Fine mapping of the major QTLs for biochemical variation of sulforaphane in broccoli florets using a DH population

Glucoraphanin is a major secondary metabolite found in Brassicaceae vegetables, especially broccoli, and its degradation product sulforaphane plays an essential role in anticancer. The fine mapping of sulforaphane metabolism quantitative trait loci (QTLs) in broccoli florets is necessary for future marker-assisted selection strategies. In this study, we utilized a doubled haploid population consisting of 176 lines derived from two inbred lines (86,101 and 90,196) with significant differences in sulforaphane content, coupled with extensive genotypic and phenotypic data from two independent environments. A linkage map consisting of 438 simple sequence repeats markers was constructed, covering a length of 1168.26 cM. A total of 18 QTLs for sulforaphane metabolism in broccoli florets were detected, 10 were detected in 2017, and the other 8 were detected in 2018. The LOD values of all QTLs ranged from 3.06 to 14.47, explaining 1.74–7.03% of the biochemical variation between two years. Finally, 6 QTLs (qSF-C3-1, qSF-C3-2, qSF-C3-3, qSF-C3-5, qSF-C3-6 and qSF-C7) were stably detected in more than one environment, each accounting for 4.54–7.03% of the phenotypic variation explained (PVE) and a total of 30.88–34.86% of PVE. Our study provides new insights into sulforaphane metabolism in broccoli florets and marker-assisted selection breeding in Brassica oleracea crops.

www.nature.com/scientificreports/ tion was suitable for constructing a permanent F 1 DH population including 176 lines for the mapping of QTLs for SF metabolism in broccoli (Fig. 1).The DH family showed differences in the distribution in florets depending on genotype, and the coefficients of variation ranged from 0 to 0.20 and 0 to 0.14 in 2017 and 2018, respectively ( Fig. 1B,C). In our previous studies, analysis of variance (ANOVA) revealed significant differences (p < 0.01) in SF contents among DH plants in both 2017 (2.03 to 183.27 mg/kg FW) and 2018 (2.17 to 187.97 mg/kg FW), which indicated the existence of heritable variation. And meanwile, there existed segregation distortion and over-parent heterosis in the DH family, thus, the suitability of the DH population for genetic analysis. Significant variance of SF in the DH population was observed at p < 0.01 level. This result demonstrated that biochemical variation of SF contents was mainly under control of genetic factors and that climate might also play a role. Additionally, the Pearson correlation test was applied to the DH family, with a correlation coefficient ranging from 0.93 to 0.98 (p < 0.01). The data showed that there was a significant correlation between 2017 and 2018. The ranges and coefficients of variation (CV/%) among the DH population were greater than those in the P 1 (1.20-1.66%), P 2 (1.57-3.51%), and F 1 (0.78-2.22%) populations, suggesting the existence of real variations in heredity and more genetic polymorphisms in the DH populations, which laid a good foundation for the subsequent genetic analysis.
The frequency distribution of SF contents in the DH population showed a continuous distribution and was difficult to group, suggesting that the SF content trait in broccoli is quantitative.

