Introduction

Cucumber, Cucumis sativus L. (2n = 2x = 14), is an economically important specialty crop and a system of choice for studying several important biological processes (Weng and Sun 2012). Cucumber is native to the Southern Asia continent (Candolle 1959; Sebastian et al. 2010). Classical taxonomic investigations have identified several botanical varieties of C. sativus which include the cultivated cucumber C. sativus var. sativus, the wild cucumber C. sativus var. hardwickii (Royle) Alef. (Royle 1835; Duthie 1903), the Sikkim cucumber, C. sativus var. sikkimensis Hook. f. (Hooker 1876; Renner and Pandey 2013), and the Xishuangbanna cucumber, C. sativus var. xishuangbannanesis Qi et Yuan (Qi 1983).

The Xishuangbanna (XIS hereinafter) cucumber, first described in 1983 (Qi 1983), is distributed in the mountainous Xishuangbanna region of Yunnan Province in southwest China (21°09′ to 22°36′N altitude, and 99°58′ to 101°50′E longitude, elevation 800–1,200 m) bordering Myanmar. The region has a typical tropical monsoon climate. The XIS cucumber is grown by local ethnic groups and is often intercropped with upland rice without trellis support. In general, XIS cucumber plants grow more vigorously than commercially cultivated cucumbers; the main vine may reach 8 m with 20–40 lateral branches and more than 900 nodes (Chen et al. 1994). The growth period may last up to 6 months, and the plant requires short day length for flowering (Bo et al. 2011). This short day length requirement has been lost in commercial cucumbers. On average, eight to ten fruits are set on each plant, and a single mature fruit weighs 2–3 kg (with the maximum of 5 kg) and contains over 1,000 seeds (Qi 1983). The mature fruits vary in shape [long, oblong, oval, or round with length to diameter ratio (L/D) from 1 to 4] and rind color (white, creamy, yellow, or brown) (Qi 1983; Shen 2009). Probably the most unique trait of the XIS cucumber is its orange flesh color (Qi 1983) which is due to the accumulation of high level of β-carotene in mature fruits (Bo et al. 2012). Unlike wild cucumbers but similar to the cultivated cucumber, the fruit of XIS cucumber is bitter-free and can be consumed immature or mature. The mature fruits are also stored as an important off-season vegetable. Because of these characteristic traits, the XIS cucumber was classified as a semi-wild variant of C. sativus (Qi 1983).

Several studies have investigated the relationship of XIS cucumber with other botanical variants of cucumber. The XIS cucumber is readily intercrossed with and shares the same 14 chromosomes as cultivated cucumber (Qi 1983). There are seven ring bivalents from chromosome pairing during meiosis, although occasionally a chromosome bridge was observed in meiotic anaphase I of XIS cucumber (Qian et al. 2003). The distribution of repetitive DNA sequences revealed by florescence in situ hybridization suggested that XIS is phylogenetically closer to the cultivated cucumber than to wild cucumber (Zhao et al. 2011). Molecular marker-based clustering analysis placed XIS cucumber in the same clade as cultivated cucumber (Zhuang et al. 2004). Further phylogenetic analysis (Lv et al. 2012; Qi et al. 2013) on worldwide cucumber collections found that the XIS cucumber is affiliated with the Indian population which has the highest genetic diversity as compared to the East Asian and Eurasian cucumber groups. In addition, the XIS accessions were genetically close to each other suggesting that a single dispersal occurred from India in the evolution of the XIS cucumber (Lv et al. 2012). While these studies have provided good information on the population structure of the XIS cucumber, the degree of differentiation between the XIS cucumber and other cucumber variants and the crop evolution history of the XIS cucumber are largely unknown. Therefore, one objective of the present study was to conduct comparative genetic mapping between the XIS and cultivated cucumbers to reveal possible chromosomal differentiations between the two botanical varieties. In cucumber, this approach has been employed to reveal structural rearrangements between cultivated and wild cucumbers (Ren et al. 2009; Yang et al. 2012), and study the syntenic relationships between cucumber and melon (C. melo L.) (Li et al. 2011).

From the cucumber breeding perspective, several traits possessed by the XIS cucumber are attractive for improving commercial cucumbers of different market classes, which may include short hypocotyl, tolerance to low light, large fruit, and high β-carotene content (orange flesh color) (Bo et al. 2012). However, due to its semi-wild nature, some traits in XIS cucumber such as photoperiod sensitivity, low percentage of female flowers, and long growth period may be potential obstacles for efficient use of this germplasm source. Understanding the genetic basis of these traits will provide insights into crop evolution and domestication processes leading to the XIS cucumber, thus facilitating its efficient use in cucumber breeding. Since most domestication or diversifying selection-related traits are quantitative in nature, the QTL (quantitative trait loci) mapping strategy has been extensively used for genetic dissecting of such traits in major crop plants such as rice, maize, sorghum, wheat, millet, and sunflower (reviewed by Doebley et al. 2006; Alonso-Blanco et al. 2009; Gross and Olsen 2010; Meyer and Purugganan 2013; Olsen and Wendel 2013; Abbo et al. 2014).

