Novel and Transgressive Salinity Tolerance in Recombinant Inbred Lines of Rice Created by Physiological Coupling-Uncoupling and Network Rewiring Effects

Abstract

The phenomenon of transgressive segregation, where a small minority of recombinants are outliers relative to the range of parental phenotypes, is commonly observed in plant breeding populations. While this phenomenon has been attributed to complementation and epistatic effects, the physiological and developmental synergism involved have not been fully illuminated by the QTL mapping approach alone, especially for stress-adaptive traits involving highly complex interactions. By systems-level profiling of the IR29 × Pokkali recombinant inbred population of rice, we addressed the hypothesis that novel salinity tolerance phenotypes are created by reconfigured physiological networks due to positive or negative coupling-uncoupling of developmental and physiological attributes of each parent. Real-time growth and hyperspectral profiling distinguished the transgressive individuals in terms of stress penalty to growth. Non-parental network signatures that led to either optimal or non-optimal integration of developmental with stress-related mechanisms were evident at the macro-physiological, biochemical, metabolic, and transcriptomic levels. Large positive net gain in super-tolerant progeny was due to ideal complementation of beneficial traits while shedding antagonistic traits. Super-sensitivity was explained by the stacking of multiple antagonistic traits and loss of major beneficial traits. The synergism uncovered by the phenomics approach in this study supports the modern views of the Omnigenic Theory, emphasizing the synergy or lack thereof between core and peripheral components. This study also supports a breeding paradigm rooted on genomic modeling from multi-dimensional genetic, physiological, and phenotypic profiles to create novel adaptive traits for new crop varieties of the 21st century.

Keywords: Omnigenic Theory; genetic network rewiring; genetic novelty; physiological and biochemical synergy; salinity stress; transgressive segregation.

FIGURE 1

Salinity tolerance during vegetative growth window (i.e., tillering stage; V4 to V8) across the minimal comparative panel representing the full phenotypic range at EC = 9 dS m^–1 (IRRI) and EC = 12 dS m^–1 (Texas). (A) Comparison of plant health in control (C) and stress (S) experiments after 6 days (144 h) at EC = 12 dS m^–1. The outliers FL510 (super-tolerant) and FL499 (super-sensitive) are highlighted compared to tolerant parent Pokkali (Saltol donor), sensitive parent IR29, and tolerant sibling FL478 (see Table 1). Differences in injury and growth were evident particularly between the transgressive FL510 and FL499. (B) Individual physio-morphometric scores were normalized and combined as Aggregate Phenotypic Score (APS), which assumes equal weights of each parameter that includes Electrolyte Leakage Index (ELI) and its ratio (ELI ratio) at first injury (72 h) and maximum injury (144 h), shoot biomass ratio (stress/control), stomatal conductance ratio (stress/control), Na⁺/K⁺ at maximum injury (Na/K 144/0), Na⁺ ratio at maximum injury (Na 144/0), K⁺ ratio at maximum injury (K 144/0), slope of K⁺ and Na⁺ changes (control/stress), change in peroxidase activity (change-POX; stress/control), and change in lipid peroxidation (change-LP; stress/control). APS was plotted against the inverse of SES for direct proportionality. (C) Heat map showing the gradients of the normalized physio-morphometric scores across the comparative panel, presented from the worst to the best SES (inverted) according to the bar graph distribution across the population in (B). From this figure, traits that drag down the APS of a given genotype are more apparent. The genotypes with an asterisk (*) indicate the most transgressive lines in the population.

FIGURE 2

Neighbor-joining dendrograms showing two layers of similarities based on the phenotypic matrix (APS and SES). Each genotype in the dendrogram is suffixed with their SES. The individual genotypes (parents, RILs) that were investigated to understand the physiological mechanisms are highlighted in blue boxes. AU (blue) refers to the adjusted P-value of clustering, and BP (orange) refers to the bootstrapping value of “pvclust” package. The y-axis of each dendrogram refers to the degree of separation (height) between groups. (A) Similarities based on the entire matrix (components of APS + SES). Dendrogram shows the poor genotypes represented by IR29, FL454, and FL499 forming distinct clades from the good genotypes represented by Pokkali, FL478, and FL510. (B) Similarities based on APS components relevant to Na⁺ sequestration (ELI, Na⁺, and K⁺ contents at control and 144 h, Na⁺/K⁺ ratio). Dendrogram shows the tendency for individuals carrying the Pokkali Saltol allele to cluster together within one large clade. In both (A,B), the super-sensitive FL499 had the earliest divergence.

