Construction of a High-Density Genetic Map and Mapping of Major Flowering Time QTLs in Rapeseed (Brassica napus L.) Based on Whole-Genome Re-sequencing

the interaction of all the in In this study, we conducted a genome-wide quantitative trait locus (QTL) analysis of a recombinant inbred line (RIL) population composed of 215 lines and constructed a high-density linkage map with 4523 recombination bin markers and of 92,627 single-nucleotide polymorphisms (SNPs). After phenotypic measurements across four different elds environments were collected for 2 years, at least 27 owering time QTLs were identied; these were distributed mainly on linkage groups (LGs) A02 (11 QTLs), A10 (1 QTL), C02 (12 QTLs), C05 (2 QTLs) and C08 (1 QTL). Of these QTLs, ve major QTLs for owering time were detected, and three of them, sqFT-A02-3, sqFT-A02-4, and sqFT-A10-1, were identied only in winter ecological conditions; thus, these QTLs were environment specic. The QTL cqFT-C02-2 was detected in all 3 spring environments, showing strong spring environmental specicity. The major QTL cqFT-C02-3, which is an environmentally stable QTL, was detected in all 4 environments and was expressed both in spring and winter ecological conditions. The winter ecological condition-specic major QTL sqFT-A02-4 was delimited to a region of approximately 216 kb, between 6.302 Mb and 6.518 Mb on A02. In this region, the BnaA02g12130D/GSBRNA2T00090951001 gene was identied as a homologue of the Arabidopsis thaliana FLOWERING LOCUS T (FT) gene by a BLASTn analysis of the candidate region. by subtracting the date of seedling emergence from the owering date. The date of cotyledon attening was recorded as the seedling emergence date of each plant. The seedling emergence in a single plot was considered the date when 75% of the plants in the plot reached seedling emergence. The date of the rst ower per plant was recorded as the owering date of each individual, and the owering time in a single plot was considered the date when 25% of the plants in the plot had owered. QTL: quantitative trait locus; SNPs: single-nucleotide polymorphisms; LGs: linkage groups; GA: gibberellic acid; WGRS: whole-genome re-sequencing; Indels: insertions/deletions; SV: structure variation; CNV: copy number variation; ANOVAs: analyses of variance; G: genotype; E: environment; h 2 : The broad-sense heritability; RIL: recombinant inbred line; DH: double-haploid; BC: back-cross; NILs: near-isogenic lines; RHLs: residual heterozygous lines; BRAD: Brassica Database; SSR: simple sequence repeat; GLM: general linear model; BWA: Burrows Wheeler Aligner; TAIR: the Arabidopsis Information Resource.

thereby promoting plant owering [12]. The product of FT expression is a owering stimulant that is translocated from the leaf to the stem tip over long distances, ultimately causing owering [13,14]. Many FT-like genes have been cloned, and their expression has been suggested to promote owering in plants [15]. B. napus and A. thaliana are cruciferous plants and are closely related. In addition, knowledge of genes related to owering in B. napus is largely based on A. thaliana via comparative genomics. B.napus BnFT (BnA2.FT, BnA7.FT.a, BnA7.FT.b, BnC2. FT, BnC6.FT and BnC6.FT.b) is homologous to the A. thaliana FT gene, which is considered the key gene for determining owering time in the photoperiod pathway [16].
Quantitative trait locus (QTL) analysis has been applied to identify potential candidate genes that control owering time in B. napus [17]. Flowering time QTLs with large phenotypic effects have been mapped onto chromosomes A02, A09, A10, C02, C03 and C08 in B. napus. Using a double-haploid (DH) population of 217 lines and a genetic linkage map with 495 markers, Liu et al. [18] detected a total of 48 QTLs related to owering time. In addition, by performing genome-wide association mapping of owering time with a 60K single-nucleotide polymorphism (SNP) array for a diversity panel comprising 523 inbred rapeseed lines, Xu et al. [19] identi ed 35 owering time QTLs. Li et al. [20] used a DH population of 348 lines and a genetic linkage map with SNP markers and detected a total of 55 QTLs related to owering time, which explained 2.99-25.96% of the phenotypic variation.
