Identification of QTLs for Domestication-Related Traits in Zombi Pea [Vigna vexillata (L.) A. Rich], a Lost Crop of Africa

Zombi pea [Vigna vexillata (L.) A. Rich] is a legume crop found in Africa. Wild zombi pea is widely distributed throughout the tropical and subtropical regions, whereas domesticated zombi pea is rarely cultivated. Plant domestication is an evolutionary process in which the phenotypes of wild species, including seed dormancy, pod shattering, organ size, and architectural and phenological characteristics, undergo changes. The molecular mechanism underlying the domestication of zombi pea is relatively unknown. In this study, the genetic basis of the following 13 domestication-related traits was investigated in an F2 population comprising 198 individuals derived from a cross between cultivated (var. macrosperma) and wild (var. vexillata) zombi pea accessions: seed dormancy, pod shattering, days-to-flowering, days-to-maturity, stem thickness, stem length, number of branches, leaf area, pod length, 100-seed weight, seed width, seed length, and seeds per pod. A genetic map containing 6,529 single nucleotide polymorphisms constructed for the F2 population was used to identify quantitative trait loci (QTLs) for these traits. A total of 62 QTLs were identified for the 13 traits, with 1–11 QTLs per trait. The major QTLs for days-to-flowering, stem length, number of branches, pod length, 100-seed weight, seed length, and seeds per pod were clustered in linkage group 5. In contrast, the major QTLs for seed dormancy and pod shattering belonged to linkage groups 3 and 11, respectively. A comparative genomic analysis with the cowpea [Vigna unguiculata (L.) Walp.] genome used as the reference sequence (i.e., the genome of the legume species most closely related to zombi pea) enabled the identification of candidate genes for the major QTLs. Thus, we revealed the genomic regions associated with domestication-related traits and the candidate genes controlling these traits in zombi pea. The data presented herein may be useful for breeding new varieties of zombi pea and other Vigna species.


INTRODUCTION
Plant domestication is a process in which wild forms of crop (cultivated) plants evolve and adapt to agricultural conditions and human needs via conscious and unconscious selections by humans and natural selection due to environmental factors. Domestication is one of the most important technological innovations in human history (Doebley et al., 2006). Although the processes and spatiotemporal requirements for domestication vary among crops, the processes generally lead to similar results regarding morphological and phenological changes to the wild forms. The changes from wild species to domesticated/cultivated species are collectively referred to as the "domestication syndrome" (Hammer, 1984). Compared with wild forms, domesticated forms exhibit a lack of or decreased seed dormancy and seed dispersal, produce larger organs (fruit, seed, leaf, and stem), flower and mature earlier, and have a more robust and firm plant architecture. During the last 20 years, plant domestication studies have become a major part of crop evolution and adaptation research. In fact, domestication is the earliest form of plant breeding (Doebley et al., 2006). The information generated in investigations of the evolutionary relationships between crops and their wild progenitors may be useful for exploiting wild plants for crop improvement (Vaughan et al., 2007). The genetic basis of the domestication syndrome of several plant species, but mainly cereals and legumes, including rice, maize, wheat, sorghum, soybean, common bean, and cowpea, has been investigated via molecular genetic analyses, especially quantitative trait locus (QTL) mapping. Several genes controlling the domestication-related traits of some of these crops have been identified and cloned (Doebley et al., 2006;Vaughan et al., 2007;Ishimaru et al., 2013).
Zombi pea [Vigna vexillata (L.) A. Rich] is an underutilized legume crop. Wild zombi pea is widely distributed in Africa, Asia, Australia, and the Americas (Dachapak et al., 2017). Because of its wide distribution, several wild forms of zombi pea have been identified and described (Maxted et al., 2004). Additionally, there are two types of cultivated zombi pea (i.e., seed and tuberous root) (Ferguson, 1954;Karuniawan et al., 2006). The cultivated seed type is grown in Africa (Ferguson, 1954), whereas the cultivated tuberous root type exists only in Bali, Indonesia, and East Timor (Karuniawan et al., 2006). The former is grown principally for its seeds, although the tuberous roots are sometimes consumed, whereas the latter is cultivated mainly for its tuberous roots. An analysis of molecular diversity based on microsatellite (i.e., simple sequence repeat) markers revealed that these two forms were domesticated independently (Dachapak et al., 2017). There exists a strong crossing barrier between the cultivated tuberous root type and the cultivated seed type/wild forms of zombi pea.
