Genetic Dissection of Drought Resistance of Common Bean at the Seedling Stage by Genome-wide Association Study

A variety of adverse conditions, including drought stress, severely affect common bean production. Molecular breeding for drought resistance has been proposed as an effective and practical way to improve the drought resistance of common bean. A genome-wide association analysis was conducted to identify drought-related loci based on survival rates at the seedling stage using a natural population consisting of 400 common bean accessions and 3832340 SNPs. The coecient of variation ranged from 40.90% to 56.22% for survival rates in three independent experiments. A total of 12 associated loci containing 89 signicant SNPs were identied for survival rates at the seedling stage. Four loci overlapped in the region of the QTLs reported to be associated with drought resistance. According to the expression proles, gene annotations and references of the functions of homologous genes in Arabidopsis, 39 genes were considered potential candidate genes selected from 199 genes annotated within all associated loci. A stable locus (Locus_10) was identied on chromosome 11, which contained LEA, aquaporin, and proline-rich protein genes. We further conrmed the drought-related function of an aquaporin (PvXIP1;2) located at Locus_10 by expression pattern analysis, phenotypic analysis of PvXIP1;2-overexpressing Arabidopsis and Agrobacterium rhizogenes-mediated hairy root transformation systems, indicating that the association results can facilitate the ecient identication of genes related to drought resistance. These loci and their candidate genes provide a foundation for crop improvement via breeding for drought resistance in common bean. An Agrobacterium rhizogenes-mediated hairy root transformation system was used to investigate the function of PvXIP1;2 in common bean under osmotic stress conditions. The K599 strain harbouring the PvXIP1;2-overexpression (PvXIP1;2-OE) and PvXIP1;2-RNA interference (PvXIP1;2-RNAi) constructs was injected into the hypocotyls of common bean seedlings for the generation of transgenic hairy roots. The transgenic hairy roots grew out of the injection sites approximately 10 days later. PvXIP1;2 expression in PvXIP1;2-OE roots was 4 times higher that in control hairy whereas that in PvXIP1;2-RNAi roots was only half of that in the The original roots were removed before the plants with hairy roots were subjected to 200 mM mannitol to simulate osmotic stress. After keeping in mannitol for 24 h, the leaves of the plants with PvXIP1;2-OE roots performed better than the control leaves; however, the leaves from the plants with PvXIP1;2-RNAi roots were more wilted than the control leaves 7b-c). The results of the RWC showed that seedlings with PvXIP1;2-OE roots had a higher water content than the control seedlings, whereas seedlings with PvXIP1;2-RNAi roots had a lower water content After keeping in mannitol for 72 h, the relative hairy root growth rates were measured and compared. The results demonstrated that the relative growth rate of PvXIP1;2-OE hairy roots was signicantly higher, whereas this parameter in PvXIP1;2-RNAi roots was signicantly lower than that of control roots (Fig. 7e-f). These results indicated that PvXIP1;2 confers resistance to drought in common bean and further demonstrated that these loci detected in this study were related to drought resistance. panel of 400 common bean accessions was employed to perform a GWAS for drought resistance at the seedling stage, and 4 of 12 candidate loci identied by GWAS overlapped with reported drought-related QTLs. A stable locus containing promising causal genes was identied in three independent experiments. A set of genes were identied as prospective candidate genes involved in drought response at the seedling stage. Importantly, a case study of PvXIP1;2 indicated the feasibility of mining candidate genes by GWAS. The associated loci and causal genes identied in this study provide insights into the genetic basis of plant mechanisms in response to drought and an important foundation for the genetic improvement of drought resistance in common bean at the seedling stage in the future.


Introduction
Common bean is the most important food legume for direct human consumption and provides proteins, vitamins, minerals and income to many people in Africa and Latin America (Beebe et al., 2013;Broughton et al., 2003). The growth, development, and production of common bean require su cient water. However, climate change is leading us towards a hotter, drier world (Gupta et al., 2020). It is generally considered that common beans have higher water requirements during the growth stage and lower resistance to drought stress than other food legumes, such as chickpea, cowpea or peanut (Broughton et al., 2003).