The development of high-throughput sequencing technology has increased the precision of species research, and the use of modern molecular methods has become a new trend in QTL research. Highthroughput sequencing technology provides new opportunities and possibilities for crop genetics and breeding [21,22]. Whole-genome re-sequencing (WGRS) is a powerful tool to identify various types of genetic variation, including SNPs, insertions/deletions (Indels), structure variation (SV), and copy number variation (CNV) [23]. The determination of these variations has been used in QTL mapping and gene cloning [24,25]. Compared with previous methods, this mapping method can signi cantly improve the e ciency of QTL mapping and marker development [26]. Page 4/22 In this study, we constructed a high-density linkage map by re-sequencing recombinant inbred lines (RILs) comprising 215 individuals that were developed from a cross between spring-type No. 3379 with an earlyowering phenotype and from a cross between spring-type No. 2839 with a markedly late-owering phenotype. The parents and RIL population were planted in three spring and winter ecological environments, and the owering times of the panel accessions were investigated. The objectives of this study were to (1) evaluate the genetic variation and identify environmentally stable and environmentspeci c QTLs for owering time based on the high-density genetic map in different ecological conditions, (2) delimit the winter ecological conditions-speci c major QTL sqFT-A02-4 to a target region, and (3) select candidate genes associated with owering time within the sqFT-A02-4 target region. The results should provide us with a better understanding of the genetic regulation of early owering.

Results
Phenotypic performance of the parents and the F1 and RIL population T tests and analyses of variance (ANOVAs) were used to compare the differences in owering time between the parents and RIL population in four environments ( These results indicate that all the bin markers were thoroughly distributed throughout the genome, and approximately 98.79% of the intervals between adjacent markers were less than 5 cM. In addition, to evaluate the quality of the genetic map, the collinearity between the genetic map and the B. napus reference genome was assessed. All bin markers were mapped to the B. napus reference genome. As shown in Additional le 6: Table S4 and in Additional le 7: Fig. S3, the Spearman correlation coe cients between the genetic map and physical map exceeded 0.9 for 17 chromosomes, excluding the coe cients corresponding to A10 (0.875) and C07 (0.871). The LGs thus have high levels of genetic collinearity with the physical map ( Fig. 1).

QTL analysis of owering time and identi cation of environmentally stable and environment-speci c QTLs
The high-density genetic linkage map combined with the phenotypic data was used to detect QTLs for owering time. A total of 27 QTLs associated with owering time were identi ed in the RIL population ( Table 2). These QTLs were distributed mainly on LGs A02 (11 QTLs), A10 (1 QTL), C02 (12 QTLs), C05 (2 QTLs) and C08 (1 QTL) and individually explained 3.25% to 23.78% of the phenotypic variance. The additive effect (from -0.79 to -9.02 days) of all the QTLs detected in the present study are negative, implying that parent No. 3379 contributed alleles for a shorter owering time. The QTLs were subsequently analysed in different ecological conditions. A total of 21 QTLs, which were located on A02, C02, C05 and C08, were detected in three spring environments (Ledu, Xining and Huzhu). A total of 6 QTLs, which were distributed mainly on A02, A10 and C02, were detected in the winter ecological conditions (Yuanmou) (Fig. 1).
