HU Fu-chu,CHEN Zhe,WANG Xiang-he,WANG Jia-bao,FAN Hong-yan,QIN Yong-huaZHAO Jie-tangHU Gui-bing

1 State Key Laboratory for Conservation and Utilization of Subtropical Agro-bioresources/Key Laboratory of Biology and Genetic Improvement of Horticultural Crops (South China),Ministry of Agriculture and Rural Affairs/College of Horticulture,South China Agricultural University,Guangzhou 510642,P.R.China

2 Key Laboratory of Tropical Fruit Tree Biology of Hainan Province/Institute of Tropical Fruit Trees,Hainan Academy of Agricultural Science,Haikou 571100,P.R.China

3 Environment and Plant Protection Institute,Chinese Academy of Tropical Agricultural Sciences,Haikou 571101,P.R.China

Abstract Litchi chinensis Sonn is widely cultivated in subtropical regions and has an important economic value.A high-density genetic map is a valuable tool for mapping quantitative trait loci (QTL) and marker-assisted breeding programs.In this study,a single nucleotide polymorphism (SNP)-based high-density linkage map was constructed by a genotyping-by-sequencing(GBS) protocol using an F1 population of 178 progenies between two commercial litchi cultivars,‘Ziniangxi’ (dwarf) and‘Feizixiao’ (vigorous).The genetic map consisted of 3 027 SNP markers with a total length of 1 711.97 cM in 15 linkage groups (LGs) and an average marker distance of 0.57 cM.Based on this high-density linkage map and three years of phenotyping,a total of 37 QTLs were detected for eight dwarf-related traits,including length of new branch (LNB),diameter of new branch (DNB),length of common petiole (LCP),diameter of common petiole (DCP),length of internode (LI),length of single leaf (LSL),width of single leaf (WSL),and plant height (PH).These QTLs could explain 8.0 to 14.7% (mean=9.7%)of the phenotypic variation.Among them,several QTL clusters were observed,particularly on LG04 and LG11,which might show enrichment for genes regulating the dwarf-related traits in litchi.There were 126 candidate genes identified within the QTL regions,55 of which are differentially expressed genes by RNA-seq analysis between ‘Ziniangxi’ and ‘Feizixiao’.These DEGs were found to participate in the regulation of cell development,material transportation,signal transduction,and plant morphogenesis,so they might play important roles in regulating plant dwarf-related traits.The high-density genetic map and QTLs identification related to dwarf traits can provide a valuable genetic resource and a basis for marker-assisted selection and genomic studies of litchi.

Keywords:Litchi chinensis, SNP genetic map,dwarf-related traits,QTL analysis,genotyping-by-sequencing (GBS)

1.Introduction

Litchi (Litchi chinensisSonn),family Sapindaceae,is one of the most important tropical and subtropical fruit trees,which has a cultivation history of more than 2 300 years.As the world’s largest litchi producer,China has planted 550 000 hectares of litchi with an annual output of 2.2 million tons,accounting for 60% of the world’s total production(Qiet al.2016).Litchi breeding initially focused on yield,and most of the commercial cultivars are obtained through selective breeding (Lal and Nirwan 1980;Campbell and Campbell 2001;Wuet al.2007;Zhanget al.2016).In recent years,controlled hybridization of litchi has been carried out to breed new cultivars with high fruit quality(Xiang 2012;Liuet al.2016).However,the efficiency of the traditional cross breeding method is still very low because of the high heterozygosity and long juvenile period of litchi.Therefore,the development and application of molecular breeding approaches can provide great opportunities for litchi breeding.