Discussion
A variety of genetic and environmental factors ultimately affect the metabolite levels of glucosinolate in Brassica crops, such as Brassica napus 35,36 , Brassica rapa 37 , and Brassica juncea 38,39 . It has been reported that this conclusion is also applicable the GRA and SF in broccoli. Moreover, different organs of broccoli, including the seeds, seedlings, sprouts, leaves, and stalks, present quite different SF contents, which are mostly determined by the genotype and its interactions with the environment 22,[40][41][42][43][44] . Similarly, in the florets and leaves that we analyzed, it was shown that there were significant variations in SF accumulation in different genotypes and organs at different developmental stages (florets at mature, buds to flowers at bolting) 3,22 . In addition, it has been proved that glucoraphanin content can be regulated and affected by several genes, such as BCAT4 (branched-chain aminotransferase 4), MAM1 (methylthioalkylmalate synthase 1), CYP79F1 (dihomomethionine N-hydroxylase), AOP2 (2-oxoglutarate-dependent dioxygenases) [45][46][47][48] . In our study, genetic analysis of SF contents in broccoli florets was firstly estimated by a DH population in both years, and the result revealed that there might be at least three major genes controlling the biochemical variation of SF contents, at the same time, the environment was also an influence factor. So our result provided a direct evidence in SF or GRA metabolism in crucifer plants, which was consistent with most previous reports. Therefore, according to previous reports, broccoli shows considerable differences in SF contents in different organs and developmental stages, suggesting that SF metabolism is regulated by different genes than glucosinolate biosynthesis. To date, most QTL mapping studies of glucosinolate have focused on Arabidopsis, Brassica napus and Brassica rapa crops, and few reports have provided QTL information on SF in broccoli florets based on a permanent DH population derived from broccoli varieties to study the underlying regulatory mechanism. Therefore, the mapping of QTLs responsible for the differences in SF metabolism in different environments is helpful to better understand the relationships among the environment, MY activity and SF content in broccoli 18 . SF plays an important role in anticancer effects and the prevention of cerebrovascular disease. Most people obtain nutrition from broccoli by consuming the florets or their extracts, so this study focused on the investigation of QTLs for SF metabolism in broccoli florets by using a permanent DH population including 176 individuals. In our study, significant QTLs for SF metabolism in broccoli florets were mapped to chromosomes 3 and 7. In previous studies, QTLs for total GLS, aliphatic GLS, GRA, progoitrin (PRO), gluconapin (NAP), glucoerucin (GER), glucobrassicin (GBS), and 4-hydroxyglucobrassicin (OHGBS) were found on chromosomes 3 and 7 ( Fig. 4) 30,49,50 . GRA is the precursor of SF, whose production is catalyzed by MY, and it belongs to the aliphatic GLS; therefore, we emphasized the mapping of GRA, aliphatic GLS and MY. In previous reports, we found that some QTLs for aliphatic GLS and GRA were located on chromosomes 1, 2, 3, 4, 7 and 9; QTLs for both aliphatic GSL and GRA were located on chromosome 7; and QTLs for GRA alone were located on chromosomes 1, 7 and 9 (Fig. 4). In our study, 18 QTLs for SF were detected on chromosomes 2, 3, 5 and 7, and 6 significant QTLs were mapped to chromosomes 3 and 7, which indicated that some QTLs identified in this study were consistent with those identified in previous reports, but our results showed more QTLs on chromosome 3, suggesting shorter genetic distance and that more detailed information will need to be obtained in future research. Therefore, to a large extent, the important QTLs for SF metabolism in broccoli florets might be located on chromosome 7. In fact, SF metabolism is determined by polygenic regulation, and major genes and microgenes both play important roles in different organs, developmental stages and environments 13,17,20,26,29,30,41,43 .
In the two years, 6 common QTLs (qSF-C3-1, qSF-C3-2, qSF-C3-3, qSF-C3-5, qSF-C3-6 and qSF-C7) were stably detected with the same flanking markers. Considering to the similar contorl of two envirnmonts, except for the slight changes in temperature, the results might provide a reliable experiment basis for studying molecular mechanism of SF regulation. At present, the research on QTL mapping for SF metabolism in broccoli is limited and is not sufficiently deep. On the basis of this study and several previous reports, we can infer that the qSF-C3-1, qSF-C3-2, qSF-C3-5, qSF-C3-6 and qSF-C7 QTLs play an important role in SF accumulation as upstream regulated genes with positive effects. The qSF-C3-3 QTL might be a negatively regulated gene in the SF synthesis pathway and could be an ESP-related gene or a secondary product-regulated gene for substrates competing with GRA. In addition to 6 common QTLs, 5 other QTLs (qSF-C2, qSF-C3-0, qSF-C3-4, qSF-C5-1 and qSF-C5-2) were detected on chromosomes 2, 3 and 5. In previous reports, QTLs for aliphatic and total GLS have been found on chromosomes 2 and 3. In the present study, we also found QTLs (qSF-C5-1 and qSF-C5-2) on chromosome 5 that have not been reported previously.
It has been reported that the AOP family plays an important role in the side chain modification of GLS. The function of the AOP2 gene is absent in broccoli, and AOP3 is associated with the apparent regulatory control of aliphatic GLS accumulation by catalyzing the production of hydroxyalkyl glucosinolate from methylsulfinylalkyl glucosinolate with C3 side chains, but the specific roles of AOP1 and AOP3 in controlling aliphatic GLS accumulation are less well known 26,[51][52][53][54] . In our study, several QTLs (qSF-C3-0, qSF-C3-4, qSF-C5-1 and qSF-C5-2) detected as special regions in broccoli might be related to the AOP family, which requires further research. www.nature.com/scientificreports/ We detected some major and special QTLs for SF metabolism in broccoli florets based on a DH population in various environments. It is believed that these QTLs can be used for marker-assisted selection breeding and fine mapping.