To verify the QTLs identi ed in multiple environments and integrate overlapping loci as consensus QTLs, BioMercator 2.1 software was used for a meta-analysis. Eighteen QTLs from the 27 QTLs detected in this study were integrated into 6 consensus QTLs as environmentally stable QTLs, and the remaining 9 QTLs were considered environment-speci c QTLs ( Table 2). The QTL cqFT-C02-3 was stable and detected in all 4 environments (one winter and three spring ecological conditions), explaining 3.83% to 21.61% of the phenotypic variance and -1.50 to -5.32 of the additive effects. A total of 4 QTLs (cqFT-A02-1, cqFT-A02-2, cqFT-C02-1 and cqFT-C02-2) were detected in the 3 spring environments; these QTLs explained 3.60-5.96%, 4.01-6.99%, 7.09-12.49% and 4.45-21.15% of the phenotypic variance, respectively. Nine QTLs for owering time with no overlapping con dence intervals in the different environments were considered environment-speci c QTLs, of which 4 were identi ed only in the spring ecological conditions and of which 5 were identi ed only in the winter ecological conditions.
The major QTLs explaining a relatively high proportion of the phenotypic variation are the focus of our research. There were 5 large-effect QTLs detected: sqFT-A02-3, sqFT-A02-4, sqFT-A10-1, cqFT-C02-2 and cqFT-C02-3 ( Table 2, Fig. 2). The QTLs sqFT-A02-3, sqFT-A02-4 and sqFT-A10-1 explained 23.78%, 22.98% and 10.70% of the phenotypic variation, respectively. Three major QTLs were detected only in the winter ecological conditions, showing strong winter environmental speci city. The QTL cqFT-C02-2 explained 13.71% of the phenotypic variance and was detected in all 3 spring environments, showing strong spring environmental speci city. The major QTL cqFT-C02-3 was detected in all 4 environments and explained 11.02% of the phenotypic variance; this QTL was expressed both in the spring and winter ecological conditions. cqFT-C02-3 was considered an environmentally stable QTL. Regardless of whether these major QTLs were environmentally stable QTLs or environment-speci c QTLs, they were all considered key QTLs for owering time for subsequent ne mapping and map-based cloning analyses.
Delimitation of the environment-speci c major QTL sqFT-A02-4 target region In this study, several environment-speci c QTLs with large-effect values were detected, and three of these major QTLs were detected only in winter ecological conditions, showing strong winter-environment speci city. The location of one of the major QTLs, sqFT-A02-4 (whose phenotypic variation exceeded 20%), was delimited to a genetic region between markers Block9955 and Block10019 (Fig. 3a), corresponding to a candidate interval of approximately 216 kb between 6.302 Mb and 6.518 Mb on A02 of the B. napus genome (Fig. 3b). Therefore, the target gene may be located within the 216 kb range near the 6.3 Mb position on A02. The sequence of the candidate region was analysed, in which a total of 27 predicted genes were included. Of these genes, the BnaA02g12130D/GSBRNA2T00090951001 gene was highly homologous to A. thaliana AT1G65480 (Fig. 3c). AT1G65480 is the FLOWERING LOCUS T (FT) gene in A. thaliana, which promotes owering with LEAFY (LFY) and is antagonistic to the TERMINAL FLOWER1 (TFL1) gene; FT mutants exhibit late owering. FT is a key gene involved in the photoperiod pathway and determines owering time [27][28][29]. Therefore, we inferred that BnaA02g12130D (BnA02.FT) was the most important candidate gene for owering time.
Discussion Rapeseed (B. napus) originated along the Mediterranean coast approximately 7500 years ago. In China, the annual production of rapeseed oil reaches 3.5 million tons, accounting for 35% of China's edible vegetable oil. In China, rapeseed is distributed mainly in southern China (winter type) and on the Qinghai-Tibet Plateau (spring type). However, in recent years, the cultivation area of southern rapeseed has been continually declining, mainly because winter-type varieties have a long owering time and a long maturity period that persists until mid-and early May of each year, which overlaps with the planting time of early rice and cotton crops grown in mid-to-late April. Eventually, most of the farmland is left idle and not used for agricultural production. The Qinghai-Tibet Plateau has an average elevation of more than 4000 metres and has a dry and cold climate, and only a short time is suitable for crop growth. Long owering time can cause delayed maturity and the formation of abnormal or incompletely mature seeds, which subsequently lead to a decline in yield. The key to solving these problems is the breeding of rapeseed varieties with early-owering as well as early-maturing phenotypes, which are urgently needed. In this study, we detected 27 QTLs in multiple environments, and the effects of these alleles may shorten the owering time by 0.79 to 9.02 days. Among these QTLs, the major QTL cqFT-C02-3 was identi ed in spring and winter ecological conditions as environmentally stable QTL that could shorten the owering time of the late-owering parent by 5.32 days in winter ecological conditions and shorten the owing time by approximately 2 days in spring ecological conditions. The major QTL sqFT-A02-3 had an effect only in the winter ecological conditions, which could shorten the owering time of the late-owering parent by approximately 9 days. Study of the molecular mechanisms of these major QTLs would provide more theoretical references for breeding early-maturing rapeseed varieties.