In cucumber, due to application of next-generation sequencing (NGS) technologies, draft genome assemblies for two cultivated cucumber inbred lines (9930 and Gy14) have been released and many other lines have been re-sequenced (Huang et al. 2009; Yang et al. 2012; Qi et al. 2013). Hundreds of simple sequence repeat (SSR) markers have been genetically mapped (Ren et al. 2009; Cavagnaro et al. 2010; Yang et al. 2013). These genetic and genomics resources have greatly facilitated genetic mapping, molecular tagging, and gene cloning in cucumber. However, only a limited number of QTLs have been reported in this crop for disease resistance (Sakata et al. 2006; Liu et al. 2008; Zhang et al. 2011, 2013; Fukino et al. 2013; He et al. 2013), flowering time, fruit size, fruit bitterness, or fruit epidermal features (Kennard and Havey 1995; Dijkhuizen and Staub 2002; Fazio et al. 2003; Yuan et al. 2008; Cheng et al. 2010; Miao et al. 2011, 2012; Qi et al. 2013). These studies utilized cultivated or wild cucumbers, but not XIS semi-wild cucumber. Therefore, the second objective of the present study was to conduct QTL mapping of domestication, or diversifying selection-related traits in XIS cucumber. We first developed an SSR-based linkage map using 124 recombinant inbred lines (RILs) derived from the cross between the semi-wild XIS cucumber and a cultivated cucumber. The resulting linkage map was compared with published cucumber genetic maps to reveal possible chromosome differentiations among the XIS, cultivated, and wild cucumbers. We next conducted multi-year and multi-location phenotyping for flowering time, fruit length, fruit diameter, and fruit weight in this RIL population and identified QTLs underlying these traits.

Materials and methods

Plant materials

For linkage map development and QTL mapping, 124 F9 RILs were developed from the cross between two cucumber inbred lines CC3 and SWCC8 through single seed descent. CC3 (maternal parent, P1) was derived from a north China fresh market type landrace ‘Beijing Jietou’ that flowers early and bears long, slim fruits (~65 cm). SWCC8 (paternal parent, P2) is a semi-wild XIS cucumber originating from southwest China which requires short day length for female flower development and bears short (~28 cm) but blocky fruits (Fig. 1).

Fig. 1
figure 1

Fruit size (length and diameter) variations of CC3, SWCC8, and their F1 (a, taken from greenhouse studies) and F9 recombinant inbred lines (b and c, taken from WI2012H experiment). b, c Fruit length and diameter distributions

Full size image

An F2 population from the CC3 × SWCC8 cross was also used to validate QTLs identified with the RIL population.

Phenotypic data collection and analysis

Phenotypic data were collected from 124 RILs plus two parental lines and their F1 in four field trials across 3 years at two locations. Details of the four experiments, NJ2009F, NJ2012S, WI2012H, and WI2013H, are presented in Table 1. Briefly, NJ2009F and NJ2012S were conducted in plastic houses at the Jiangpu Experiment Farm of Nanjing Agricultural University in Nanjing, China (32°03′N and 118°47′E, 30/18 °C day/night temperature, 14/10 h day/night photoperiod in May and June; 30/17day/night temperature, 11/13 h day/dark photoperiod in September and October). On each plant, only one self-pollinated fruit was allowed to develop, and fruit data were collected approximately 40 days after pollination. The experiments WI2012H and WI2013H were performed in an open field at the University of Wisconsin Experiment Station in Hancock, Wisconsin, USA (44°08′N, 89°31′W; 15/9 h day/dark photoperiod in August and 28/15 °C day/night temperature in July that is the warmest month of the year). The WI2012H experiment also included 540 CC3 × SWCC8 F2 plants of which 436 were able to grow to maturity and bear fruits on which data were collected. Individual RIL or F2 plants were spaced 40 cm apart in rows placed 80 cm apart. A honey bee colony was placed near the experimental field and fruits (3–10 per plant) were produced from open pollination. Fruit data were collected from ten or more mature fruits from each RIL, or two to five fruits from each F2 plant. In all experiments, no female flowers were observed for SWCC8 at the time of data collection. For comparison purpose, fruit data of SWCC8 used in the present study were obtained from historical data.

Table 1 Details of four environments used in QTL mapping for domestication or diversifying selection-related traits in the semi-wild Xishuangbanna cucumber
Full size table

In each experiment, data were collected for the flowering date of the first female flower (FFT), the length (FL), diameter (FD), and weight (FW) of mature fruits from each plant. For each trait, family means were calculated from four plants in NJ2009F, three plants per replication in NJ2012S, and five plants per replication in WI2012H and WI2013H (no replication) (total 15 plants per RIL). For FFT, CC3 was the earliest among the two parents, F1 and 124 RILs. FFT was calculated as days after the flowering date of CC3 which was set as 1. Statistical analysis of phenotypic data was performed using SAS v9.3 (SAS Institute Inc., Cary, NC, USA). Pearson’s correlation coefficients among FFT, FL, FD, and FW data were estimated with the PROC CORR function based on grand means of each RIL across all four experiments, or measurement of individual F2 plants.

Linkage map development and comparative analysis

Cucumber or melon SSR markers described in Ren et al. (2009), Cavagnaro et al. (2010), and Yang et al. (2012) were used for polymorphism screening between CC3 and SWCC8. Polymorphic markers were used to genotype 124 RILs. Selected markers were also used to genotype F2 plants to validate a major-effect QTL in chromosome 1 for fruit length. DNA extraction, PCR amplification of molecular markers, and gel electrophoreses followed Li et al. (2011). For each marker, χ 2 test for goodness of fit was performed against the expected 1:1 segregation ratio in the RIL population. Linkage analysis was carried out using JoinMap 4.0 software. Linkage groups were determined with a minimum LOD score of 4.0. Genetic distance was calculated with the Kosambi mapping function.

