tolerance

2 Highlights: 26 We identified QTL for salt tolerance response for two different developmental stages of rice 27 plant and detected a significant contribution of cytoplasm-nuclear genome interaction for few 28 traits.

the chromosome that is being tested for QTL mapping. We also tested cytoplasm as an additive 2 3 6 and interactive covariate in the QTL model using likelihood ratio tests and retained factors in the 2 3 7 models when significant. To test whether our selection of a test cohort imposed significant 2 3 8 population structure for the reproductive stage treatment we assigned each F 2 family into one of 2 3 9 the following categories: tolerant, intermediate and sensitive. We incorporated selection cohort 2 4 0 as a covariate while building QTL model for traits at reproductive stage. However, we did not 2 4 1 find any significant effect of selection cohort on QTL models and therefore did not include 2 4 2 selection cohort as a covariate for further analysis. Significance thresholds for QTL were 2 4 3 determined for each trait by 1000 permutations (alpha=0.05) and QTL peaks that passed the 2 4 4 threshold were considered for further analysis. Permutations were stratified by cytoplasm for the 2 4 5 QTL models where cytoplasm was considered as a covariate. We also evaluated the normality of 2 4 6 the QTL model residuals and found this assumption has been violated of for trait TK. Confidence intervals (1.5 LOD drop) for each QTL were calculated using the lodind function of 2 4 8 the R/qtl package (Broman et al., 2003) expanding it to a true marker on both sides of the QTL. Codes for QTL analysis are available in GitHub repository (Haque, 2019). In order to identify candidate gene models within a given QTL interval, we integrated the genetic 2 5 3 and physical map based on the marker order of genetic map. We first pulled out the genetic 2 5 4 markers flanking a given QTL confidence interval and their basepair positions to define the 2 5 5 physical interval on the genome for that QTL. Gene models in these physical QTL intervals were 2 5 6 retrieved using the structural gene annotation of the rice Nipponbare reference genome from 2 5 7 Phytozome 9. We used the Gene ontology (GO) annotated for each gene model of this reference 2 5 8 (Phytozome 9) for GO enrichment analysis. We then tested for the enrichment of GO terms for 2 5 9 each QTL interval using the classical Fisher's exact test available in the topGO (Alexa and  Phenotypic traits vary between cross direction in both developmental stages: 2 6 3 . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 3, 2020. ; https://doi.org/10.1101/2020.03.01.971895 doi: bioRxiv preprint 1 0 We used a bi-directional F 2 : 3 experimental design to examine the effect of salinity on various 2 6 4 growth, yield and physiological parameters of rice as well as the role of cytoplasm on these 2 6 5 traits. Rice is most susceptible to salinity during seedling and reproductive growth stages.

6 6
Therefore, we focused on phenotypes at these two stages that can potentially mediate stress 2 6 7 during salinity treatment. For the seedling stage, we scored traits that were related to survival, Spikelet Fertility (SF).

