Unique Salt-Tolerance-Related QTLs, Evolved in Vigna riukiuensis (Na+ Includer) and V. nakashimae (Na+ Excluder), Shed Light on the Development of Super-Salt-Tolerant Azuki Bean (V. angularis) Cultivars

Wild relatives of crops have the potential to improve food crops, especially in terms of improving abiotic stress tolerance. Two closely related wild species of the traditional East Asian legume crops, Azuki bean (Vigna angularis), V. riukiuensis “Tojinbaka” and V. nakashimae “Ukushima” were shown to have much higher levels of salt tolerance than azuki beans. To identify the genomic regions responsible for salt tolerance in “Tojinbaka” and “Ukushima”, three interspecific hybrids were developed: (A) azuki bean cultivar “Kyoto Dainagon” × “Tojinbaka”, (B) “Kyoto Dainagon” × “Ukushima” and (C) “Ukushima” × “Tojinbaka”. Linkage maps were developed using SSR or restriction-site-associated DNA markers. There were three QTLs for “percentage of wilt leaves” in populations A, B and C, while populations A and B had three QTLs and population C had two QTLs for “days to wilt”. In population C, four QTLs were detected for Na+ concentration in the primary leaf. Among the F2 individuals in population C, 24% showed higher salt tolerance than both wild parents, suggesting that the salt tolerance of azuki beans can be further improved by combining the QTL alleles of the two wild relatives. The marker information would facilitate the transfer of salt tolerance alleles from “Tojinbaka” and “Ukushima” to azuki beans.


Introduction
The azuki bean (Vigna angularis) is one of the most important legumes in East Asia, particularly in Japan and China [1]. It is an important source of protein in this region and is cooked with rice in times of celebration. This bean has also been used as an ingredient for making traditional Japanese confectionary (called "Wagashi") and is attracting attention to health-conscious consumers in the U.S.A. and Europe. To expand the production area of this crop, it is necessary to improve its ability to tolerate saline soil. Azuki bean is ranked in the most salt-susceptible group IV [2].
To improve the salt tolerance in azuki beans, identifying promising genes is necessary. In our study, we screened 69 azuki bean landraces together with cross-compatible wild relatives consisting of 150 accessions in 8 species [3]. The primary screening was performed in soil culture, followed by secondary screening in hydroponic culture, neither of which revealed a good source of salt tolerance from azuki bean landraces. In contrast, two accessions, V. riukiuensis "Tojinbaka" and V. nakashimae "Ukushima", were identified as the most promising sources of salt tolerance, with each having different tolerance mechanisms. "Tojinbaka" accumulated Na + throughout the whole plant body (Na + includer), while "Ukushima" prevented Na + accumulation in leaves by filtering Na + in the roots and stems (Na + excluder) [3,4]. In addition, "Ukushima" and "Tojinbaka" showed improved growth in a sea-water-damaged field from a tsunami in Fukushima, Japan, in which the local soybean cultivar, "Tachinagaha", could not survive [3].
From DNA barcoding sequence analysis of rDNA-ITS and atpB-rbcL by Takahashi et al. [5], it was observed that V. riukiuensis and V. nakashimae are different species but are genetically closely related. V. riukiuensis is adapted to the coastal saline environment in the Ryukyu islands, Japan, whereas V. nakashimae is adapted to the northeastern region of East Asia and is mainly found in and around the Korean Peninsula. Among the V. nakashimae populations, some populations such as "Ukushima" adapted to the coastal saline environment and acquired salt tolerance [6]. V. riukiuensis and V. nakashimae are considered to have evolved from the common ancestral species V. minima [7]. V. minima originated in mainland Southeast Asia, and V. riukiuensis and V. nakashimae might have evolved from V. minima in the course of the expanding geographical distribution to East Asia. The time of divergence between V. riukiuensis and V. nakashimae was estimated to be only 60,000 years ago [2]. It is interesting that two genetically closely related species, "Tojinbaka" and "Ukushima", have different salt tolerance mechanisms.
Genetic mapping is a powerful tool for detecting the loci involved in various biological processes [8]. The identification of genes that control morphological and physiological traits is crucial for understanding the mechanisms of adaptive evolution and for plant breeding. Simple sequence repeat (SSR) markers are highly reproducible, have codominant inheritance, are relatively abundant, show high levels of polymorphism and are easy to genotype, making them particularly useful for gene characterization and advantageous for mapping studies [9][10][11][12]. Additionally, a genome-wide microsatellite DNA marker database for three species within the Vigna species, including azduki bean, has been reported [13]. Recently, several SSR-based azuki bean linkage maps have been reported [14][15][16].
The advancement of next-generation sequencing (NGS) technologies has accelerated whole-genome sequencing and genotyping methods. Whole-genome sequences for 14 Vigna species have been constructed, including azuki bean (V. angularis) [17][18][19][20][21][22][23]. Genotyping using high-throughput NGS technologies reduces the complexity of genome analysis and allows for the development of a large number of SNP markers for any species of interest. Restriction-site-associated DNA sequencing (RAD-seq), including double-digested RADseq (ddRAD-seq), is one of the preferred NGS technologies for high-throughput genotyping. The RAD-seq method, first described by [24], has been used in many species without a reference genome, including the Vigna species V. vexillata (L.) A. Rich., V. angularis and V. unguiculata (L.) Walp [16,[25][26][27]. While RAD-seq can be applied without a reference genome, a reference-guided RAD-seq approach can improve the accuracy of SNP detection. The high-quality Vigna genomes of 10 species including the azuki genome are available in the following database (VigGS: http://viggs.dna.affrc.go.jp/ (accessed on 1 March 2023)) [28]. Because the azuki genome is highly accurate, it can be used to generate draft genomes of closely related species or as a reference genome. The azuki genome can also be used for genetic analysis such as SNP marker detection in closely related wild species.
The objectives of this study were to (1) confirm the differences in salt tolerance mechanisms between "Tojinbaka" and "Ukushima", (2) detect the genetic loci of salt tolerance in "Tojinbaka" and "Ukushima", which could function on the azuki bean genetic background, and (3) investigate the possibility of achieving higher salt tolerance in azuki beans by pyramiding non-allelic QTLs from "Tojinbaka" and "Ukushima".