However, approximately 60% of common bean-producing areas around the globe suffer from different degrees of drought stress (Asfaw et al., 2013;Funk et al., 2008;Muñoz et al., 2006). In China, common bean is mainly produced on a small scale and is generally cultivated on relatively infertile land, which is susceptible to abiotic and biotic stresses. Under such conditions, due to the lack of irrigation equipment, common bean production depends on natural precipitation, so drought is the main reason for bean yield losses. Climate models predict that the main production area for common beans in China will become successively drier in the future (Piao et al., 2010). In addition, agricultural water demand accounts for more than two-thirds of global water consumption, which has increased sharply in recent decades and will further increase in the future (Wada et al., 2011;Fang et al., 2015). It is estimated that the world's population will increase from 7.5 billion presently to 9.7 to 10 billion by 2050, at which time an additional 1 million hectares of arable land will be needed to ensure food security, so the agricultural water demand will accordingly be doubled. However, climate change could give rise to a 50 percent drop in the availability of fresh water (Gupta et al., 2020). Molecular breeding for drought resistance has been proposed to be an effective and practical way to ensure sustained bean productivity yield traits, and eight drought-related SNPs were detected. However, all these drought-related studies mentioned above in common bean were conducted at the late growth stage, and the genetic basis of drought resistance at the early stage remains to be elucidated.
Drought can occur during the whole period or at speci c stages of crop development, including the seed germination, seedling growth, owering and grain lling stages. There are differences in drought resistance mechanisms in different growth stages (Levitt, 1972). In the common bean production area of China, drought stress often occurs at the early growth stage of common bean, which directly affects the rate of germination and seedling growth vigour of common bean and even results in the death of seedlings, severely restricting yield increases . Therefore, identifying stable droughtrelated QTLs or alleles at early growth stages is also essential for crop improvement in common bean. Wu et al. (2021a, b) conducted a GWAS using 438 common bean accessions to identify drought-related SNPs associated with root and germination traits at the bud stage. A series of candidate SNPs were detected, which greatly enriched the genetic information of drought resistance and root traits at the bud stage, and a set of candidate genes were identi ed, including transcription factors and protein kinases (Wu et al., 2021a, b). However, genetic information associated with drought resistance of common bean at the seedling stage is lacking. Different characteristics were used to evaluate drought resistance in different growth stages; for instance, yield and phenological traits were used at the late growth stage Mukeshimana et al., 2014), and the relative germination rate and germination index were used at the bud stage (Tan et al., 2017). In the seedling stage, the survival rate of seedlings after repeated drought is frequently used to evaluate the drought resistance of different genotypes Li et al., 2015;Wang et al., 2015). Qin's research group evaluated the tolerance of maize to severe drought stress at the seedling stage and then performed GWAS based on 368 maize inbred lines and 560000 SNPs to identify a set of drought resistance-related genes (Liu et al., 2013;Mao et al., 2015;Wang et al., 2016). Few studies have dissected the genetic basis of drought resistance in common bean at the seedling stage.
In this study, a natural population consisting of 400 common bean accessions was used to evaluate drought resistance at the seedling stage, and the survival rate after repeated drought was measured. Combining 3832340 SNPs, a GWAS was performed to identify candidate regions associated with drought resistance. Our objective is (a) to identify SNPs associated with drought resistance at the seedling stage, (b) to identify potential drought-related genes within candidate regions and (c) to perform preliminary functional validation of potential candidate genes.

Materials And Methods
Population materials and genotypic data used for GWAS A total of 400 common bean accessions representing broad genetic diversity, including 329 from China and 71 from the world, were used for the GWAS (Table S1). This panel is derived from a whole-genome resequencing panel of 683 accessions as previously described (Wu et al., 2020).
A set of SNPs generated from the whole-genome resequencing project contained 4811097 SNPs (Wu et al., 2020). For association analysis, the SNP dataset was further ltered based on the exclusion of data with a missing rate ≥20% and minor allele frequency <0.05 to obtain 3832340 quali ed SNPs.
