Genetic diversity for drought and low-phosphorus tolerance in rice (Oryza sativa L.) varieties and donors adapted to rainfed drought-prone ecologies

Drought and phosphate availability are two major abiotic factors limiting productivity of rice in rainfed upland areas. There has been a constant need for new improved donor with tolerance to multiple abiotic stress conditions for rainfed rice breeding. In the present study, a set of 32 popular rice varieties and landraces were evaluated for drought and low-phosphorus (P) tolerance, and also characterized using grain yield under reproductive drought QTLs (DTY QTLs) and Pup1 linked/specific molecular markers. Twenty-seven genotypes were identified as tolerant to moderately tolerant to drought. The SSR markers linked to ten DTY QTLs classified the genotypes into two groups corresponding to aus and indica. The tolerant genotypes were distributed under both groups. Based on the core markers of Pup1 locus, complete tolerant haplotype was recorded in nine genotypes other than the tolerant check Dular. Nine more genotypes showed the incomplete tolerant haplotypes. The rice genotypes showed significantly high genetic variability for low-P tolerance in hydroponic study. A few genotypes revealed non-Pup1 type tolerance which needs further confirmation.

www.nature.com/scientificreports/ relying on fertilizer application. Identification of Phosphorus Uptake 1 (Pup1) which is associated with increased P uptake and a significant yield advantage under P-deficient environments 10,11 . Pup1 was identified from the donor variety Kasalath as a major QTL on rice (Oryza sativa) chromosome 12 explaining around 70% of the phenotypic variance for tolerance to phosphorus deficiency in soil 10,12 . Molecular markers were designed for different candidate genes within the Pup1 locus and effectively been used for detecting Pup1 haplotypes in diverse rice genotypes 11,13 . The Pup1 protein kinase gene, OsPupK46-2, was functionally characterized and named as PSTOL1 (Phosphorus Starvation Tolerance 1) 14 . Interestingly, PSTOL1 was found to be absent from the rice reference genome (Nipponbare, a temperate japonica variety), where it falls within a ~ 90 kb insertion-deletion polymorphism (indel) on chromosome 12. PSTOL1 haplotype analysis in Oryza gene pool suggested a longterm selective maintenance of functional alleles, but with repeated evolution of loss-of-function alleles that have evolved convergently in multiple rice species and cultivated rice varieties 15 .
Identifying new donors for abiotic stress tolerance with desirable background is essential for rice breeding under rainfed environments. The donors of many of the DTY QTLs were landraces such as Aday Sel., Dhagad Deshi, Kali aus, Moroberekan, N22, etc. 7 Similarly, the Pup1 locus was found in more that 80% of the drought tolerant varieties and breeding lines 11 . In this context, the present study was undertaken using grain yield under drought stress (DTY) QTL-linked SSR and Pup1-gene-based markers to assess the diversity based on DTY-QTL makers, and to evaluate Pup1 haplotypes locus in a diverse set of released varieties and landraces mostly adapted to rainfed drought-prone ecologies. The knowledge of the extent of genetic variation, diversity and presence of stress tolerance genes will lead to exploit the traits for rainfed rice improvement.

Results
Genetic diversity analysis based on SSR markers linked to DTY QTLs. Multiple alleles were observed for each of the DTY QTL-linked SSR markers (Supplementary Table S1). A total of 88 alleles were detected at the 24 SSR markers across 32 genotypes, ranging from two alleles (RM28199, qDTY 12.1 ) up to five alleles in RM262 (qDTY 2.1 ), RM211 (qDTY 2.2 ) and RM204 (qDTY 6.1 ), with an average of 3.67 alleles. The polymorphism information content (PIC) ranged from 0.27 (RM279) to 0.70 (RM262) with an average PIC of 0.54. The mean expected heterozygosity (He) was 0.60. The most common allele at each locus ranged from 34% (RM262) to 84% (RM279).