Map Construction and Linkage Analysis of Population A and B Using SSR Markers
Among the 377 SSR markers screened in each population, 370 (99.2%) and 374 (98.1%) SSR markers amplified a fragment in populations A and B, respectively ( Table 1). The percentage of SSR markers that amplified fragments ranged from 71.4% for cowpea SSR markers to 99.7% for azuki bean SSR markers. Among the 370 and 374 amplified SSR markers, 237 (62.9%) and 285 (75.6%) showed polymorphisms between parents in populations (A) and (B), respectively. The percentage of polymorphic SSR markers ranged from 37.5% for common bean SSR markers to 80.9% for azuki bean SSR markers (Table 1). In total, 193 (for population A) and 313 (for population B) clear polymorphic SSR products were mapped to linkage maps ( Table 2, Figures S1 and S2). Markers could be assigned to 11 linkage groups (LGs), which coincide with the haploid chromosome number. The linkage maps spanned 560.2 cM for population A and 648.6 cM for population B in total length, with a mean distance between adjacent markers of 3.4 and 2.4 cM, respectively. In population A, the number of markers in the LGs ranged from 10 (LG7) to 41 (LG1) and the length of the LGs ranged from 31.8 cM (LG7) to 108.8 cM (LG1). In population B, the number of the markers ranged from 18 (LG9) to 59 (LG1) and the length of the LGs ranged from 35.5 cM (LG11) to 82.3 cM (LG1).

Map Construction and Linkage Analysis of Population C Using RAD Markers
In total, 13,052,980 RAD-tag 85 bp reads of "Ukushima" were mapped onto the azuki genome to construct a custom pseudo-reference "Ukushima" genome (Table S1). Further, 6,632,164 reads of "Tojinbaka" and 515,437,838 reads of F 2 individuals were mapped onto the "Ukushima" genome. The average number of reads per F 2 individual was 1,662,703. These mapped reads were used in SNP detection. In total, 74.1% ("Ukushima"), 62.0% ("Tojinbaka") and 55.4% (F 2 ) of the total number of reads were mapped onto the custom "Ukushima" genome. The mean coverage depths were 36.8 ("Ukushima"), 32.4 ("Tojinbaka") and 17.4 (F 2 ). The total number of detected SNPs was 25,350. Among them, 15,438 sites supported by Ukushima reads were reference-type alleles in Ukushima, and 5,469 of them were polymorphic in Tojinbaka. These sites were selected as bi-allelic RAD Plants 2023, 12, 1680 4 of 17 markers and were used in subsequent analysis. In F 2 , 2485 sites in Chr. 1 to 11 and 19 sites in scaffolds were available (Table S1). A total of 625 markers could be assigned to 11 linkage groups (LGs) ( Table 2, Figure S3), which coincide with the haploid chromosome number. The LGs spanned 471.8 cM in total length, with a mean distance between adjacent markers of 0.9 cM. The number of markers in the LGs ranged from 55 (LG9) to 175 (LG1) and the length of the LGs ranged from 29.8 cM (LG3) to 67.8 cM (LG1) ( Table 2).