The genetic basis of zombi pea domestication is particularly interesting because the two cultivated types were likely domesticated from the same or different wild zombi pea species via different processes (Dachapak et al., 2017). The seed-type zombi pea plant produces relatively few branches and exhibits a determinate growth habit. In contrast, the tuberous roottype zombi pea is a highly branched, vine-like plant with an indeterminate growth habit. Dachapak et al. (2018) identified QTLs for the domestication of the tuberous root form of zombi pea. However, because of the sterility of the mapping population, the QTLs of key domestication-related traits, including pod dehiscence (pod shattering) and seed dormancy, were not detected (Dachapak et al., 2018). Therefore, identifying the QTLs controlling domestication-related traits in the seed-type zombi pea may provide important insights regarding the genetic basis of the domestication of this species and other Vigna species, which may facilitate the identification of new candidate genes for crop improvement. The objectives of this study were to (i) detect the QTLs associated with domestication-related traits and (ii) identify candidate genes for these traits in the seedtype zombi pea.

Mapping Population
An F 2 population comprising 198 individuals developed from a cross between TVNu 240 and TVNu 1623 was used in this study. Specifically, TVNu 240 is a cultivated zombi pea (V. vexillata var. macrosperma) accession from the Central African Republic, whereas TVNu 1623 is a wild zombi pea (V. vexillata var. vexillata) accession from Nigeria. The F 2 population was previously used for developing high-density linkage maps and for the mapping of QTLs for bruchid (Callosobruchus spp.) resistance . Briefly, the F 2 plants and 10 plants of each parent were grown under field conditions, with 0.75 m × 0.75 m spacing, at Kasetsart University, Kamphaeng Sean Campus, Nakhon Pathom, Thailand between December 2017 and March 2018.

Measurement of Domestication-Related Traits
Thirteen domestication-related traits were assessed in the F 2 and parental plants ( Table 1). All of the traits, except for leaf area (LFA), were analyzed in previous studies on Vigna species, including mung bean (Isemura et al., 2012), azuki bean (Isemura et al., 2007), rice bean (Isemura et al., 2010), cowpea (Kongjaimun et al., 2012), and zombi pea (Dachapak et al., 2018). The LFA of nine fully expanded leaves was measured with the LI-3100C area meter (LI-COR Biosciences, United States) at 50 days after sowing. Stem length (STL) and stem thickness (STT) were recorded at harvest. The number of days from sowing to first flowering (FLD) and the number of days from sowing to harvesting of the first pod (PDDM) were also recorded. The mature pods and seeds of each plant were harvested separately to analyze the pod-related and seed-related traits. Specifically, pod length (PDL) was measured and the number of seeds per pod (SDNPPD) was determined based on 10 pods. Pod shattering was evaluated by counting the number of twists along the pod (PDT) of five pods maintained at room temperature. The seed length and width (SDL and SDW, respectively) were recorded as the average value of five seeds. The 100-seed weight (SD100WT) was measured twice using intact seeds. Seed dormancy was assessed by measuring seed water absorption (SDWA). Specifically, 40 intact seeds harvested from each plant were immersed in water in a plastic container and incubated at 25 • C for 7 days. The number of seeds that imbibed water was recorded.

Heritability Estimation and Correlation Analysis
The broad-sense heritability (H 2 ) (Allard, 1999) of each trait was calculated using the following formula: where σ 2 F2 , σ 2 TVNu240 and σ 2 TVNu1623 are the variances of the F 2 population, TVNu 240, and TVNu 1623, respectively. The correlation between traits was calculated with the R program (version 10.12.0).

Quantitative Trait Loci Mapping
The F 2 population was previously used for constructing a single nucleotide polymorphism (SNP)-based linkage map and for mapping QTLs for bruchid resistance . The SNP markers were generated via a specific locus amplified fragment sequencing method (Sun et al., 2013). The map consisted of 11 linkage groups (LGs) and comprised 6,529 SNP markers . The linkage map and genotypic (SNP marker) data were used for analyzing the QTLs of domestication-related traits in the present study.
The QTLs were analyzed according to the inclusive composite interval mapping (ICIM) method (Li et al., 2007) of the QTL IciMapping 4.2 software 1 , with 0.1 cM steps. A probability of

Identification of Candidate Genes for Domestication-Related Traits
Candidate genes at the major QTLs [i.e., phenotypic variance explained (PVE) ≥10%] for domestication-related traits were identified based on the nearest DNA markers flanking the QTLs on the cowpea reference genome sequence (Lonardi et al., 2019) 2 . All of the annotated genes between the markers were identified, after which the genes with functions potentially associated with the specific trait were selected as candidate genes.