Marker-assisted selection (MAS) technology can greatly improve breeding efficiency,and it has been widely applied in crop improvement to enhance crop yield,quality,and tolerance to abiotic or biotic stresses (Heet al.2014).Construction of a linkage map and mapping of trait-related quantitative trait loci (QTL) are required as essential tools to assist MAS approaches (Wuet al.2014;Chenet al.2015;Wang 2016).In litchi,only a few linkage maps have been published.The first molecular map of litchi was constructed with 107 random amplified polymorphic DNA (RAPD)markers in 13 linkage groups covering a map distance of about 1 982.5 cM using 76 F1progenies of the cross between‘Maguili’ and ‘Jiaohesanyuehong’ (Liu and Mei 2003).Later,amplified fragment-length polymorphism (AFLP) markers and sequence-related amplified polymorphism (SRAP)markers were used to increase the marker density,yielding a maternal ‘Maguili’ map containing 179 loci in 16 linkage groups covering 610.55 cM with an average distance of 3.75 cM,and a paternal ‘Jiaohesanyuehong’ map consisting of 82 loci allocated into 15 linkage groups covering 548.09 cM with an average distance of 8.18 cM (Liuet al.2010;Zhaoet al.2010;Zhouet al.2012).However,there have been no QTLs analyses in litchi due to the low-density genetic maps based on small F1population sizes.

The vast sequencing data generated by high-throughput sequencing technologies have created great opportunities for generating high numbers of single-nucleotide polymorphisms(SNPs) (Ganalet al.2009).SNP markers have been increasingly used for the construction of high-density genetic maps and QTL mapping in many crops due to their high abundance and relatively even distribution across a genome(Kobayashiet al.2014;Chenet al.2015;Weiet al.2016).Several new methods have been used for identifying SNPs such as restriction site associated DNA sequencing (Pfenderet al.2011) and genotyping-by-sequencing (GBS) approach(Elshireet al.2011).GBS utilizes restriction endonucleases to digest the genome into fragments,which are subsequently sequenced on high-throughput platforms.GBS has been developed as a rapid and robust approach for construction of high-density linkage maps in many organisms,including several fruit trees such as peach (Bielenberget al.2015),cranberry (Daverdinet al.2017),and pear (Kumaret al.2017).Moreover,GBS has also been used to identify QTLs in chickpea (Vermaet al.2015),oil palm (Pootakhamet al.2015),and Chinese chestnut (Jiet al.2018).However,high-density genetic maps and QTLs identification using GBS technology has not been reported in litchi.

Dwarfism is a desirable characteristic for fruit trees,which can reduce the high cost of labor for pruning.At present,the mechanisms on litchi dwarfism are still poorly known (Huet al.2018).Therefore,in this study,we used an F1population derived from ‘Ziniangxi’ (dwarf cultivar)and ‘Feizixiao’ (vigorous cultivar) for construction of a highdensity genetic linkage map based on SNP markers using GBS.Moreover,the preliminary QTLs for several dwarfrelated traits were identified by combining this high-density genetic map with phenotypic data collected over three successive years.The results of the present study will help to establish a molecular breeding system to enhance genetic studies and breeding programs for litchi.

2.Materials and methods

2.1.Plant materials

The F1mapping population contained 178 progenies which were generated from two commercial cultivars,Litchi chinesisSonn.dwarf cultivar ‘Ziniangxi’ (ZNX,female) and vigorous cultivar ‘Feizixiao’ (FZX,male).The F1population was developed in 2013 and all the plant materials were soil-planted at 4 m×5 m at the Yongfa Fruit Research Station,Institute of Tropical Fruit Trees,Hainan Academy of Agricultural Science,in 2014.All seedlings were managed identically in terms of fertilizer applications,pruning and pest control.

2.2.Phenotypic analysis

Eight traits of the F1population and their parents were collected in 2015,2016 and 2017 in late autumn,when the new shoots and leaves were fully matured.Length of new branch (LNB),diameter of new branch (DNB),length of common petiole (LCP),diameter of common petiole (DCP),length of internode (LI),length of single leaf (LSL),width of single leaf (WSL),and plant height (PH) were evaluated as dwarf-related traits in this study,and the investigation and measurement methods are according to Wuet al.(2016).