Flowering time is an important agronomic trait. Constructing a mapping population is the key and foundation for QTL analysis. There are many kinds of mapping populations that can be used to construct linkage maps. According to their characteristics, there are two types of mapping populations: temporary and permanent ones. On the basis of the accuracy of QTL positioning, there are primary positioning genetic populations and ne positioning genetic populations. Speci cally, they are divided into F2, backcross (BC), DH, and RIL as well as ne positioning populations of near-isogenic lines (NILs) and residual heterozygous lines (RHLs). RIL populations have been used to construct genetic linkage maps in rice [30]. RIL populations are constructed from F2s via single-grain transmission after multiple rounds of meiosis, and the degree of recombination is relatively high. Compared with DH populations, which are subject to severe partial segregation, RIL populations have higher accuracy and can be tested in a long-term and repeated manner. Likewise, we constructed a RIL population of 215 lines and planted them in both spring and winter ecological conditions. We ultimately used the genetic map constructed via this RIL population to identify multiple environmentally stable consensus QTLs.
QTL mapping of owering time traits is one of the most effective ways to elucidate the genetic basis of owering time and identify potential candidate genes underlying QTLs. There are currently many reports on QTL mapping of owering genes in rapeseed, mainly in B. napus. Because of the differences in genome coverage, marker intensity, and speci c markers applied to the maps, the genetic maps constructed by different populations signi cantly differ in terms of their QTL loci in different studies. For instance, Liu et al.
[18] detected 3 consensus QTLs related to owering time on the LG C2, but only cqDTFC02a was a major locus that explained 23.0% of the phenotypic variance. We submitted the anking marker sequence of the cqDTFC02a locus to the Brassica Database (BRAD) and compared it with the B. oleracea C02 chromosome; the results showed that cqDTFC02a was located at approximately 0.1 Mb from the end of the C02 chromosome in Brassica oleracea. Wei et al. [31] used a DH population of 261 lines and a genetic linkage map with simple sequence repeat (SSR) markers and detected QTLs (qFT10-5 and qFT11-7) for owering time; these QTLs explained more than 20% of the phenotypic variance in the same environment for two consecutive years. We used the same method above for BLAST analysis and found that qFT10-5 (qFT11-7) was located at approximately 4.0 Mb from the end of chromosome CO2 in B. oleracea. Moreover, we detected a major QTL, cqFT-C02-3, in all environments and conducted BLAST analysis on the C02 chromosome of B. oleracea; our results showed that cqFT-C02-3 is located at approximately 8 Mb from the end of the C02 chromosome in B. oleracea. Therefore, cqFT-C02-3 is different from the position of cqDTFC02a and qFT10-5 (qFT11-7). With respect to the A02 chromosome, Li et al. [20] compared the QTLs for owering time in different studies and found that the QTL Li-cqFT.A2-2 is a spring-speci c QTL that colocalizes in the common genomic region with Long-qFT2-6 [20,32], and those authors found that this QTL is located within the A02 genomic region from 6.27-6.32 Mb and from 6.28-6.28 Mb, respectively. Our A02 chromosome analysis revealed two major speci c QTLs in the winter ecological conditions: one of the major QTLs, sqFT-A02-4, was narrowed to a genomic region between 6.302 Mb and 6.518 Mb on chromosome A02 of B. napus. The genomic region of sqFT-A02-4 is close to that of both cqFT.A2-2 and Long-qFT2-6, so it is possible that sqFT-A02-4 is located in a common genomic region (Li-cqFT.A2-2 and Long-qFT2-6).
