Strategies to identify and dissect trade‐offs in plants

Trade‐offs between traits arise and reflect constraints imposed by the environment and physicochemical laws. Trade‐off situations are expected to be highly relevant for sessile plants, which have to respond to changes in the environment to ensure survival. Despite increasing interest in determining the genetic and molecular basis of plant trade‐offs, there are still gaps and differences with respect to how trade‐offs are defined, how they are measured, and how their genetic architecture is dissected. The first step to fill these gaps is to establish what is meant by trade‐offs. In this review we provide a classification of the existing definitions of trade‐offs according to: (1) the measures used for their quantification, (2) the dependence of trade‐offs on environment, and (3) experimental designed used (i.e. a single individual across different environments or a population of individuals in single or multiple environments). We then compare the approaches for quantification of trade‐offs based on phenotypic, between‐individual, and genetic correlations, and stress the need for developing further quantification indices particularly for trade‐offs between multiple traits. Lastly, we highlight the genetic mechanisms underpinning trade‐offs and experimental designs that facilitate their discovery in plants, with focus on usage of natural variability. This review also offers a perspective for future research aimed at identification of plant trade‐offs, dissection of their genetic architecture, and development of strategies to overcome trade‐offs, with applications in crop breeding.

research interest in the study of plant trade-offs has steadily increased, resulting in over 1660 peer-reviewed publications in the last decade alone. 1The knowledge about trade-offs in plants have been summarized in recent reviews that discuss growth-defence trade-offs (He et al., 2022) and involved mechanisms (Panda et al., 2021), trade-offs in the context of plant-microbe interactions (Khanna et al., 2021) as well as biodiversity and yield-related tradeoffs (Gong et al., 2022).
Despite the growing interest in plant trade-offs, there seem to be conceptual gaps related to what is meant by trade-off situations, how trade-offs are quantified, and how these quantifications are used to determine the genetic and molecular mechanisms underlying trade-offs.This is largely due to the lack of interdisciplinary effort that is required to integrate methodologies and data across different levels of biological organizations, from genomics, genetics, and physiology to molecular and evolutionary biology.In addition, this is also due to the differences in understanding the type of tradeoffs that can be identified based on particular experimental design and the possibility of dissecting their genetic architecture.
In this review, we provide a critical view on the existing definitions and classifications of trade-offs.We then focus on reviewing the measures used to identify and quantify trade-offs from data on different traits.To this end, we make the distinction between tradeoffs quantified in a single individual across different environments and trade-offs measured in a population in one or multiple environments, based on application of mixed effect modelling framework (Arnold et al., 2019;Careau & Wilson, 2017;Dingemanse & Dochtermann, 2013;Houslay & Wilson, 2017).In doing so, we also draw an important, but not yet explicitly stated, connection between trade-offs and plasticity of traits.We then discuss recent evidence concerning the genetic architecture of plant trade-offs and provide a perspective for future research targeted at mining of the existing and newly gathered data sets to identify the genetic underpinning of plant trade-offs.

| DEFINITIONS AND CL A SSIFIC ATIONS OF TR ADE-OFFS
A trade-off situation denotes the dependence between traits whereby increase in one of the traits leads to decrease in at least one of the others.Before presenting examples of studies that have dissected the genetic underpinning of plant trade-offs, it is important to highlight and discuss the nuanced differences in the existing definitions of trade-offs, the assumptions quantification of trade-offs relies on, and the implications they have in practice.We would like to stress that in the following we use the term individual and genotype synonymously, unless otherwise stated, assuming that the individual refers to a clone or isogenic line (as is the case in many studies with plants).

| The concept of dependence reaction norm
A reaction norm for a trait depicts the trait values expressed by an individual over multiple environments (Figure 1a).Therefore, the reaction norm captures the extent to which the individual responds to changes in the environment with respect to the trait, referred to as phenotypic plasticity (Laitinen & Nikoloski, 2019).For instance, the two traits in Figure 1a remain constant and do not show plasticity across environments E3-E5, but they show pronounced plasticity when considering any other combination of environments.Reaction norms can differ between different individuals (i.e., genotypes), denoting the presence of genotype-by-environment (G x E) interaction for the studied trait.As a result, phenotypic plasticity of a focal trait has been considered a trait itself, and considerable recent efforts have started to unravel its genetic underpinnings (see references in Table 1) (Laitinen & Nikoloski, 2019;Ronnegard & Valdar, 2012).
The dependence between two traits in a single individual over different environments can be described by a dependence reaction norm (Figure 1b), alluded in Stearns (1989) (therein called trade-off function).A dependence reaction norm for two traits in an individual can readily be obtained by projecting the reaction norms of the respective traits (Figure 1a) in two-dimensional coordinate systems in which the axes correspond to the traits (Figures 1b).Every point on the dependence reaction norm then denotes the specific combination of traits in a different environment.Clearly, such a diagram can be expanded to include dependence reaction norms for multiple individuals (lines of different colours in Figure 1c).
From the concept of dependence reaction norm, the following can be easily deduced: 1.For trade-offs to be manifested in an individual, the traits must exhibit plasticity (Figure 1a).Specifically, if one of the tradeoff traits does not change over the considered environments, then no dependence between the traits can be established.
For instance, the two traits in Figure 1a do not exhibit plasticity in environments E3-E5, and therefore do not lead to a trade-off situation in these environments (Figure 1b).
2. The dependence between the traits may change with the considered environments, leading to fleetingness of trade-offs.
Specifically, the trade-off in a single individual may only be manifested in one set of environments, but not in another.For instance, trade-off between the two traits in Figure 1a cannot be identified in environment E3-E5 due to the absence of plasticity in the traits in these environments, as stated above (Figure 1b).
3. Individuals (i.e., genotypes) may exhibit differences in the extent to which given traits are in trade-offs depending on the extent of genotype-by-environment interaction for the respective traits.
For instance, the dependence reaction norms of different individuals do not have to coincide (Figure 1c).This leads to differences in genetic correlations determined from data in different environments (see Section 3).
The visualization facilitated by the dependence reaction norm is instructive and can be extended in three dimensions; however, dependence reaction norms cannot be visualized for more than three traits.