2.3.DNA extraction and library construction

Young leaves of the 178 F1individuals and the parents were collected,immediately frozen in liquid nitrogen and preserved at -80°C until use.The genomic DNA was isolated using an improved cetyltrimethylammonium bromide (CTAB) method (Puchooa 2014).DNA purity was checked using the NanoPhotometer®spectrophotometer(IMPLEN,CA,USA).DNA concentration was measured using Qubit®DNA Assay Kit in a Qubit®2.0 Flurometer (Life Technologies,CA,USA),and the DNA samples were diluted to 20 ng µL-1for library construction.

The library construction was performed by GBS protocol.First,the genomic DNA was incubated withMseI (New England Biolabs,NEB),T4 DNA ligase (NEB),ATP (NEB),andMseI Y adapter N containing barcodes at 37°C,and then heated at 65°C to inactivate the enzymes.To further decrease the complexity and increase the sequencing depth and genome coverage,two additional enzymes,HaeIII andEcoRI (NEB),were simultaneously added into theMseI digestions to further digest the fragments at 37°C.Second,the digested fragments with ligations were purified with Agencourt AMPure XP (Beckman,California,USA) and subjected to PCR amplification using Phusion Master Mix(NEB) after adding a universal primer and index primers to each sample.The PCR products were purified using Agencourt AMPure XP (Beckman) and pooled,then run out on a 2% agarose gel.Fragments of 350 to 400 bp in size (with indexes and adaptors) were isolated using a gel extraction kit (Qiagen,Valencia,CA).These fragments were then purified using the Agencourt AMPure XP System,then diluted for sequencing.Finally,150-bp paired-end sequencing was performed on the selected tags on the Illumina HiSeq 4000 platform (Illumina,USA) by Novogene Bioinformatics Institute,Beijing,China.

2.4.Data analysis

The sequences of each sample were sorted according to the barcodes in the raw reads.To ensure that reads were reliable without artificial bias in the subsequent analysis,raw data in FASTQ format were first processed through four stringent filtering steps:(1) Removing the reads with≥10% unidentified nucleotides (Ns);(2) Removing the reads with ＞50% of bases having a Phred quality ＜5;(3)Removing the reads with ＞10 nt aligned to the adapter,allowing ≤10% mismatches;and (4) Removing the reads containing theHaeIII orEcoRI enzyme sequences.Then,the Burrows-Wheeler Aligner (BWA) Software was used to align the clean reads from each individual against the litchi reference genome (data provided by South China Agricultural University) with settings ‘mem-t-4-k 32-MR’,where t is the number of threads,k is the minimum seed length,M is an option used to mark shorter split alignment hits as secondary alignments,and R is the read group header line (Li and Durbin 2009).Alignment files were converted into BAM files using the SAMtools Software by the sort setting (Liet al.2009).

2.5.SNP calling and genotyping

SNP calling was performed for all samples using the SAMtools Software (Liet al.2009).The number of SNPs and types of transitions or transversions were counted.Then,a custom Perl script was used to filter the SNPs that had more than two genotypes,and those exhibiting segregation distortion or sequencing errors were discarded.Polymorphic markers between the two parents were detected and classified into eight segregation patterns(ab×cd,ef×eg,hk×hk,lm×ll,nn×np,aa×bb,ab×cc,and cc×ab) according to the CP (cross-pollinated) Model in JoinMap 4.0 Software (Van Ooijen 2011).Among them,hk×hk,lm×ll,ef×eg,ab×cd,and nn×np were selected for construction of the genetic map of the F1population.

2.6.Genetic map construction

Markers were filtered by JoinMap 4.0 if they contained abnormal bases,showed significantly distorted segregation(P＜0.001),or did not have enough integrity (＜90%).Markers in the nn×np segregation patterns (the female parent is homozygous,and the male parent is heterozygous) were used for the male parent genetic map construction,while markers in the lm×ll segregation patterns (the female parent is heterozygous,and the male parent is homozygous) were used for the female parent map using JoinMap 4.0.Linkage analysis was performed under the parameters set for the cross-pollinated (CP) population type.The regression algorithm,three times circulation sequence,and Kosambi mapping function (Kosambi 1943) were used in marker distance calculations.The logarithm of the odds (LOD)value was set to 2-20.The integrated map for both the male and female parents was computed using the combine group for map integration function in MergeMap Software.A Perl script SVG was used to visualize the exported maps,and heat maps were constructed to evaluate the maps.