The physical locations of all mapped markers in the Gy14 scaffold and draft genome assemblies (Version 1.0, Yang et al. 2012) were used to verify their genetic map locations. Inference of chromosomal locations of molecular markers on the map was performed with BLASTn or in silico PCR according to Cavagnaro et al. (2010). Chromosome assignment (Chr1 to Chr7) of the seven linkage groups followed Yang et al. (2012).

To examine possible chromosome structural rearrangements between cultivated cucumber and the XIS semi-wild cucumber, the order of mapped loci on the CC3 × SWCC8 RIL map from this study was compared with that of previously developed SSR-based cucumber genetic maps including the inter-subspecific Gy14 × PI 193967 (C. s. var. hardwickii) RIL map (Ren et al. 2009, 995 SSR loci) and the intra-varietal Gy14 × 9930 F2 map (Yang et al. 2012, 781 SSR loci). The cultivated cucumber consensus map (1,681 loci) (Yang et al. 2013) was also used to validate marker or scaffold orders in the regions of interest. Alignment of each linkage group was based on shared markers and marker-associated scaffolds of the Gy14 draft genome assembly (Yang et al. 2012).

QTL analysis

A whole genome scan was performed to map the QTLs using the composite interval mapping (CIM) function of WinQTL Cartographer Version 2.5 (Zeng 1994; Wang et al. 2012) with the default settings (Model 6 with a walking speed of 1 cM, a window size of 10 cM, and the inclusion of 6 maximum background marker loci in a stepwise forward regression procedure). The significance of each QTL interval was tested by a likelihood-ratio statistic (LOD). The LOD threshold for declaring significant QTLs for each trait (P = 0.05) was determined using a permutation test with 1,000 repetitions. The QTL was named according to its chromosome location and trait name [female flowering time (FFT), fruit length (FL), fruit diameter (FD), fruit weight (FW)]. For example, fft1.1 and fl3.1 designated the first QTL for female flowering time and fruit length in cucumber chromosomes 1 and 3, respectively.

Results

Linkage map construction

We screened 1,344 cucumber or melon SSR primer pairs between CC3 and SWCC8 and identified 303 polymorphic ones (22.5 %), of which 269 were mapped with 124 RILs. The resulting genetic map is illustrated in Fig. 2, and the main statistics of the map are presented in supplemental Table S1 (online materials). A majority of the marker loci fitted the expected 1:1 segregation ratio, while 55 markers (20.4 %) (those with asterisks in Table S2) showed distorted segregation in χ 2 tests (P < 0.05). Nearly one-third (81/269) markers on this map were not mapped in previous studies, which may be useful for the cucurbit research community. Therefore, the detailed information (marker names, map locations, Gy14 cucumber scaffold and draft genome assembly locations, and primer sequences) of all mapped markers is provided in supplemental Table S2 (online materials).

Fig. 2
figure 2

The SSR-based CC3 × SWCC8 linkage map and chromosomal locations of QTLs for female flowering time (FFT, stripped), length (FL, dotted), diameter (FD, waved), and weight (FW, checked) of mature fruits from four experiments (NJ2009F, NJ2012S, WI2012H, and WI2013H) over 3 years. QTLs for FL, FD, and FW detected in Yuan et al. (2008) (solid filled black bars), Miao et al. (2011, 2012) (solid filled dark gray bars), Cheng et al. (2010), and Qi et al. (2013) (solid filled light gray bars) are also shown. Numbers to the left of each chromosome (Chr) are map length in centiMorgan (cM). For QTLs detected from the present study, vertical bars represent 2-LOD support interval of each QTL and black filled circles are QTL peak locations. The experiment detecting the QTL is listed above each bar; the name of consensus QTL detected across multiple years (environments) is below or alongside the vertical bar(s). LOD support intervals of QTLs in all other studies are not available; thus the lengths of vertical bars for these QTL do not represent the confidence intervals. QTL symbols from original publications were used. For all studies, bold-faced and underlined QTLs are major-effect QTLs that explained more than 10 % genotypic variations. Map locations of QTLs from other publications were inferred from in silico PCR or BLASTn using primer sequences of QTL-associated molecular markers and are approximations

Full size image

This genetic map covered 705.9 cM in seven linkage groups, which is similar to previously published cucumber linkage maps (for example, Yang et al. 2012; He et al. 2013). Considering Gy14 draft genome scaffolds (Yang et al. 2012) anchored by these markers, this map seemed to physically cover the majority of the cucumber genome (data not shown). The marker orders were also highly consistent with their physical locations in the Gy14 scaffolds (Table S2). The mean marker interval of this map was 2.6 cM with only one gap larger than 10 cM in chromosome 4 from 52.8 to 63.5 cM (Table S2). Therefore, this high-quality genetic map was suitable for subsequent QTL mapping.