Variation and Heritability of Domestication-Related Traits
The data for the domestication-related traits of the F 2 population (mean and range) and the traits of the parents (mean) are summarized in Table 2. The values for the following traits were higher for the cultivated parent (TVNu 240) than for the wild parent (TVNu 1623): germination (SDWA), organ size (SD100WT, LFA, SDL, PDL, and STL), yield (SDNPPD), and earliness (FLD and PDDM). In contrast, the STT and branching (BRNPP) values were higher for the wild parent than for the cultivated parent. The parents had very similar PDT and SDL values. The SDWA, SD100WT, LFA, SDW, STT, BRNPP, and SDNPPD mean values of the F 2 population were between those of the parents ( Table 2). However, the PDT and SDL mean values were similar among the F 2 population and both parents. The F 2 population mean values for PDL and STL were almost the same as those of the wild parent and cultivated parent, respectively, whereas the SDNPPD mean value was lower for the F 2 population than for either parent ( Table 2). All traits exhibited a continuous distribution (Figure 1), indicative of a quantitative trait inheritance. Specifically, SDNPPD exhibited a bimodal distribution, suggesting that this trait may be controlled by a few loci. Transgressive segregation was clearly detected for several traits, including PDT, SDWA, LFA, PDL, STT, STL, BRNPP, FLD, PDDM, and SDNPPD (Figure 1 and Table 2), implying that both parents possessed alleles that positively and negatively affected these traits. Most of the traits exhibited moderate to relatively high heritability (Table 2), with the heritability value exceeding 0.70 for seven traits.
The correlations among the examined traits are listed in Supplementary Table S2. There were significant and positive correlations between related traits, including SD100WT and SDW or SDL, PDL and SDNPPD, and FLD and PDDM. Additionally, SD100WT was positively correlated with PDT, STT, STL, BRNPP, and LFA, but negatively correlated with SDNPPD. Interestingly, LFA was significantly and positively correlated with all of the other traits, with the exception of SDWA  and PDT. Moreover, SDWA was significantly and negatively correlated with PDT, FLD, PDDM, and BRNPP. The data also revealed a significant correlation between BRNPP and all other analyzed traits.

QTLs and Candidate Genes for Domestication-Related Traits
A total of 62 QTLs were identified for the 13 analyzed traits, with 1-11 loci per trait ( Table 3 and Figure 2). These 62 QTLs were distributed in all LGs, except for LG10. The number, location, and effect of the QTLs identified for each trait are summarized in Table 3.
The wild and cultivated parents differed considerably regarding SDWA (23.20 and 59.80%, respectively). The three QTLs detected for this trait, qSdwa1.1+, qSdwa3.1-, and qSdwa5.1+, were in LGs 1, 3, and 5, respectively (Table 3 and Figure 2). The QTL in LG3, qSdwa3.1-, had the largest effect (PVE of 22.2 %). Interestingly, at this QTL, the cultivated zombi pea alleles decreased the seed coat permeability (decreased germination). The qSdwa1.1+ and qSdwa5.1+ QTLs both accounted for about 9% of the total variation. At these two QTLs, the cultivated zombi pea alleles increased the seed coat permeability. The qSdwa3.1-region consisted of approximately 179.0 kb on cowpea chromosome 3 and included 16 genes (Supplementary Table S3). Of these genes, Vigun03g339300, encoding a calmodulin-binding protein-like protein, was identified as a candidate gene for this QTL. The qSdwa5.1+ region spanned about 1.74 Mb on cowpea chromosome 5 and contained 38 genes (Supplementary Table S3), with Vigun05g170600 and Vigun05g170700, both of which encode gibberellin 2-beta-dioxygenase 1, identified as candidate genes for this QTL.
