Genetic analysis and traits association study in marker-assisted multi-drought-traits pyramided genotypes under reproductive-stage moisture stress in rice (Oryza sativa L.)

Reproductive-stage drought-stress is a major production constraint in rainfed rice ecosystem. Emergence of marker-assisted breeding strategies for developing drought-tolerant rice varieties are being optimized through exploiting adaptive-traits for their increased contribution towards grain-yield under recurring-drought. Grain-yield is a complex-trait; requires knowledge of genetics and association among yield contributing component-traits. Current study was undertaken using 21 marker-assisted multi-drought-traits pyramided genotypes response for genetic variability and association of traits for grain-yield under aerobic and reproductive-stage drought conditions. Field evaluation was carried-out in two seasons and data was collected on various parameters. Path-coefficient analysis was used as a selection criterion to select yield contributing-traits and found nine phenotypic traits were having a positive direct-effect on grain-yield during both and/or at least one season under both moisture-regimes. The data from summer and Kharif seasons have been pooled within their respective moisture-regimes due to the non-significance of Levene’s test of homogeneity of variances and estimated BLUP values. ANOVA based on BLUP values revealed significant differences for moisture-conditions and also among genotypes. Phenotypic variation via. box-plots and histogram depicted mean phenotypic differences of traits under two moisture-regimes. Majority of the traits possessed high PCV, GCV with high heritability and GAM indicating higher trait expression and additive gene action lead to effectiveness of selection under drought/moisture stress. Grain-yield possessed a positive correlation with all the component-traits under consideration during both moisture-regimes. Selection of genotypes based on these component-traits were rewarding and seems to be better selection-criteria. Finally, we can end-up with superior-genotypes suitable for intermittent-drought conditions.


Introduction
Grain production is the basis of global food security and is indispensable for feeding mankind worldwide; it needs a significant increase in output in the coming decades to sustain the food and feed requirements (Mesterhazy et al. 2020). As sustainability is the key factor in the production, globally rice, wheat, and maize are the three leading food crops; together they directly supply more than 50% of all calories consumed by the entire human population (FAO 2016). Among these, rice is a major cereal food crop in Asia and feeds more than two-third of a global population (Dowling et al. 1998;Fahad et al. 2019). Asia is the biggest producer and consumer of rice, accounting for 90% of the world's production and consumption (Khahani and Hittalmani 2015). About 55% of the rice area is irrigated and accounts for 75% of total rice production across the globe (Fahad et al. 2019). It signifies that rice is a water intensive crop and requires 5000 L of water per kg rice grain production (Hittalmani 2010;Jana et al. 2018). Rice production is dwindling due to many abiotic and biotic factors (Korrens et al. 2017;Rawte et al. 2021). To meet the evergrowing food demands of the population, it is critical to increase the productivity levels of rice by combating abiotic and biotic stresses (Roberts and Mattoo 2018).
Among the abiotic stresses, drought is the period of no rainfall or irrigation that affects plant growth and is a major production constraint for about 50% of the world production area of rice (Fahad et al. 2017). It is a major prevailing limiting factor affecting rice productivity in the Asian-Pacific region and reported to affect the area of 56 million hectares under different rice ecosystems (46 Mha: rainfed lowland and 10 Mha: rainfed upland) (Pandey et al. 2007;Bhandari et al. 2020). Varying intensities of drought stress at different developmental stages of crop via. seedling, vegetative and reproductive stages are quantified (Price and Courtois 1999;Tripathy et al. 2000;Nguyen and Bui 2008;Xu et al. 2011). In Asia itself, drought is reported to affect approximately 42 Mha of rice production (34 Mha: rainfed lowland and 8 Mha: rainfed upland) (Huke and Huke 1997). Drought causes more damage during reproductive stage moisture stress (RMS), especially during flowering stage (Venuprasad et al. 2007;Serraj et al. 2009) reducing the number of grains per panicle and increasing grain sterility; thereby reducing the grain yield of 34-53% under moderate stress and 65-88% under severe stress in comparison to irrigated condition (O'Toole 1982;Kumar et al. 2008;Torres and Henry 2018;Bhandari et al. 2020). Globally, rice production is predicted to be further challenged by an erratic and increasing frequency; even more severe drought due to climate change (Wassmann et al. 2009). Targeted higher productivity with climate resilience is thereby essential to stabilize rice production by creating climatesmart varieties for adverse ecologies.
The efforts are needed to address the problem of water scarcity by breeding new genotypes tolerance to moisture stress conditions. The alternate option is the growing of aerobic rice varieties, which can be grown in rainfed conditions (Hittalmani and Shivashankar 1987;Atlin et al. 2004;Zhao et al. 2010;Babu et al. 2011;Liu et al. 2019). Considerably, the incidences of pest and disease are less in aerobic rice as compared to wet land conditions. The pests like stem borer, leaf folders and brown plant hoppers and diseases like blast brown spots are the major ones. All the pests that require water to perpetuate and proliferate are precluded and the diseases that require high humidity and temperature do not proliferate. Also, the diseases that are transmitted through water are prevented (https:// www.kisankraft.com/wp-content/uploads/2020/06/ Aerobic-Rice). Moreover, before taking any breeding programme, a breeder should have a thorough knowledge of nature and magnitude of genetic variability, heritability, genetic advance as per cent mean (GAM) and traits associations with grain yield in crop species under different moisture regimes (Lone et al. 2019). The multi-traits pyramided genotypes are thought to serve the problem of moisture stress conditions to a greater extent. Most of the economically important traits are complex and are the result of the interaction of a number of component traits. Understanding the relationship between grain yield and its component traits is most important step for making the best use of these relationships in selection. Grain yield is a complex trait; the direct selection of genotypes based on yield is misleading (Eltaher et al. 2021). Hence, the selection of genotypes suitable for reproductive stage moisture stress conditions based on component traits is most meaningful. The genotypes possessing higher genetic divergence would be responsible for improvement as they were likely to produce required recombinants upon crossing of such genotypes.
Breeding efforts at University of Agricultural Sciences, GKVK, Bengaluru has led to the development of set of F 9 generation marker-assisted multitraits pyramided (MAP) lines with different combinations of gene/QTLs for drought tolerance, bacterial blight and blast disease resistance. However, these lines have not been characterised at phenotypic level under aerobic and reproductive stage drought stress. The present investigation was thus aimed to (i) assess the genetic variability among the MAP genotypes under aerobic and RMS conditions (ii) study the traits association for grain yield and related traits in MAP genotypes under aerobic and RMS conditions.