Comparative analysis of linkage maps among cultivated, semi-wild, and wild cucumbers

To investigate possible chromosome structural rearrangements between the XIS cucumber and cultivated (C. s. var. sativus) or wild (C. s. var. hardwickii) cucumbers, we aligned the XIS RIL map developed herein with the Gy14 × 9930 F2 inter-varietal map (Yang et al. 2012, 2013) and the inter-subspecific Gy14 × PI 183967 RIL map (Ren et al. 2009). The order of shared loci or marker-associated Gy14 draft genome scaffolds between the XIS map and the other two maps were compared. Among the 269 markers placed on the XIS map, 62 were shared across all three maps, and 102 and 156 shared with the Gy14 × 9930 F2 map and the Gy14 × PI 183967 map, respectively. Chromosome by chromosome alignment of shared markers is presented in supplemental Fig. S1 (online materials). Putative structural rearrangements represented by Gy14 draft genome scaffolds or shared SSR markers between XIS and cultivated cucumber, as well as between XIS and the wild cucumber are illustrated in Fig. 3. While no substantial changes were found between the XIS and cultivated cucumber genomes, there were two putative inversions in chromosome 1 (scaffold00953 vs. scaffold01357 on the top, and scaffold01063 vs. scaffold00598 in the distal end of the long arm) (Fig. 3, Fig. S1). In chromosome 6, there were three blocks in which the orders of scaffolds or molecular markers were inconsistent suggesting possible inversions between the two variants.

Fig. 3
figure 3

Putative structural rearrangements in chromosomes 1, 4, 5, 6, and 7 of the semi-wild XIS cucumber (C. sativus var. xishuangbannesis, center, dark gray) as compared with the wild (C. sativus var. hardwickii, left, light gray) and cultivated (C. sativus var. sativus, right, black colored) cucumbers. Structural changes were inferred from orders of Gy14 draft genome scaffolds (S, scaffold) or shared markers. Only scaffolds or SSR markers involved in the putative rearrangements are listed. The bins are clusters of mapped loci in the wild cucumber, and their map locations (in cM) are shown to the left of the chromosome (Ren et al. 2009). Dotted lines connect the same scaffolds or markers between two maps under comparison. Markers or scaffolds on the cultivated cucumber chromosomes 4, 5, and 7 are collinear with those in the respective XIS cucumber chromosomes and are not shown

Full size image

Significant non-collinearity of markers and scaffolds was identified in chromosomes 1, 4, 5, 6, and 7 between XIS cucumber and the wild cucumber (Fig. 3). The inversions in chromosomes 1 and 6 seem to be consistent with those between XIS and cultivated cucumber suggesting these rearrangements might be specific to the XIS cucumber. Inversions between the XIS cucumber and the wild cucumber in chromosomes 4, 5, and 7 were evident (Fig. 3). It is known that two, three, and one inversion differentiated the chromosomes 4, 5, and 7 of cultivated and wild cucumbers (Yang et al. 2012). The locations of the rearrangements identified from the present study were consistent with those found between the cultivated and wild cucumbers. This suggests that, except for the putative XIS-specific small inversions in chromosomes 1 and 6, no substantial structural changes occurred during crop evolution of the XIS cucumber.

Phenotypic data analysis

In four environments across 3 years, we recorded the flowering dates for the first female flowers (FFT), the length (FL), diameter (FD), and weight (FW) of mature fruits of CC3, SWCC8, their F1, and 124 RILs. In WI2012H, 436 F2 plants were also included in the trial for data collection. The phenotypic means, standard derivation, and range of the four traits in all experiments are presented in Table 2. Since analysis of variance (ANOVA) indicated significant effects of genotypes and genotype × season interactions for all traits examined (data not shown), QTL mapping was conducted on the basis of RIL means of each experiment.

Table 2 Phenotypic means and range of first female flowering time (FFT), fruit length (FL), diameter (FD), and weight (FW) in two parental lines (CC3 and SWCC8), their F1, F2, and 124 recombinant inbred lines (RILs) in four experiments (NJ2009F, NJ2012S, WI2012H, WI2013H)
Full size table

In all experiments, the parental line CC3 flowered earlier than most RIL lines for both male and female flowers. The XIS cucumber SWCC8 (in all experiments) and some F2 plants (in WI2012H) did not have female flowers, which was due probably to the lack of adequate short day length in the experimental locations. The flowering time of the F1 was similar to CC3 (Table 2) suggesting early flowering of CC3 was dominant over late flowering in SWCC8. As compared with CC3, SWCC8 had shorter, more cylindrical, and heavier fruits, and the fruit length, diameter, and weight of F1 plants were in general close to the mid-parental value indicating the quantitative nature of these traits (Table 2; Fig. 1). In the F2 population, while the FD and FW means among the 436 F2 plants were almost the same as in F1, the mean fruit length (FL) of F2 plants was longer (40.1 cm) than the RIL mean in any experiment. In general, the range of variations among F2 plants for each trait was also wider than that of the RILs and both parents (Table 2).

The frequency distribution of RIL and F2 means in the four experiments is illustrated in Fig. 4 (for RILs) and supplemental Fig. S2 (for F2, online materials), and was largely normal except for FFT of WI2013H. The RIL means and frequency distributions among the three experiments, NJ2009F, NJ2012S, and WI2012H, were highly consistent. For the WI2013H field trial, the temperature was lower than normal in the early growing season, and there were several days with very low night temperatures (<15 °C) in June and July. A few RILs did not flower before conclusion of the experiment. We believe this was the reason for the longer FFT and smaller fruits (lower means of FL, FD and FW) in the WI2013H dataset (Table 2). Despite this, major QTLs were detected with the WI2013H data that were largely consistent with those identified with other data sets (see below).