The leaves of the cultivated parent were larger than those of the wild parent. The only QTL detected for LFA, qLfaA9.1− (PVE of 18.2%), was mapped to LG9 (Table 3 and Figure 2). Moreover, the cultivated parent alleles at this QTL decreased the LFA value. The qLfa9.1− region spanned 201.1kb on cowpea chromosome 9 and comprised 17 genes (Supplementary Table S3). Among these genes, Vigun09g018300 and Vigun09g017000, encoding a WD40 repeat-containing protein and falz-related bromodomaincontaining protein/transcription factor GTE1, respectively, were identified as candidate genes for this QTL.  The cultivated zombi pea pods were approximately 3.5 cm longer than the wild zombi pea pods ( Table 2). Five QTLs related to PDL were detected on different LGs (Table 3 and Figure 2). At QTLs qPdl3.1+ and qPdl7.1+, the cultivated zombi pea alleles increased the PDL value. Additionally, at QTLs qPdl5.1−, qPdl6.1−, and qPdl9.1−, the wild zombi pea alleles increased the PDL value. The qPdl5.1− QTL in LG5 had a PVE of 37.7%, which was considerably higher than the PVE for the other four QTLs (<5%). The qPdl5.1− region spanned 860.4 kb on cowpea chromosome 5 and contained 41 genes (Supplementary Table S3). The Vigun05g125800 gene, encoding pectate lyase 1, was chosen as a candidate gene for qPdl5.1−.
The cultivated zombi pea plants had shorter stems (the first 10 internodes), but more branches, than the wild zombi pea plants. The four QTLs for STL belonged to different LGs (Table 3 and Figure 2). The PVE of each of the QTLs for STL was ≤10%, but was highest for qStl9.1− in LG9. The cultivated parent alleles at three of the four QTLs decreased the internode length. Similarly, four QTLs on different LGs were detected for BRNPP, and the PVE of each QTL was ≤12%, but was highest for qBrnpp5.1− in LG5 (Table 3 and Figure 2). The cultivated parent alleles at three of the four QTLs decreased the number of branches. Only one marker flanking qStl9.1−, Marker293030, matched a sequence in the cowpea reference genome. The Vigun09g037900 gene, encoding a MOB-like protein phocein (MOB1), which was detected approximately 113 kb from Marker293030, was considered to be a candidate gene for qStl9.1−. The qBrnpp5.1− region was less than 1.0 kb on cowpea chromosome 5 and included Vigun05g082200, which encodes an auxin efflux carrier family protein (Supplementary Table S3).
Domestication generally leads to cultivated crops that flower and mature earlier than their wild progenitors. The cultivated zombi pea plants flowered and matured approximately 10 and 8 days earlier, respectively, than the wild zombi pea plants ( Table 2). Four QTLs on three LGs were identified for FLD (Table 3 and Figure 2). The effects of QTLs qFld2.1− in LG2 and qFld5.1− in LG5 on FLD were similar (PVE of 16.5 and 18.6%, respectively). With the exception of qFld5.2+, the cultivated parent alleles at these QTLs enhanced flowering. The qFld2.1− region spanned 391.5 kb on cowpea chromosome 2 and comprised 28 genes FIGURE 2 | Quantitative trait locus plots for 13 domestication-related traits in the zombi pea F 2 population developed from a cross between cultivated zombi pea (TVNu 240; V. vexillata var. macrosperma) and wild zombi pea (TVNu 1623; V. vexillata var. vexillata). The x-axis indicates the linkage groups, whereas the y-axis indicates the logarithm of odds (LOD) scores. The gray line horizontal to the y-axis indicates the LOD significance threshold. Table S3). Both Vigun02g051100 (WRKY transcription factor 57) and Vigun02g052800 (ring finger and CCH-type zinc finger domain-containing protein) were selected as candidate genes for qFld2.1−. An analysis of qFld5.1− revealed that this QTL spanned 90.5 kb on cowpea chromosome 5 and contained eight genes (Supplementary Table S3), of which Vigun05g043100, encoding a bHLH transcription factor PRE1related protein, was identified as a candidate gene.

DISCUSSION
Compared with wild zombi pea, cultivated zombi pea has undergone phenotypic changes due to domestication. In this study, we identified 62 QTLs controlling 13 domesticationrelated traits in zombi pea. The QTLs were distributed in all LGs, except for LG10. The number of QTLs identified for individual traits varied from one to ten, with an average of five QTLs ( Table 3). Large-effect QTLs (PVE > 15%) were detected for all examined traits, except for STL, PDDM, and BRNPP (Table 3). These results imply that the zombi pea domestication-related traits are controlled by one or only a few major QTLs along with a small number of minor QTLs. This is consistent with the results of earlier investigations of other Vigna species, including mung bean (Isemura et al., 2012), azuki bean (Isemura et al., 2007), rice bean (Isemura et al., 2010), and cowpea (Kongjaimun et al., 2012;Lo et al., 2018). However, in a recent study, Dachapak et al. (2018) identified an average of 2.3 QTLs for zombi pea domesticationrelated traits, which is about twofold less than the number of QTLs detected in our study. This difference may have been due to the sterility of the mapping population used in the earlier study.