Measurement of Na + Content in Leaves, Stems and Roots of Parental Materials
To confirm the differences in Na + accumulation patterns among "Kyoto Dainagon", "Tojinbaka" and "Ukushima", the Na + concentration in each of the following parts of the plants was measured: the 1st leaf to 5th (newest) leaf, the 1st to 3rd (newest) stem node and the roots. Yoshida et al. reported that Na + concentration was measured, including the wilted leaves, and "Tojinbaka" was characterized as a "Na + includer" and "Ukushima" as a "Na + excluder" under 200 mM and 100 mM NaCl stress [3]. Noda et al. reported that Na + allocation was measured in plants before wilting using radio-isotope 22 Na + with 100 mM NaCl treatment under artificial environmental conditions [4]. To confirm whether this phenomenon is reproduced in the greenhouse environment where the genetic analysis is performed, each parental plant was grown in a salt-treated environment and then divided into roots, stems and leaves to measure Na + concentrations.
In general, the findings of Yoshida et al. [3] and Noda et el. [4] were confirmed. "Tojinbaka" was once again observed as a "Na + includer", while "Ukushima" showed a "Na + excluder" type of Na + accumulation pattern ( Figure 1). In all of the plant parts, Na + concentrations in "Tojinbaka" were higher than in "Ukushima". However, it was revealed that "Tojinbaka" also had a Na + filtering ability, since the Na + concentration became lower in stems, particularly in the upper leaves, compared to the roots. On the contrary, "Kyoto Dainagon" did not show any Na + filtering ability, even under mild salt stress, and accumulated more Na + in the upper leaves.
"Tojinbaka" continued to produce 4th and 5th leaves, even after a considerable amount of Na + had accumulated in the leaves (Figure 1). On the contrary, "Ukushima" stopped growing despite lower Na + concentrations in the leaves relative to "Tojinbaka". excluder" type of Na accumulation pattern ( Figure 1). In all of the plant parts, N centrations in "Tojinbaka" were higher than in "Ukushima". However, it was r that "Tojinbaka" also had a Na + filtering ability, since the Na + concentration becam in stems, particularly in the upper leaves, compared to the roots. On the contrary, Dainagon" did not show any Na + filtering ability, even under mild salt stress, an mulated more Na + in the upper leaves.
"Tojinbaka" continued to produce 4th and 5th leaves, even after a consi amount of Na + had accumulated in the leaves (Figure 1). On the contrary, "Uku stopped growing despite lower Na + concentrations in the leaves relative to "Tojinb Figure 1. Na + concentrations in different plant tissues of azuki bean "Kyoto Dainagon" and t tolerant wild relatives, V. riukiuensis "Tojinbaka" and V. nakashimae "Ukushima". Leaves vided by stem node position. Stems were divided into three parts: 1 is base to primary nod top to two nodes from the end of stem; "2" is the middle part of stem other than 1 and 3. were collected 6 days after 100 mM NaCl treatment (9 days after salt treatment). All leaves w green at sampling. Error bars indicate standard deviations (n = 3). (ppm: parts per million).

QTL Analysis
2.4.1. Positive and Negative Effects of "Tojinbaka" Alleles on Salt Tolerance in QT tected from Population A (V. angularis "Kyoto Dainagon" × V. riukiuensis "Tojinba In population A, 119 plants (62.6%) survived for one month under 100 mM conditions ( Figure 2A), but only 24 F2 plants (12.6%) had no wilting leaves like "Toji ( Figure 2D). No F2 plants died earlier than "Kyoto Dainagon". QTL analysis id three QTLs for days to wilt (qDtW1-1, qDtW2-1 and qDtW8-1) and three QTLs for p age of wilt leaves (qPWL1-1, qPWL2-1 and qPWL8-1) in population A (Table 3). Th for DtW and PWL on Chr 1 showed the highest LOD and PVE values. qDtW1explain 15.0% of the variation in DtW, and qPWL1-1 could explain 19.1% of the va in PWL. QTLs on Chr 8 showed the second-highest LOD and PVE values. qDtW8explain 12.0% of the variation and qPWL8-1 could explain 15.9%. Each of the thre for DtW and PWL detected at the adjacent region on the same chromosome (Ch considered to be the same allele because of the close mapped positions and the s rection of the allelic effect (Table 3, Figure 3). "Tojinbaka" alleles increased salt to Figure 1. Na + concentrations in different plant tissues of azuki bean "Kyoto Dainagon" and two salt-tolerant wild relatives, V. riukiuensis "Tojinbaka" and V. nakashimae "Ukushima". Leaves were divided by stem node position. Stems were divided into three parts: 1 is base to primary node; "3" is top to two nodes from the end of stem; "2" is the middle part of stem other than 1 and 3. Samples were collected 6 days after 100 mM NaCl treatment (9 days after salt treatment). All leaves were still green at sampling. Error bars indicate standard deviations (n = 3). (ppm: parts per million).