7 4
We found striking difference between the two parents for many of our measured traits across 2 7 5 both stages of salt treatment (Table 1). As reported in previous studies, we have found that Horkuch is more tolerant to salinity (measured by SES score) compared to IR29 which is highly 2 7 7 sensitive for salt at the seedling stage (p-value < 0.01). For reproductive stage salinity treatment, Horkuch parent had higher number of ET and also higher SF, FGW and HI. It has been reported that complex cytoplasm-nuclear interaction can alter plant phenotypes 2 8 1 (Joseph et al., 2013a). Therefore, we aimed to test whether cytoplasm has a role on the 2 8 2 performance of this mapping population under salinity stress. We found many traits were 2 8 3 significantly different as a result of cytoplasmic background (Table 1). Tchlr, TNa, TK and K/Na 2 8 4 were found to differ significantly among the Horkuch and IR29 cytoplasms at the seedling stage  The generation of extreme phenotypes in a crossing population, or transgressive segregation, has 2 8 9 been reported in many plants which may be due to the effect of complementary genes, over-2 9 0 dominance or epistasis (Rieseberg et al., 1999). Here, we further investigated the role of cytoplasm 2 9 1 on the transgressive segregation of these traits. In this experiment we found that SES, SRWC, 2 9 2 TChlr, TK, FGN, FGW clearly segregate transgressively (Table 1). Interestingly, TK, TChlr, 1 1 FGN and FGW showed significant differences in the two reciprocal crosses with respect to the 2 9 4 segregation pattern. The distribution of TK showed a strong bimodal pattern with very little 2 9 5 overlap of distribution between cytoplasmic backgrounds (Figure 1). This observation is 2 9 6 indicative of an important role of cytoplasm on transgressive segregation for this population. To understand the partitioning of genetic variation we used principal component analysis of 130 2). This indicated substantial genetic correlation among these traits in this population. mechanisms of salinity responses that has some trade-off between two different treatment stages. In our previous study we constructed a linkage map for this reciprocal mapping population by 3 1 0 applying ddRAD technique. Unfortunately, we had high genotyping error and inflated 3 1 1 segregation distortion for the two parental alleles for a given locus. Therefore, we failed to 3 1 2 capture the genome-wide linkage map which resulted in the removal of the entire chromosome 5 3 1 3 for QTL mapping. In this study, we genotyped this reciprocal mapping population using to reduce the complexity of the genome and has been optimized for various plant species to 3 1 8 achieve best complexity reduction. We used this platform for our mapping population to generate 3 1 9 genotype information at ~10 thousand loci that are well-distributed in the rice genome. For analyses, we filtered all the DArTseq SNP loci to obtain polymorphic homozygous SNPs for the 3 2 1 parents and retained only loci that had minimum 50% representation in the population. With 3 2 2 these filter criteria we obtained 2,230 high quality SNPs for this mapping population. Among 3 2 3 . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 3, 2020. ; https://doi.org/10.1101/2020.03.01.971895 doi: bioRxiv preprint 1 2 these loci, 956 markers that showed significant segregation distortion by chi-square test [P value 3 2 4 < 1e-2] were removed. More distorted loci were skewed towards the Horkuch parent than the 3 2 5 IR29 parent. 739 markers that were similar were dropped using a minimum recombination  traits in both stages of treatment therefore we aimed to test (Chi-square test of independence, see Method section) for cytoplasm-nuclear association for each marker in this genetic map. We  We measured eight traits that reflect the survival performance of rice seedlings under salinity 3 4 5 stress. We found six QTL for three traits: SL, RL and TK (Table 2, Figure 3). We found three 3 4 6 significant QTL for SL occurring at qSL.1@183 (reporting a QTL for SL located at chromosome 3 4 7 1 at 183 cM), qSL.3@218, and qSL.5@160. For these QTL, the positive alleles were from 3 4 8 Horkuch parent and we detected no significant cytoplasmic effects or interactions. The QTL at qSL.1@183 had a large effect corresponding to a ~3.5 cm increase in seedling length and was qSL.5@160 had a very larger confidence interval (~ 9 Mbp) with small effect size.

5 3
. CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 3, 2020. ; https://doi.org/10.1101/2020.03.01.971895 doi: bioRxiv preprint 1 3 We identified one large effect QTL for RL at qRL.2@167 with a confidence interval of ~ 4 Mb.