2.7.QTLs mapping

Dwarf-related traits were checked whether they were normally distributedviathe Shapiro-Wilk test using R language (Royston 1982).If they did not display a normal distribution,the values were transformed to show a normal distribution by transformation.MapQTL®6.0 (Van Ooijen 2009) was used to map dwarf-related QTL,and the interval mapping (IM) and multi-QTL mapping (MQM) methods were used to map QTLs.IM was initially performed to reveal general information on the candidate QTLs.Then,the MQM was used to identify multiple QTLs for every linkage group,while the loci with the largest LOD scores in the other linkage groups were selected as co-factors.The LOD scores in each linkage group were fixed after searching the co-factors in all other linkage groups.The threshold of LOD scores for evaluating the statistical significance of QTL effects was determined by using 1 000 permutations for each trait and a thresholdP-value of 0.05 with the MQM permutation test function (Churchill and Doerge 1994).Those with LOD above the threshold were identified as QTL.The location of each QTL was determined according to the location of its LOD peak and the surrounding region.

2.8.Statistical analyses

Statistical analyses were performed using SPSS 20.0 statistical software (SPSS Inc.,Chicago,USA) with the one-way ANOVA method.The normality of distributions was tested using the Shapiro-Wilk test,assuming significant difference from a normal distribution whenP-value＜0.05.To obtain the correlation matrix among traits,we calculated the Pearson coefficient with the R package“Hmisc”with“corrplot”.

3.Results

3.1.Phenotyping of the F1 population

By statistical analysis of data from 2015 to 2017,several interesting traits related with dwarf morphology segregated in the population.Except for the trait of PH,the other traits such as LNB,DNB,LCP,DCP,LI,LSL,and WSL did not show significant differences between years,suggesting that the variances of these traits might be mostly influenced by genotype (Table 1).The averages of the three-year’s data for each tree were used for QTL analysis (Appendix A).

Table 1 The analysis of variance of eight phenotypic traits in three years

The F1hybrid progenies have significant separation in those dwarf-related traits such as length of internode,length of new branch,and leaf size (Fig.1;Appendix B).LNB ranged from 11.00 to 33.63 cm,with an average of 19.19 cm.DNB ranged from 3.79 to 11.32 mm,with an average of 6.66 mm.LCP ranged from 54.70 to 146.23 mm,with an average of 90.30 mm.DCP ranged from 1.69 to 3.54 mm,with an average of 2.41 mm.LI ranged from 17.86 to 63.50 mm,with an average of 34.25 mm.LSL ranged from 72.24 to 148.15 mm,with an average of 114.98 mm.WSL ranged from 22.68 to 51.26 mm,with an average of 38.61 mm.PH ranged from 68.75 to 280.88 cm,with an average of 167.31 cm.The values for LCP,LSL,WSL,and PH followed a normal distribution,but values for LNB,DNB,DCP,and LI did not (Fig.2),so they were transformed for normality by using the formula:y=lnx2,wherexrepresents trait values before transformation andyrepresents trait values after transformation.

Fig.1 Compound leaf characteristics of the parents and some F1 progenies.FZX,Litchi chinesis Sonn.vigorous cultivar ‘Feizixiao’;ZNX,L.chinesis Sonn.dwarf cultivar ‘Ziniangxi’.

The correlations between traits are presented in Fig.3,and Pearson or relation coefficients between dwarf-related traits ranged from 0.065 to 0.663 (Appendix C).As expected,significant correlations were found between LNB and LI,common petiole (LCP,DCP) and single leaf (LSL).A positive correlation was also detected between DCP and WSL.

Fig.3 Correlation matrix between the measured traits.The scale represent the values of Pearson coefficient between the traits using the mean value across the blocks for each variable.LNB,length of new branch;DNB,diameter of new branch;LCP,length of common petiole;DCP,diameter of common petiole;LI,length of internode;LSL,length of single leaf;WSL,width of single leaf;PH,plant height.