Positive and Negative Effects of "Tojinbaka" Alleles on Salt Tolerance in QTLs
Detected from Population A (V. angularis "Kyoto Dainagon" × V. riukiuensis "Tojinbaka") In population A, 119 plants (62.6%) survived for one month under 100 mM NaCl conditions ( Figure 2A), but only 24 F 2 plants (12.6%) had no wilting leaves like "Tojinbaka" ( Figure 2D). No F 2 plants died earlier than "Kyoto Dainagon". QTL analysis identified three QTLs for days to wilt (qDtW1-1, qDtW2-1 and qDtW8-1) and three QTLs for percentage of wilt leaves (qPWL1-1, qPWL2-1 and qPWL8-1) in population A ( Table 3). The QTLs for DtW and PWL on Chr 1 showed the highest LOD and PVE values. qDtW1-1 could explain 15.0% of the variation in DtW, and qPWL1-1 could explain 19.1% of the variation in PWL. QTLs on Chr 8 showed the second-highest LOD and PVE values. qDtW8-1 could explain 12.0% of the variation and qPWL8-1 could explain 15.9%. Each of the three QTLs for DtW and PWL detected at the adjacent region on the same chromosome (Chr) was considered to be the same allele because of the close mapped positions and the same direction of the allelic effect (Table 3, Figure 3). "Tojinbaka" alleles increased salt tolerance at QTLs detected on Chr 1 and Chr 2, while the "Tojinbaka" allele decreased salt tolerance at a QTL detected on Chr 8. at QTLs detected on Chr 1 and Chr 2, while the "Tojinbaka" allele decreased salt tolerance at a QTL detected on Chr 8.   at QTLs detected on Chr 1 and Chr 2, while the "Tojinbaka" allele decreased salt tolerance at a QTL detected on Chr 8.
2.4.3. "Tojinbaka" Allele Increases Salt Tolerance at QTLs Detected from Population C (V. nakashimae "Ukushima" × V. riukiuensis "Tojinbaka") To clarify the combined effects of salt tolerance QTLs from "Tojinbaka" and "Ukushima", F 2 population C was developed. As expected, most F 2 plants from population C exhibited higher salt tolerance compared to plants from populations A and B ( Figure 2). Therefore, the NaCl concentration was increased from 200 mM to 250 mM 25 days after salt treatment began. Under 250 mM salt stress conditions, the wild parent with the higher salt tolerance, "Tojinbaka", could survive up to 55 days after salt treatment ( Figure 2C). In addition, strong transgressive segregation was observed in population C. Among the 348 F 2 segregating plants, 72 plants (24.0%) could survive longer than "Tojinbaka". Strong transgressive segregation was also observed for PWL. One hundred F 2 plants (32.8%) showed lower PWL compared to "Tojinbaka" ( Figure 2F). Of these, 15 F 2 plants (12.1%) showed no wilting leaves (PWL = 0).
Three QTLs for DtW (Chr 4, 5, 11) and two QTLs for PWL (Chr 4, 11) were detected in population C (Table 3). At all of the detected QTLs, "Tojinbaka" alleles increased salt tolerance. The QTLs detected on Chr 4 for DtW (qDtW4-2) and PWL (qPWL4-2) were positioned close to one another and were considered to be the same locus (Table 3, Figure 3). QTLs on Chr 4 showed the highest LOD and PVE values. This QTL explained 11.8% of the variation in DtW and 10.8% of the variation in PWL.
2.4.4. Transgressive Segregation for Na + Contents in Leaves in Population C (V. nakashimae "Ukushima" × V. riukiuensis "Tojinbaka") Since "Ukushima" is considered as a "Na + excluder" and "Tojinbaka" is a "Na + includer" (Figure 1), the Na + concentration in the primary leaves of the F 2 population was measured 17 days after the commencement of salt treatment for QTL analysis. The Na + concentration in the primary leaf of "Ukushima" was still lower than 250 ppm at that time ( Figure 4). However, "Ukushima" wilted completely (dying) three days after the Na + concentration was measured (20 days after salt treatment). The Na + concentration in the primary leaf of "Tojinbaka" was 2166 ppm, and this accession could survive up to 55 days under 250 mM saline conditions (Figure 2). In the F 2 population, strong transgressive segregation was also observed in Na + accumulation. There was a weak but significant correlation between Na + concentration in primary leaves 17 days after salt treatment and days to wilt (p < 0.0001) in the F 2 population ( Figure S4).

Discussion
In this research, three interspecific hybrids were developed, three l on SSR or RAD markers were constructed using F2:3 or F2 derived from three parents, Vigna angularis "Kyoto Dainagon", V. riukiuensis "T nakashimae "Ukushima", and QTL analysis was conducted. All ma groups ( Table 2, Figures S1-S3), which corresponded to the haploid chr of these species [29]. Variant detection using RADseq allowed us to ob at once without prior screening, including SSR markers (Table 1). Altho be able to obtain enough polymorphic markers in very genetically c breeding cultivars) inbred populations within a species [30], we recon very powerful tool for large-scale genetic analysis in Vigna interspeci trast, small-scale marker analysis such as marker-assisted selection (M ers that can be analyzed individually, such as SSR markers. Therefore develop a small number of SNP markers that can be used in the futu the MAS to introduce salt-tolerant QTLs in azuki beans.
The different Na + allocation patterns in three accessions used in th viously reported by Yoshida et al. [3] and Noda et al. [4]. Yoshida et al concentrations in leaves, stems and roots under 200 mM salt concent after salt treatment, in which "Kyoto Dainagon" completely died [3]. that "Ukushima" was the excluder type, with low Na + concentrations fr leaves, while "Tojinbaka" was an includer with a high Na + concentrati the leaf. However, since 200 mM NaCl is fairly strong salt stress for azu tolerant species, Na + distributions were observed in dying plants. Noda the radioisotope 22 Na + distribution under 100 mM salt stress conditio days after salt stress with radioactive 22 Na + , they exposed whole plants 22 +