5 4
Here, the Horkuch parent contributed the positive allele. For TK, cytoplasm was a significant 3 5 5 covariate that interacted with two QTL that were detected at qTK.2@45 and qTK.3@203. In 3 5 6 addition, the QTL qTK.2@45 is co-localized with the cluster of genetic loci that showed 3 5 7 significant cytoplasm-nuclear association. The other QTL for TK, qTK.3@204 did not overlap 3 5 8 with the association cluster in chromosome 3 but resided in close proximity. This evidence 3 5 9 suggests a possible role of cytoplasmic-nuclear interaction for this trait. contributed the positive allele. We found significant cytoplasm-nuclear interaction for QTL 3 6 6 model where the IR29 allele had positive effects only for IR29♀ ( Figure 4). However, this QTL 3 6 7 had a very wide confidence interval of ~6 Mbp. We found two co-localized QTL in this same 3 6 8 region including a QTL for FGW and another for SF at qFGW.10@58.5 and qSF.10@59 respectively. These two QTL also had significant interactions with cytoplasm in their for SF this allele not only had positive effect for IR29♀ but also had a negative effect for Horkuch♀. We also found a QTL for HI at qHI.10@104 for which the positive allele was from 3 7 4 IR29. However, this model only had an additive effect of cytoplasm. this QTL model. However, the positive allele from Horkuch performed better in IR29♀. We also 3 7 8 found three PH QTL occurring at qPH.1@215, qPH.3@211 and qPH.5@144, for which the 3 7 9 positive alleles were from the Horkuch parent. The first two QTL models showed only additive 3 8 0 contribution of cytoplasm but the third showed also an interactive effect of cytoplasm. Overall, we found a hotspot of QTL on chromosome 10 for multiple parameters related to yield. The 3 8 2 significant correlation of these traits may reflect related metrics of yield performance in rice.
CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made In this study, we found that some QTL intervals of various traits overlapped, therefore we 3 8 5 annotated these overlapping intervals as QTL clusters. We detected a co-localized QTL at parent. However, TK showed no significant correlation with PH which was causal for the co- where the positive alleles were from Horkuch parent. The fourth QTL Cluster (QC 4) at 3 9 5 10@58:107 was found for four yield related traits including FGN, FGW, SF and HI for which the 3 9 6 positive alleles were from the IR29 parent. To understand the molecular mechanism of salt tolerance, we further investigated the function of For traits at reproductive stage, the QTL qPH.1@215 was significantly enriched with the GO 4 1 5 terms anion and potassium ion transmembrane transport, divalent metal ion transport, 4 1 6 pollination, reproduction process, endoplasmic reticulum and organelle sub-compartment. Another PH QTL, qPH.3@211 showed enrichment for GO terms such as cellular nitrogen 4 1 8 compound metabolism, organic acid transport, mitochondrial membrane and protein complex. The third QTL, qPH.5@144 also showed significant enrichment for nitrogen compound  were significantly enriched with processes such as chromatin assembly, amide biosynthesis, CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 3, 2020. ; https://doi.org/10.1101/2020.03.01.971895 doi: bioRxiv preprint 1 6 and various sodium symporters were enriched for molecular function. DEGs for root tissues were terms such as oxidoreductase, hydrolase activity and voltage gated potassium channel activity.

7
For the reproductive stage of salinity treatment, the expression profile of shoot and root tissues at interaction between cytoplasm and treatment was very low (5 significant DEGs) for root tissue 4 5 0 therefore we only tested enrichment for DEGs in shoot tissue. DEGs for shoot tissues for symporter activity including organic acid:sodium symporter, bile acid:sodium symporter activity.

5 7
These were also enriched for DEGs for the seedling stage shoot tissue. Many of these symporters we had many significantly enriched GO terms for the cellular component of mitochondria.  further tested for enrichment of GOs for the same. We found these genes to be significantly  In this study, we explored the responses of rice to salinity stress at two different growth stages 4 7 0 with a reciprocal mapping population. Among the 14 QTL that we reported, 8 QTL models 4 7 1 . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 3, 2020. ; https://doi.org/10.1101/2020.03.01.971895 doi: bioRxiv preprint 1 7 showed significant effect of cytoplasm. This finding underlines the importance of considering 4 7 2 both organelle and nuclear genome for complex traits such as salinity tolerance.

7 3
Cytoplasmic background may play an important role in trait genetic architecture by itself or 4 7 4 through complex interactions with the nuclear genome. (Joseph et al., 2013b;Lovell et al., 2015;4 7 5 Moison et al., 2010;Tang et al., 2013). Gregorio and Senadhira (1993)  framework where the cytoplasm-nuclear interaction was also considered as a contributor to  Cytoplasmic genome can influence the interaction of alleles from nucleus and cytoplasm and can 4 9 7 favor the evolutionary co-adaptation of high-fitness. In the current study we found a significant 4 9 8 association of cytoplasm for some traits and therefore further tested for non-random interaction of alleles for nucleus and cytoplasm. We found that the QTL qTK.2@45 was a hotspot of (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 3, 2020. ; https://doi.org/10.1101/2020.03.01.971895 doi: bioRxiv preprint 1 8 hotspot on chromosome 3. Both of these QTL models showed significant effect for cytoplasm.