FIGURE 3

Real-time growth profiling of the sensitive parent IR29, tolerant parent Pokkali, sensitive RIL FL454, tolerant RIL FL478, super-sensitive RIL FL499, and super-tolerant RIL FL510 at EC = 9 dSm^–1. (A) Growth curves as a function of changes in plant area and plant height (y-axis, as cm² and cm, respectively) during the 18-day period of stress (x-axis, plotted as the vegetative growth stages V2 to V12), calculated based on pixels captured by RGB and hyperspectral camera. Note the time of forking between the control and stress curves and the angle of the fork, reflective of growth plateauing. (B) Variation in hyperspectral variance at 243 wavelength bands ranging from 550 to 1,700 nm. Heat maps of hyperspectral variances are potential indicators of overall plant health and stress injuries. The overall patterns in hyperspectral variances mirror the patterns revealed by the growth curve analysis.

FIGURE 4

Shotgun LC-MS/MS metabolite profiles across the representative phenotypic classes at the point of maximum stress (7 days) at EC = 12 dS m^–1. (A) Principal component analysis (PCA) of metabolites with common occurrence regardless between control and stress in sensitive parent IR29, tolerant parent Pokkali, sensitive FL454, tolerant FL478, super-sensitive FL499, and super-tolerant FL510. PC-1 and PC-2 separated the good RILs FL478 and FL510 from the poorest RIL FL499, accounting for 27.15% and 21.38% of phenotypic variance, respectively. The PCA also shows high similarities between the two parents. (B) PCA of metabolites with differential abundances between control and stress at p < 0.05. Based on PC-1 and PC-2, which accounted for 26.89% and 22.14% of phenotypic variance, respectively, the super-sensitive FL499 was quite distant from the super-tolerant FL510, despite the high level of similarities between the parents, hence non-additive and transgressive. The tolerant FL478 was more similar to Pokkali than to IR29, and any of its siblings. (C) K-means clustering heat maps and dendrograms showing the patterns of metabolite co-abundances across genotypes. Heatmap highlights a cluster that is highly abundant in each genotype. Clustering by genotype shows the similarity of IR29 and FL510, especially in clusters 1, 2, and 4. Cluster 5, highlights the differences that are contributory toward contrasting phenotypes.

FIGURE 5

Temporal profiles of salinity stress (EC = 12 dS m^–1) transcriptomes in parents (IR29 = sensitive; Pokkali = tolerant) and their RILs (FL454 = sensitive; FL499 = super-sensitive; FL478 = tolerant; FL510 = super-tolerant). Parallel comparison of transcriptomes across genotype shown as K-means++ clusters included 14,696 unique transcript loci. The super-tolerant FL510 is the most unique, having gradual changes in expression. The other genotypes have large clusters of co-expressed genes with drastic upregulation or downregulation across time.

FIGURE 6

Comparison of transcriptional co-expression networks across parents and their RILs (FL454, sensitive; FL499, super-sensitive; FL478, tolerant; FL510, super-tolerant). Models represent the networks for: (A) OsCML27, (B) OsHKT7, (C) OsMTI4A, and (D) OsMPS, with important roles in developmental and stress-related responses. The temporal co-expression plots of the bait or core genes (red line) with their cohort genes (gray lines) are given for each network. Connectivity of cohort genes was based on Pearson Correlation Coefficients by mutual rank. Cohort genes are represented by nodes and co-expression is reflected on the edges.

FIGURE 7

Relatedness of the genotypes according to the gene networks in Figure 6. Similarities among the genotypes were assessed by coefficient of correlation among median expression values of network components (A), and by the number of shared cohort genes (B).

FIGURE 8

Metabolic similarities based on transcriptome profiles. KEGG models of metabolic pathways are shown on the left panel for glycolysis (A), TCA cycle (B), starch/sucrose metabolism (C), and nitrogen assimilation (D). Transcripts for enzymes in each step were mapped against each pathway. Pathway activities and the pattern of similarities across genotypes are depicted by hierarchical clustering dendrograms based on maximum log₂-fold change in transcript abundances at the time-point with the widest difference from control. FL510 and Pokkali share similarities in repression of the carbon metabolism pathways, while nitrogen assimilation is induced in both genotypes. In comparison, the poor genotypes cluster together.

FIGURE 9

Hypothetical model of physiological coupling and uncoupling in transgressive segregants for salinity tolerance across the IR29 × Pokkali recombinant inbred population based on macro-physiological, biochemical, and molecular profiles according to de los Reyes (2019). This model proposes that the novelties of FL510 and FL499 are due to the coupling in the progeny of the good properties coming from either parent or uncoupling of bad properties from the good properties from the same parent. On top of the core mechanisms that contribute to a large proportion of phenotypic variance for defense potentials, each parent has their own characteristics that may or may not be beneficial under stress. Benefit from IR29 would be its superior growth and development potentials. Pokkali offers many stress defense mechanisms including salt exclusion. Combining the physiological potentials of parents with the reconfigured (non-parental) properties led to positive or negative coupling and uncoupling effects in RILs.