In Vigna species, the major QTLs for domestication-related traits, especially organ size, have always been mapped to the same LG (Isemura et al., 2007(Isemura et al., , 2010(Isemura et al., , 2012Kongjaimun et al., 2012;Lo et al., 2018;Yundaeng et al., 2019). In our study, the major QTLs for the domestication-related traits of zombi pea, including SD100WT, SDL, PDL, STL, BRNPP, FLD, PDDM, and SDNPPD, were clustered in LG5, corresponding to cowpea chromosome 5. In an earlier study of cowpea, which is the legume crop most closely related to zombi pea, the major QTLs for domestication-related traits, such as pod shattering, seed dormancy, seed size (weight, width, and length), pod size (length and width), stem size (thickness and length), and leaf size (width and length), were clustered in LG7 (Kongjaimun et al., 2012), which corresponds to chromosome 5. These findings suggest that the genetics underlying the effects of domestication on the organ size (e.g., seed size, pod length, and stem length) of these two African Vigna crops may be the same. However, the genetic basis of seed dormancy and pod shattering, which are key domestication-related traits, varies between these two species.
Domestication results in a decrease in or a loss of seed dormancy, enabling the uniform germination of seeds, which is relevant for agricultural production. The seed dormancy of cultivated zombi pea is suppressed. In this study, we identified one major and two minor QTLs for the seed dormancy-related trait (Table 3). Specifically, qSdwa3.1− was detected as the major QTL, accounting for 22.2% of the phenotypic variation. Because the wild alleles at qSdwa3.1− positively affected SDWA (i.e., decreased dormancy), their introgression into cultivated zombi pea may be useful for decreasing seed dormancy. The Vigun03g339300 gene, encoding a calmodulin-binding proteinlike protein, was identified at qSdwa3.1−. Calmodulin (CaM) and CaM-like (CML) proteins are the major Ca 2+ -binding proteins. The CaM/CML signals are involved in the ABAinduced inhibition of seed germination and seedling growth (Zhou et al., 2018). In Arabidopsis thaliana (Arabidopsis), IQM4, which is a CaM-binding protein, was recently reported to affect seed dormancy by modulating ABA biosynthesis and ABA signaling during seed maturation and germination (Zhou et al., 2018). In cowpea, Sdp1.1+ is the major QTL related to seed dormancy, explaining about 40% of the trait variation (Kongjaimun et al., 2012). The markers flanking Sdp1.1+ suggest that this QTL is located on chromosome 3 between positions 45,527,961 and 51,968,469. Thus, qSdwa3.1− and Sdp1.1+ are different loci. The minor QTL qSdwa5.1+ accounted for about 9% of the phenotypic variation, and included Vigun05g170600 and Vigun05g170700, both of which encode gibberellin 2-betadioxygenase 1 (GA2OX1). Gibberellin is a key plant hormone that regulates seed dormancy and germination by antagonistically suppressing ABA-triggered seed dormancy (Shu et al., 2016). Loss-of-function mutations to GA2ox2, GA2ox7, and GA2ox8, which encode GA2OX, reportedly enhance seed germination in Arabidopsis (Schomburg et al., 2003;Yamauchi et al., 2007).
Preventing or limiting pod dehiscence (pod shattering) may minimize pre-harvest yield losses, resulting in a more efficient harvest. Thus, pod indehiscence may be an advantageous trait during the harvesting of seeds, making it an important consideration during crop domestication. A major QTL in LG11 and a minor QTL in LG5 were related to pod twisting ( Table 3). The major QTL, qPdt11.1+, explained 17.6% of the dehiscence variation, and included two genes, Vigun11g053700, encoding UGE1, and Vigun11g054200, encoding the glycosyltransferase GT14. Earlier studies confirmed that UGE is a nucleotide sugar interconversion enzyme that mediates the interconversion of UDP-D-glucose and UDP-D-galactose and contributes to cell wall biosynthesis (Barber et al., 2006;Rösti et al., 2007). Glycosyltransferases in plants are mainly involved in the biosynthesis of cell wall polysaccharides and glycoproteins (Hansen et al., 2012). Although the functions of plant GT14 remain uncharacterized, two GT14 genes identified in a hybrid aspen species (Populus tremula × P. tremuloides) reportedly exhibit xylem-specific expression (Aspeborg et al., 2005). Accordingly, GT14 may participate in secondary cell wall biosynthesis. In cowpea, three major QTLs for pod shattering have been identified (Kongjaimun et al., 2012;Suanum et al., 2016;Lo et al., 2018). Candidate genes for these QTLs include Vigun02g095200 (cellulose synthase), Vigun03g306000 (NAC domain transcription factor), Vigun03g302600 (C2H2 zinc finger protein), Vigun05g262100 (MYB46 transcription factor), Vigun05g273500 (MYB26 transcription factor), and Vigun05g266600 (beta glucosidase) (Suanum et al., 2016;Lo et al., 2018;Takahashi et al., 2019;Watcharatpong et al., 2020). Thus, the major QTLs/genes which control pod shattering differ between zombi pea and cowpea.
