Guoyun Liu,Wenfeng Peib,,Dn Li,Jinjing M,Yupeng Cui,Nuohn Wng,Jikun Song,Mn Wu,Libei Li,Xinshn Zng,Shuxun Yu,*,Jinf Zhng*,Jiw en Yu,*

a State Key Laboratory of Cotton Biology,Cotton Institute of the Chinese Academy of Agricultural Sciences,Key Laboratory of Cotton Genetic Improvement,Ministry of Agriculture,Anyang 455000,Henan,China

b Xinjiang Research Base,State Key Laboratory of Cotton Biology,Xinjiang Agricultural University,Urumqi 830001,Xinjiang,China

c Department of Plant and Environmental Sciences,New Mexico State University,Las Cruces,NM 88003,USA

A B S T R A C T Cotton fiber is the m ost im portant natural raw m aterial for the textile industry,and fiber length(FL)is one of the m ost im portant traits in cotton.Quantitative trait locus(QTL)m apping based on high-density genetic m aps is an efficient approach to identify genetic regions for FL.In our study,two backcrossed inbred lines(BILs)were chosen as parents to construct a high-density genetic m ap in F2 w hich w as used to fine m ap FL QTL in F2:3 population.The genetic m ap had a total size of 3462.8 c M,containing 9182 singlenucleotide polymorphisms(SNPs)based on genotyping-by-sequencing.Two FL related stable QTL w ere identified on tw o chrom osom es(qFL-A08-1 on A08 and qFL-D03-1 on D03),and qFL-A08-1 w as confirm ed by a m eta-analysis.Utilizing previously obtained RNA-seq data for the tw o BILs and q RT-PCRanalysis,tw o candidate genes annotated as cytochrome b5(CB5,Gh_A08G1729)and microtubuleend-binding 1C(EB1C,Gh_D03G0232)that m ay regulate FL during the fiber elongation stage w ere identified.In addition,nine recom bination hotspots in this population w ere found.The results of this study w ill provide an im portant foundation for further studies on the m olecular and genetic regulation of fiber elongation.

Keywords:Gossypium Single-nucleotide polym orphism Fiber length Quantitative trait locus

1.Introduction

The cotton(Gossypium spp.)fiber is the most important natural raw material for the textile industry.Fiber quality traits,including fiber length(FL),fiber strength(FS),fiber micronaire,fiber elongation,and fiber uniform ity,are complex quantitative traits controlled by multiple genes and environmental factors.Longer fibers are satisfying for the textile industry to produce fine yarns.Developing high-yielding cultivars with good fiber quality rem ains a challenge for cotton breeding[1,2].Until now,a series of m olecular studies have focused on genes w hich play an important role during fiber developm ent[3,4].How ever,their roles in genetically controlling fiber quality traits are poorly understood.

G.barbadense has much longer,stronger and finer fibers than G.hirsutum,and introgression lines or backcrossed inbred lines(BILs)with improved fiber quality traits introduced into G.hirsutum from G.barbadense have been developed by advanced backcrossing[5-7].To better understand the genetic and m olecular mechanism s related to fiber quality(especially FL),researchers have created num erous genetic populations,including F2populations,recom binant inbred line populations,backcross populations and natural populations from existing germplasm lines.Recently,1198 and 1084 QTL related to fiber quality,yield,seed quality and stress tolerance were identified based on 62 intraspecific G.hirsutum and 31 G.hirsutum×G.barbadense populations,respectively[8,9].Since the genomes of G.hirsutum and G.barbadense were sequenced[10,11],natural populations have been used to study fiber quality and yield traits based on genome-w ide association analysis using SNPm arkers.Fang et al.[12]evaluated 318 w orldw ide cotton accessions and detected 15 fiber quality-related genomic regions.In another study,352 diverse accessions w ere analyzed,and 93 fiber qualityrelated putative dom estication sw eeps w ere discovered[13].Ma et al.[14]resequenced 419 accessions and identified 11,026 SNPs associated with fiber traits.However,most of these identified QTL have large confidence intervals and are not useful for m arker-assisted selection or map-based cloning of candidate genes for QTL.Therefore,exploring high-density markers for high-resolution genetic mapping and fine mapping of fiber quality QTL in cotton is important.