Fig. 4
figure 4

Frequency distribution of first female flowering time (FFT), fruit length (FL), fruit diameter (FD), and fruit weight (FW) among 124 CC3 × SWCC8 recombinant inbred lines in four experiments (NJ2009F, NJ2012S, WI2012H, and WI 2013H). Arrows indicate corresponding values of CC3, SWCC8, and their F1 in each experiment based on means across all experiments (FFT data for SWCC8 were not available)

Full size image

We analyzed the correlation among the four traits using RIL means across the four environments, as well as the F2 data from WI2012H. The Pearson’s correlation coefficients (r) among FFT, FL, FD, and FW are listed in Table 3. From RIL data, insignificant correlation was found between FFT and FD or FW; FFT was negatively correlated with FL (r = −0.1667 at P = 0.05). However, with the F2 data, FFT was significantly and negatively correlated with FL, FD, and FW. In both RIL and F2 data, no correlation was found between FL and FD; in contrast, FW were highly, significantly, and positively correlated with FL and FD (P < 0.001), which seem to be consistent with the QTL locations underlying each trait (Fig. 2; Table 4).

Table 3 Pearson’s correlation coefficients among first female flowering time (FFT), fruit length (FL), diameter (FD), and single fruit weight (FW) in the CC3 × SWCC8 RIL and F2 populations
Full size table
Table 4 QTLs for female flowering time (FFT), fruit length (FL), diameter (FD), and weight (FW) detected with 124 RILs of CC3 × SWCC8 in four experiments (NJ2009F, NJ2012S, WI2012H, and WI2013H)
Full size table

QTL analysis

The RIL means of FFT, FL, FD, and FW from the four experiments were used in QTL analysis. For each trait, the LOD threshold to declare significance of QTL was determined with 1,000 permutation tests (P = 0.05) which ranged from 2.7 to 3.3. A global view of all QTLs detected across the seven chromosomes is provided in supplemental Fig. S3 (online materials). Details of each detected QTL including map location, LOD value, percentages of total phenotypic variances explained (R 2), additive effect, and 2-LOD support interval are provided in Table 4. Their chromosomal locations are visually illustrated in Fig. 2. Among the 39 QTLs identified for 16 traits across four environments, 11, 9, 12, and 7 were detected by NJ2009F, NJ2012S, WI2012H, and WI2013H, respectively. The WI2012H experiment detected the most QTLs which may reflect the most accurate and complete set of data collected among the four due to the large number of plants per RIL and the number of fruits per plant that could be used for data analysis.

QTL of first female flowering time (FFT)

Two QTLs were detected for FFT: one was located in chromosome 1 (fft1.1), and the other in chromosome 6 (fft6.1) (Table 4). The major-effect QTL fft1.1 was reproducibly identified in all four environments with highly consistent peak locations on the genetic map (at 96.5 cM, Fig. S3 A1). In NJ2009F, NJ2012S, and WI2012H trials, fft1.1 could explain >50 % total phenotypic variations. In WI2013H, this QTL had R 2 value of 25.8 %, which was due probably to the lower than normal temperature and slower growth of plants during the growing season. The minor QTL fft6.1 (at 42.7 cM on chromosome 6, R 2 ≈ 6 %) was detected in two seasons (NJ2009F and WI2012H). Both QTLs contributed to early flowering (negative additive effects).

QTL of fruit length (FL)

Five QTLs in five chromosomes were identified for fruit length (Table 4; Fig. 1). The QTL fl1.1 and fl7.1 were detected in all four experiments; fl3.1 was detected in three seasons (NJ2012H, WI2012H, and WI2013H); fl4.1 and fl6.1 each was identified in two seasons. The three QTLs, fl1.1, fl3.1, and fl7.1, each could explain 7.5–22.5 % phenotypic variations depending on the season. While fl6.1 had negative additive effect (reduction of fruit length), all other QTLs contributed to increasing fruit length.

Since the major-effect QTL fl1.1 (R 2 > 15 %) had a highly consistent peak location across all four experiments, as a test case we validated this QTL with F2 data from the WI2012H trial. We genotyped 394 of 436 F2 plants with the marker SSR12331 at 53.7 cM where fl1.1 peaked. The frequency distribution of fruit length among the 394 F2 plants, as well as their corresponding genotypes (AA, AB and BB) at the SSR12331 locus, is presented in supplemental Fig. S4 (online material). It is clear that plants with longer fruits were enriched with the A allele (from CC3) and those with short fruits carried the B allele (from SWCC8). For example, there were 197 plants with an average fruit length of >40.0 cm (the mean FL of F2, Table 2). Among the 197 plants, 97 carried A allele (genotype AA), 98 were heterozygotes (H, genotype AB), and only two carried the SWCC8 allele (B allele, genotype BB). Among the 93 plants with FL >45 cm, 55 and 48 had the A and H genotypes, respectively (none carried B allele). This result provided further evidence that fl1.1 was a major QTL for fruit length.

QTL of fruit diameter (FD)

Three QTLs, fd1.1, fd4.1, and fd6.1, were detected; the first two were consistently identified in all four experiments with high LOD support (Table 2, Fig. S3C). All FD QTLs showed negative additive effects (reduction of fruit diameter). The major QTL in chromosome 1 (fd1.1 at 43.2 cM) had the largest effect: the R 2 varied from 12.3 to 31.7 % among four environments and the peak locations of fd1.1 in the four trials were highly consistent (Fig. S3C1). The QTL fd4.1 that was also detected in all four experiments seemed to have moderate effects on fruit diameter (R 2 ≈ 5~15 %) (Table 4).

