Localization of quantitative trait loci for cucumber fruit shape by a population of chromosome segment substitution lines

Cucumber fruit shape, a significant agronomic trait, is controlled by quantitative trait loci (QTLs). Feasibility of chromosome segment substitution lines (CSSLs) is well demonstrated to map QTLs, especially the minor-effect ones. To detect and identify QTLs with CSSLs can provide new insights into the underlying mechanisms regarding cucumber fruit shape. In the present study, 71 CSSLs were built from a population of backcross progeny (BC4F2) by using RNS7 (a round-fruit cucumber) as the recurrent parent and CNS21 (a long-stick-fruit cucumber) as the donor parent in order to globally detect QTLs for cucumber fruit shape. With the aid of 114 InDel markers covering the whole cucumber genome, 21 QTLs were detected for fruit shape-related traits including ovary length, ovary diameter, ovary shape index, immature fruit length, immature fruit diameter, immature fruit shape index, mature fruit length, mature fruit diameter and mature fruit shape index, and 4 QTLs for other traits including fruit ground and flesh color, and seed size were detected as well. Together our results provide important resources for the subsequent theoretical and applied researches on cucumber fruit shape and other traits.

; Supplementary Table S4) based on the following criteria: (I) the most of genomes displayed high-level homozygosity with RNS7, except one to two substitutions from CNS21; (II) the selected individuals harbored less CNS21derived chromosomal segments, which were able to cover the whole genome of CNS21 with overlapping regions between different ones. To obtain the desired CSSLs, 60 BC 4 F 1 lines were self-pollinated and the resulted 1980 BC 4 F 2 plants were further investigated by marker-assisted selection (MAS) based on cucumber 9930 V3.0 draft genome. A total of 71 independent BC 4 F 2 substitution lines were kept as the cucumber CSSL population (Fig. 2).   Table S1). Among these 71 CSSLs, 66 lines harbored only one CNS21-derived chromosomal segment, 5 harbored two segments, and none harbored three or more segments (Fig. 3). Numbers of the substituted segments were 20, 12, 12, 6, 10, 7 and 9 in chromosome 1 to chromosome 7, respectively (Supplementary Table S1).  Table S1). Among these CNS21-derived substitutions, 24 segments were smaller than 5 Mb, 37 were 5-10 Mb, and 15 were over 10 Mb (Fig. 4). The recovery ratio of the 71 CSSLs ranged from 95.63 to 99.03% (Fig. 5).

Identification of QTLs for fruit shape.
To identify the chromosomal segments involved in fruit shape, phenotypic variations of fruit shape related parameters including fruit length, fruit diameter and the ratio of    Table S3). The fruit shape was markedly different between the two parents, and the L/D index of CNS21 was consistently greater than that of RNS7 19 . In the CSSL population, the fruit length and fruit diameter segregated significantly at immature fruit stage, ranging from 44 to 111 mm and 36.5 to 80 mm, respectively (Table 1; Supplementary Table S3; Fig. 7). The similar results were found at anthesis and mature fruit stages as well (Table 1; Supplementary Table S3). Of these CSSLs, fruit shape of 10 lines were dramatically different from RNS7 (Table 1). By analyzing the CSSL population, we detected 21 fruit shape related QTLs on chromosomes 1, 2, 3, 5 and 6, which included 2 responsible for ovary length (OL), 2 for ovary diameter (OD), 2 for ovary shape index (OSI), 2 for commercial fruit length (FL), 1 for commercial fruit diameter (FD), 4 for commercial fruit shape index (FSI), 2 for mature fruit length (MFL), 3 for mature fruit diameter (MFD) and 3 for mature fruit shape index (FSI) (Fig. 6 Identification of QTLs for seed shape and fruit color. In addition to fruit shape, the two parents showed significant differences in other fruit traits such as commercial fruit ground color (FGC), fruit flesh color (FLC) and seed shape 19 . Using CSSL2-7, two QTLs for seed length (SDL) and seed width (SW) were identified in the region of 5.  Table S2).