In our previous study,a BILpopulation developed from a cross betw een a G.hirsutum cultivar as the recurrent parent and a G.barbadense cultivar,followed by tw o generations of backcrossing and then four generations of self-pollination,was constructed[5].A total of 28 fiber quality-associated QTL including four QTL for FL,w ere identified using 392 polymorphic simple sequence repeat(SSR)markers.However,due to the limited number of m arkers em ployed,a large portion of the phenotypic variation in the BIL population w as uncounted for,leaving many QTL unidentified.The objectives of this study were,(1)to construct a high-density genetic m ap for an F2population from a cross of tw o BILs w ith significantly different FL and FS using SNPs based on genotyping-by-sequencing(GBS),w hich w as used to locate QTL for FL in the F2:3progeny;and(2)to combine above results with QTL m eta-analysis,RNA-seq analysis and RT-PCR analysis to identify candidate genes w ithin stable FLQTL regions.

2.Materials and m ethods

2.1.Plant materials

Two BILs,i.e.,NMGA-062(named“Long”)w ith FLallele introgression from G.barbadense and NMGA-105(nam ed“Short”)w ith significantly different FL were chosen as parents,and they are 95.4%identical based on 2349 SSR and SNP markers.The BILs w ere generated from a cross betw een G.hirsutum SG 747 as the recurrent parent and G.barbadense Giza 75 followed by two generations of backcrosses and then four generations of selfpollination at New Mexico State University,Las Cruces,NM,USA[5].The BILs were hybridized to produce F1,and one of the F1plants was self-pollinated to produce F2seed.The F2population of 148 plants w as grow n for DNA extraction and generation advancem ent at the experimental farm,Institute of Cotton Research,Chinese Agricultural Academ y of Science(CAAS),Anyang,Henan,China,in 2015.The parents and its 148 F2:3progeny w ere planted follow ing a randomized complete block design in Anyang w ith three replicates(as three different environments)at three different planting dates on April 26th,May 10th,and May 20th in 2016.

2.2.Measurement of fiber quality

Maturefibersof the parentsand F2:3population wereharvested in November 2016.Because the tw o BIL parents significantly differed in FL and FS(Table 1),only these tw o traits w ere investigated based on results from a High Volume Instrument(HVI)900 by the Test Center of Cotton Fiber Quality affiliated w ith the Ministry of Agriculture and Rural Affairs,Institute of Cotton Research,CAAS,Anyang,Henan,China.

2.3.GBSlibrary construction

Genom ic DNA w as extracted from young leaves of the 148 plants in the F2population and tw o parents using a rapid CTAB m ethod[15].GBS libraries w ere constructed as described in previous studies[16,17].Mse I w as used to digest genom ic DNA,and DNA fragm ents at a size of 350-500 bp w ere then ligated to adapters using T4 DNA ligase,follow ed by PCR.The PCRproducts w ere purified and sequenced using the Illum ina HiSeq2500 platform,w ith 150-bp paired-end reads generated for further analysis.The reads w ere deposited at NCBIunder accession num ber PRJNA431664.