QTL of fruit weight (FW)

Among the three QTLs for fruit weight, the major QTL fw6.1 (at 42.7 cM, with R 2 ranging from 10.0 to 28.7 %, Fig. S3E1) was identified in all four experiments. The QTL fw4.1 in chromosome 4 (at 26.2 cM) was detected in three seasons with moderate effects (R 2 = 5.1–9.1 %). A minor QTL, fw2.1 with R 2 ≈ 5.3 % (peaked at 58.3 cM in chromosome 2), was detected in WI2012H. However, as shown in Fig. S3E, the LOD curves from the data of NJ2009F and NJ2012S also supported the presence of this minor QTL at the same location despite the non-significant LOD support scores.

Consensus QTL for FFT, FL, FD, and FW across the four experiments

Among the 39 QTLs for the four traits, all but one (fw2.1) (but see Fig. S3E) were detected in at least two environments. QTLs for the same trait often peaked at the same or closeby map locations (Table 4) suggesting that these QTLs probably belong to the same locus for the trait. As such, by synthesizing information from Table 4, 13 QTLs could be recognized with 2 (fft1.1, and fft6.1), 5 (fl1.1, fl3.1, fl4.1, fl6.1, and fl7.1), 3 (fd1.1, fd4.1, and fd6.1), and 3 (fw2.1, fw4.1, and fw6.1) for FFT, FL, FD, and FW, respectively. Information on these 13 QTLs is summarized in Table 5.

Table 5 Consensus map locations of QTLs for female flowering time (FFT), fruit length (FL), fruit diameter (FD), and fruit weight (FW) detected from four experiments over 3 years
Full size table

Among the 13 QTLs, 3, 3, and 4 were located in chromosomes 1, 4, and 6, respectively; chromosomes 2, 3, and 7 each harbored one QTL, and chromosome 5 had none. The QTLs for FD and FW were co-localized in chromosomes 4 and 6 (Fig. 2), which may explain the significant correlation between the two traits in the present study (Table 3).

Discussion

Chromosome differentiation in XIS cucumber

Classical taxonomic studies recognized four cucumber botanical variants in C. sativus including the cultivated cucumber (var. sativus), the wild cucumber (var. hardwickii (Royle 1835; Duthie 1903), the Sikkim cucumber (var. sikkimensis) (Hooker 1876), and the XIS cucumber (var. xishuangbannesis) (Qi 1983). Morphological variations among these variants are large with cultivated and wild cucumbers representing the two extremities (Kirkbride 1993; de Wilde and Duyfjes 2010). Yang et al. (2012) identified significant differences in the amount and distribution of heterochromatins, as well as six inversions in chromosomes 4, 5, and 7 between C. sativus var. sativus and C. sativus var. hardwickii; the results support the subspecies status of these two cucumber taxa and C. sativus var. hardwickii as the progenitor of cultivated cucumber.

In the present study, alignment of the XIS genetic map with the cultivated cucumber map (Yang et al. 2012, 2013) suggested no major gross structural rearrangements between the two variants except for five small inversions in chromosomes 1 and 6 (Fig. 3). Since the orders of shared marker loci or draft genome scaffolds involved in these inversions were consistent between cultivated and wild cucumbers, these structural changes might be specific to the XIS cucumber. Unlike the wild cucumber map (Ren et al. 2009), no clustering of markers was found in chromosomes 4, 5, and 7 on the XIS map (Fig. 3). This is consistent with the cultivated cucumber map (Yang et al. 2012) suggesting that the XIS and cultivated cucumbers shared common ancestors. That is, the origin of the XIS cucumber was after domestication of cucumber from C. sativus var. hardwickii. Therefore, the characteristic morphological traits such as large and heavy fruits, dark green leaves, vigorous vine growth, and the orange flesh color in the XIS cucumber were likely the result of diversifying selection for adaptation to the local environment where the XIS cucumber grows today. This may also imply that the XIS-specific inversions in chromosomes 1 and 6 identified herein (Fig. 3) occurred after its divergence with the cultivated cucumber. In plants, lineage- or population-specific inversions are believed to play important roles in evolution (for example, local adaptation and speciation) (Hoffmann and Rieseberg 2008; Kirkpatrick 2010; Lowry and Willis 2010). However, it is not known if the XIS-specific inversions identified herein exist in all XIS populations or these inversions have any adaptive significance. On the other hand, the XIS cucumber requires short day length for flowering. It is appropriate to treat the XIS cucumber as a semi-wild botanic variant (Qi 1983).

QTL mapping of traits under domestication and diversifying selection in XIS cucumber

Our QTL mapping effort focused on four traits that were characteristic of XIS cucumber. Thirteen QTLs were identified for the first female flowering time (FFT), the length (FL), diameter (FD), and weight (FW) of mature fruits (Table 5). Several previous studies have also identified QTLs for fruit length, diameter and weight, as well as flowering time in cucumber (Kennard and Havey 1995; Yuan et al. 2008; Dijkhuizen and Staub 2002; Fazio et al. 2003; Cheng et al. 2010; Miao et al. 2012, 2011). For convenience of discussion, the locations of QTLs detected by those and the present studies are placed onto the linkage map developed herein (Fig. 2); but for some early studies (Kennard and Havey 1995; Dijkhuizen and Staub 2002; Fazio et al. 2003), the chromosomal locations of mapped QTL were difficult to infer due to the nature of markers used and were not included in Fig. 2.