Discussion
In this research, three interspecific hybrids were developed, three linkage maps based on SSR or RAD markers were constructed using F 2:3 or F 2 derived from the combination of three parents, Vigna angularis "Kyoto Dainagon", V. riukiuensis "Tojinbaka" and V. nakashimae "Ukushima", and QTL analysis was conducted. All maps had 11 linkage groups ( Table 2, Figures S1-S3), which corresponded to the haploid chromosome number of these species [29]. Variant detection using RADseq allowed us to obtain many markers at once without prior screening, including SSR markers (Table 1). Although RAD may not be able to obtain enough polymorphic markers in very genetically close (e.g., between breeding cultivars) inbred populations within a species [30], we reconfirmed that it is a very powerful tool for large-scale genetic analysis in Vigna interspecific crosses. In contrast, small-scale marker analysis such as marker-assisted selection (MAS) requires markers that can be analyzed individually, such as SSR markers. Therefore, it is necessary to develop a small number of SNP markers that can be used in the future at a low cost for the MAS to introduce salt-tolerant QTLs in azuki beans.
The different Na + allocation patterns in three accessions used in this study were previously reported by Yoshida et al. [3] and Noda et al. [4]. Yoshida et al. measured the Na + concentrations in leaves, stems and roots under 200 mM salt concentrations for 16 days after salt treatment, in which "Kyoto Dainagon" completely died [3]. The authors found that "Ukushima" was the excluder type, with low Na + concentrations from the roots to the leaves, while "Tojinbaka" was an includer with a high Na + concentration from the root to the leaf. However, since 200 mM NaCl is fairly strong salt stress for azuki and related salt-tolerant species, Na + distributions were observed in dying plants. Noda et al. investigated the radioisotope 22 Na + distribution under 100 mM salt stress conditions [4]. Three or six days after salt stress with radioactive 22 Na + , they exposed whole plants to a storage phosphor screen for 24 h to visualize the 22 Na + distribution patterns in plants and then separated the root, stem and leaves and dried the samples to measure the 22 Na + radioactivity. At 3 days under 100 mM salt treatment, "Ukushima" and "Tojinbaka" mainly accumulated 22 Na + in the stem, while "Ukushima" maintained lower 22 Na + in the leaves compared with "Tojinbaka" [4]. Six days after 100 mM salt treatment, "Ukushima" showed 22 Na + filtering ability by the roots, resulting in the highest 22 Na + concentration in the roots and lowest 22 Na + concentration in the leaves. These results indicate that the Na + allocation varies with salt concentration and salt stress period. In the present study, we cultivated these three accessions for 6 days under 100 mM salt treatment after 3 days under 50 mM salt treatment, dried them immediately and measured the Na + concentration in each plant part based on the dry weight. Our results showed that the Na + concentration was high and almost the same from the root to the first leaf in "Kyoto Dainagon". This result indicates that "Kyoto Dainagon" does not have any Na + filtering ability under 9 days of salt treatment (50 mM 3 days and 100 mM 6 days). "Ukushima" and "Tojinbaka" still retain the ability to filter Na + from the stems and leaves upon 9 days of salt treatment, while the filtering ability was higher in "Ukushima". This suggests that QTL analysis using populations A and B could detect salt-tolerant QTLs related to Na + filtering ability from the stem and the leaf. Noda et al. [4] speculated that this mechanism of filtering and reducing the Na + concentration in leaves might rely on Na + /K + homeostasis in "Ukushima" and not in "Tojinbaka". "Kyoto Dainagon" and "Ukushima" stopped growing after salt treatment started, whereas "Tojinbaka" could maintain growth even with Na + accumulation in the leaves. The uppermost leaves of all accessions were the 3rd leaf at the beginning of salt treatment but only "Tojinbaka" could develop the 5th leaf when the Na + concentration in each plant tissue was measured (9 days after salt treatment, Figure 1). This strongly suggests that "Tojinbaka" has a mechanism not only to filter Na + allocation to the leaves, but also has a mechanism to detoxify Na + in the leaves. In contrast, "Ukushima" may not have a mechanism to detoxify Na + because it stopped growing, even though the Na + concentration in leaves was lower compared to "Tojinbaka" (Figure 1). Thus, in population C, the detected QTLs might be related to filtering ability in reducing Na + allocation to the leaves, as well as QTLs related to the Na + detoxification mechanism. In addition, in population A, it is also possible to detect QTLs related to the detoxification mechanism.
Generally, the detoxification mechanism is considered to be related to the vacuolar membrane transport systems, which lead to Na + sequestration in vacuoles [31][32][33][34][35][36][37][38][39][40][41]. However, Noda et al. reported that there was no difference in the development and features of vacuoles between salt-sensitive V. angularis cv. "Erimoshouzu" and "Tojinbaka" after 3 days of 100 mM salt (NaCl) treatment with 5 kBq 22 Na + [42]. The only difference they could find between "Erimoshouzu" and "Tojinbaka" was that "Tojinbanka" contained well-developed starch granules in the chloroplast, whereas "Erimoshouzu" did not. In addition, 22 Na + was co-localized around starch granules developed in chloroplasts in the leaves of "Tojinbaka" and was not detected in other parts in the leaf cell. Therefore, they speculated that "Tojinbaka" has a Na + trapping system using Na + -binding starch granules, which may be related to salt tolerance [42]. Common reed (Phragmites australis), a well-known salt-tolerant plant, also produced a large number of Na + -binding starch granules at the shoot base when salt stressed, and it was proposed as a novel salt tolerance mechanism not reported before [43]. The QTLs detected in population A might be related to genes involved in the formation of Na + -binding starch granules in the chloroplasts of the leaf cells, which could be part of the mechanism for salt tolerance in "Tojinbaka". Interestingly, Noda et al. reported that "Ukushima" also contained starch granules in the chloroplasts [42]. However, they could not observe the co-localization of 22 Na + with starch granules in the chloroplasts of "Ukushima". Therefore, they proposed the possibility that the starch granules in "Ukushima" may lack Na + binding ability. Further studies are needed to confirm this hypothesis.
Four QTLs related to Na + filtering were detected based on the Na+ concentration in the primary leaf in population C (Table 3). Among the four QTLs, the "Tojinbaka" alleles of QTLs (qNa8-1 and qNa11-1) increased the Na + concentration. In population B, no salt-tolerance-related QTL was detected in these two regions. In population A, salttolerance-related QTLs were detected near the qNa8-1 region as qDtW8-1 and qPWL8-1, and no polymorphic markers were present in the qNa11-1 region (Figure 3). Thus, these two QTLs are likely due to the effect of the "Tojinbaka" allele on increasing Na + concentration in the primary leaf. The other two QTLs of "Tojinbaka" alleles (qNa2-1 and qNa8-2) decreased the Na + concentrations in the primary leaf. This indicated that "Tojinbaka", which was characterized as a "Na + includer" in previous studies [3,4], had QTLs with opposite effects of suppressing Na + accumulation in the leaves. Surprisingly, these two "Tojinbaka" alleles at these two QTLs with Na + filtering effects (qNa2-1 and qNa8-2) showed higher LOD (8.1 and 7.1) and PVE (11.1% and 9.7%) values (Table 3). In contrast, no QTLs were detected for the effect of the "Ukushima" allele to reduce the Na + concentration in primary leaves. These results suggest that the effect of Na + filtering genes in "Ukushima" might be small enough to be detected as QTLs or might be obscured by interactions with other QTLs. It is also possible that "Ukushima" and "Tojinbaka" have the same genes (QTLs) for Na + filtering from leaves.
The "Tojinbaka" allele in the QTL region (qNa8-1) increased the Na + concentration in population C and decreased the salt tolerance (qDtW8-1 and qPWL8-1) in population A (Table 3, Figure 3). These three QTLs mapped around the 5 Mb region on Chr. 8, detected in both populations A and C, could be considered as the same locus and are thought to have decreased salt tolerance in population A due to increased Na + in leaves by the "Tojinbaka" allele. In population A, a genetic interaction was observed between QTLs qDtW8-1/qPWL8-1 and qDtW1-1/qPWL1-1 ( Figure S5). When the genotype of qDtW8-1/qPWL8-1 has the "Tojinbaka" allele (BB), which is effective on increased Na + accumulation in leaves, DtW dramatically decreased when the genotype of qDtW1-1/qPWL1-1 has the "Kyoto Dainagon" allele (AA). However, even when the genotype of qDtW8-1/qPWL8-1 has the "Tojinbaka" allele (BB), which is effective on increased Na + accumulation in leaves, DtW does not decrease when the genotype of qDtW1-1/qPWL1-1 carries the "Tojinbaka" allele (BB). This phenomenon suggests that the "Tojinbaka" allele of qDtW1-1/qPWL1-1 might be involved in the detoxification of Na + in the leaves. It is possible that these QTLs were involved in the Na + trapping system using starch granules [42]. In population C, the "Tojinbaka" allele of qNa8-1 had no significant effect on DtW and PWL but an increased Na + concentration in leaves (Table 3, Figure 3). This suggests that other QTLs are more effective on Na + detoxification than qDtW1-1 in the genetic background of wild species with high Na + filtering ability.