Phenotyping common bean drought resistance at the seedling stage To evaluate the drought resistance of common bean at the seedling stage, survival rate (SR) tests were performed in a greenhouse in Beijing, China, including three independent repeats in 2017 (from September to December) and 2018 (from March to June and from September to December). Each independent repeat contained two replicated assays. A completely random design was applied in this experiment. In each assay, a total of 40 plastic boxes (65 × 40 × 20 cm, length × width × depth) were used for planting, which were lled with 13 kg mixed soil containing uniform topsoil, vermiculite and nursery substrate (mass ratio, 4: 3: 3). Before sowing, each box was watered until the relative soil water content reached approximately 40%, as measured by SU-LB (Mengchuangweiye Technology Co., Ltd). Then, each box was divided into 12 rows (Fig. S1). The rst and twelfth rows were designed as guard rows, so each box contained rows to accomodate 10 materials. All of the genotypes were randomly planted, and 12 plants of each genotype were grown per row in each assay before being covered with 2 kg of mixed soil. The relative soil water content was maintained until all genotypes developed trifoliate leaves, and then the number of seedlings was recorded. Subsequently, water was withheld, and the relative soil water content was measured by SU-LB every day. Approximately 6 d after the relative soil water content reached approximately 0%, when all seedlings were severely wilted, each box was rewatered until the relative soil water content reached approximately 40%. Three days after rewatering, the number of viable plants of every accession was recorded. Plants with green leaves and vigorous stems were regarded as survivors. Then, water was withheld again, and each box was rewatered again approximately 6 d after the relative soil water content reached approximately 0%. Three days after rewatering, the number of viable plants of every accession was recorded again.
In each independent repeat, the seedling SR was calculated as follows: SR = (N 1 /N × 100% + N 2 /N × 100%) × 2 -1 , where SR represents the survival rate in an independent repeat; N 1 represents the mean of the number of viable plants in two assays after the rst drought treatment; N 2 represents the mean of the number of viable plants in two assays after the second drought treatment; N represents the mean of the total number of seedlings in two assays. The three independent repeats were denoted as SR1, SR2, and SR3. SR1, SR2, SR3 were used for statistical analysis. Descriptive statistical analysis and signi cance testing were performed using SPSS 19.0 software (SPSS, Inc., Chicago, IL, USA), and correlations among independent repeats were calculated using the function cor.test() in the psych package of R 3.5.1 based on Pearson's correlation coe cient (Best and Roberts, 1975).

Genome-wide association study
To identify potential related SNPs, GWAS was performed using the compressed mixed linear models (CMLM) program of Tassel 5.0 (Bradbury et al., 2007). Principal components (PCs) of the association panel were calculated by PLINK 1.9 software based on all SNPs after ltration (Purcell et al., 2007). The rst three PCs were used as the PC matrix for the associated method. The kinship matrix (K matrix) was calculated using Tassel 5.0. The P value threshold of the entire population was 1 × 10 -5 . The chromosome region with an associated SNP ± LD decay distance was considered a candidate segment or associated locus (represented by pre x "Locus_"), and the LD decay distance of common bean was 107 kb (Wu et al., 2020). The lead SNPs (SNPs with the lowest P value) for each locus were de ned as candidate SNPs.
Locus identi cation, candidate gene prediction and candidate sequence analysis The physical region of drought resistance-related QTLs was determined by the physical position of the left border marker and the right border marker or 107 kb around each leak SNP, which were obtained by searching marker information in the NCBI database (https://www.ncbi.nlm.nih.gov/). If the physical position of the candidate segment overlapped with the physical region of the reported drought-related QTL, this candidate segment was considered an overlapping locus.
According to their physical positions, we selected genes of all candidate segments from the gene structure annotation le (gff3 format) of common bean accession G19833 v1.0 (Schmutz et al. 2014). All selected genes were annotated using the gene function annotation le of common bean accession G19833 v1.0 (Schmutz et al. 2014). Genes that were stress-responsive transcription factors, such as a NAC, MYB, or bHLH transcription factors or whose homologous genes were involved in the drought response, osmotic response, and abscisic acid (ABA) response in A. thaliana in a previous study, were considered candidate genes. Combining the expression data changes of common bean seedlings after drought stress (

Plant materials and growth conditions for the functional investigation of candidate genes
The common bean cultivar YUEJINGDOU was used for the expression level assay. Seeds were germinated on soil in a greenhouse equipped with a supplemental lighting and cooling system. The environmental settings for common bean growth were 23°C at night and 25°C at daytime with a photoperiod of 16 h:8 h (day:night). Uniformly developed trifoliate leaves were selected for drought treatment, and water was withheld from common bean seedlings grown in soil for the durations indicated. For mannitol, ABA, and NaCl treatments, seedlings were cultured by hydroponics until they developed trifoliate leaves, then seedlings were transplanted in a solution containing 300 Mm mannitol, 100 μM ABA, or 150 mM NaCl for the durations indicated.