Discussion
Fruit shape/size, an important quality trait in cucumber, is often affected by both genetic composition and environmental conditions. To date, there is little information available on the genetic mechanisms of fruit shape/ size. CSSLs are ideal materials to detect QTLs and evaluate their contributions to the trait of interest as a single Mendelian factor. CSSLs have extensively been applied for the identification of genes that control important agronomic traits in rice 46,47 , maize 28,29 , Brassica rapa 22,36 , tomato 40,41 , and so on. However, thus far, only three sets of cucumber CSSLs have already been constructed. The first set of CSSLs was created through a cross of the wild cucumber PI183967 (donor) and the cultivated line Xintaimici (receptor), providing new resources for utilization of valuable genes from wild cucumber 48 . The other set was adopted to detect powdery mildew (PM) resistance-related genes 49 . The third set is in the present study (Fig. 2). Polymorphic marker density across whole genome profoundly influences the quality of CSSLs and thus plays crucial roles in the creation of CSSLs 50 . The CSSLs constructed by Li et al. 48 only contain 31 lines including 10 lines harboring two substitution segments, and  However, 114 InDel markers that were distributed on the 7 chromosomes relatively evenly were adopted for the construction of CSSLs in the current study ( Supplementary Fig. S1). Moreover, 66 of the 71 CSSLs contained single substituted segment and the other 5 lines were identified to contain two substituted segments (Fig. 3). The length of substituted chromosomal segments in each line ranged from 1.73 to 19.31 Mb, and the average value of these segments was approximately 7.19 Mb (Supplementary Table S1). So these lines harbored a high recovery rate of 95.63-99.03% of the recurrent parent genome and simultaneously the genetic background noise was tremendously decreased (Fig. 5), thus being considered as a powerful tool to identify, map and validate QTLs of interest.
Using the CSSL population, totally, 21 QTLs responsible for cucumber fruit shape were identified and of which, eight QTLs were detected in the region of 22.73-28.27 Mb on chromosome 1 ( Fig. 6; Table 1), where numerous QTLs for OL, OD, OSI, FL, FSI, MFL, MFD and MFSI were detected in previous studies ( Fig. 6; Supplementary Fig. S4) 19 . As the best candidate of FS1.2, CsSUN was located in this region, being a major QTL of fruit shape (Fig. 7a) 14,19 . Comparing with previously reported regions on chromosome 1 19,51 , the size of estimated QTL region in our research was much smaller (Supplementary Fig. S4). The QTLs for OL, OD, OSI, FL, FSI, MFL, MFD and MFSI were identified at the long arm of chromosome 2 in the present study ( Fig. 6; Supplementary  Fig. S4). The chromosome 2 region harboring these detected QTLs displayed overlapping, but much smaller than the previous three reports by Weng et al. 2 , Gao et al. 19 and Pan et al. 51 . This QTL(s) on chromosome 2 was (were) uncovered as (a) major one(s) by Pan et al. 14 , and our data provided a direct evidence for FS2.1 being a major QTL ( Fig. 7b; Table 1). Wu et al. 15 reported that there were 10 possible candidate genes in FS2.1 locus such as CsTRM5, an ortholog of tomato TRM5 that was able to balance the OVATE and SlOFP20-mediated cell division patterns to determine the final tomato shapes. It was thus that CsTRM5 was regarded as the most possible candidate gene for FS2.1, but they did not give direct genetic evidence of sequence difference or gene expression between the two parental lines used in their research 15 . Our resequencing results revealed that some single nucleotide polymorphisms (SNPs) or InDels in CsSUN and CsTRM5 genes, which were just located on the region of 22.73-28.27 Mb on chromosome 1 and the region of 5.10-14.23 Mb on chromosome 2 respectively, between RNS7 and CNS21 (data not shown). We thus speculate that CsSUN and CsTRM5 genes could be the candidate genes for QTLs on chromosomes 1 and 2, respectively. However, more experimental evidence should be provided in future study to support this assumption. In addition, SF1 was localized in the region of 5.10-14.23 Mb on chromosome 2 in which FS2.1 was mapped in our study and previous studies ( Supplementary Fig. S4) 17,19 . However, there is no difference in protein sequence and gene expression of SF1 between RNS7 and CNS21. The identification of major QTLs (R 2 > 10%) on shortened regions of chromosomes 1 and 2 in the present study (Table 1; Supplementary Fig.  S4) indicated that CSSLs could be an advantageous tool for fine mapping stable QTLs and give more information about these QTLs under different cucumber genetic backgrounds. www.nature.com/scientificreports/ www.nature.com/scientificreports/ Furthermore, the CSSLs harboring a single segment substitution make it feasible to mine minor-effect QTLs 21,25,[52][53][54] . In the current study, five minor-effect QTLs (R 2 ≤ 10%) associated with fruit shape were detected on chromosomes 3, 5 and 6 (Figs. 6, 7c-f; Table 1) and displayed a relatively complex relationship with previous studies 2, 12,13,19,51,55 . In most cases, the identified minor-effects regions on the three chromosomes were well consistent with those described previously 2,12,13,51,55 , while the inconsistency was also revealed for FD3.2 and FSI3.3 on chromosome 3 with the detected effect regions on the same chromosome by Wei et al. 55 and Pan et al. 13 , possibly due to the differences in genetic background, traits of interest or environmental conditions ( Fig. 6; Supplementary  Fig. S4). Up to date, none of them has yet been fine mapped and cloned because that it is scarcely possible to fine map or clone these minor QTLs using the F 2 , F 3 , BC or RIL populations. However, the CSSLs that we constructed in this present study provided an opportunity for isolating the minor QTLs related to cucumber fruit shape.