Plant material used and design of the experiment
The present study comprised of 21 F 9 generation marker assisted multi-traits pyramided genotypes obtained by crossing divergent parental lines 'RB6' (derived from cross RpBio 9 BL 122, resistant to leaf blast and bacterial blight diseases with fine grain quality) and 'QRT25' (derived from the cross IR64 9 Azucena, drought-tolerant, high water use efficiency, high yielding with medium bold grain quality). F 1 plants were obtained and forwarded through marker assisted selection (MAS), genotypes were selected in each generation based on grain yield per plant (Haradari 2013;Uday 2016). The list containing the gene/QTLs possessed by the pyramided lines along with their parents (RB6 and QRT25) have been presented in Table S1. Twenty-one marker assisted multi-traits pyramided genotypes along with the parents (RB6 and QRT25) and check (MAS 946-1) were evaluated for grain yield and other component traits under two moisture regimes namely aerobic (A) as control and reproductive stage moisture stress (RMS) as test environment during summer-2017 and Kharif-2017-18 with two replications in randomized complete block design (RCBD) at experimental plots of Genetics and Plant Breeding, 'K' block, UAS, GKVK, Bengaluru. The experimental plots mainly consisted of sandy loam type of soil profile. MAS 946-1 (Sharada) is used as check; it is a medium fine grain aerobic rice variety, which was released and widely adopted in Karnataka. The variety which is the progeny of the cross between IR64 and Azucena (drought tolerant genotype). It has good root system, plant stand, vigour and tolerate brief water stress at both vegetative and reproductive stages.
Reproductive stage moisture stress (RMS) The standard crop management practices were followed to grow the crop during both seasons. The weather data during the crop growth seasons has been presented in Table S2. The crop management practices were same till 75 days after sowing (DAS) in both moisture regimes. The water supply was withheld from 75 DAS in RMS condition to drain out available soil moisture and reproductive stage moisture stress was imposed at 80 DAS for the period of 15 days ( Fig. 1) but in the aerobic condition, need based irrigation has been given throughout crop cycle in both the seasons (Fig. 1). The data pertaining to moisture content at different intervals under aerobic and RMS was presented in Table S3.

Phenotypic evaluation
The observations were recorded on five randomly selected plants from each marker assisted pyramided genotype in each replication from both aerobic and RMS conditions. Mean values were utilized for statistical analysis and the traits observed for recording the information are: plant height-PH (cm), days to 50% flowering-DFF, days to maturity-DM, productive tillers per plant-PT, panicle length-PL (cm), panicle exertion-PE (cm), flag leaf length-FL (cm), spikelet fertility-SF (%), 1000-grain weight-1000-GW (g), biomass yield-BM (kg/ha), harvest index-HI (%) and grain yield-GY (kg/ha).