Arabidopsis thaliana ecotype Columbia (Col-0) was used as the wild-type (WT) control in the present study, and all transgenic lines were generated in the background of Col-0. Seeds were sterilized with chlorine, vernalized for 3 d at 4°C, and sown on half-strength Murashige and Skoog (MS) medium. Plants were grown in a growth chamber under continuous light (70 μmol m -1 s -1 ) with a photoperiod of 16 h:8 h day:night at 22°C. For the root length assay with the young seedlings, four-day-old seedlings were transplanted to MS medium supplemented with 100, 200, 300 mM mannitol, 50, 100, 150 μM ABA or 50, 100, 150 mM NaCl for 10 days. Then, photos were taken, root length was measured, and MS medium was used as a control. For drought stress resistance analysis, one-week-old seedlings were transplanted to pots lled with a mixture of soil and sand (1:1) for 3 additional weeks. Water was withheld from the treatment group for 13 days before rewatering, photos were taken, and the SR was recorded.

Constructs and generation of transgenic Arabidopsis thaliana
Full-length cDNA of PvXIP1;2 with Hind and Spel restriction sites was ampli ed with the primers PvXIP1;2_1F/1R (Table S2). The ampli ed fragments were cloned into the modi ed pCAMBIA1300-GFP expression vector to generate the 35S::PvXIP1;2-GFP fusion vector. The fusion vector was transferred into Agrobacterium tumefaciens GV3101 and transformed into Col-0 by the oral dip method (Clough and Bent, 1998). Transgenic plants were selected with MS medium containing 50 mg/L hygromycin and were identi ed by PCR ampli cation with the primers PvXIP1;2_2F/2R (Table S2). T3 or T4 homozygous lines were used for further functional investigation.

Subcellular localization
The pCAMBIA1300-PvXIP1;2-GFP fusion vector was used for Agrobacterium-mediated transient expression in Nicotiana benthamiana (Sheludko et al., 2007), and pCAMBIA1300-GFP was used as a control. Seedlings were cultivated in the dark for one day and then under normal conditions for one day after in ltration, and then uorescence signals were captured using a confocal microscope (LSM880 ZEISS, Germany).  (Table S2). Experiments were carried out with three replications.

Ion leakage, proline, and malondialdehyde measurements
Four-week-old Arabidopsis thaliana seedlings were well watered or not watered for 13 days, and then the leaves were sampled to measure the malondialdehyde (MDA) and proline contents as well as ion leakage (IL). The MDA content was measured using an MDA Assay Kit (Solarbio, Beijing, China) according to the manufacturer's instructions. The proline content was measured using a Micro Proline (Pro) Content Assay Kit (Solarbio, Beijing, China) according to the manufacturer's instructions. IL was measured according to a previously described method (Jiang and Zhang, 2001). Collected leaves were cut into wellproportioned strips and incubated in 10 ml distilled water at 25°C for 8 h. The initial conductivity (C1) was measured by a conductivity metre (SevenExcellence TM , Switzerland). Then, the samples were boiled for 10 min. The conductivity (C2) was measured again when the samples cooled to 25℃. IL was calculated as follows: IL = C1/C2 × 100%.