Methods
Plant materials. A DH of broccoli was developed from F 1 plants resulting from the cross of parents from inbred lines 86,101 and 90,196, and there were 176 plants (genotypes) in this DH family generated from F 1 cultivated by a pollination method as our previously described 55 . Actually, DH is normally used to retain the desired alleles in the genome and a quicker way to produce homogenous line, instead of self-pollination over generations to produce inbred lines. In this study, all plant materials were bred and planted at the same farm of the Institute of Vegetables and Flowers (39°90′N, 116°29′E), Chinese Academy of Agricultural Sciences (Beijing) (IVF-CAAS). These 176 DH lines, their parents (individual 30 plants), and the F 1 (30 plants) were all grown in autumn 2017 (environment 1) and 2018 (environment 2) in Beijing (IVF-CAAS), separately. All plants were sown on July 6-8 in 2017 and 2018, and were planted in the field after one month. The two environments included 266 plants with three repeats (n = 3, total 798 plants) at random, and the plants were spaced 30 cm × 50 cm apart with 15 plants in each row. For the DH population, the experimental plots were surrounded by two additional rows planted to serve as a protective buffer. There were similar control and management for environment 1 and environment 2, the difference was that the monthly average temperature in september and october 2018 was 18.6 °C to 35 °C and 10.2 °C to 21.6 °C, which were a little higher than corresponding month in 2017 (17.2 to 32.3 °C and 8.1 °C to 19.2 °C).
Line 86,101 showed very early maturity (55 days after planting in the field) and exhibited some clovers in its small florets a yellow-green broccoli head color (Fig. 5). Inbred line 90,196 also showed early maturity (60 days after planting in the field) but exhibited no clovers in its middle florets, and its broccoli head color was green and changed to deep purple under freezing temperature. The DH family presented differences in the phenotypes of traits such as head color, shape, size, and the presence of clovers (Fig. 5).  www.nature.com/scientificreports/ Pretreatment and genetic analysis of SF. When the broccoli plants were mature, the florets were harvested, and the plant materials of the parents, F 1 hybrid, and each DH line were collected and cut into small pieces 5 cm in diameter. All the samples were immediately frozen in liquid nitrogen and stored at − 80 °C. Then, the frozen samples were dried in a lyophilizer (BETA 2-8 LD plus, Christ). The dried samples were powdered using an IKA-A10 (IKA-Werke GmbH & Co. KG) mill, and the fine powder was used for SF extraction and quantitative analysis by RP-HPLC according to methods described in our previous reports 3,22,56 . According to mixed major gene plus polygene inheritance analysis, genetic analysis of sulforaphane content in the DH population and parental lines was performed following our previous report 57 . The maximum likelihood method based on the iterated expectation conditional maximization (IECM) algorithm, was used for estimating the distribution parameters. The genetic analysis (parameters) were carried out by a least-squares method in the optimal model choosen by the Akaike information criterion (AIC).
Genotype analysis and QTL mapping. Genomic  www.nature.com/scientificreports/ oleracea expressed sequence tags (ESTs) were obtained from the National Center for Biotechnology Information database (978 pairs) 58 . For PCR, the reaction volume of 10 μL contained 5 μL of a 2X reaction mix, 0.5 μL of the forward primer, 0.5 μL of the reverse primer, 2 μL of genomic DNA template and 2 μL of ddH 2 O. The cycling conditions were as follows: 5 min 94 °C; 40 cycles of 30 s at 94 °C, 30 s at 55 °C and 1 min at 72 °C; and a final extension of 10 min at 72 °C. Thereafter, 6% denaturing polyacrylamide gel electrophoresis (PAGE) was used to separate the PCR products. QTL analysis was carried out via inclusive composite interval mapping (ICIM-ADD) with QTL IciMapping version 4.2 software (http:// www. isbre eding. net). The critical LOD score for a significant QTL was set at 3.0, and the walking speed for the genome-wide scan was set at 1 cM, through which both the additive and dominant effects of a QTL can be estimated 34,59 . The LOD threshold for each significant QTL was calculated via 1000 permutations at p < 0.05. Conditional QTL analysis between 86,101 and 90,196 in the DH population was conducted using the software QGAstation2.0 based on a mixed model for the complex quantitative traits 60 .
Statistical analysis. The calculation of descriptive statistics, frequency distributions and one-way ANOVA was performed using SPSS 19.0 software (http:// www. spss. com). Additionally, Microsoft Office Excel 2010 software was used for data entry and simple analysis.
Human and animal rights. This article does not contain any studies with human participants or animals performed by any of the authors.

Data availability
Data supporting the current study can be obtained by contacting the corresponding author (lizhansheng@caas. cn). www.nature.com/scientificreports/