In most cases, domestication increases organ sizes, especially the harvested or consumed or exploited organs. Seed size is a key determinant of seed yield in legume crops as well as evolutionary fitness. It is a polygenic trait affected by various cellular processes. Twenty-two QTLs were identified for seed size-related traits, of which the QTLs with the largest effects on seed weight (qSd100wt5.1+) and seed length (qSdl5.1+) were mapped to the same position of LG5 (Table 3). Two genes were identified at these two QTLs, namely Vigun05g139100, encoding an auxin-responsive protein-related protein, and Vigun05g140300, encoding DUF538. Auxin is important for seed development, with specific effects on the embryo, endosperm, and seed coat, all of which influence seed size and yield (Cao et al., 2020). Genes encoding auxin-related proteins control seed size (Schruff et al., 2006;Ishimaru et al., 2013;Liu et al., 2015;Sun et al., 2017). In Arabidopsis mutant line F08314, the overexpression of RAFL06-12-J05 (encoding DUF538) results in seeds that are larger than the wild-type seeds (Akiyama et al., 2014) 3 . In addition to qSd100wt5.1+, qSd100wt5.2+ also substantially affects seed weight. We identified Vigun05g277600, which encodes an IAAamino acid hydrolase ILR1-like 1-related protein, as a candidate gene for qSd100wt5.2+. Earlier studies revealed ILR1 is involved in auxin conjugation and affects auxin homeostasis (Rampey et al., 2004;and see Cao et al., 2020 for review). In rice, TGW3, encoding an IAA-glucose hydrolase, controls the rice grain weight and yield (Ishimaru et al., 2013). The QTL with the largest effect on seed length, qSdl9.1+, was included in LG9. Both Vigun09g215300 (WRKY transcription factor 40related protein) and Vigun09g215400 (WRKY DNA-binding protein) were detected as candidate genes for qSdl9.1+. In plants, WRKY proteins have diverse biological functions related to seed development as well as developmental and hormonecontrolled processes (Bakshi and Oelmüller, 2014). For example, MINISEED3, which encodes WRKY10, regulates seed size in Arabidopsis (Luo et al., 2005) and Loose Panicle1, which encodes a WRKY transcription factor, regulates seed size by increasing the seed length and width in foxtail millet [Setaria italica (L.) P.
In the current study, qLfa9.1−, with a PVE of 18.2%, was the only QTL identified for leaf size ( Table 3). The Vigun09g018300 gene (WD40 repeat-containing protein) was detected as a candidate gene for this QTL. In cucumber, LITTLELEAF, which encodes a WD40 repeat domain−containing protein, regulates the size of various organs, including leaves (Yang et al., 2018). We also identified Vigun09g017000, encoding the bromodomaincontaining protein GTE1, as a candidate gene. In Arabidopsis, GTE4 and GTE6 encode bromodomain-containing proteins that regulate leaf shape (and thus leaf size). Specifically, GTE4 controls the mitotic cell cycle during plant development (Airoldi et al., 2010), whereas GTE6 induces the expression of ASYMMETRIC LEAVES1, encoding a MYB-domain protein that regulates proximodistal leaf patterns (Chua et al., 2005).
