Bitter gourd (Momordica charantia; 2n = 2x = 22) is a tropical and subtropical vine in the family Cucurbitaceae, which is widely cultivated in Asia, Africa, and the Caribbean for its edible and medicinal fruit (Grover and Yadav, 2004; Marr et al., 2004; Van Wyk, 2015). Bitter gourd fruit provides a good source of phytonutrients like carbohydrates, minerals, and vitamins (Behera et al., 2010). The pharmacological composition and properties of bitter gourd have also been widely investigated (Tan et al., 2016). Although bitter gourd is consumed and cultivated, it remains an underutilized crop and scant efforts have been made for molecular improvement.

The construction of a genetic map is a common approach to detect QTLs and conduct gene mapping. The first genetic map of bitter gourd was constructed based on amplified fragment length polymorphism (AFLP) markers (Kole et al., 2012). A year later, the second genetic map was constructed, and it consisted of combined markers including simple sequence repeats (SSR), AFLP, and sequence-related amplified polymorphism (SRAP) markers (Wang and Xiang, 2013). Recently, two RAD-based genetic maps were constructed using two F₂ populations derived from the same parents (Matsumura et al., 2014; Urasaki et al., 2017), allowing for preliminary attempts at next-generation sequencing (NGS). Urasaki et al. (2017) reported a bitter gourd reference (OHB3-1 reference) and anchored 255 scaffolds to the genetic map. RAD markers are short fragments of DNA adjacent to each instance of a particular restriction enzyme recognition site (Baird et al., 2008). Combined with NGS, RAD-seq provides an inexpensive platform that allows high-density SNP discovery and genotyping in large populations (Davey et al., 2011).

In order to conduct marker-assisted selection (MAS) in bitter gourd breeding, it is necessary to target genetic loci underlying important horticultural traits. Similar to cucumber and melon, bitter gourd shows different sexual morphs. Gynoecy has potential applications in heterosis breeding and hybrid seed production. The gene that controls gynoecy (gy) was reported as a single recessive gene (Ram et al., 2006; Behera et al., 2009; Matsumura et al., 2014). An identified SNP marker linked to this gene has been mapped at a distance of 5.46 cM (Matsumura et al., 2014). In addition, bitter gourd fruits show extensive variation in epidermal color and structure, which directly effects exterior quality. Immature bitter gourd fruits have a broad color spectrum, ranging from white to dark green (Cheng et al., 2012). Several studies have shown that the green fruit color is monogenically dominant over the white fruit color (Srivastava and Nath, 1972; Hu et al., 2002; Dalamu et al., 2012). However, a more recent study indicated that this trait is controlled by quantitative genes (Huang and Hsieh, 2017). Bitter gourd fruit has either a smooth or a distinct warty exterior, and has broken or continuous ridges. The warty exterior is monogenically dominant over the smooth (Kole et al., 2012); however, research devoted to identifying genetic loci responsible for these important traits is limited.

As part of our ongoing efforts to sequence the bitter gourd genome, an efficient genetic map is instrumental for anchoring and orienting the assembled scaffolds. Here, we report a RAD-based linkage map of bitter gourd using an F₂ mapping population with 423 individuals derived from two cultivated inbred lines, the gynoecious line ‘K44’ and the monoecious line ‘Dali-11.’ Another major goal of the study was to map genetic loci underlying horticulturally important traits such as sex expression, fruit epidermal structure, and immature fruit color. To reach this objective, we carried out an elaborate investigation of phenotypic characters in three environments. The results of this study will be beneficial to research attempting to develop linked molecular markers in MAS breeding and to clone these genes in the future.

One of the aims of this research was to construct a bitter gourd genetic map to anchor assembled scaffolds. RAD-seq has been widely used to construct genome wide genetic maps in ryegrass (Pfender et al., 2011), barley (Zhou et al., 2015), peanut (Gupta et al., 2015), apple (Sun et al., 2015), and soybean (Wang W. et al., 2016). In this study, we successfully created a RAD-based genetic map for bitter gourd, comprising 1,009 SNP markers that distributed on 11 linkage groups. The majority of assembled sequences (85.48%) can be anchored to this map, indicating a high genomic coverage of SNP markers. This map provided a basis for integrating assembled scaffolds into a draft genome of bitter gourd, which served as an important step in our ongoing work on sequencing the bitter gourd genome.

This RAD-based genetic map had a mean marker density of 0.46 (maker/cM), which is higher than the 0.30 mean marker density reported by Matsumura et al. (2014) and 0.42 by Urasaki et al. (2017). Nonetheless, it is far from being high resolution. Our study likely recovered a higher mean marker density due to the choice of restriction enzyme (Baird et al., 2008). We used EcoRI to digest genomic DNA for RAD-seq analysis, but Matsumura et al. (2014) and Urasaki et al. (2017) used PacI and AseI, respectively. In the future, we recommend the use of additional restriction enzymes when attempting to construct a high density or saturated RAD-based genetic map. Furthermore, the recombination rate is a critical consideration for fine mapping and breeding, and we estimated an overall value of 8.77 cM/Mb for bitter gourd. This recombination rate was significantly higher than 3.28 cM/Mb for cucumber (Huang et al., 2009) and 2.30 cM/Mb for watermelon (Ren et al., 2012), but lower than 13.19 cM/Mb for pumpkin (Zhang et al., 2015). Because a high recombination rate can indicate a large genetic distance over a narrow physical distance, our finding suggests that it is beneficial to use fine mapping methods to determine QTLs/genes for bitter gourd compared with the methods used for species with low recombination rates like cucumber and watermelon.