The whole-genome of B. napus has been sequenced. When the B. napus genomes was compared with the genomes of its progenitors B. rapa and B. oleracea, the B. napus genome was found to contain a remarkable evolutionary feature by which extensive DNA sequence exchange occurred between the A and C subgenomes [33]. Moreover, subgenomes A and C exhibit high synteny with each other, and these homologous genomic regions descend from a common ancestral genome and largely retain the same or similar genes, although some of them have undergone functional differentiation [34]. Studies have shown that the QTLs for owering time are distributed on chromosomes A02 and C02 in B. napus. Moreover, the QTLs display collinearity of the con dence intervals for chromosomes A02 and C02. For instance, Zhang et al. [35] performed a co-regional alignment of the con dence intervals of the detected owering time QTLs on chromosomes A02 and C02, and multiple QTLs were localized in the A02-C02 syntenic regions and were subsequently referred to as syntenic QTLs. Therefore, the QTLs within coregions on chromosomes A02 and C02 may indicate that these two syntenic regions are conserved and contain genes with the same or similar functions. For instance, Chen et al.
[36] detected two major QTLs on 2 LGs (A02 and C02). The two QTLs were subsequently delimited to the 86 kb and 201 kb regions via two separate near-isogenic populations. As a candidate gene, BnFLC was identi ed in both regions (A02 and C02) [36]. In the present study, ten QTLs were located on LG A02, and 13 QTLs were located on LG C02; the QTLs localized within the A02-C02 homologous regions will be analysed in subsequent studies.
Integrating QTLs with overlapping con dence intervals can help identify environmentally speci c QTLs or multi-environmentally stable QTLs. In this study, 21 and 6 QTLs were detected in the spring ecological conditions and winter ecological conditions, respectively. In addition, QTLs were stably detected in the winter and spring ecological conditions. These environmentally stable QTLs could be bene cial for the breeding of new rapeseed varieties with broad adaptability. Environmentally stable QTLs have been reported in several studies; for instance, Liu et al.
[18] discovered several owering time QTLs in B. napus that were stably detected in ve micro-environments. These QTLs were stable in different environments and are more suitable for ne mapping. Environment-speci c QTLs have also been reported; for instance, Li et al. [20] investigated QTLs for owering time in B. napus, and two major speci c QTLs were detected in spring ecological conditions. These QTLs were expressed in a speci c environment and were greatly affected by temperature and day length. In the present study, the major QTL cqFT-C02-3 was detected in all 4 environments and was considered environmentally stable. The major QTL cqFT-C02-2 was detected in all 3 spring environments, showing strong spring environmental speci city. Therefore, whether these major QTLs were environmentally stable or environment speci c, they were considered key QTLs for owering time for future ne mapping and map-based cloning analyses.