Legume domestication results in increased pod length. Five QTLs were identified for pod length, but qPdl5.1− was the only major QTL, accounting for nearly 38% of the trait variation ( Table 3). Pod elongation requires changes to fiber cell structures and components. The Vigun05g125800 gene, encoding pectate lyase 1, was selected as a candidate gene for qPdl5.1−. Pectin is a major polymer in the plant primary cell wall. Pectate lyase is a pectin-degrading enzyme that can degrade demethylesterified homogalacturonan (HG), which is a pectic polymer (Palusa et al., 2007). The methylesterification status of HG can regulate cellular growth and cell shape, thereby influencing plant growth and development (Wolf et al., 2009;Peaucelle et al., 2011). A previous study regarding common bean (Phaseolus vulgaris L.), which is a legume closely related to zombi pea, revealed that the major changes in cell wall polysaccharides during pod development involve pectic polymers and that the deposition of galactose-rich pectic polymers in cell walls increases in elongating pods (Stolle-Smits et al., 1999). Moreover, during pod elongation, the HG level steadily increases. Accordingly, pectate lyase may help mediate the elongation of zombi pea pods. In rice, a mutation to DEL1, which encodes a pectate lyase precursor, decreases the total pectate lyase activity, leading to an increase in the methylesterified HG content and altered cell wall compositions and structures, with implications for plant height, root length, flag leaf length, grain size, and panicle and internode lengths. In cotton, the downregulated expression of pectate lyase genes GhPEL48_Dt and GhPEL76 decreases fiber length by impeding pectin degradation (Wang et al., 2009;Sun et al., 2020).
Most domesticated legume crops exhibit greater determinate growth, with less branching, than their wild ancestors. The qStl9.1− QTL explained about 10% of the variability in the stem length (plant height) ( Table 3). We identified Vigun09g037900, encoding MOB1, as a candidate gene for this QTL. Arabidopsis MOB1 proteins are critical for plant development because they promote auxin signaling  and regulate jasmonate accumulation (Guo et al., 2020). Mutations to the MOB1 gene in Arabidopsis negatively affect plant height (Xiong et al., 2016;Guo et al., 2020). We identified Vigun05g082200, encoding an auxin efflux carrier protein, as a candidate gene for qBrnpp5.1−, which controls the number of branches in zombi pea. The Vigun05g082200 sequence is similar to that of the PIN-LIKES 6 transporter (AT5G01990). The PIN proteins form a family of auxin efflux carriers. Auxin is well known for its ability to regulate numerous plant growth and developmental processes such as meristem formation and lateral organ formation and patterning (Wang and Jiao, 2018). For example, PIN6 contributes to auxin signaling-mediated developmental processes, including lateral/adventitious root organogenesis, primary/lateral root development and growth, and shoot apical dominance (Cazzonelli et al., 2013;Nisar et al., 2014;Simon et al., 2016). In rice, PIN1 and PIN2 are involved in tillering (Xu et al., 2005;Chen et al., 2012).
Generally, wild ancestors of domesticated crops are sensitive to short-day conditions. In this study, we identified four QTLs related to days-to-flowering in zombi pea, of which two are large-effect QTLs, qFld2.1− and qFld5.1− (PVE of 16.5 and 18.6%, respectively). We selected Vigun02g051100 (WRKY57 transcription factor) and Vigun02g052800 (RING finger and CCCH-type zinc finger protein) as candidate genes for qFLD2.1− and qFLD5.1−, respectively. The WRKY transcription factors are involved in physiological changes and responses to biotic and abiotic stresses. In Arabidopsis, WRKY12, WRKY13, WRKY71, and WRKY75 regulate flowering time (Li W. et al., 2016;Yu et al., 2016;Zhang et al., 2018). Additionally, WRKY75 interacts with DELLA proteins, which are the major components involved in gibberellic acid perception and signaling (Zhang et al., 2018). A mutation to WRKY75 reportedly delays flowering, whereas the overexpression of this gene has the opposite effect (Zhang et al., 2018). Plant zinc finger proteins comprise a large protein family and have diverse functions related to plant development and resistance to biotic and abiotic stresses. The Arabidopsis genes AtC3H17, AtZFP1, AtKHZ1, and AtKHZ2 as well as the Medicago sativa gene MsZFN, encoding CCCH zinc finger proteins, influence flowering time (Chao et al., 2014;Seok et al., 2016;Yan et al., 2017;. Four QTLs were detected for days-to-flowering in zombi pea, of which qPddm2.1− and qPddm5.1− had a similar PVE of about 10% (Table 3). We identified Vigun02g050200, encoding the 14-3-3-like protein GF14 IOTA, and Vigun05g050300, encoding 3-deoxy-8-phosphooctulonate synthase/phospho-2-keto-3-deoxyoctonate aldolase, as the candidate genes for qPddm2.1− and qPddm5.1−, respectively. The 14-3-3 (GF14) proteins are critical components of various cellular signaling processes and are important regulators of several physiological processes in plants. The GF14 genes are associated with fruit development and/or ripening in banana (Musa acuminata L.) (Li et al., 2012;Li M. et al., 2016) and pear (Pyrus pyrifolia Nakai.) (Shi and Zhang, 2014). Phospho-2-keto-3-deoxyoctonate aldolase is also known as 3-deoxy-d-manno-2-octulosonic acid-8-phosphate synthase (KDO8P), which has a key function related to the biosynthesis of KDO, which is a sugar present only in the rhamnogalacturonan II pectic fraction of the primary cell walls of higher plants (Delmas et al., 2003). In the common bean, which is a legume closely related to zombi pea, pod elongation and maturation are accompanied by changes to rhamnogalacturonan (Stolle-Smits et al., 1999).