Agrobacterium rhizogenes-mediated transformation of seedlings and mannitol treatment For the overexpression vector, full-length cDNA of PvXIP1;2 with NcoI and BstE restriction sites was ampli ed with the primers PvXIP1;2_3F/3R (Table S2) and then cloned into the pCAMBIA3301 vector. For the RNAi vector, a fragment with NcoI and BstE restriction sites was synthesized, including the sense and antisense fragments of PvXIP1;2, and cloned into the pCAMBIA3301 vector. The pCAMBIA3301 vector was used as control. The common bean cultivar YUEJINGDOU and the A. rhizogenes strain K599 were used for transformation. Transgenic common bean hairy root composite plants were generated by a previously described method (Estrada-Navarrete et al., 2007). When hairy roots developed, the seedlings with hairy roots were moved to water for 2 days for recovery. For measurement of the relative water content (RWC), the seedlings with uniform hairy roots were transferred to 200 mM mannitol or water for

Results
Phenotypic analysis of the survival rate of 400 common bean accessions The seedling SR can re ect plant drought resistance mechanisms and cellular responses. It is less affected by environmental uctuation, which helps to identify the underlying genetic determinants (Mao et a., 2015). Therefore, the drought resistance of 400 common bean accessions was assayed by calculating the SR after severe drought stress at the seedling stage. A large range of variation was detected, with the coe cient of variation (CV) varying from 40.90% for SR3 to 56.22% for SR2 (Table S3). The SR at different time points ranged from 0.00 to 100.00%, and the means were 50.81, 52.04, and 58.22, respectively (Table S3). All SR data showed a uniform distribution (Fig. S2). These results indicated that the drought resistance of the 400 common bean accessions was different at the seedling stage. Correlation analysis showed that pairwise positive correlations (p < 0.01) were detected for SR1, SR2 and SR3 (Table S4) Association analysis of drought resistance at the seedling stage To identify drought resistance-related association loci, a reported genotypic dataset consisting of 3832340 SNPs was used for conducting GWAS for the SR. According to the de nition described in the methods, a total of 12 associated loci containing 89 SNPs were identi ed (Fig. 1, Table S5), and 12 leak SNPs were detected, which were distributed on Chr. 2, 3, 5, 6, 7, 10, and 11 (Table S6). For SR1, 4 associated loci were detected, which were distributed on Chr. 6, 10, and 11 (Table S6). For SR2, 5 associated loci were detected, which were distributed on Chr. 5, 7, and 11 (Table S6). For SR3, 5 associated loci were detected, which were distributed on Chr. 2, 3, and 11 (Table S6) Table S6). It is worth noting that Locus_10, located on Chr. 11, was simultaneously associated with SR1, SR2 and SR3 and contained a signi cant SNP with the lowest P value of 4.36E-08. These results indicated that these associated loci were related to drought resistance.

Prediction of candidate genes associated with drought resistance
Among all physical positions of the 12 associated loci, 199 genes were annotated (Table S7) (Table S7). Phvul.003G124000, located at Locus_3, is associated with SR3 and encodes a WRKY family transcription factor, and the expression of the gene was downregulated in common bean after drought treatment with a Log 2 (FC) = -2.76 (Table S7, Wu et al., 2014). We subsequently searched the NCBI database to determine the functions of these genes in A. thaliana. Two genes, Phvul.002G278200 and Phvul.002G279300, located at Locus_1 and associated with SR3 (Fig. 1, Table S7) encode a protein kinase superfamily protein and calciumdependent protein kinase 2, whose homologous genes in Arabidopsis are AtSNRK2. 4 (Zhu et al., 2007). A total of six genes were annotated in Locus_2; among these genes, Phvul.003G067700 encodes a homeobox-leucine zipper protein, whose homologous gene in Arabidopsis (AtHB13) is involved in drought resistance (Cabello et al., 2012), and Phvul.003G067800 encodes MYB domain protein 88, whose homologous gene in Arabidopsis (AtMYB88) is also involved in drought resistance (Xie et al., 2010) (Fig. 1, Table S7). Another gene (Phvul.007G220900) located at Loucs_8 (Fig. 1, Table S7) encodes a RING/U-box superfamily protein whose homologous gene in Arabidopsis (LOG2) is involved in drought resistance (Kim et al., 2013). There were 39 candidate genes screened from all annotated genes (Table S8) according to the following rules: (1) it was a stress-responsive transcription factor, such as an NAC, MYB, or bHLH transcription factor; (2) its homologous genes in A. thaliana were identi ed in a previous study as being involved in the drought and ABA response and ABA signal transduction; and (3) its expression was upregulated or downregulated after drought treatment with |Log 2 (FC)| ≥ 2. These results further showed that these signi cant SNPs were related to drought resistance, and potential genes associated with drought resistance were identi ed.