Compared with reports on the QTL mapping for owering time in B. napus, there are fewer reports concerning the cloning of genes responsible for owering time. Currently, more than 180 genes related to owering time have been detected in A. thaliana, and the majority of related research has focused on the FLC, CO, and FT genes. The A and C genomes of B. napus are remarkably similar to those of A. thaliana. Based on the genome synteny between A. thaliana and B. napus, many A. thaliana homologous genes have been cloned in B. napus [37,38]. For instance, Tadege et al. [39] cloned a homologue of A. thaliana FLC in B. napus (BnFLC1-5); BnFLC1-5 was subsequently transformed into A. thaliana, and the owering time of the transgenic plants was strongly delayed. Hou et al. [40] delimited the major QTL to an 80 kb interval and cloned a homologue of A. thaliana FLC in B. napus (BnFLC.A10). Similarly, Zheng et al. [41] cloned a homologue of A. thaliana CO in B. napus (Bn1CON19), and the similarity between the Bn1CON19 gene sequence and the AtCO gene sequence reached 99.8%. Robert et al. [42] identi ed three homologues of A. thaliana CO in B. napus (BnCOA1, BnCOA9, BnCOB1 and BnCOB9), and Wang et al. [16] identi ed three FT homologues in B. napus (BnFTA2, BnFTC6a and BnFTC6b) and 6 homologues of A. thaliana CO in B. napus: BnA2.FT, BnA7.FT.a, BnA7.FT.b, BnC2.FT, BnC6.FT and BnC6.FT.b. In addition, those authors reported that the variation in the BnFT allele might affect differences in owering time between the winter and spring. FT is considered a key gene that determines the owering time in the photoperiod pathway, and FT production leads to owering-stimulating substances. When A. thaliana receives an appropriate photoperiod induction, the CO gene initiates transcription and activates FT expression; afterward, the FT transcript (mRNA) moves from the leaves to the apical tissue, where the post-translational FT protein and transcription factor protein FD interact to promote the expression of the downstream owering gene AP1, thereby promoting owering [14,15,43]. In the present study, we selected BnA02.FT as a key candidate gene at a position close to 6.3 Mb on chromosome A02 for further ne mapping and sequence analyses.

Conclusions
Five major novel QTLs for owering time were discovered in spring Brassica napus: 1 environmentally stable QTL and 4 environment-speci c QTLs. The major QTL sqFT-A02-4 was delimited to a candidate region of approximately 216 kb, and BnaA02g12130D was determined to be a potential gene.

Phenotypic evaluations
The owering time was calculated by subtracting the date of seedling emergence from the owering date. The date of cotyledon attening was recorded as the seedling emergence date of each plant. The seedling emergence in a single plot was considered the date when 75% of the plants in the plot reached seedling emergence. The date of the rst ower per plant was recorded as the owering date of each individual, and the owering time in a single plot was considered the date when 25% of the plants in the plot had owered.

Statistical analysis of phenotypic data
Various modules of the Statistical Analysis System (SPSS 19.0) software package were used to analyse the phenotypic data. Broad-sense heritability (h 2 ) was calculated as σ 2 g / (σ 2 g + σ 2 ge / n + σ 2 e / nr) × 100%, where σ 2 g is the genetic variance, σ 2 e is the error variance, σ 2 ge is the interaction variance between genotypes and environments, n is the number of environments, and r is the number of replications in each environment. The estimates of σ 2 g , σ 2 ge and σ 2 e were obtained from a two-way ANOVA using the general linear model (GLM) procedure in SAS 8.1. In the variance analysis model, the variances of the genotypes, environments and interactions between the genotypes and environments were considered random effects. Correlations were calculated for the owering time between the four environments by SAS 8.1 software.

Library construction and high-throughput sequencing
The genomic DNA of the 215 RILs and their parents was extracted from fresh leaves (100 mg) using the cetyl-trimethylammonium bromide (CTAB) method. The quality of the DNA samples was measured, and the fragments was randomly interrupted by sonication to a size of 200-500 bp. The sequencing libraries were constructed by terminal repair followed by the addition of a 3'A and a sequencing adapter. The resulting fragments were further puri ed and ampli ed by PCR. After quality inspection, the constructed library was subjected to paired-end sequencing by an Illumina HiSeq 2500 system (Illumina, San Diego, CA, USA) following the manufacturer's instructions. Re-sequencing was performed by the Biomarker Technologies Company (Beijing, China).