3.2.Genotyping-by-sequencing (GBS)

We obtained 67.85 G raw data containing 235 558 698 paired-end reads by sequencing 178 F1progenies(accession number:SUB6 289239;Appendix D).The parents were sequenced at a higher level to improve the chances of detecting more SNP markers,and 30.03 Gb of raw data containing 198 270 762 paired-end clean reads were obtained from maternal parent ZNX and paternal parent FZX.After data filtering,the clean data ranged from 176 536 224 to 564 476 256 bp for the F1progenies,with an average of 381 128 680 bp.In addition,13 729 883 400 and 16 010 730 900 bp clean data were obtained for ZNX and FZX,respectively (Appendix E).The Q30 was 85.19%,and the guanine-cytosine (GC) content was 36.74%.

Cleaned reads were mapped against the litchi genome(unpublished).As a result,91 532 556 clean reads were obtained for ZNX and 106 738 206 clean reads for FZX.For the F1progenies,an average of 2 646 727 clean reads were aligned to the reference genome.As shown in Appendix F,98% of these clean reads could be aligned to unique locations on the genome and used for the subsequent SNP calling and genotyping.The average sequencing depth was 16.88 for ZNX,15.33 for FZX,and 10.00 for each progeny.

3.3.SNP discovery and genotyping

In total,8 923 004 and 1 142 816 SNPs were detected in the female and male parents,respectively.The frequency of the single nucleotide substitutions observed here was about 1 SNP in 921 bp.For the F1progenies,an average of 275 836 SNPs was discovered for each progeny.Among them,the heterozygosity of the female parent was 56.65% and the male parent was 51.37%.The average heterozygosity of F1individuals was 50%.After filtering out the SNPs lacking parental information,8 654 911 polymorphic SNPs were detected between the two parents.As shown in Fig.4,these SNPs could be classified into eight segregation patterns according to the CP model in JoinMap 4.0.Among them,nn×np,hk×hk,lm×ll,and aa×bb accounted for nearly 99.62%,while the other four patterns,cc×ab,ef×eg,ab×cd,and ab×cc,only accounted for 0.38%.In this study,the five segregation types of hk×hk,lm×ll,ef×eg,ab×cd,and nn×np were selected for genotyping in F1individuals.

Fig.4 Statistics of the genotyped genotyping-by-sequencing (GBS) markers in eight segregation patterns.

3.4.Linkage map construction

A total of 5 943 SNPs markers was used for the linkage map construction,including 5 039 markers with nn×np,779 markers with lm×ll,106 markers with hk×hk,and 19 markers with ef×eg segregation types.Finally,the genetic map of the F1population consisted of 3 027 SNP markers distributed over 15 linkage groups (LG1-LG15)(Fig.5;Table 2).The map spanned a cumulative distance of 1 711.97 cM,with each linkage group ranging from 61.22 cM (LG10) to 184.43 cM (LG13),with an average of 114.13 cM.The number of SNP markers mapped in each linkage group ranged from 82 (LG8 and LG10) to 612(LG7),with an average of 201.8 SNPs per linkage group.The average distance between two adjacent markers was 0.57 cM across the 15 linkage groups (Table 2).

Fig.5 Distribution of the SNP markers on 15 linkage groups (LG1-LG15).

The highest density group was LG7,with 612 markers spanning 122.40 cM length,and the average marker density was 0.20 cM.The longest group was LG13,which contained 262 markers spanning 184.43 cM length,with an average genetic distance of 0.70 cM.The shortest group was LG10,containing 82 markers with an average genetic distance of 0.75 cM.A total of 3 014 gaps were detected in these 15 LGs,and the average percentage of markers having a gap distance of ≤5 cM was 98.47%.Thirty-seven gaps were between 5 and 10 cM,and nine gaps were larger than 10 cM,which were on LG14,LG12,LG13,LG7,and LG9,respectively.There was no gap distance of ≥5 cM on LG4 or LG5.The largest gap was 17.13 cM,which was located on LG7 (Table 3).