2.4.Data analysis

Low-quality reads in the raw data were removed in reference to the follow ing quality control standards:1)no＜10%unidentified nucleotide reads w ere rem oved;2)reads w ith＞50%bases having Phred quality＜5 were removed;3)reads with＞10 nucleotides aligned to the adapter and＞10%of mismatches w ere removed;and 4)putative PCRduplicates generated by PCRamplification in the library construction process w ere rem oved.The clean reads from each sample were then aligned to the reference genome using BWA(Burrow s-Wheeler Aligner)softw are(set at:m em-t 4-k 32-M-R)[18].Reads w ith a sequence depth of＜8 in the parents and＜3 in the F2populations w ere filtered out in further analysis.GATK software was used to call SNPs for all of the samples[19].Polymorphic SNP m arkers w ere classified into eight groups:ab×cd,ab×cc,cc×ab,ef×eg,hk×hk,nn×np,lm×ll,aa×bb.How ever,only SNP m arkers with segregation patterns of aa×bb in the parents w ere used for linkage m ap construction.

2.5.Linkage map construction and mapping of fiber quality

SNP m arkers w ith no＞20%m issing data in the F2population and a P value of segregation distortion of＜0.05 w ere selected to construct a linkage m ap.The SNP m arkers w ere first divided into 26 groups according to their positions m apped onto the 26 chromosom es of the reference genome of TM-1.Next,the m arkers w ere ordered by Join Map(version 4.0)using the regression approach w ith a logarithm of the odds(LOD)threshold at 6[20].Inclusive com posite interval m apping(ICIM)w as em ployed to detect QTL using QTL IciMapping(version 4.1)[21].The param eters w ere set to 1 c M of the step,and 1000 perm utations w ere taken as the LOD threshold.QTL at the sam e location for the sam e traits in different environments(replications)w ere regarded as stable QTL.QTL w ere nam ed according to Mc Couch et al.[22].Marey Map w as applied to construct a recom bination m ap,w hich displayed a sm ooth curve w ith the Loess m ethod[23].Regions no＜20 c M Mb-1were regarded as recom bination hotspots[24].

Table 1-Perform ance and analysis of fiber length and strength in tw o parents and F2:3 populations.

2.6.Candidate gene selection and expression pattern analysis

The expression levels of genes located within stable FL QTL regions were compared in the two parents using 10-days postanthesis(DPA) fibers and RNA-seq data deposited in NCBI under accession number SRP039385. Differentially expressed genes were then annotated using information from the sequenced reference genome of TM-1 [10]. Co-expressed genes were then detected using the database of coexpressionnetworks with functional modules for diploid and polyploid Gossypium (http://structuralbiology.cau.edu.cn/gossypium/). Total RNA of ovules at 0 and 3 DPA and fibers at 5, 7, 10, 15, 20, and 25 DPA was extracted from the two parents.

The One-Step SYBR Primer Script Plus RT-PCR kit (Takara,Beijing, China) was used according to the manufacturer's instructions to conduct qRT-PCR of candidate genes.

The cotton His3 gene (i.e., histone 3) was used as an internal control[25]. All primers are listed in Table S1.

3.Results

3.1.Determination of FLand FSin thetw o BILparents and F2:3 population

The traits of the F2:3population and parents are sum m arized in Table 1.Both FL and FS show ed a significant difference betw een the tw o parents,except FS in 2016A.The FL and FS exhibited a transgressive segregation in the segregating population,w ith longer FL and higher FS in some of the progeny of the population than in the parents.FLand FSw ere also significantly and positively correlated(r=0.7247,P＜0.01),indicating a possible tight linkage betw een QTL or a pleiotropic effect of the sam e QTL for the tw o traits.An analysis of variance show ed significant variation in FL and FS in the F2:3population.

3.2.Genotyping of the F2 population based on sequencing

Using GBS libraries constructed w ith the Illum ina HiSeq2500 platform,12,916,840 and 11,384,934 clean reads w ere obtained for the tw o parents,i.e.,“Long”and“Short”,respectively.After m apping to the sequenced TM-1 genom e[10],13,273 SNPs w ere found to be hom ozygous betw een the parents;and these SNPs w ere then selected for further analysis in the F2population.Results of SNP distribution am ong the 26 chrom osom es(Table 2)show ed that som e chrom osom es(i.e.,D01,D11,and D13)carry few er than 50 SNPs,w hich is indicative of a low level of polym orphism betw een the tw o BIL parents for these chromosomes.SNPs w ith distorted segregation and＞20%m issing sequences in the F2population w ere rem oved before linkage m ap construction and QTLanalysis.As a result,only 11,454 SNPs w ere used for the follow ing linkage m apping.