0 2
For qTK2@45, the effect of cytoplasm was mostly additive where Horkuch♀ contributed large 5 0 3 positive effect. On the contrary, for qPH3@211, cytoplasm had an interactive effect. Horkuch 5 0 4 nuclear allele had a positive effect on PH but the effect was even higher for IR29♀ 5 0 5 (Supplementary Figure 3). Taken together, this suggests a significant interaction of nuclear  We detected significant cytoplasm-nuclear linkage of a few markers that overlapped with some 5 1 0 QTL intervals. Therefore, careful consideration is needed in order to select these loci for QTL 5 1 1 pyramiding.

1 2
One important finding in this study is that we have detected multiple co-localized QTL within where multiple trait QTL co-localized. Co-localized QTL can impose constraints on selection for 5 1 6 QTL pyramiding. As an example, we found that QTL cluster 1 had a positive effect for the 5 1 7 Horkuch parental allele for PH and SL. However, a taller plant is not the desired plant 5 1 8 architecture for a breeding program for high-yielding rice varieties since this will lead to over- (2015) reported that percent of shoot length reduction under saline treatment is highly co-related 5 2 1 to saline sensitivity. This conditional relationship between traits results in some possible trade-5 2 2 offs between favorable and undesirable traits. The same logic is applicable for QTL cluster 2 5 2 3 where traits (SL, TK and PH) for these clusters are positively correlated but increased PH is not for different developmental stage. In addition to that, we need to consider the fact that selection 5 2 8 on multiple traits may not be orthogonal due to the complex mechanisms of salt adaptation.

2 9
To understand the molecular mechanism of salt response and the effect of cytoplasm for salt 5 3 0 tolerance we tested for enrichment of GO functions for genes within QTL confidence intervals.

3 1
. CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 3, 2020. ; https://doi.org/10.1101/2020.03.01.971895 doi: bioRxiv preprint 1 9 Both the QTL intervals for TK were enriched with various transmembrane transporter activity, 5 3 2 and potassium ion transmembrane transporter. K + is involved in numerous metabolic process in 5 3 3 plants and excess Na + interferes with the K + homeostasis during salinity stress. To maintain the 5 3 4 cellular homeostasis of K + various potassium transmembrane transporters have been reported 5 3 5 that showed increase salt tolerance in various glycophytes (Tester and Davenport, 2003). For the 5 3 6 reproductive stage, we found most of the QTL intervals for PH, ET, FGW, FGN, SF and HI were 5 3 7 enriched with mitochondria, cytoplasm and organelle related GOs. This supported the 5 3 8 observation that these QTL models also showed significant interaction with cytoplasm. Thus a tolerance or sensitivity to it. This evidence also suggests a plausible explanation why we found 5 5 4 cytoplasm as a covariate in QTL models for this study. These are likely candidates for future 5 5 5 functional genomic studies of salinity tolerance in rice.

6
In this QTL analysis framework, we applied linear mixed model which can handle cytoplasm genotype a moderate number of SNPs that are well-dispersed in rice genome and aimed to select SNPs close to gene space of the rice genome. In our previous study, we had generated a genetic 5 6 1 map on this mapping population by ddRAD technique which failed to capture a significant space 5 6 2 . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made reproductive stage salinity treatment, we were able to detect additional five QTL for PH, ET and 5 6 7 SF. We have also detected one big effect QTL for FGN and FGW in a different chromosome in 5 6 8 this current study due to the fact that in our previous study we failed to capture markers at that region. In addition to that, this framework provided QTL with higher likelihood and tighter 5 7 0 confidence interval and provided better estimation of effect size of each QTL for a given trait. Therefore, this additional detected QTL with high LOD scores and tighter confidence intervals In this study, we aimed to identify genetic loci for salinity tolerance of a rice landrace, Horkuch, 5 7 7 at two sensitive developmental stages. We found 14 QTL for 9 traits under salinity treatment.

7 8
We detected some overlap in the genomic regions affecting traits across developmental stages.

7 9
One chief finding of this study was the significant contribution of cytoplasm on many traits and pyramiding of QTLs that were detected in this study can pave a way to generate high yielding  (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Harvest Index***(ratio) HI 0.26 0.12 0.23 0.32 . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 3, 2020. ; https://doi.org/10.1101/2020.03.01.971895 doi: bioRxiv preprint 2 8 of QTL alleles with cytoplasm for traits such FGW, ET. . CC-BY-NC-ND 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted March 3, 2020. ; https://doi.org/10.1101/2020.03.01.971895 doi: bioRxiv preprint