The distance-based results using UPGMA clustering revealed two major groups in the present rice germplasm (Fig. 1a). Twelve genotypes including the known aus cultivars and three released varieties (Bala, Poornima and Vandana) corresponds to aus group, while 19 indica varieties correspond to the indica group. The tropical japonica cultivar Moroberekan cluster separately from both indica and aus groups. Similarly, the principal coordinate analysis revealed a clear separation between indica and aus, although Vandana and Poornima showed some distinctiveness from the primary aus cluster (Fig. 1b). The lowest genetic distance was observed between a b Phenotypic evaluation for low P tolerance. The low phosphorus (P) treatment was effective in producing P deficiency stress, as shown by reduced shoot biomass. Significant differences among the genotypes were observed for all three root and shoot traits studied. Additionally, the interaction between genotype and P treatment was significant (P < 0.0001) indicating that the impact of P levels on the growth traits depends on genotype (see Supplementary Table S2 online). Under low P conditions, root length increased significantly. However, root length did not show significant differences between high-and low P conditions in plants lacking the PSTOL1 gene (Fig. 4). In contrast, the plants with PSTOL1 produced longer roots under P deficient conditions. Both shoot and root growth reduced significantly under low P treatment regardless of presence or absence of PSTOL1 (Fig. 4). Root length ranged from 7.3 cm (Kalinga III) to 13.5 cm (Moroberekan) under high P, whereas under low P it varied between 6.8 cm (IR64) to 15.1 cm (Moroberekan). The root length increased up to 41.1% (Kalinga III) with an average of 12.2% with the genotypes under low P. Overall, seven genotypes did not show longer roots under P deficiency (see Supplementary Fig. S1 online). The mean shoot dry weight of PSTOL1 genotypes was similar to that of the genotypes without PSTOL1 under P deficient conditions. However, under high P conditions, root dry weight was considerably higher in genotypes carrying PSTOL1. The shoot dry weights under low P against those under high P was plotted to identify genotypes with high vigour under both P levels (Fig. 5).

Discussion
The DTY QTLs have a genetic gain of 10-30%, with a yield advantage of 150-500 kg ha −1 under reproductive stage moisture stress 16 24 . The present analysis also indicates that Kalakeri, a highly drought tolerant landrace from the Indian state of Odisha, could be utilized further in drought breeding programmes. Likewise, other aus cultivars www.nature.com/scientificreports/ such as Black gora, Bhutmuri, Asanlaya, Lalnakanda 41 and AUS 257 clould also be utilized for identification of novel QTLs for drought tolerance. Genetic diversity analysis of the 32 rice genotypes using the drought QTL-linked SSR markers revealed two major groups, and based on the previous reports, we could designate one group as indica and the other as aus.
The aus group included all drought tolerant aus cultivars along with the varieties having aus in their parentage. Earlier reports also indicated a common ancestry of drought-tolerant rice genotypes, mostly from the aus group 25 . In the present study, the DTY QTL markers effectively revealed the ancestry and genetic differentiation within the genotype panel.