3.3.High-throughput linkage map construction

The 11,454 SNPs w ere divided into 26 linkage groups (LGs) to construct a linkage m ap containing 9182 SNP m arkers. The genetic distances of the 26 LGs ranged from 2.5 c M (D01) to 291.8 c M (D08), w ith a total genetic distance of 3462.8 c M for the linkage map and an average distance of 0.49 c M betw een adjacent m arkers (Fig. S1, Table 2). The num ber of m arkers m apped to each LG varied from 18 (in D02) to 1994 (in D08),w ith an average of 353 m arkers per LG. In addition, the maximum gap of the 26 LGs ranged from 0.74 c M (D01) to 43.91 c M (D07). Am ong the 26 LGs, D01, D11, and D13 contained no ＞25 SNPs. The genetic distances of A02, D01,D09, and D13 w ere ＜50 c M, but the genetic distances of A05,A11, and D08 w ere longer than 200 c M. Because the two lines w ere selected from a BIL population based on FL differences, it is understandable that these tw o lines show ed m ore sim ilarities in m any chrom osom es w ith few er polym orphic SNP m arkers.Therefore,unlike other linkage maps,markers are distributed non-random ly am ong the 26 LGs,resulting in different genetic distances for the 26 LGs in the linkage m ap and a low er level of total genetic distance.

Table 2-Detailed inform ation on the high-d ensity genetic m ap.

Collinearity w as m easured to assess the quality of this genetic m ap(Fig.1,Table 2).The results indicated that m ost LGs in this new ly constructed linkage m ap have a high level of collinearity w ith the physical m ap of the G.hirsutum reference TM-1 genome.The lack of SNPs on D01,D11,and D13 resulted in low collinearity betw een the genetic m ap and the physical m ap for these chrom osom es.

The distribution of genome-wide variation of recom bination rates w as then m easured,and as show n in Fig.1,the average genom e-w ide recom bination rate w as non-random in that distal chrom osom al regions show ed higher recom bination rates than did proximal regions.Additionally,the recom bination rate w as found to vary across chrom osom es.For m ost chrom osom es,the recom bination rate w as low in pericentrom eric regions,but recom bination spikes w ere found in centromeric regions of A01,A07,D05,and D08.The average recom bination rate for each chrom osom e also show ed significant differences,ranging from 0.04 to 4.43 c M Mb-1(Table 2),w ith an overall genom e-w ide recombination rate of 1.82 c M Mb-1.On average,the recom bination rate w as higher on Dt chrom osom es than on At sub-genome,indicating m ore recom binations for Dt than for At.This result w as consistent w ith a previous study[26].The genetic length and recom bination rate of D04,D06,D07,and D08 w ere significantly higher than their respective hom eologous chromosomes A04,A06,A07,and A08,w hereas A01,A11,A12,and A13 exhibited a significantly higher recom bination rate than their hom eologous chrom osom es D01,D11,D12,and D13,respectively.In this population,9 recombination hotspots w ere identified on seven chrom osom es D01,D03,A07,D07,D09,D12,and D13(Fig.1).Overall,m ore recom bination hotspots w ere found in the Dt sub-genom e than in the At sub-genom e,w hich is consistent w ith the result that the overall recombination rate for the Dt sub-genom e is higher than that of the At sub-genom e,as reported by Shen et al.[26].