A phylogenetic tree based on XIPs from different species was constructed (Fig. S3a), indicated that the XIPs from legumes clustered together. The results of sequence alignment (Fig. S3b) showed that two highly conserved 'NPA signature motifs' were found in PvXIP1;2, but the rst NPA motif was replaced with SPV, similar to that in soybean. The four residues forming the ar/R selectivity lter of PvXIP1;2 are valine from TMH2, isoleucine from TMH5, and alanine and arginine from R3 and R4, which are identical to those in soybean (Bienert et al., 2011, Fig. S3b). The results demonstrated that PvXIP1;2 is an aquaporin and belongs to the X-intrinsic protein (XIP) subfamily, which is absent in Arabidopsis ( . However, the function of PvXIP1;2 remains to be elucidated. Therefore, Phvul.011G025800 was assumed to be a promising candidate gene for further functional investigation.

Expression patterns of PvXIP1;2
To investigate the expression of PvXIP1;2 in different organs of common bean, total RNA was extracted from roots, leaves, trifoliates and stems for quantitative real-time RT-qPCR analysis. The results showed that PvXIP1;2 was expressed in various tissues of common bean (Fig. 3a).
The expression le data indicated that PvXIP1;2 transcript levels were strongly induced by drought stress (Wu et al., 2014;Pereira et al., 2020). To verify the expression data, we performed quantitative real-time RT-qPCR using RNA isolated from drought-treated common bean, and the results con rmed the expression data (Fig. 4a). Furthermore, various stress treatments were applied to common bean with trifoliates, including salt, mannitol and ABA treatments, to determine the transcriptional response of PvXIP1;2 to abiotic stress. The results demonstrated that the expression of PvXIP1;2 was induced in leaves and roots after salinity stress, simulated drought and ABA treatments (Fig. 4b-c). These results suggested that the PvXIP1;2 transcript level was affected by various stress treatments.
To determine the subcellular localization of the PvXIP1;2 protein, its ORF was introduced into the pCAMBIA1300-GFP vector upstream of the GFP gene to create a PvXIP1;2-GFP fusion construct. Then, we transformed the PvXIP1;2-GFP fusion construct into Nicotiana benthamiana by injection. A strong uorescent signal derived from GFP alone was observed in the cytoplasm and nuclei (Fig. 3b), whereas transformed cells carrying PvXIP1;2-GFP showed a strong green uorescence signal in the plasma membrane, indicating the plasma membrane localization of PvXIP1;2.

Overexpression of PvXIP1;2 increases the drought resistance of Arabidopsis
To further understand the function of PvXIP1;2, we constructed a 35S::PvXIP1;2 vector. After oral-dip transformation of Arabidopsis, three homozygous lines (L6, L8 and L10) were selected from the T4 generation for further functional investigation (Fig. 5a). To investigate the drought resistance of transgenic Arabidopsis overexpressing PvXIP1;2, WT and transgenic Arabidopsis were grown for 3 weeks in soils before water was withheld for 13 d and then rewatered. Most transgenic plants remained turgid, and their leaves remained green, whereas the WT plants wilted, and their leaves became yellow (Fig. 5b). When data from three different experiments were analysed, approximately 80% of the transgenic plants survived stress after rewatering, which was approximately four times higher than that of WT plants (Fig. 5c). Osmotic adjustment and antioxidant defence systems are the main pathways of drought tolerance-associated mechanisms (Fang et al., 2015). Therefore, the MDA and proline contents and IL were measured in WT and transgenic plants under drought stress and wellwatered environments to identify the function of PvXIP1;2 in these physiological processes. MDA, which causes oxidative injury to the cytomembrane, accumulates in leaves when plants experience stress. No evident difference was observed in the MDA content of WT and transgenic plants under well-watered conditions, whereas a signi cantly lower MDA content was observed in transgenic plants than in the WT under drought treatment (Fig. 5d). Consistent with these results, a lower IL was observed in transgenic plants than in the WT under drought treatment, whereas there was no signi cant difference in WT and transgenic plants under well-watered conditions (Fig. 5e), suggesting that the more severe membrane damage after water was withheld in the WT could at least partly be attributed to the inability of these plants to e ciently eliminate MDA due to the disruption of antioxidant defence systems under severe drought stress. In addition, proline, which is an essential osmotic substance, accumulated more in transgenic plants than in WT plants under drought stress conditions, while no signi cant difference was observed in WT and transgenic plants (Fig. 5f). These results indicated that heterologously overexpressing PvXIP1;2 in Arabidopsis improved resistance to drought stress due to reduced lipid peroxidation and membrane injury and increased accumulation of osmotic substances. 