Three Interspecific Hybrid Populations
To develop genetic linkage maps and to conduct QTL analysis, three interspecific hybrid populations were developed: (A) V. angularis "Kyoto Dainagon" × V. riukiuensis "Tojinbaka", (B) V. angularis "Kyoto Dainagon" × V. nakashimae "Ukushima" and (C) V. nakashimae "Ukushima" × V. riukiuensis "Tojinbaka". In all cross combinations, an accession with higher salt tolerance was used as the male parent. For population A, the F 2 and F 2:3 population consisting of 190 individuals was used to develop a genetic linkage map and to phenotype salt tolerance. For population B, the F 2 population consisting of 211 individuals was used to develop a genetic linkage map, and 87 F 2:3 lines (with 1 to 10 plants per line) were used to phenotype salt tolerance. For population C, the F 2 population consisting of 348 individuals was used to develop a genetic linkage map and to phenotype salt tolerance.

Growth Conditions and Salt Stress Treatments
The F 2 plants and F 2:3 lines of hybrid populations A and B were used for salt tolerance phenotyping, respectively. Seeds were sown in trays with Seramis clay granules (SERAMIS ® ) and were supported and kept in a growth chamber at 25 • C for 10 days. Each plant was transplanted to a hydroponic culture in a greenhouse of the Genetic Resources Center, Tsukuba, Japan (36 •  . The final solution was adjusted to an EC of 100 mS/m with water. Three weeks after transplanting, NaCl treatment was initiated. NaCl concentration was 50 mM for the first four days and was increased by 50 mM at four-day intervals until a final concentration of 200 mM was reached. After that, the NaCl concentration was maintained at 200 mM for 16 days (until 28 days after commencement of salt treatment). Accessions that survived the salt treatment were grown under lower salt conditions after the experiment and were used for F 4 seed production.
In the case of hybrid population C, the F 2 seeds were sown and transplanted as described above. Three weeks after transplanting, NaCl treatment was initiated. NaCl concentration was 50 mM for the first three days and was increased by 50 mM at three-day intervals until a concentration of 200 mM was reached. After that, the NaCl concentration was maintained at 200 mM for 16 days (25 days after salt commencement of treatment). The primary leaves were sampled 17 days after salt treatment began, and Na + concentration in the primary leaves was measured by the method later described in "Na + concentration measurement in parental lines". Potentially because both wild parents have high levels of salt tolerance, many F 2 individuals were still healthy 25 days after salt treatment began. Therefore, NaCl concentration was further increased to 250 mM and was maintained at 250 mM until 76 days after commencement of salt treatment.