Statistical analysis
Path analysis was carried out for both the seasons (summer-2017 and Kharif-2017-18) and both moisture regimes (aerobic and RMS) separately employing Windostat v8.1 statistical software. Levene's test has been carried-out using SPSS v26 software to test the homogeneity of variances among both seasons' data under two different moisture regimes. If the test is significant, the variances of both the seasons differ significantly; if it is non-significant then both the season's variances are approximately equal or same. Hence, the decision on combining/pooling of data from both the seasons can be done. Due to nonsignificance of Levene's test, the summer and Kharif seasons data have been pooled within their respective moisture regimes and the Best Linear Unbiased Predictor (BLUP) values have been estimated via. Multi Environment Trail Analysis with R (META-R) statistical software. Further statistical analyses have been carried-out using BLUP values. BLUP-based analysis of variance (ANOVA) was done further to decide whether both the conditions (aerobic and RMS) data can be pooled or not and also to know statistical significance among the genotypes for various characters using Cropstat v7.2 statistical software. Phenotypic variation via. box plots, frequency distribution curves, estimation of genetic variability parameters [Phenotypic and Genotypic Coefficient of Variation (PCV and GCV), heritability (h 2 ) and genetic advance as percent mean (GAM)] and correlation plots have been deduced using 'R v4.1.0' statistical tool.
Formulae used for estimation of variability parameters are: a. PCV and GCV were estimated as per Burton and Devane (1953) PCV b. Heritability in broad sense (h 2 (b)) was estimated as per Allard (1960) c. Genetic advance was estimated as per Johnson et al. (1955) GA Further, the GA as percent of mean (GAM) was estimated where r2P = Phenotypic variance, r2G = Genotypic variance, X = General mean of character, K = Selection differential at 5% (2.06) selection intensity, h 2 = Broad sense heritability estimate, rp = Phenotypic standard deviation.

Results and discussion
Drought stress has emerged as an essential part of research and improvement of drought tolerance in plants is a thought-provoking task due to the complexity of these traits. Yield is one of the most important and complex trait; primarily affected by drought and leads to greater reduction in grain yield. It is both regulated by genes via quantitative trait loci (QTLs) and even influenced by external environmental factors (Wang et al. 2012;Zeng et al. 2017). Selection of genotypes for prevailing environment solely based on yield is misleading. To overcome this complexity of decision, it is necessary to know association of traits for grain yield and it can act as strong evidence and most precise selection criteria.
Path coefficient analysis as a selection criterion for growth and yield contributing traits: Path analysis was carried out for both the seasons' data under both the moisture regimes (Aerobic and RMS) separately for all the studied characters using phenotypic coefficient to ascertain the direct and indirect effect of the yield components on grain yield as suggested by Wright (1921) and Dewey and Lu (1959). Path analysis is a powerful tool to understand the association among different variables more clearly as recorded by simple correlation values by partitioning the overall association of particular variables with dependent variables into direct and indirect effects. The results of path coefficient analysis for studied traits under aerobic (A) and reproductive stage moisture stress (RMS) during summer-2017 (Table S4) and Kharif-2017-18 (Table S5)  hereafter; mainly to focus on core contributing traits and to restrict our discussion to emphasise more on these traits.

Levene's test of homogeneity of variances
The results of Levene's test for selected traits have been presented in Table 1. Levene statistic in the table (1) represents F-test value at df1 and df2 degrees of freedom. Non-significance of Levene's statistic based on p-value ([ 0.005) has been observed for all the phenotypic traits i.e., DFF, DM, PT, PE, SF, 1000-GW, BM, HI and GY across the seasons in both aerobic and RMS conditions separately. The positive sign of acceptance of Levene's homogeneity of variances, the trait variances of summer and Kharif seasons found to have equal variances in their respective moisture regimes (aerobic and RMS) (Levene 1960;O'Neill et al. 2002;Kim et al. 2018). Hence, the data from the summer and Kharif seasons have been pooled within their respective moisture regimes separately.
Estimation of BLUP value for pooled data BLUP, a standard method for estimating random effects of a mixed model (Robinson 1991); was used for estimating genetic merits. Due to the non-significance of Levene's test, the pooled data has been used to estimate combined BLUP values for all the phenotypic traits i.e., DFF, DM, PT, PE, SF, 1000-GW, BM, HI and GY under aerobic and RMS conditions separately (Table S3 and S4 respectively). The estimated BLUP values have been used for further analyses hereafter.