35S::PvXIP1;2 transgenic Arabidopsis is resistant to osmotic, salt and ABA stresses PvXIP1;2 is a mannitol-, salt-and ABA-induced gene (Fig. 4b-c), indicating that PvXIP1;2 might also be necessary for plant responses to osmotic, salt and ABA stresses. To further identify the results, four-dayold WT and transgenic seedlings were transformed on MS medium containing different concentrations of mannitol, NaCl and ABA for 10 days. There was no signi cant difference in WT and transgenic plants (Fig. 6a) on MS medium without treatments. The addition of mannitol, NaCl and ABA signi cantly inhibited root development, and the degree of inhibition increased with increasing concentrations of mannitol, NaCl and ABA (Fig. 6b-f). The root development of WT plants was more inhibited than that of transgenic plants in all treatments except 150 mM NaCl (Fig. 6b-f). These results demonstrated that the overexpression of PvXIP1;2 improved the resistance of Arabidopsis to osmotic, salt and ABA stresses.
PvXIP1;2 improves stress resistance in transgenic common bean hairy roots An Agrobacterium rhizogenes-mediated hairy root transformation system was used to investigate the function of PvXIP1;2 in common bean under osmotic stress conditions. The K599 strain harbouring the PvXIP1;2-overexpression (PvXIP1;2-OE) and PvXIP1;2-RNA interference (PvXIP1;2-RNAi) constructs was injected into the hypocotyls of common bean seedlings for the generation of transgenic hairy roots. The transgenic hairy roots grew out of the injection sites approximately 10 days later. PvXIP1;2 expression in PvXIP1;2-OE roots was 4 times higher than that in control hairy roots, whereas that in PvXIP1;2-RNAi roots was only half of that in the control (Fig. 7a). The original roots were removed before the plants with hairy roots were subjected to 200 mM mannitol to simulate osmotic stress. After keeping in mannitol for 24 h, the leaves of the plants with PvXIP1;2-OE roots performed better than the control leaves; however, the leaves from the plants with PvXIP1;2-RNAi roots were more wilted than the control leaves (Fig. 7b-c). The results of the RWC showed that seedlings with PvXIP1;2-OE roots had a higher water content than the control seedlings, whereas seedlings with PvXIP1;2-RNAi roots had a lower water content (Fig. 7d). After keeping in mannitol for 72 h, the relative hairy root growth rates were measured and compared. The results demonstrated that the relative growth rate of PvXIP1;2-OE hairy roots was signi cantly higher, whereas this parameter in PvXIP1;2-RNAi roots was signi cantly lower than that of control roots (Fig. 7e-f). These results indicated that PvXIP1;2 confers resistance to drought in common bean and further demonstrated that these loci detected in this study were related to drought resistance.

Discussion
To the best of our knowledge, this is the rst attempt to conduct a genome-wide association study to identify drought resistance-related QTLs in common bean at the seedling stage. In the present study, the largest SNP dataset consisting of 4.8 M SNPs, which was generated by resequencing a panel of 683 common bean accessions collected from China and elsewhere, including landrace and breeding lines representing both Andean and Mesoamerican gene pools (Wu et al., 2020), was used for conducting the GWAS. A total of 400 common bean accessions derived from the resequencing population by random sampling were used to evaluate drought resistance at the seedling stage. Both the number of individuals and the density of markers are relatively large in genome-wide association analysis of common bean. In the present study, 12 association loci containing 89 signi cant SNPs were identi ed to be associated with SR. Four loci overlapped with reported drought-related QTLs. One locus located on Chr. 11 was stably detected in three independent repeats (Table S6). Seven genes whose homologous genes in Arabidopsis were involved in the response to drought or ABA were found to be located in these loci (Table S7), suggesting the excellent reliability of the genetic loci of drought resistance identi ed in this study. More importantly, a case study of PvXIP1;2 demonstrated that the candidate loci identi ed in this study are highly valuable for further identi cation of the causal genes associated with drought resistance, and some of these loci may be used for genetic improvement of drought resistance in common bean at the seedling stage.