3.4.QTL mapping of fiber length and strength in F2:3 population

Based on this high-density genetic m ap,tw o stable FL QTL w ere identified on chromosomes A08 and D03,as show n in Table 3.qFL-A08-1 at the interval of 155-163 c M on A08 w as detected in the F2:3population in three environm ents,each explaining 8.52%-22.44%of the phenotypic variation.Interestingly,a FS QTL qFS-A08-1 w as also detected in this region based on BLUP and com bined analysis of the F2:3population.The results indicated that this QTL region(for qFL-A08-1 and qFS-A08-1)affects both FL and FS.The other stable FL QTL(qFL-D03-1)at 15-20 c M on D03 w as detected in the F2:3population also in tw o environm ents and com bined and BLUP analyses,each explaining 9.24%-13.59%of the phenotypic variation.There may be another QTL at the 15-37 c M interval for FLin a close proxim ity(detected in the 2016Ctest),but it w as not verified by a com bined and BLUPanalysis.

Fig.1-Correlation am ong genetic and p hysical m ap s,estim ated local recom bination rates and their distribution in chrom osom al rearrangem ent regions.The red shadow rep resents recom bination hotspots.

3.5.Meta-QTL analysis of the stable fiber length QTL

To further understand w hether the stable QTL w ere com m on to these reported in previous G.hirsutum×G.barbadense populations and to identify new QTL,w e co-located the two stable FL QTL w ith 170 FL QTL from Cotton QTLdb Release 2.3(Jan.24,2018,http://www.cottonqtldb.org/)[9]and FL quantitative trait nucleotides(QTNs)reported by Fang et al.[12],Wang et al.[13]and Ma et al.[14]on the sequenced TM-1 genom e[10].qFL-A08-1 co-localized w ith previously identified 10 QTL within the sim ilar region on the sam e chrom osom e(Fig.2).The results indicated that q FL-A08-1 is a comm on FL QTL in different genetic populations.How ever,no previously reported QTL or QTN co-localized w ith the other stable FL QTL,i.e.,qFL-D03-1(Fig.2),indicating that this QTLmight be a new one identified in our secondary population developed from tw o BILs derived from the sam e interspecific population.

Table 3-Detailed inform ation of three com m on QTL for fiber length(FL)and strength(FS)in F2:3 pop ulation.

3.6.Identification of candidate genes for FL QTL

Regions for qFL-A08-1 at 156.5-162.5 c M and for FL-D03-1 at 2.43-2.54 c M(Table 3)(detected in no less than tw o environm ents)were selected to identify genes for FL.Based on the sequenced genom e of TM-1,fragments of 980 kb length(for stable QTL)w ere selected as candidate gene regions related to FL,including the sm allest(w ith 110 kb)region for qFL-D03-1.We then analyzed the function of genes located w ithin the tw o stable FLQTL(qFL-A08-1 and qFL-D03-1)regions,involving 105 and 8 predicted genes,respectively.

Fig.2-Stable QTL of fiber length(FL)on the physical map.Stable FL QTL are marked in red.QTL detected in previous studies are m arked in black.

To further identify FL related genes,w e evaluated the transcriptom e in 10 DPA fibers of the tw o parental BILs.Our hypothesis is that only these differentially expressed genes that contained SNPs may be genetically associated w ith FL.First,a total of 86 genes w ere identified to be expressed in at least one parent,but only 14 genes exhibited significantly different expression(DE)betw een the tw o parents(Table S2).Genes co-expressed w ith these 14 DE genes w ere then analyzed,leading to the identification of 216 positive and 82 negative co-expressed genes(Table S2).Second,w e analyzed the SNPs in the 14 DE genes betw een the tw o parents.Six of the 14 DEgenes contained 15 SNPs between the tw o parents,and 7 SNPs belonging to five genes had non-synonym ous m utations(Table S2).