3.5.QTLs for dwarf-related traits

A total of 37 QTLs was identified,including 4 for LNB,3 for DNB,5 for LCP,6 for DCP,6 for LI,4 for LSL,6 for WSL,and 3 for PH (Table 4;Appendix G).The LOD scores of the QTL varied from 3.05 to 5.79,with an average of 3.73.The phenotypic variance explained by these QTLs ranged from 8.0 to 14.7%,with an average of 9.7%.For LNB,DNB,LCP,DCP,LI,LSL,WSL,and PH,the QTLs explained 44.3,28.1,48.6,60.7,50.1,35.4,66.2,and 25.9%,respectively,of the total phenotypic variance.The phenotypic variance explained by these QTLs was more than 20% of the total variation,indicating that the major loci could be detected.

Table 2 Distribution of the SNP markers on the linkage groups

Table 3 Gaps on the linkage groups

Table 4 Characteristics of quantitative trait loci for dwarf-related traits

These 37 QTLs were distributed across eight linkage groups on the map,including LG1,LG4,LG5,LG6,LG7,LG9,LG11,and LG14 (Fig.6).Seven QTLs related to dwarf-related traits were localized on LG4 and LG11,which indicated that LG4 and LG11 could show enrichment for genes regulating dwarf-related traits.Several clusters of QTLs were detected in the narrow region on some LGs.The first major cluster containing seven QTLs (qDw-4,qDw-7,qDw-12,qDw-17,qDw-24,qDw-33,andqDw-37)was located on LG11 between 44.135 and 81.358 cM.The second cluster containing six QTLs (qDw-1,qDw-9,qDw-13,qDw-20,qDw-29,andqDw-36) was located on LG4 between 59.29 and 94.84 cM.The third cluster containing five QTLs (qDw-2,qDw-10,qDw-14,qDw-26,andqDw-30)was located on LG5 between 35.097 and 45.592 cM.In addition,three other clusters were situated between 68.272 and 79.973 cM of LG6 (qDw-11,qDw-15,qDw-22,andqDw-31),between 24.981 and 36.534 cM of LG7 (qDw-5,qDw-23,andqDw-27),and between 66.682 and 76.954 cM of LG14 (qDw-18andqDw-34).

Fig.6 Linkage distribution of QTLs.Colored bars show the locations of QTLs,and the names of the QTLs are listed in Table 4.DCP,diameter of common petiole;DNB,diameter of new branch;LI,length of internode;LCP,length of common petiole;LNB,length of new branch;PH,plant height;LSL,length of single leaf;WSL,width of single leaf.LG1-LG14,linkage groups.

In particular,multiple QTLs associated with the dwarfrelated traits were identified at the same site.For example,QTL of DCP (qDw-13) and QTL of WSL (qDw-29) were both on LG4,while QTL of DCP (qDw-18) and QTL of WSL (qDw-34) were on LG14,and QTL of LNB (qDw-03) and QTL of DCP (qDw-16) were on LG9.

3.6.The expression patterns of the candidate genes in the QTL interval

Base on the RNA-seq data using mature leaves and apical buds of FZX and ZNX (Huet al.2018),126 candidate genes were identified within the QTL regions,55 of which are differentially expressed between FZX and ZNX (Fig.7).There were 26 DEGs (9 down-regulated and 17 upregulated) between the ZNX-leaves and FZX-leaves,and 24 DEGs (8 down-regulated and 16 up-regulated) between the ZNX-buds and FZX-buds.Among them,21 genes showed significant differential expression (7 down-regulated and 14 up-regulated) in both leaves and apical buds between ZNX and FZX.Regarding KEGG pathways,36 genes wereassigned to 39 KEGG pathways,such as ‘phenylpropanoid biosynthesis’,‘cyanoamino acid metabolism’,‘porphyrin and chlorophyll metabolism’,‘starch and sucrose metabolism’,and ‘glycerolipid metabolism’ (Appendix H).Furthermore,some important pathways related to plant growth,development,and morphogenesis,including‘other glycan degradation’,‘amino sugar and nucleotide sugar metabolism’,‘plant hormone signal transduction’,‘circadian rhythm’,and others,have also been annotated successfully.These DEGs were found to participate in regulation of cell development,material transportation,signal transduction,and morphogenesis of plants,so they might play important roles in regulating plant dwarfrelated traits.