ANOVA based on BLUP values for yield and yield attributing traits
The results of ANOVA based on BLUP values under aerobic and RMS conditions revealed the existence of highly significant differences between aerobic and RMS conditions for all the phenotypic traits i.e., DF, DM, PT, PE, SF, 1000-GW, BM, HI and GY ( Table 2). The prevailing moisture conditions for yield and yield attributes were adversely affected on the genotypic performance and indicating significant variability among the pyramided genotypes (Uaday 2016; Bhandari et al. 2020). The mean performance among the genotypes varying significantly for all the traits  deducing the presence of higher variability; the presence of variability is a pre-requisite factor for any plant breeding activity (Uday 2016;Bhandari et al. 2020;Harijan et al. 2021). Even the genotype 9 condition interaction also showed significant differences for the traits i.e., PT, PE, 1000-GW, HI and GY indicating the genotypic performance is better restricted to particular moisture condition (Uday 2016) (Table 2). Significant differences among the genotypes due to differential expression of gene/QTLs in MAP genotypes under aerobic and RMS conditions (Yadav et al. 1997;Davierwala et al. 2001;Shen et al. 2001;Sun et al. 2001;Wang et al. 2001;Zhang et al. 2001;Steele et al. 2006;Haradari 2013;Uaday 2016;Haradari et al. 2017;Bhandari et al. 2020). Selection can be operated on available variability to select the suitable MAP genotype for prevailing/intermittent drought conditions; thereby achieving the sustainable production of grain yield under adverse climatic conditions. The significant results of BLUP-based ANOVA for the condition is strong evidence to conclude that the data on aerobic and RMS conditions have to be maintained separately and hence further analysis is carried-out individually for aerobic and RMS conditions. From among the pyramid combinations, the line with qRT2 ? qRT9 ? qWUE2 ? Pi-1combination has given higher grain yield, whereas the combination with xa5gene alone in the pyramid combination was found to yield lower grain yield based on BLUP values under both the moisture regimes. The differential performance of the pyramid combinations emphasises the new insights into their response under different moisture regimes.

Phenotypic variation of the traits among MAP genotypes under aerobic and RMS conditions
Distribution of the measured phenotypic traits via.

Box plots
Box-plots based on BLUP values for the phenotypic traits i.e., DF, DM, PT, PE, SF, 1000-GW, BM, HI, and GY presented in Fig. 2. The plots depicted the mean phenotypic differences under two moisture regimes (i.e., aerobic and RMS) and the performance of genotypes varied significantly except for days to 50% flowering and days to maturity (Fig. 2). Overall, the traits range was higher in aerobic compared to RMS; the traits PT, PE, 1000-GW, and GY exhibited symmetric distribution while the traits DF, DM, SF, BM, and HI exhibited skewed distributions under aerobic condition (Bhandari et al. 2020). Under RMS, all the traits values decreased in comparison to aerobic condition; the traits PT, 1000-GW and GY exhibited symmetric distribution under RMS also (Fig. 2). The mean consistent performance of genotypes under aerobic and RMS conditions is due to the presence of drought responsive gene/QTLs in the marker Fig. 2 Boxplots of nine phenotypic variables within MAP genotypes under aerobic and RMS conditions assisted pyramided genotypes (Haradari 2013;Uday 2016). The traits like PT, 1000-GW and GY distributed symmetrically under both moisture regimes can serve as strong selection criteria to select the genotypes for adverse conditions.