Drought resistance studies in common bean have mainly focused on the late developmental stage, and a few drought-related QTLs have been identi ed using different molecular markers (Schneider et  Limited studies have reported drought-related QTLs in common at the early growth stage. Wu et al.
(2021) evaluated the drought resistance of a common bean natural-variation population at the bud stage based on nine root traits and conducted a GWAS to identify drought-related loci. A set of loci and candidate genes were found to be associated with drought resistance at the bud stage. Among these loci, three loci associated with roots were collocated with Locus_1, Locus_7 and Loucs_8 identi ed in the present study, indicating that good root development after germination was related to the drought resistance of common bean at the seedling stage. In addition, a SNP associated with days to owering and the reproductive period was collocated with Locus_12 identi ed in this study (Villordo-Pineda et al.,

2015;
). The overlapping loci were detected by different mapping populations at different growth stages, indicating the possibility of nding common regulatory genes for the whole growth period.
Predicting candidate genes around leak SNPs within LD decay distances could be an effective method for identifying causal genes (Mao et a., 2015; Wu et al., 2020). However, despite the relatively low LD decay of 107 kb in common bean (Wu et al., 2020), one association locus in this study contained more than ten genes on average. A total of 199 genes were annotated among all associated loci (Table S7), so it is rather di cult to pinpoint the causal genes for these loci. The combined analysis of the expression pro les, gene annotations and references of the functions of homologous genes in Arabidopsis is a feasible and e cient way to narrow down the candidate genes. Among these genes, seven were reported to be involved in osmotic development, root development, the ABA-activated signalling pathway and drought resistance and were located at Locus_1, Locus_7, Locus_8 and Locus_10 (Table S7). Locus_1 identi ed in this study was also associated with root traits (  ., 2013). These results further demonstrated that these loci identi ed in the study were related to drought resistance and that the candidate genes are worth further functional study. In addition to the seven genes reported in Arabidopsis, many unreported loci and genes for drought resistance were detected in our association analysis. There were 39 candidate genes screened from all annotated genes (Table S8) according to their expression pro les, gene annotations and references to the functions of homologous genes in Arabidopsis. For example, Phvul.011G026700 encodes proline-rich protein 2 and was downregulated after drought treatment with Log 2 (FC) of 5.99, and Phvul.011G025800 encodes aquaporins and was upregulated after drought treatment with a Log 2 (FC) of 8.2, indicating that these are promising causal genes for further functional investigation. Nevertheless, the functions of these genes in response to drought stress need to be con rmed by molecular experiments.
Aquaporins are channel proteins that are responsible for water transport during seed germination, cell elongation, stomatal movements and abiotic stress responses (Maurel et al., 1997 and. Some members of the aquaporin family were identi ed to enhance resistance to drought, such as MaPIP1 (Xu et al., 2014) and ScPIP1 (Wang et al., 2019). These aquaporins were induced by NaCl and water de ciency treatment, increased primary root elongation, reduced membrane injury and accumulated osmotic substances under drought stress. PvXIP1;2 is an aquaporin and belongs to the Xintrinsic protein (XIP) subfamily, which is absent in Arabidopsis ( In conclusion, our study provides relatively rich genetic information on drought resistance in common bean at the seedling stage. A panel of 400 common bean accessions was employed to perform a GWAS for drought resistance at the seedling stage, and 4 of 12 candidate loci identi ed by GWAS overlapped with reported drought-related QTLs. A stable locus containing promising causal genes was identi ed in three independent experiments. A set of genes were identi ed as prospective candidate genes involved in drought response at the seedling stage. Importantly, a case study of PvXIP1;2 indicated the feasibility of mining candidate genes by GWAS. The associated loci and causal genes identi ed in this study provide insights into the genetic basis of plant mechanisms in response to drought and an important foundation for the genetic improvement of drought resistance in common bean at the seedling stage in the future.       Values are the means (± SE) of three replications, and ten plants were tested in each experiment. * indicates statistically signi cant differences at the level of P < 0.05, ** indicates statistically signi cant differences at the level of P < 0.01, Student's t test, two-tailed.