Another aim of this research was to map genetic loci of three horticulturally important traits involved in sex expression, fruit epidermal structure, and immature fruit color. These traits were evaluated using a combination of qualitative and quantitative data. The repeated QTLs/genes were identified in three environments, and both genetic and physical intervals for each locus were delimited. Currently, we are sequencing a bitter gourd genome dependent on this genetic map, and once the gene annotation is completed, fine mapping of these loci combined with candidate gene analysis will assist in cloning the relevant genes more efficiently.

The inheritance of gynoecy was not completely consistent with previous reports (Ram et al., 2006; Behera et al., 2009; Matsumura et al., 2014); it appeared to be controlled by two pairs of genes in the F₂ population in Haikou in spring 2014. Both correlation and inheritance analyses indicated that sex expression in bitter gourd is easily affected by the environmental factors in which it is grown, such as the temperature and photoperiod, which can regulate the sex of cucumber flowers (Yamasaki and Fujii, 2003). Matsumura et al. (2014) located the gy locus at the distal end of the linkage group, and a similar result was found in this study. The reported SNP marker (GTFL-1) that mapped at a distance of 5.46 cM to gy has been aligned at the site of scaffold44_7277273, consistent with our confidence interval (scaffold44_3793313–scaffold44_7318231). Because of its importance, this target region is necessary for further fine mapping. In melon and cucumber, the mechanism of sex determination has been studied in depth. Sexual morphs of melon plants are determined by a combination of alleles at gynoecious (G), andromonoecious (M), and androecious (A) sex loci, controlled by CmWIP1, CmACS7, and CmACS11, respectively (Boualem et al., 2015). Gynoecy in melons is governed by a combination of Cmwip1/Cmwip1, CmACS7/-, -/- (the minus symbol indicates any allele at the locus) at the G, M, and A loci, respectively. CmWIP1 encodes the zinc finger protein WIP1; CmACS7 and CmACS11 encode two 1-aminocyclopropane-1-carboxylic acid synthases. Interestingly, no homologs of these genes were annotated in this QTL (gy) region. Unlike for cucumber and melon, gibberellin (GA3) is the most effective growth regulator for increasing femaleness for bitter gourd (Ghosh and Basu, 1983). These findings imply that the gynoecy of bitter gourd may be controlled by other genetic mechanisms. Further work should be performed to unveil the genetic determination of sexual morphs in bitter gourd.

Fruit warts and ridges are distinct in shape and epidermal distribution, both of which act as identifying features of bitter gourd. The width of the fruit ridge is correlated with the presence of fruit warts, which provides a novel measurement to dissect the genetic loci of fruit epidermal structure. Our results indicated that the two characters of fruit epidermal structure were possibly controlled by a same dominant gene (Fwa/Wr), suggesting a pleiotropic effect. In cucumber, fruit warts consist of spines (non-glandular trichomes) and tubercules governed by csgl1/mict (Li et al., 2015; Zhao et al., 2015) and Tu (Yang et al., 2014), respectively, and csgl1/mict has an epistatic effect on Tu (Yang et al., 2014). Recently, tril, required for the initiation of trichomes, was reported to show epistasis on the csgl1/mict gene (Wang Y.L. et al., 2016). The fruit warts on bitter gourd have no spine, while trichomes can be produced on stems, leaves, fruit stalks, tendrils, and floral organs. Consequently, cloning the genes underlying both fruit warts and trichomes of bitter gourd is essential to understanding the molecular mechanisms that control the formation of fruit warts. Undoubtedly, our results will accelerate the cloning of the Fwa/Wr gene.

Immature fruit color is another important trait directly affecting exterior quality. We used data from our observations of fruit color (green and white) and two indirect measurements to delimit the determined gene, w, into a narrow physical interval (∼936.70 kb). In cucumber, the w gene was confirmed as an APRR2 (two-component response regulator-like) gene. The white color is caused by a single-base insertion in the w gene, which may disable its function in chlorophyll accumulation and chloroplast development (Liu et al., 2015, 2016). Similarly, the white color of bitter gourd fruit is a visible result of a lack of chlorophyll synthesis or disabled development of the chloroplast. The annotation of the bitter gourd genome anchored by the genetic map in the present study will help to directly analyze candidate genes in this physical interval.

In this paper, we constructed a bitter gourd genetic map using the RAD-seq method. This map successfully assisted in anchoring most of the assembled scaffold sequences. The genetic map and anchored assembly revealed very high recombination rates in bitter gourd. We further used this genetic map and component characters to detect the genetic loci controlling sex expression, fruit epidermal structure, and immature fruit color in bitter gourd. Three reliable QTL/gene loci responsible for these traits were identified through multiple phenotypic data and environments. This study provides the foundation for the assembly of the bitter gourd genome and for further dissecting these horticulture-associated loci.

JuC, JiC, and KH conceived and designed the experiments. SL, YN, RH, QW, JS, and NM contributed to field management and conducted the DNA extractions. WH and ZD performed the bioinformatic analyses. JuC wrote the manuscript and KH revised the manuscript. All authors read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.