Finally,according to their annotations,SNPs and expression levels,tw o genes(Gh_A08G1729 and Gh_D03G0232)w ere selected as possible candidate genes to perform qRT-PCRanalysis during the fiber elongation stage.The cytochrome b5(CB5)gene(Gh_A08G1729)for qFL-A08-1 show ed a higher level of expression in developing fibers in“Short”than in“Long”.This result indicates that the expression of this gene might suppress fiber elongation(Fig.3-a).For qFL-D03-1,the gene(Gh_D03G0232)encoding the microtubule end-binding 1C (EB1C)protein expressed at a higher level in“Long”than in“Short”,especially in 0-15 DPA fibers,though expression was low after 15DPA(Fig.3-b).According to the co-expression netw ork database,3 and 35 genes showed positive co-expression and three and four negative co-expression w ith Gh_A08G1729 and Gh_D03G0232,respectively.The expression levels of the co-expressed genes w ere then analyzed using transcriptome data,and only one gene(Multiple C2 domain and transmembrane region protein 3,MCTP3,Gh_A06G0987)w as positively co-expressed w ith Gh_A08G1729(Fig.3-c).For Gh_D03G0232,28 genes displayed a positive coexpression trend,suggesting that this gene may participate in many pathways(Fig.3-d).

4.Discussion

To fine map FLrelated genes,tw o lines w ere selected as parents from a BIL population.The fiber quality of the tw o lines w as measured,and no significant difference betw een the tw o w as observed,except for the mature FL and FS.Differences in the genetic backgrounds betw een the tw o parents have already been identified in previous studies using SSR marker analysis.Although the two lines were developed from BILs with an expected average of 87.5%recovery of G.hirsutum SG747,a highly similar genetic background(95.4%based on 2349 markers)was identified[27].Nonetheless,based on high-resolution SNP markers,many m ore polymorphic markers betw een the tw o BILs were identified,facilitating QTL detection.Several studies had been reported using introgressed lines to study fiber quality.Previously,Nie et al.[28]analyzed an introgression based BIL population derived from G.hirsutum and G.barbadense,and identified four markers on four chrom osom es associated w ith FL.Chen et al.[29]identified four stable FLQTLon chrom osom e A03,A07,D05,and D07,based on tw o G.hirsutum×G.barbadense introgression populations.In another research,a cross of G.mustelium×G.hirsutum was used to identify 42 QTL for fiber strength and fineness[30].The further study of introgressed populations also provided a strategy for improving upland cotton fiber quality by introgressing genomic regions from other Gossypium species.

The genetic map contains 9128 SNP markers spanning 3462.85 c M,w ith an average of 0.49 c M betw een adjacent SNP m arkers.This finding w as consistent w ith these of previous studies[31-35].With regard to the tw o G.barbadense introgression BILs,the results of the present study also reflect the relationship betw een FL and the genom ic difference betw een G.barbadense and G.hirsutum.Additionally,tw o backcrosses resulted in a non-average distribution of aa×bb genotype SNPs on the 26 chromosomes in the tw o BILs.As show n in Fig.1,gaps w ere found on A01,A02,A07,D01,D09,D10,D11,and D13,indicating poor polym orphic SNP m arkers in these regions.The length of the LGs constructed using the SNP m arkers also show ed significant differences:the lengths of A02,D01,D09,and D13 w ere＜50 c M,but the lengths of A05,A11,and D08 w ere over 200 c M.The m ost likely reason for this is the high sim ilarity of the genetic background betw een the tw o parents.

A total of 9 recom bination hotspots w ere identified in our genetic m ap.Previous studies have reported that recom bination in Arabidopsis[36],m aize[37],and Triticum aestivum[38]usually occurs in intergenic regions and around transposons.In cotton,researchers have identified som e relatively high recom bination rates in the distal region of chrom osom es A04,A05,homeologous A10 and D10,and hom eologous A07 and D07[26],w hereas such events for D03,D09,and D12 w ere first identified in our study.The lack of polym orphic m arkers for D01 and D13 m ight induce a significant effect on the discovery of recom bination events.