For detection of tolerance alleles at the Pup1 locus in the present genotypes, the genotypes along with low P tolerant and sensitive checks were surveyed with 14 gene-specific markers. The frequency of different Pup1 genes varied widely across the genotypes. Pup1 is a major locus on rice chromosome 12 conferring tolerance to P deficiency in soils. This locus contributed around 70% of the total phenotypic variance for P uptake 10,12 . This locus is prevalent in varieties and landraces adapted to rainfed drought-prone and acidic soil environments 13,26 . Our study also indicated a similar trend that Pup1 is present in most of the drought tolerant genotypes either completely or partially. Therefore, it is safe to note that, the breeders have directly or indirectly selected for the Pup1 QTL while developing resilient varieties for the rainfed environments. A cluster analysis using six core Pup1 markers grouped the genotypes based on the presence or absence of respective Pup1 marker loci and provided details on the different haplotypes present in the current genotypes. Altogether, ten genotypes (including tolerant check Dular) had the Kasalath allele at all six loci, and these genotypes along with nine others possessing the tolerant haplotype partially formed a separate cluster from the genotypes that are completely devoid of the Kasalath haplotypes. We got a variable result with the three markers targeting the hypothetical gene OsPupK29. The frequency of tolerant allele these marker loci varied from 48% (K29-3) to 64% (K29-2). Chin et al. 13 reported that OsPupK29 is an unstable gene with possible role in the evolution of Pup1. The protein kinase gene OsPupK46 was characterized as the candidate gene PSTOL1 underlying the Pup1 locus 14 . Genotyping with the markers www.nature.com/scientificreports/ specific to PSTOL1 and OsPupK46 revealed a similar result with a frequency of 61% in the current genotypes. The tolerant genotypes identified in this study with the null Pup1 will be utilized in future investigations on the low P tolerance mechanisms. Swamy et al. 26 also reported null Pup1 tolerant genotypes in rice collections from north-eastern India. Phenotypic evaluation of the genotypes under low-and high P conditions revealed both genetic variation in root and shoot traits and the genotypic response to P deficiency. When shoot dry weight under low P was plotted as a function of that of high P, a wide variation in plant vigour was observed. Genotypes such as Swarna, Kalinga III, Vanaprabha, Heera, Way Rarem, Vandana, CR Dhan 40, and the aus cultivars were vigorous under both P levels. Among these genotypes a few (Rasi, Vanaprabha and CR Dhan 40) were devoid of the tolerant Pup1 haplotype suggesting a non-Pup1 type tolerance. However, the result needs to be revalidated under field evaluation. The aus cultivars showed considerable Pup1 type tolerance to low P. The aus genotypes are well recognized as a source of abiotic stress tolerance in rice breeding [27][28][29] .
In conclusion, the present study had characterized a set of popular rainfed rice varieties and landraces for tolerance to two most important abiotic stresses, drought and low P, widespread under rainfed drought-prone ecologies in India. The results of genotyping with DTY QTL-linked and Pup1-specific markers will help rice breeders in selecting suitable parents for varietal development.  Table S4 online. The genotypes Apo, Kali aus, Moroberekan, N22, Vandana and Way Rarem were reported donors of several DTY QTLs 6 . The variety Dular was used as positive check for Phosphorus uptake 1 (Pup1) survey. It had Kasalath-type alleles for all the Pup1 genes 13 . The experiments on these plant materials were done as a part of inhouse research project of ICAR-National Rice Research Institute for developing resilient rice varieties for rainfed drought-prone ecologies.

Methods
Drought screening. The rice genotypes were screened under rainout shelter screening facility at CRURRS for drought tolerance during August-October 2018 wet season as per the protocol described in Verulkar et al. 30 The genotypes were dry seeded under upland conditions in rows spaced at 25 cm in well prepared levelled fields. The experiment was conducted in a randomized complete block design with three replications. Inorganic NPK fertilizer dose of 60-40-40 was applied. P and K were applied as a single basal dose at sowing, whereas N was applied in three equal splits, at sowing, 28 and 50 days after sowing. Weeds were controlled by application of the post-emergence herbicide bispyribac sodium (0.02 kg a.i. ha −1 ) at five days after sowing. Two hand weedings were done at later stages to keep the plots weed-free. The crop was flood-irrigated twice a week until 50 DAS. At 50 DAS, stress was imposed by stopping irrigation and protecting the trial from rainfall by covering it with a www.nature.com/scientificreports/ shelter during rainfall events. Genotypes were scored for leaf drying at the peak stress period (about three weeks after the withdrawal of irrigation) using the IRRI Standard Evaluation System 31 . With appearance of severe leaf rolling and drying on the susceptible check (Kalinga III), at about 21 days after withdrawal of irrigation, the stress was relieved by one flood irrigation. A second cycle of stress was then imposed until the maturity of the crop.