We identified a major QTL, fft1.1 located at the distal end of cucumber chromosome 1 that could explain >50 % observed phenotypic variations; the minor QTL fft6.1 was mapped in chromosome 6 (Fig. 1). Both alleles carried by the XIS cucumber exhibited negative additive effects, which is consistent with its very late flowering nature (photoperiod sensitive for flowering initiation). The QTL fft1.1 seems to be co-localized with the major QTL Da1.1 (days to anthesis of first female flower) detected by Miao et al. (2012). Interestingly, the mapping population used by Miao et al. (2012) was RILs derived from a cross between the north China fresh market cucumber 9930 and the European greenhouse type cucumber 9110Gt, which only have 2 days’ difference in anthesis of the first female flowers. Since XIS cucumber has the same flowering habit (requirement of short day length for flowering initiation) as the wild cucumber (C. sativus var. hardwickii), this may suggest that the late flowering QTL fft1.1 is a trait under selection during domestication of cucumber. Consistent with this, through analysis of genome-wide genetic variations between cultivated and wild cucumbers, Qi et al. (2013) identified 112 putative domestication sweeps, 4 of which (DS14, 15, 16 and 17) were located in the fft1.1 2-LOD interval between SSR05723 and SSR16995 (Fig. 2).

Five QTLs were identified for fruit length with three major-effect QTLs (fl1.1, fl3.1, and fl7.1) (Table 5). Kennard and Havey (1995) also detected five fruit length QTLs from an F2:3 mapping population derived from Gy14 × PI 183967 (wild cucumber), but the locations of these QTLs in the early study are difficult to infer. None of the three major-effect QTLs showed consistent map locations with previously detected fruit length QTLs (Yuan et al. 2008, Cheng et al. 2010; Miao et al. 2011) (Fig. 2). The locations of the two minor-effect QTLs (fl4.1 and fl6.1) seem to be consistent with those detected by Yuan et al. (2008) (fl4.1 major-effect QTL on chromosome 4) and Miao et al. (2011) (major-effect QTLs fl4.1, sfl6.1 in chromosomes 4 and 6, respectively) (Fig. 2). In an RIL population developed from 981 (north China type) × PI 183967, five fruit length QTLs were detected in four chromosome regions (1, 3, 4 and 6) (Cheng et al. 2010; Qi et al. 2013), which were believed to be under selection during domestication (Qi et al. 2013). Among the five FL QTLs detected in this study, only fl4.1 showed consistent map location with fl4.2 by Cheng et al. (2010) (Fig. 2). For the fruit diameter QTLs (fd1.1, fd4.1 and fd6.1) detected herein, their chromosome locations were largely consistent with those identified in Yuan et al. (2008) and Miao et al. (2012). Lastly, the locations of fruit weight QTLs fw2.1, fw4.1, and fw6.1 (major-effect QTL) from this study were close to those detected by Yuan et al. (2008) (fw3.1 on chromosome 2, fw4.1 on chromosome 4) and Miao et al. (2012) (sfw6.1 and sfw6.1 on chromosome 6) (Fig. 1). The significant positive correlation of FD and FW (Table 3) could be explained by the co-localization of FD and FW major-effect QTLs on chromosome 4 and the close linkage in chromosome 6 (Fig. 1).

While the QTLs at the same or close locations across different studies may imply common mechanisms underlying the fruit shape and size during domestication or diversifying selection, the discrepancies in the number, location, and magnitude of effect of QTLs for the same trait in different studies could be explained in several ways. The most reasonable explanation is that the cucumber lines used in these studies belong to different taxonomic groups or market classes. The populations used for QTL mapping were derived from crosses between wild and cultivated (Kennard and Havey 1995; Dijkhuizen and Staub 2002; Cheng et al. 2010; Qi et al. 2013), semi-wild and cultivated (this study), or between cultivated cucumber lines of different marker classes (Yuan et al. 2008; Miao et al. 2011, 2012). It is possible that these traits have undergone domestication or diversifying selection for specialized market classes, and the genes underlying these traits may be different targets of the selection. In addition, the ability to detect QTLs of fruit-related traits may depend on the growth stages (e.g., commercial harvest stage vs. mature fruits) of data collection (e.g., Miao et al. 2011). Of course, the criteria of trait phenotyping, the seasons (spring, summer or fall) for data collection, and environments (open field, greenhouse or protected plastic houses) may all contribute to the different results in these QTL mapping studies.

Fruit shape QTL in cucumber

The length and diameter of cucumber fruits are economically important traits. For example, at commercial harvest stage, North American pickling cucumbers should have length-by-diameter (L/D) ratios of approximately 3.0. In other crops such as tomato (Solanum lycopersicum) and melon, the L/D ratio is often called fruit shape or fruit shape index. In cucumber, L/D has been considered as an independent trait, and QTLs for this trait have been identified in several studies (Kennard and Havey 1995; Dijkhuizen and Staub 2002; Fazio et al. 2003; Yuan et al. 2008; Miao et al. 2011).