The F2:3populations show ed transgressive segregation in FL and FS.Tw o FL related stable QTL w ere identified.One stable FL QTL(qFL-A08-1)w as also associated w ith FS QTL,which may explain the significant and positive correlation analysis betw een the tw o traits.In previous studies,170 FL QTL and m any QTNs w ere detected in natural populations[9,12-14].Most FL related genomic regions are located on chrom osom es A03,A07,D05,and D12,am ong others,in G.hirsutum×G.barbadense populations[8].Based on our present study,tw o stable FL QTL on A08 and D03 w ere identified.Ten QTL identified in previous studies,w hich overlapped w ith qFL-A08-1.However,qFL-D03-1 w as a new QTL identified in this study,possibly because(1)SNP m arkers have m uch higher resolution than do SSRm arkers used in m ost previous studies,and(2)this QTL might be a rare QTL in certain G.barbadense that w ere not detected in natural populations.

Moreover,14 DEgenes w ere identified by utilizing RNA-seq data for 10 DPA fibers from the tw o parents[27].How ever,only five genes containing seven non-synonymous mutations w ere detected.Finally,tw o candidate genes w ere identified as DE during fiber elongation:a CB5 gene on qFL-A08-1 and an EB1C gene(Gh_D03G0232)on qFL-D03-1.CB5 w as expressed at a higher level in“Short”than in“Long”,specifically at the early fiber elongation stage(5 DPA to 20 DPA).The CB5 protein w as found to interact w ith m ultiple C2 dom ain and transm embrane region protein 3(MCTP3,Gh_A06G0987)and exhibit a positive co-expression relationship.Previous studies in Arabidopsis have identified that MCTP3 is expressed in young leaves and vascular tissues[39].Another study found that in Arabidopsis,mutation in MCTP3 and FTIP4 can result in early term ination of m ain inflorescence apices and increase the num ber of disorganized vascular bundles and cells;thus,MCTIP3 m ay regulate cell division and term ination[40].EB1C expression w as higher from 0 to 7 DPA and decreased from 10 DPA to 25 DPA.By affecting cell w all deposition,m icrotubules determ ine the grow th direction of cells[41-43],and the higher expression level of EB1C m ay be beneficial for fiber elongation via m icrotubule distribution regulation.More experim ents are needed to confirm w hether the two candidate genes coexpress w ith other genes and regulate fiber elongation.Although further experim ents,such as overexpression or CRISPR-Cas9 silencing,are needed to confirm roles in fiber elongation,the present QTL analysis and q RT-PCR analysis show ed that these tw o genes m ay be im portant for fiber elongation.

Fig.3-The ex pression trend of cand idate genes and their co-expressed genes in the tw o parents.(a,b)q RT-PCRanalysis of the transcript levels of cand idate genes d uring fiber elongation in the tw o p arents.The Y-axis rep resents the relative exp ression level,and the X-axis rep resents different DPA during fiber elongation.The error bars denote the standard error.(c,d)Heatmap analysis of co-ex pressed genes in the tw o parents at 10 DPA.

5.Conclusions

In our study,tw o BILs w ere chosen as tw o parents to construct F2and F2:3populations to detect FL-related QTL and genes.Based on GBS technology,two FL related stable QTL were identified.A m eta-analysis w as undertaken to confirm the stable QTL.The further use of our previous RNA-seq data for the tw o parents,follow ed by q RT-PCR analysis,detected tw o candidate genes that m ay regulate FL during the fiber elongation stage.Furtherm ore,nine recom bination hotspots w ere identified in our population.These results provide an important foundation for further studies on the molecular and genetic regulation of fiber elongation.

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2018.11.005.

Acknow ledgm ents

The research w as supported by the National Natural Science Foundation of China(31621005),the National Key Research and Development Program of China(2016YFD0101400),the National Transgenic Research Program of China(2016ZX08005005),and the New Mexico Agricultural Experim ent Station.