Identi cation of SNPs and genotyping
The original reads were processed in accordance with the following quality control steps: 1) the sequences of the adapter were removed; 2) reads with 10% or more unidenti ed nucleotides were removed; and 3) reads having more than 50% bases with a phred quality lower than 10% were removed. Burrows Wheeler Aligner (BWA) was subsequently applied to align the clean reads of each material to the B. napus reference genome (http://www.genoscope.cns.fr/brassicanapus/). Potential PCR duplications were eliminated using the MARK duplicate tool of Picard [44]. The GATK software toolkit 3.8 was then used to detect potential SNPs among the lines [45]; this process included base recalibration, variation calling, and strict ltering of SNPs to obtain a nal SNP cluster. SnpEff software was used for variation (SNP and small InDel) annotation and for the prediction of variation effects [46].
To ensure the quality of the genetic map, markers were screened in accordance with the following three criteria: 1) only the aa´bb genotype was kept based on the SNPs of two parents, 2) the SNPs with a depth ≥ 1 in each parent were selected, and 3) the SNPs that were not located on chromosomes were removed.

Linkage map construction
Filling and correction of SNPs were carried out using a sliding window analysis, with 15 SNPs per window and 1 SNP per increment. The genotype of the window was identi ed as aa when there were 11 or more SNPs with the aa genotype in a sliding window. When the genotype of 11 or more SNPs in the sliding window was bb, the genotype of this window was classi ed as bb. Except for the above 2 cases, the genotype of a window was otherwise identi ed as ab.
Bin partitioning was subsequently performed according to the sub-reorganization. Each sample was arranged according to its physical position on the chromosome. When there was a genotype transition in any sample, a recombination breakpoint appeared. The SNP between the recombination breakpoints was included in the bin, and it was supposed that no recombination events occurred in the bin. The bins were subsequently used as mapping markers for genetic map construction [47].
According to the distribution of markers, bins with a length less than 10 kb were removed from LGs A09, C03, C04, and C07, and bins with a length less than 5 kb were removed from the other LGs. The markers showing severe segregation distortion were ltered by speci c parameters. The ltration parameter for A01, C03, C04 and C07 is 0.001; the parameter for A03, A04, A05, A10, C01, C05, C06 and C09 is 0.01; the parameter for A06, A07, A08 and C08 is 0.00001; and the markers on the remaining chromosomes were not ltered.
All bin markers of the linkage map corresponded to the physical sequences of the reference genome of B. napus. The collinearity between the genetic location and physical location was determined by mapping the relationship between the genetic marker positions (cM) and the physical positions (Mb).
QTL identi cation and analysis QTL mapping was performed by the composite interval mapping (CIM) method in Windows QTL Cartographer 2.5 software [48]. The LOD thresholds for determining signi cant loci were estimated from 1000 permutations [49]. The LOD threshold was set to 2.5 to detect signi cant QTLs for owering time.
The con dence interval for each QTL was de ned by a LOD change from the peak position -when multiple peaks are close to each other and within a half LOD distance. The identi ed QTLs were named according to the different environments and chromosome numbers [50,51]. The consensus QTLs were integrated by a meta-analysis by BioMercator 2.1 software [52]. If a QTL was identi ed only in one environment, then that QTL was considered an environment-speci c QTL. If a QTL was identi ed simultaneously in the spring and winter ecological conditions or in multiple spring environments, those QTLs were thus located at a designated position on the same chromosome and were treated as environmentally stable consensus QTLs. A QTL explaining more than 10% of the phenotypic variance was considered a major QTL.

Select candidate gene
The candidate interval of the major QTLs with overlapping physical regions in speci c environments were selected for subsequent analysis. Based on the positions of the anking bin markers, all of the genes within the con dence interval were identi ed as candidates, and the sequences of the candidate regions were subsequently submitted to the B. napus genome resource (http://www.genoscope.cns.fr/brassicanapus/), the BRAD (http://www.brassicadb.org/brad/) and the Arabidopsis Information Resource (TAIR) (https://www.arabidopsis.org/) databases for BLAST analysis.
The candidate genes of the target loci regions were predicted by combining the gene annotation of the BnaA02g12130D reference genome and the homologous gene annotation related to the owering time of A. thaliana.

Availability of data and materials
All the data generated and analysed during this study are included within the article and its additional les.