simultaneously in the RIL and TC populations under the summer maize-growing region (Fig.2).Previous studies also found few QTL forper setraits detected in common between inbred and TC progeny,indicating that MAS using QTL identified in inbred could not guarantee that hybrids with superior performance would be selected [4,5,49].

Because the evaluation of all potential hybrids is resourceintensive,only a few elite hybrids are actually identified in field trials[4].Thus,accurate prediction of hybrid performance is a critical aspect of selection of parents for breeding maize and recent studies have used DNA markers and QTL for successful genomic prediction of the performance of hybrids in maize[4,15],sunflowers[36],and rice [10,58].Empirical testcrossing has been the method conventionally used to identify parents of superior hybrids during plant breeding,and testcrosses have often been used to detect QTL controlling various traits of interest to crop breeders [5,59].We used TCs to a RIL population with sufficiently high recombination to confine QTL to narrower intervals.We mapped 11 QTL responsible for FT performance in hybrids under summer conditions,some of which could also be detected using a FaSTLMM algorithm with an overdominance model,as exemplified byqFH10.The finding that over 80%of QTL showed overdominance in the TC populations(Table 1)indicates that overdominance at single loci influences FTrelated traits in maize hybrids [12].To best exploit heterosis,the parental inbreds for developing maize hybrids are normally selected from distinct heterotic groups that have been characterized by maize breeders [12].However,in the present study,although relationship among Chang 7-2 and each of the 328 RILs exhibited extensive genetic diversity,no correlation between interparental genetic distance and heterosis for FT in hybrids was detected(Fig.1D,E).This finding is consistent with those of previous studies[8,58,60]inArabidopsisand rice that also found no relationships between the genetic distance between parents and the degree of heterosis expressed by hybrids.In the present study,four QTL for FT heterosis,including the major QTL clusterhDS10/hDT10overlapping withqFH10,were detected with both additive and overdominance models.This contrasting result suggests that QTL for heterosis will probably be accompanied by strong differences in hybrid performance.

ZmCCT10is orthologous to the rice photoperiod response regulatorGhd7[54] and is a crucial controller of variation in flowering time that has enabled the dispersion of maize from the tropics to temperate zones [45,57].In our study,the likely function ofZmCCT10as the causal gene in the major FT hybrid performancerelated QTLqFH10(Fig.3) was tested.In theZmCCT10promoter region,the allele in Qi319 carrying no TE insertion showed lower cytosine methylation at CG,CHG,and CHH than did the alleles carrying theCACTA-like TE insertion in Ye478 and Chang 7-2 under SD conditions(Fig.4C).This finding is consistent with previous studies showing high cytosine methylation in maize line HZS(TE-positive)compared 1145 (TE-negative) under LD condition [45].Although Ye478 and Chang 7-2 carry the TE insertion in theZmCCT10promoter,the DMR (-700 to -1520) showed higher methylation in Chang 7-2 than in either Ye478 or Qi319.Indeed,lower CG methylation of theZmCCT10promoter in this DMR was associated with higher expression ofZmCCT10in Qi319 and Ye478 than in Chang 7-2(Fig.4A).This finding suggests that CG methylation differences in the DMR is more important for regulatingZmCCT10expression than the intact promoter.

Recent studies [30,61] have shown that DNA methylome interactions between methylated parental epialleles may result in nonadditive interactions in F1hybrids and heterotic phenotypes.Such methylation interactions includetrans-chromosomal methylation(TCM) and demethylation (TCdM),whereby the DNA methylation of an allele in one parent is altered by that of the allele in the other parent [30,62].Studies in the model plantArabidopsisand in rice have identified nonadditive alterations in DNA methylation at DMRs[30,62,63].As a result,epigenomic interactions might confer nonadditive transcriptional and epigenetic activities on genes in F1progeny compared to their parental lines.Our analyses also identified methylation interactions at the DMR of theZmCCT10promoter in hybrids.In our study,although Ye478 and Qi319 exhibited similar degrees of methylation at the DMR,when they were testcrossed to Chang 7-2,their hybrid offspring exhibited different methylation patterns(Fig.4D).In the CQ hybrid,methylation interactions at the DMR were attributed to TCM in which methylation in the F1is higher than the MML,a non-additive pattern.But because methylation of the high parent allele from Chang 7-2 is not maintained in the CY hybrid,methylation decreases overall.This finding might be explained by allele-specific DNA methylation at this DMR without methylation interactions or parent-of-origin effects leading to MML expression ofZmCCT10in CY [22].The differences in CG methylation in the DMR among the F1hybrids also correlated negatively with theZmCCT10expression (Fig.4A).Though these results do not prove that differential methylation causes the QTL effects associated withZmCCT10,they do suggest a clear association between the insertion of aCACTA-like transposon element and increased methylation at this locus.This alteration could reduceZmCCT10transcription and eventually affect flowering time in hybrids [45,64].More evidence,such as a study of histone modifications in the region,might bolster these conclusions and clarify whether the differential methylation observed at the two alleles affects chromatin conformation at theZmCCT10promoter and how such a chromatin state would affectZmCCT10expression.

CRediT authorship contribution statement

Zhiqiang Zhou:Conceptualization,Formal analysis,Investigation,Writing -original draft,Writing -review &editing.Xin Lu:Conceptualization,Validation,Writing -review &editing.Chaoshu Zhang:Investigation.Mingshun Li:Resources.Zhuanfang Hao:Resources.Degui Zhang:Resources.Hongjun Yong:Resources.Jienan Han:Resources.Xinhai Li:Conceptualization,Supervision,Project administration,Funding acquisition,Writing-review &editing.Jianfeng Weng:Conceptualization,Supervision,Writing -review &editing.

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.

Acknowledgments

This research was jointly funded by the National Natural Science Foundation of China (31971963) and Agricultural Science and Technology Innovation Program of CAAS.

Appendix A.Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2023.05.006.