Fig.7 Heatmap of relative expression levels of candidate genes in Litchi chinesis Sonn.dwarf cultivars ‘Ziniangxi’ (ZNX;dwarf)and vigorous cultivar ‘Feizixiao’ (FZX;vigorous).

4.Discussion

Only one genetic linkage map has been developed from data in litchi.However,that genetic map was mainly constructed using RAPD,AFLP,and SRAP markers (Liu and Mei 2003;Liuet al.2010;Zhaoet al.2010) and the map density was very low because of the limitations of these markers.Zhaoet al.(2010) constructed the litchi genetic map based on 76 F1individuals covering 610.55 cM with an average distance of 3.75 cM.In the present study,a high-density linkage map of litchi using SNP markers was developed by GBS technology.A larger F1population (178 F1population between ZNX and FZX) and more markers (5 943 SNP markers) were used in this study.Thus,our genetic linkage map has the shortest average genetic interval (0.57 cM/marker) (Fig.5) and constitutes the highest-density genetic map of litchi constructed to date.Moreover,the length of the integrated genetic map,which spans 1 711.97 cM,is much longer than the one obtained by Zhaoet al.(2010).

GBS has been successfully used in the construction of high-resolution genetic maps for many different plants(Glaubitzet al.2014).It generates large numbers of SNPs for use in genetic analyses and genotyping (Beissingeret al.2013).Compared with microsatellites,SNPs are considered to be a better tool for rapidly constructing high-density genetic maps because they are more abundant and can be genotyped on a much larger scale (Slateet al.2009).The genetic map in this study is the first litchi genetic map constructed using SNP markers developed from GBS.The number of SNP markers used in this study was more than 3 000.More SNPs will be developed because of the highly heterozygous genetic background of the hybrid population(Wuet al.2016).FZX is a vigorous cultivar,with long,sparse,strong branches,and large leaves,and ZNX is a dwarf cultivar,with relatively low,thin,short,and open spreading branches.Therefore,GBS has been successfully implemented in our F1litchi population,and allowed for the construction of a high-density genetic map with a greater genetic depth and more accurate QTL detection.

QTLs for dwarf-related traits are undoubtedly important for fruit trees as they can be used for selecting dwarf cultivars for modern orchard production systems.In apple,a major dwarfing locus (Dw1) was first identified on LG5 by a QTL analysis of a rootstock population derived from the cross between ‘M9’ and ‘R5’ (non-dwarfing) (Pilcheret al.2008).Later,a smaller-effect QTL affecting dwarfing(Dw2) was identified on LG11 (Fazioet al.2014).The combination ofDw1andDw2has the strongest influence on rootstock-induced dwarfing,andDw1has a stronger effect thanDw2(Fosteret al.2015).In addition to apple,genetic markers linked to the dwarfing trait were detected in several other fruit trees,such as pear (Wanget al.2011),peach (Hollenderet al.2016;Luet al.2016),and oil palm(Leeet al.2015).However,to date,the mechanisms for the genetic factors controlling tree size and shape of litchi are still poorly known,and the inheritance of dwarf traits in litchi has not been clearly characterized.In the present study,dwarf-related traits of F1progenies by crossing dwarf cultivar ZNX and vigorous cultivar FZX mostly showed a normal or approximately normal distribution (Fig.2),and exhibited continuous variance,indicating that these characteristics are quantitative.The high-density linkage map developed in this study enabled us to identify QTLs with greater accuracy and higher resolution.A total of 37 QTLs for eight dwarfrelated traits of litchi were detected in this study which explained phenotypic variance ranging from 25.9 to 66.2%,indicating that dwarf-related traits are complex (Table 4).The phenotypic variances explained by these QTLs were relatively high (with an average of 44.9%),suggesting that the major loci were detected and the dwarf-related traits could be regulated by several major genes.