Measurement of Na + Concentration in Parental Lines
To compare the pattern of Na + accumulation in different plant parts, three seeds each from "Kyoto Dainagon", "Tojinbaka" and Ukushima" were sown and transplanted to hydroponic solution, as described above. Three weeks after transplanting, NaCl treatment was initiated. NaCl concentration was 50 mM for the first four days and was increased to a final concentration of 100 mM. Six days after the NaCl concentration reached 100 mM (nine days after salt treatment began), the leaves, stems and roots were harvested separately. The leaves were further divided according to the stem nodes on which they grew, from the primary node up to the fifth node. The stems were also divided into three parts: (1) from the base of the stem to the node of the primary leaf, (2) from the first node to second node and (3) the remaining part of the stem above the second node.
Measurement of Na + concentration was conducted using the ion meter method [2], with slight modifications. The harvested plant samples with three replications were dried at 80 • C for 24 h and were ground into a powder. Roughly 20 mg of powder was collected into 2 cc tubes, and 700 µL of extraction solution (1 N ammonium acetate) was added and mixed at 90 • C for 1 h. After centrifugation at 7000 rpm (4400× g) for 5 min, 200 µL of supernatant was analyzed with a LAQUA twin Na + meter (HORIBA Advanced Techno Co., Ltd. Kyoto, Japan). Before measurements, the LAQUA twin Na + meter was calibrated using the corresponding calibration solution. All measurement procedures were conducted at room temperature (approx. 20 • C).

Evaluation of Salt Tolerance
Two salt-tolerance-related traits were recorded for each plant or line in populations A and B, respectively. The first trait, called 'Days to Wilt' (DtW), recorded the number of days from salt treatment until the plant wilted. The second trait, known as 'Percentage of Wilt Leaves' (PWL), measured the proportion of leaves that had wilted, a symptom of the plant dying. PWL was visually scored using eight categories: 0% = normal healthy leaves, 15% = 1-15% of leaves wilted, 30% = 16-30% of leaves wilted, 45% = 31-45% of leaves wilted, 60% = 46-60% of leaves wilted, 75% = 61-75% of leaves wilted, 90% = 76-90% of leaves wilted and 100% ≥ 90% of leaves wilted or plant completely dead. PWL scores were recorded 19 days after salt treatment began.
In addition to DtW and PWL, the Na + concentration of primary leaves was measured in population C 17 days after NaCl treatment began (9 days after 200 mM). The method of Na + measurement was the same as described above.

SSR Marker Analysis and Construction of Linkage Maps of Populations A and B
DNA was extracted from young leaf samples using the EZ1 DNA Tissue kit (QIAGEN, Valencia, CA, USA). The extracted DNA samples were used for PCR amplification using 330 azuki bean SSR primer sets [44], 40 common bean SSR primer sets [45,46] and 7 cowpea SSR primer sets [47].
PCR amplification was performed with a 5 µL reaction mixture containing 5 ng DNA template, 10x PCR buffer, 2 mM dNTP, 25 ng of each primer and 0.125 U Ex Taq polymerase enzyme. The thermal cycler protocol consisted of denaturation at 94 • C for 30 s, 40 cycles of 94 • C for 30 s, 55 • C for 60 s and 72 • C for 60 s, followed by 72 • C for 10 min cooling at 10 • C. PCR was performed in multiplex PCR reactions. Subsequently, PCR product size was analyzed on an ABI Prism 3100 genetic analyzer using GENEMAPPER ver. 3.0 (Applied Biosystems). The two linkage maps were constructed using AntMap [48].

Construction of V. nakashimae "Ukushima" Pseudo-Reference Genome for RADseq
To improve the mapping quality of reads from wild species to the azuki genome, a pseudo-custom V. nakashimae "Ukushima" genome was constructed. The raw RAD reads of "Ukushima" were first processed via the Trimmomatic-0.30 to remove the reads with sequencing adaptors and of low quality (phred quality < 5) [49]. Then, the processed reads were mapped onto the azuki genome (Vangularis_v1: http://viggs.dna.affrc.go.jp/ (accessed on 4 August 2015)) using Burrows-Wheeler Aligner (BWA) v.0.7.9a [50]. For local realignment and base quality score recalibration of the mapped reads, the Realign-erTargetCreator and IndelRealigner tools from the GATK v3.0.0 software package [51] were applied. All variants were named using GATK with HaplotypeCaller. The genomic sequence of azuki was replaced with the variant of V. nakashimae using GATK with FastaAlternateReferenceMarker, and the result was used as a custom "Ukushima" pseudo-reference genome.