Distribution of the measured phenotypic traits via. histogram
Phenotypic traits variation via. histogram analysis among the MAP genotypes under aerobic and RMS conditions have been depicted separately using BLUP values and also represented the parental (RB6 and QRT25) and check MAS946-1 values in the histograms (Figs. 3 and 4). The frequency distributions for all the studied traits are found to be normal and following a standard normal distribution. The number of days required for flowering and maturity were found to be quite less as compared to parents (RB6 and QRT25) and check (MAS946-1) under both moisture regimes (Figs. 3 and 4). As earlier reports suggest that the significant delay in days to flowering and days to maturity under reproductive stage drought as compared to aerobic rice conditions (Lafitte et al. 2006;Saikumar et al. 2016). The earliness in flowering and maturity is due to the recovery of transgressive segregants among the MAP genotypes (Haradari 2013). Earliness is one of the mechanism for drought escape and/or tolerance, hence these MAP genotypes can serve as a good source of breeding for adverse conditions (Haradari 2013;Uday 2016;Torres and Henry 2018). From among the parents and check, the mean performance of QRT25 is found to be better in aerobic as well as in RMS conditions since it is possessing higher mean values for the majority of the traits. Even though the flowering time of QRT25 is very late but better performance could be due to the drought responsive QTLs as it possesses root related QTLs (qRT2, qRT7 and qRT9) and water use efficiency QTL (qWUE2) (Yadav et al. 1997;Shen et al. 2001;Zhang et al. 2001) compared to other parent (RB6) and check (MAS946-1). Overall, the frequency distribution for all the traits is normal indicating to possess all kinds of combinations of gene/QTLs among the MAP genotypes (Haradari 2013). Hence, the histogram based selection of MAP genotypes for adverse conditions is additional support to the results of box plots. In any crop improvement programme, the availability of genetic variability is a first and foremost prerequisite (Venkataramana and Hittalmani 2003;Uday 2016;Bhandari et al. 2020;Harijan et al. 2021). The economically important characters like grain yield; a well-established complex character and influenced by the large number of genes that are greatly influenced by environmental factors. The genetic variability observed is the average of total hereditary effects of concerned genes as well as the environmental influence. Hence, the variability is partitioned into heritable and non-heritable components with suitable genetic variability components such as phenotypic coefficient of variation (PCV %), genotypic coefficient of variation (GCV %), heritability in a broad sense (h 2 bs %) and genetic advance as per cent mean (GAM). The estimation of these variability parameters assists the breeder in achieving the projected crop improvement by operating selection on available variation.
In the current study, the genetic variability parameters viz., PCV, GCV, heritability and GAM for the phenotypic traits i.e., DF, DM, PT, PE, SF, 1000-GW, BM, HI and GY under both aerobic and RMS conditions have been presented in Table 3. Among all the studied characters, the highest PCV is observed than GCV; which indicated the role of environment on the expression of characters (Manjappa et al. 2014;Uday 2016;Harijan et al. 2021). The traits like productive tillers, panicle exertion, 1000-grain weight, biomass yield, harvest index and grain yield possessed high PCV and GCV with moderate to high heritability and high GAM under both moisture regimes. The spikelet fertility possessed moderate to high PCV and GCV with high heritability and GAM. This high heritability coupled with high GAM indicates that most likely the heritability is due to additive gene effects and the selection for aforesaid characters in drought/moisture stress conditions is found to be highly effective (Courtois et al. 2003;Singh and Narayanan 2017). These traits are highly amenable for selection leading to high genetic variability and transmissibility while taking efforts towards crop improvement for grain yield under reproductive stage moisture stress condition (Singh et al. 2011;Haradari 2013;Manjappa et al. 2014).

Association of yield and yield attributing traits under aerobic and RMS conditions among the MAP genotypes
The correlation coefficient estimates the degree and direction of association between pair of characters and  Fig. 5a and b). The existence of negative correlation among these traits is due to the prolonged period of flowering and maturity. Non-emergence of panicle from leaf sheath has an adverse effect on spikelet fertility, 1000-GW thereby on grain yield (Chakrabort and Chaturvedi 2014; Khahani and Hittalmani 2015). The selection of genotypes based on these component traits (DF, DM, PT, PE, SF, 1000-GW, BM, and HI) in addition to grain yield was rewarding; we can robust the selection criteria and end up with superior genotypes suitable for an adverse climatic condition to attain sustainability in production.

Conclusions
Drought is particularly damaging during the reproductive stage and reduces the grain yield drastically. Hence, the selection of genotypes solely based on grain yield is misleading. In our findings, path analysis identified nine phenotypic traits (days to 50% flowering, days to maturity, productive tillers, panicle exertion, spikelet fertility, 1000-grain weight, biomass yield, harvest index and grain yield), one can rely upon these traits to select drought tolerant genotypes under moisture stress conditions. Non-significance of Levene's test of homogeneity of variances among the data from summer and Kharif seasons lead to data pooling within their respective moisture regimes and estimation of BLUP values. ANOVA based on BLUP values revealed significant differences for moisture conditions and genotypes for all the nine phenotypic traits. Phenotypic variation of traits via. box plots and histograms depicted the mean phenotypic differences for all the nine phenotypic traits among the MAP genotypes under both moisture regimes. The transgressive segregants have been observed among the MAP genotypes for early flowering and maturity. Majority of the traits possessed moderate to high GCV and PCV with moderate to high heritability and high GAM under both moisture regimes. Grain yield possessed a positive correlation with all the phenotypic traits under consideration. The selection based on these component traits is rewarding and we can end up with superior genotypes suitable for diverse climatic conditions to attain sustainability in production.