In cucumber, the fruit develops from an enlarged inferior ovary. In the pickling cucumber cultivar ‘Vlaspik’, fruit elongation begins almost immediately after pollination, with the most rapid increase occurring approximately 4–12 days post-pollination (dpp); the rapid increase in cell size mirrors the rapid increase in fruit length. The increase in fruit diameter is somewhat lagging behind the length which occurs primarily between 4 and 16 dpp (Ando and Grumet 2010; Ando et al. 2012). Consistent with these observations, the correlation between length and diameter was not significant in both RIL and F2 populations (P = 0.05) in the present study (Table 3) suggesting that elongation of fruit (FL) and increase of diameter (FD) might be under different genetic mechanisms. The non-significant correlation between FL and FD was also found in melon (Eduardo et al. 2007). L/D is a composite trait (calculated from FL and FD) for which the mechanisms and QTL mapping strategy are not well understood (Li et al. 2010). It is obvious that the L/D value is influenced by the larger of the two component traits (L and D). For example, in melon lines with contrasting fruit length, the fruit shape (FS = L/D) was often highly correlated with FL (Perin et al. 2002; Monforte et al. 2005; Eduardo et al. 2007). But in tomato when the difference of FD was dominant between two parental lines, the FS is significantly correlated with FD (e.g., Lippman and Tanksley 2001). It is not surprising that, in cucumber, the QTL location for L/D often co-localized with either an FL or FD QTL, whichever had the larger effect (Kennard and Havey 1995; Fazio et al. 2003; Yuan et al. 2008; Miao et al. 2011), or in some case, no L/D was detected (Miao et al. 2011). In the present study, the locations of L/D major QTLs were largely consistent with fruit length QTLs in chromosomes 1 and 7 (Fig. S3D). It is not known if QTLs for L/D truly exist. The biological interpretation of L/D QTLs is also unknown. This is the reason we did not list L/D QTLs in the present study. L/D is an important selection criterion in cucumber breeding in all major market classes, but caution should be exercised in using markers linked with L/D QTLs in marker-assisted selection. Probably, both fruit length and diameter QTLs should be considered in decision making.

Candidate genes for major-effect fruit shape and size QTLs

In tomato, six genes or QTLs controlling fruit shape and size have been cloned which include CNR/FW2.2 and SlKLUH/FW3.2 controlling fruit size (weight), SUN and OVATE controlling elongated shape, as well as FASCIATED (FAS) and LOCULE NUMBER (LC) controlling fruit locule number and flat shape (reviewed in Rodriguez et al. 2011; Monforte et al. 2014). CNR/FW2.2 encodes a member of the cell number regulator (CNR); SlKLUH/FW3.2 encodes a member of a subfamily of cytochrome P450 A78 class (CYP78A) and the ortholog of KLUH; SUN encodes a protein that is a member of the IQ domain family; OVATE encodes a protein in the ovate family protein (OFP); FAS encodes a protein that is a member of the YABBY family, whereas LC is probably encoded by the ortholog of the A. thaliana gene WUSCHEL, which is a member of the WOX family (reviewed in Rodriguez et al. 2011; Monforte et al. 2014). Using tomato or Arabidopsis thaliana CNR, CYP78A, OFP, SUN, WOX, and YABBY gene family sequences as queries, Monforte et al. (2014) identified 74 homologs of the six gene families in the melon genome and found that QTLs for fruit weight co-localized frequently with members of the CNR/FW2.2 and KLUH/FW3.2 families, and fruit shape QTLs co-localized with the OFP family members.

Both cucumber and melon belong to the genus Cucumis with similar fruit development patterns, and the genomes of the two species share high degree of chromosomal synteny and sequence homology (e.g., Li et al. 2011; Garcia-Mas et al. 2012; Yang et al. 2013, 2014). We hypothesize that the genes for fruit shape and size cloned in tomato or Arabidopsis may also be responsible for QTLs we identified herein in cucumber. We used the 74 melon fruit size and shape-related sequences identified in Monforte et al. (2014) as queries to BLAST against the Gy14 cucumber draft genome (Yang et al. 2012) and found all of them had homologs in the cucumber genome. The melon homologs located within or very close to the region defined by 2-LOD interval of a QTL on the genetic map are shown in supplemental Table S2. Among the 74 candidate genes, CmOFP-14, CmOFP-15, and CmOFP-17 were co-localized with the fruit length QTL fl1.1, and fruit diameter QTL fl1.1; CmYABBY-4 was also within the 2-LOD interval of fl1.1. In addition, both CmOFP-16 and CmWOX-8 were located in the vicinity of fruit diameter QTL fd6.1. As compared with findings in melon (Monforte et al. 2014), co-localization of tomato candidate genes with mapped QTLs was not as common as found in melon. The reason may be multifold. First, the mechanisms underlying fruit length, width, and weight in cucumber may be different from those cloned in tomato or melon. Second, the QTL locations in either melon or cucumber need be refined through fine genetic mapping. Lastly, since we used melon homologs as queries to BLAST the cucumber genome, some sequences not presented in the melon draft genome may not be detected. A more detailed examination of those genes could be performed by looking into the genome annotations in those QTL-residing regions. Nevertheless, these candidate genes provide important clues for future fine mapping and cloning of these fruit shape and size QTLs.

Author contributions

YW and JC designed the experiment. KB performed the research. ZM collected data for the NJ2012S experiment. YW and KB analyzed the data and wrote the manuscript with inputs from JC. All authors reviewed and approved this submission.