RADseq Analysis and Construction of Linkage Map for Population C
Genomic DNA was extracted from parents and 348 F 2 plants from population C (V. riukiuensis × V. nakashimae) using the CTAB method [52]. Each genomic DNA sample (100-300 ng) from parents and F 2 individuals was simultaneously digested with BamHI-HF and NlaIII and purified. Adaptor-1 and adaptor-2 were ligated to the digested DNA samples and purified. Adaptor-ligated biotinylated DNA samples were collected using streptavidincoated magnetic beads (Dynabeads M270, Dynal). Adaptor-ligated DNA on the beads was amplified by PCR using Phusion High-Fidelity DNA polymerase (Thermo Fisher Scientific, Waltham, MA, USA) and the adapter primers. The size of the PCR-amplified fragments (smear between 200 and 1000 bp) was checked using electrophoresis in 1% agarose gel. The 96 purified PCR products were pooled and sequenced using the Illumina HiSeq2000 system. Sequence reads were mapped to custom pseudo-reference "Ukushima" genome, and local realignment and SNP detection were carried out as mentioned above. Error correction was performed based on the reference genome sequence.
The genetic linkage group was constructed using R/qtl ver. 1.36.6 software [53]. For potential RAD markers, we selected bi-allelic markers based on parents" data, markers with genotyped plants more than 100 F 2 individuals and dropped segregation distortion markers (chi-squared test, p < 10 −5 ). Suitable values for recombination fractions and LOD scores were estimated using the "est.rf" command. After these steps, initial linkage group construction was carried out using the "Form Linkage Groups" command, with a maximum recombination fraction of 0.25 and a LOD threshold of 15. Robustness of linkage groups was checked using the "plot.rf" command. Marker order within linkage groups was estimated using a program called TMAP [54]. Some RAD markers mapped at loci less than 0.1 cM were not used for linkage map construction.

QTL Analysis
Single QTL analysis was run using the standard interval mapping method, as described in R/qtl [39]. The significance thresholds were calculated using 1000 permutation tests for each phenotype. Only those with an LOD score of three or more were adopted as QTLs in order to achieve a more stringent level. Interactions between QTLs and additional QTLs were examined using the "addint" and "addqtl" functions. The final multiple-QTL model fit was determined by drop-one-qtl analysis using the "fitqtl" function.
To compare QTL positions across the three populations, the physical positions of SSR markers mapped on the linkage maps of the V. angularis × V. riukiuensis and V. angularis × V. nakashimae populations of the azuki reference genome sequence (Vangularis_v1: cultivar Shumari in VigGS: https://viggs.dna.affrc.go.jp/ (accessed on 18 June 2015) [29]) were determined by using BLASTN on the SSR primer sequences. The SSR, RAD markers and detected QTLs were distributed into 11 chromosomes according to their physical positions on the azuki reference genome.

Conclusions
This study demonstrated that the mechanisms underlying salt tolerance vary among azuki beans and their wild relatives. The salt-sensitive azuki bean V. angularis "Kyoto Dainagon" lacked Na + filtering ability, while the salt-tolerant wild relatives V. nakashimae "Ukushima" and V. riukiuensis "Tojinbaka" were able to maintain lower Na + concentrations in their leaves and had Na + filtering ability. V. nakashimae had particularly strong Na + filtering ability, but growth was stopped after salt treatment. In contrast, although V. riukiuensis accumulated Na + in its leaves (about half to two-thirds concentration of azuki bean), growth was sustained. We developed three interspecific hybrid populations and identified 21 QTLs for salt tolerance. In population A (cross between azuki bean and V. riukiuensis), three QTLs were detected for DtW and PWL (qDtW1-1/qPWL-1, qDtW2-1/qPWL2-1 and qDtW8-1/qPWL8-1). In population B (cross between azuki bean and V. nakashimae), five QTLs were detected (qDtW 4-1, qDtW5-1/qPWL5-1, qDtW 7-1, qPWL 4-1 and qPWL7-1). In population C (cross between V. nakashimae and V. riukiuensis), four QTLs were detected (qDtW4-2/qPWL4-2, qDtW5-2, qDtW11-1 and qPWL11-1). Additionally, four QTLs (qNa2-1, qNa8-1, qNa 8-2 and qNa 11-1) that were detected in the leaf Na + concentration analysis performed only in population C did not match the location of QTLs for DtW and PWL in population C. However, qDtW8-1/qPWL8-1 with the effect of the V. riukiuensis allele on reduced salt tolerance, which was detected in population A, was located on the same region as qNa8-1. This suggests that qDtW8-1/qPWL8-1/qNa8-1 has an effect on reduced salt tolerance by promoting Na + accumulation in the leaves. Additionally, a genetic interaction was observed between qDtW8-1/qPWL8-1/qNa8-1 and qDtW1-1/qPWL1-1. This suggests that qDtW8-1/qPWL8-1/qNa8-1 is associated with Na + accumulation ability in the leaves in V. riukiuensis, and qDtW1-1/qPWL1-1 is involved in the detoxification mechanism of Na + that allows for growth under the presence of Na + in the leaves. Furthermore, in population C, transgressive segregation was observed, indicating that the salt tolerance mechanisms in the two wild species might be different and that stronger salt tolerance could be acquired by combining genes conferring these mechanisms. Finally, this study not only provides useful marker information for subsequent fine mapping of candidate genes involved in salt tolerance and elucidating the genetic basis of Na + filtering and detoxification mechanisms in the future but also contribute to developing super-salt-tolerant azuki beans by combining the salt-tolerant mechanisms of V. riukiuensis and V. nakashimae.