Line Chr Position (Mb) a Substituted region Trait (mm) % variation (R 2 ) Add
We also detected QTLs that were associated with seed size and fruit color in the present study (Supplementary Table S2). The QTLs for SDL and SW were identified in the same region for OL, OD, OSI, FL, FSI, MFL, MFD and MFSI on chromosome 2, suggesting that FS2.1 might have pleiotropic effects (Supplementary Fig.  S2; Supplementary Fig. S4; Table 1; Supplementary Table S2). More recently, two consensus QTLs (CsSS2.1 and CsSS2.2) associated with seed size have been reported on chromosome 2 in a review paper by Guo et al. 56 , and CsSS2.1 displays overlapping, but larger than the identified QTLs for SDL and SW in the present study. The smaller QTL regions in this study will facilitate the future fine-mapping for genes responsible for seed size. The QTLs for FGC and FLC were observed in the distal region of chromosome 3 ( Supplementary Fig. S3; Supplementary Table S2), being consistent with the previous results reported by Liu et al. 57,58 and Tang et al. 59 . The w gene controlling white immature fruit color was localized in this region of chromosome 3, but no difference in coding sequence (CDS) of w was observed between RNS7 and CNS21. It will be very intriguing to reveal more candidate genes responsible for fruit color in future studies. In addition, QTLs related to other agronomic traits could be identified with these CSSLs.
In summary, we created a set of CSSLs that resulted from a cross between RNS7 (a round-fruit line) and CNS21 (a long-stick-fruit line) using 114 InDel markers covering the whole cucumber genome (9930 V3.0). Using these CSSLs, we identified 25 QTLs related to fruit shape, fruit color and seed size. Our study provides a powerful tool to isolate the QTLs for fruit shape, especially the minor ones, and other agronomic trait QTLs.

Materials and methods
Plant materials and growth conditions. Two parents CNS21 and RNS7, were used to construct the CSSL population 19 . Seeds of two parents were germinated in darkness at 28 °C overnight in petri dishes and grown in a growth chamber that was programed as photoperiod of 16 h, air temperature of 25 °C over light course and of 18 °C over dark course. Cucumber seedlings were transferred to a greenhouse of Shandong Agricultural University when they were grown to two-leaf stage. Standard field managements were carried out over cucumber cultivation course.

Molecular marker development.
A total of 114 InDel markers that were distributed evenly throughout the cucumber (Chinese Long) 9930 V3.0 genome (https ://cucur bitge nomic s.org/organ ism/20) were developed from the data of sequenced genomes (Supplementary Fig. S1) 19 . And 18,16,19,14,19,17 and 11 InDel markers were located on chromosome 1 to chromosome 7, respectively ( Supplementary Fig. S1). The average distance was approximately 1.85 Mb between two neighboring markers on the same chromosome. The primers used in the present study were listed in Supplementary Table S4. www.nature.com/scientificreports/ Construction of CSSLs. The schematic illustration for construction of CSSLs was displayed in Fig. 1 DNA extraction and genotype analysis. Genomic DNAs were extracted from unexpanded young leaves of each plant following the CATB protocol reported by Murray and Thompson 60 . Then the above-mentioned 114 InDel markers were applied to detect the individuals over foreground and background selections. The target DNA segments were amplified on a ABI PCR machine (Thermo Fisher Scientific, USA) with the correspond InDel markers. The resulted products were separated on a 3.5% (W/V) agarose gel and photographed with a FR-980A image analysis system (Shanghai Furi Science and Technology, China).
Phenotypic analysis. Phenotypic data of CSSLs and RNS7 were recorded in the solar greenhouse of Shandong Agricultural University over three years (2016, 2017 and 2018). Two self-pollinated fruits were allowed on each plant. Fruit length (L), fruit diameter (D) and the ratio of length to diameter (L/D) were determined at three developmental stages: ovary length (OL), ovary diameter (OD) and ovary shape index (OSI) at anthesis; commercial fruit length (FL), fruit diameter (FD) and fruit shape index (FSI) at 10-12 days post pollination (dpp), and mature fruit length (MFL), mature fruit diameter (MFD) and mature fruit shape index (MFSI) at 45-55 dpp. At least 5 biological repeats were performed to collect all data. For each repeat, five to ten typical fruits at anthesis, three to five typical fruits at immature fruit stage, and two typical fruits at mature fruit stage, respectively, were selected for statistical analysis of phenotypic parameters including fruit length (L), fruit diameter (D) and the ratio of length to diameter. Seed length (SDL) and seed width (SW) were collected from at least 20 seeds. Data analyses were performed with statistical algorisms installed in MICROSOFT Excel 2013.
QTL mapping. Given there were significant differences in the average value of a trait between a CSSL and RNS7, the existence of QTLs was further estimated. The detection of QTLs was performed on the basis of the t-test results that were derived from the difference comparison between the mean values of each CSSL and RNS7 (P value ≤ 0.05). The additive effect of individual QTL was evaluated by following the formula below 53 : Additive effect = 1/2 × (value of CSSL-value of RNS7). The observed phenotypic variance (R 2 ), a parameter commonly adopted to evaluate the effect strength of a given QTL, was calculated for these detected QTLs by using the QTL IciMapping V4.1 software with previously introduced settings 19 . The QTLs with over 10% of R 2 were defined as major-effect ones and the others were defined as minor-effect ones according to the previous study 19 .