In the present study,these SNP markers were easily located on the reference genome sequence and could be used to screen candidate genes related to traits on those corresponding chromosomal regions.Hollenderet al.(2016)identified a candidate gene fordwas a non-functional allele of the gibberellic acid receptor PpeGID1c by further sequence analysis and fine mapping,indicating that PpeGID1c serves a primary role in vegetative growth and elongation of peach.Pootakhamet al.(2015) identified two candidate genes co-locating with SNP markers linked to the cluster of a QTL affecting trunk height,GAI1 and GA2ox2,which have both been implicated in the regulation of plant height through the gibberellin signaling pathway.In our study,126 candidate genes were initially identified within 32 QTL regions,and 55 of them had significant differences in expression between ZNX (dwarf) and FZX (vigorous).These DEGs were found to participate in the regulation of cell development,material transportation,signal transduction,and morphogenesis of the plants.Moreover,the clustering of genes with similar functions is universal,as it was observed for peach (Eduardoet al.2011),papaya (Blaset al.2012),and Chinese chestnut(Jiet al.2018).Interestingly,we found two genes which have been reported to be associated with dwarfing out of 55 DEGs.One of them isLitchi_GLEAN_10023629for the QTL of WLS (qdw-29)as the transcription factor GAMYB,which was a locus on LG4.GAMYB mediates dwarf-related traits of plants by regulating key genes of GA synthesis(Gubleret al.2002;Xianget al.2014).The other one isLitchi_GLEAN_10026086for the QTL of WLS (qdw-33) as the NAC transcription factor,which was a locus on LG11.NAC transcription factors have been implicated in controlling dwarf traits,such as inArabidopsis thaliana(Katoet al.2010) and apple (Jiaet al.2018).

In addition,QTLs for highly correlated traits were mapped to the same or adjacent LG regions (Changet al.2014;Jiet al.2018).In this study,several QTLs were identified at the same site.qDw-6 for DNB andqDw-32 for WSL shared genetic loci on LG9 at 10.941 cM (Table 4).Stable QTLs are valuable for MAS breeding programs and have been used in fruit trees (Changet al.2014;Pottset al.2014).Interestingly,seven QTLs clustered on LG4 and LG11 (Fig.6),indicating that LG4 and LG11 perhaps contain multiple dwarf-related genes.As expected,11 DEGs were mapped in LG4 and 20 DEGs were mapped in LG11.In our study,several QTLs for dwarf-related traits were found to be stable and 55 candidate genes in the QTL interval were differentially expressed between the parental lines in the three-year study,which could be useful in MAS programs for finding the genetic factors controlling plant size and shape and facilitating the early selection for litchi molecular breeding.

5.Conclusion

We used GBS technology to develop SNP markers based on an F1population derived from the cross of ZNX (dwarf)×FZX(vigorous),and constructed the first high-density SNP map with an average distance of 0.57 cM between two adjacent markers across 15 linkage groups in litchi.Based on this genetic map,37 QTLs for dwarf-related traits were detected,including 4 for LNB,3 for DNB,5 for LCP,6 for DCP,6 for LI,4 for LSL,6 for WSL,and 3 for PH.A total of 55 candidate genes were identified within the QTL regions,which were significantly differentially expressed between ZNX and FZX.The high-density genetic map and expression patterns of the candidate genes in the QTL intervals will help us understand the dwarfism mechanism and provide an efficient tool for molecular-assisted breeding of litchi in the future.

Acknowledgements

This study was funded by the Key-Area of Research and Development Program of Guangdong Province,China (2018B020202011),the National Natural Science Foundation of China (31701885),the China Agriculture Research System of MOF and MARA (CARS-32-05),the State Key Laboratory for Conservation and Utilization of Subtropical Agro-bioresources,China (SKLCUSA-b201716),and the YangFan Innovative &Entrepreneurial Research Team Project,China (2014YT02H013).

Declaration of competing interest

The authors declare that they have no conflict of interest.

Appendicesassociated with this paper are available on http://www.ChinaAgriSci.com/V2/En/appendix.htm