The number of seeds per pod is one of the yield determinants of legumes. Some domesticated legumes have more seeds per pod than their wild ancestors. Up to nine QTLs were detected for this trait in zombi pea (Table 3). Among these QTLs, qSdnppd3.1−, qSdnppd5.1−, qSdnppd7.1−, and qSdnppd7.2+ had relatively large effects (PVE between 15 and 33%). Because the wild alleles at qSdnppd3.1−, qSdnppd5.1−, and qSdnppd7.1− were observed to increase the number of seeds per pod, they may be useful for increasing the seed yield of cultivated zombi pea. We identified Vigun03g187300 (ABA-insensitive 5-like protein 6) as a candidate gene for qSdnppd3.1−. The encoded protein is an ABA-responsive element (ABRE)-binding factor that regulates ABRE-dependent gene expression (Nakashima and Yamaguchi-Shinozaki, 2013). In Arabidopsis, an ABA-deficiency reportedly decreases the number of seeds per silique (Cheng et al., 2014). The Vigun05g126900 gene, encoding MALE STERILE 5, was selected as a candidate gene for qSdnppd5.1−. In an earlier investigation on Arabidopsis, mutations to MS5, which encodes MALE STERILE 5, resulted in the formation of "polyads" (i.e., tetrads with more than four pools of chromosomes following male meiosis) (Glover et al., 1998). Plants that are homozygous for the ms5 recessive allele reportedly exhibit stunted growth and produce empty siliques, whereas in plants that are heterozygous for MS5, silique elongation and seed set are less inhibited (Glover et al., 1998). We identified Vigun07g094201 (B3-like transcription factor) and Vigun07g096300 (B3 DNA-binding domain protein) as the candidate genes for qSdnppd7.1−. The B3 domain family proteins are associated with ovule development (Skinner and Gasser, 2009). A recent study on Arabidopsis investigated the roles of three Reproductive Meristem (REM) genes, REM34, REM35, and REM36, which encode B3 domain family transcription factors affecting female and male gametophyte development (Caselli et al., 2019). This previous study proved that the silencing of one or more of these genes via RNA interference increases the number of unfertilized ovules, thereby decreasing the number of seeds per silique. Additionally, the Vigun07g094201 sequence is highly similar to that of Arabidopsis REM39 (AT3G18990). In the current study, Vigun07g074900, which encodes the transcription factor TCP21, was selected as a candidate gene for qSdnppd7.2+. In Arabidopsis, several TCP transcription factors, including TCP21, are involved in ovule development via their interactions with SPOROCYTELESS (Chen et al., 2014;Wei et al., 2015), which is required for sporogenesis in both male and female organs (Yang et al., 1999). An earlier study involving cowpea identified a gene encoding TCP5 as a candidate gene for the number of seeds per pod (Lo et al., 2018).

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
PS and LW conceptualized the study and secured research funding. PS designed and supervised the study and revised the manuscript. KA and KL developed population and conducted data collection and analysis. PS and KA wrote the manuscript. All authors read and approved the accepted manuscript.

ACKNOWLEDGMENTS
We are thankful to the Joint Legume Research Center between Chinese Academy of Agriculture Sciences and Kasetsart University for molecular lab facilities. We thank Liwen Bianji, Edanz Editing China (www.liwenbianji.cn/ac) for editing the English text of a draft of the manuscript.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fgene. 2020.00803/full#supplementary-material TABLE S1 | Number of markers, marker length, and distance between markers in each linkage group of the zombi pea genetic map used for detecting the QTLs of domestication-related traits in the F 2 population derived from a cross between TVNu 240 and TVNu 1623. The markers are SNPs generated by specific locus amplified fragment sequencing. The map was constructed by Amkul et al. (2019).