Low P screening. Sensitivity of the genotypes to P-deficient growing conditions was measured by root growth and shoot mass under low-and high P treatments under hydroponic experiments in the greenhouse. The seeds were surface sterilized with 1% sodium hypochlorite solution for 1 h with agitation and then rinsed five time with sterile doubled distilled water. Surface sterilized seeds were then sown on moist paper towel soaked with sterile water for 72 h at 28 °C in an incubator prior to planting. Healthy germinated seeds were then individually placed in the holes (1.5-2.0 cm in diameter) made on a Styrofoam sheet (50 cm × 60 cm) floated in a plastic tank containing 40 L tap water. A total of 96 holes were made per sheet. The experiment was laid out in a completely randomized design with three replications. In each replication, three germinated seeds of each variety were sown in three adjacent holes. After seven days, the seedlings were evaluated in two hydroponic solutions, one with deficient P (LowP) and the other with sufficient P (HighP). Both Low P and High P conditions were established by maintaining the NaH 2 PO 4 concentration in the hydroponic media 32 as 10 µM and 100 µM, respectively. The nutrient solution was changed every 3 days and maintained at a pH of 5.0 by adding 1 M NaOH or HCl as needed. The plants were sampled at 14 days after treatment. The root length of each plant was recorded. Plants were then partitioned into shoots and roots and dried in an oven at 70 °C for 72 h and weighed. Rice genotypes Annada and NDR 1045-2 were not included due to nonavailability of pure seeds.
Genotyping with DTY QTL and Pup1 markers. Twenty-four SSR markers were selected for detection of the DTY QTLs (Table 1). The peak and flanking markers for each DTY QTLs were included based on the information summarized in Kumar et al. 6 . For genotyping of Pup1 locus and PSTOL1, 15 gene-specific markers were used (Table 1). Positive checks were included in both the assays. The total genomic DNA was extracted from the ground leaf tissue using Qiagen DNeasy Pant Mini Kit (Qiagen, India). The extracted DNA was quantified using a UV-Vis spectrophotometer (NanoDrop 2000c, Thermo www.nature.com/scientificreports/ Scientific, USA) and was adjusted to a final concentration of 25 ng µL −1 . Polymerase chain reaction (PCR) was carried out in a reaction mix containing 50 ng template DNA, 0.5 μM of each primer, 12.5 μL PCR Master Mix (2X) (Thermo Scientific, USA) and double-distilled water to make the total volume up to 25 μL. Amplification was carried out using thermal cycler with the following thermal conditions: 94 °C for 5 min, followed by 35 cycles of 94 °C for 30 s, primer-specific annealing temperature for 30 s, 72 °C for 1 min, and final extension step was included for 72 °C for 7 min. The amplified products were visualized by SYBR™ Safe DNA Gel (Thermo Scientific, USA) stained agarose gels (1.0-4.0% based on amplicon size) in gel documentation system. In case of DTY QTL-linked SSR markers, the band with the lowest molecular weight was assigned allele number 1 and the progressively heavier bands were assigned numbers incrementally. For Pup1-specific markers, the molecular sizes (in base pair) of the amplified alleles were determined based on migration relative to the DNA ladder. The reproducible and well amplified DNA fragments were considered for analysis.
Data analysis. The drought and P-deficient screening data were subjected to the statistical analysis in R version 3.5.2 (The R Foundation for Statistical Computing, Vienna, Austria) using the packages "tidyverse" and "rstatix". One-and Two-way analysis of variance (ANOVA) were used to examine the influence of treatments on dependant variables and interactions between genotype and treatment level. The mean leaf drying scores of 32 genotypes were compared by the Duncan's Multiple Range Test (DMRT). The number of alleles locus −1 , major allele frequency, gene diversity, and polymorphism information content (PIC) were estimated using PowerMarker V3.25 33 . The genetic distance between all possible pairs of accessions were calculated and used to visualize relationships between accessions, and a phylogenetic tree was constructed using neighbour joining clustering using the C.S. Chord 1967 distance method implemented in PowerMarker V3.25. The allele frequency data from PowerMarker was used to export the data in binary format (allele presence = 1, and allele absence = 0) for analysis with NTSYS-pc V2.2 34 . A similarity matrix was calculated with the SIMQUAL subprogram using the Jaccard's coefficient and was also used for principal coordinate analysis with the DCENTER, EIGEN and MXPLOT subprograms in NTSYS-pc.

Data availability
Available on request from the authors.