| Classification of definitions and types of trade-offs
The dependence between traits implied by the intuitive definition of Regarding the first criterion, trade-offs can be quantified based on measures of linear or nonlinear associations.This leads to two principal classes -of linear and nonlinear trade-offs (Figure 2a).
For instance, linear trade-offs for two traits can be quantified by Pearson's correlation coefficient (Kirch, 2008).In this regard, linear trade-offs between two traits correspond to negative phenotypic correlations, in line with the intuitive description of trade-offs above (see Section 3 for critical details).The nonlinear trade-offs can further be subdivided into concave or convex, depending on the properties of the mathematical function that describes the dependence between the traits (Figure 2a).Convex trade-offs have been related to life history strategies in plants and other organisms (Bernardes et al., 2021;Stearns, 1989).
Trade-offs can also be classified according to the degree to which they depend on the environment -namely as absolute or relative trade-offs.A trade-off is considered absolute if the respective dependence between the traits holds over all conceivable environments.In contrast, a trade-off is relative if the dependence holds only over a specified set of environments (Figure 2b).This classification has implications on the capacity of statistical approaches to identify trade-offs.
Finally, regarding the third criterion, trade-offs can be classified based on whether they are identified for traits of a single individual over multiple environments, referred to as intraindividual trade-offs (Stearns, 1989), and for individuals in a population in a single environment.The intraindividual trade-offs (reflecting correlated plasticities) may differ between individuals (Figure 2c, left panel), which suggests a genetic component to these types of trade-offs.The trade-offs determined over an entire population can involve traits scored in a single or multiple environments (Figure 2c, right panel).Furthermore, trade-offs can be investigated between individuals from same or different generations (i.e., parents and offsprings), leading to the classification of tradeoffs into within-generation and intergenerational (Dingemanse & Dochtermann, 2013;Stearns, 1989), respectively.
F I G U R E 1 (a) Reaction norms of two traits (y-axis) measured in a single individual in five environments, E1-E5.Each point corresponds to the trait score in the respective environment (x-axis).The reaction norms show that both traits are plastic over the considered environments, with no plasticity E3-E5 (plateauing curves) and plasticity for any other combination of environments.(b) the projection of the reaction norms from panel a onto the space of the two traits leads to a dependence reaction norm.Each point is determined by the trait values in the respective environment.(c) Dependence reaction norms for three individuals, marked in different colours.The black and grey points correspond to the trait values in two respective environments, Ei and Ej.Genetic correlations, marked by dotted lines, depend on the environment, with trait values in environment Ei yielding negative genetic correlations, while those in environment Ej resulting in positive genetic correlations From a biological perspective, three types of trade-offs are considered fundamental depending on the processes that determine trade-offs; namely, trade-offs that result from allocation of resources, trade-offs in mortality due to duration of resource acquisition, and trade-offs due to specialization in a particular environment (Angilletta et al., 2003) (Figure 2d).Although several authors have stressed the need to develop models and theories to explain how these trade-offs arise due to the inherent constraints on the limited resources in relation to the environment (Reznick, 1985;Roff & Fairbairn, 2007;Stearns, 1989), little effort has been made in this area of research, predominantly focused on black-box models or specific cellular systems (e.g., metabolism, see Hashemi et al. (2022)).

| INFEREN CE OF TR ADE-OFFS FROM DATA
Focal traits are characteristics of individuals (genotypes) that are determined by the genetic and environmental factors, and are inherited from parent to offspring.A focal trait can also vary in response to different environmental cues, termed as phenotypic plasticity.
The different response of the genotypes to an environmental cue indicates the presence of genotype-by-environment (G x E) interaction.This indicates that phenotypic plasticity has a genetic basis and can be studied as a trait on its own (Bradshaw, 1965;Laitinen & Nikoloski, 2019).Analogously to phenotypic plasticity, we propose that a trade-off, reflecting the dependence between focal traits in an individual over a gradient of environmental cues, can be seen as a trait itself.This is the case if there is a genetic variability for intraindividual trade-offs in a population.
As a result, the problems of identifying a trade-off from determining the genetic architecture of the trade-off may or may not be decoupled, depending on the experimental design employed.For

| Inference of trade-offs using correlationbased approaches
As noted in Section 2, negative phenotypic correlation between two traits is indicative of a linear trade-off.The phenotypic correlation

White clover
Life history strategies Allelic correlations (Wright et al., 2022) between two traits, captured by the random variables X and Y, is Since traits are shaped by the genotype, environment, and the GxE interaction (Lynch & Walsh, 1998), it is also important to understand the factors that contribute to phenotypic correlations.To this end, when data over multiple individuals and environments are available, multivariate mixed effect models, popularly applied in ecology and behavioural biology research as well as quantitative genetics research (Careau & Wilson, 2017;Dingemanse & Dochtermann, 2013;Houslay & Wilson, 2017), can provide insights in the factors that shape phenotypic correlations.In the following, we provide a brief account of decomposition of phenotypic correlations in the bivariate case, in an attempt to describe the factors that shape such correlations in a population of individuals.
In a univariate setting, the phenotype, z X,ij , for the trait X scored in replicate q for individual p can be modelled as where X denotes the grand mean for trait X, i X,p is the deviation of the mean of individual p from the grand mean, and e X,pq represents the residual error.In mixed effect models, i X,p is referred to as a random intercept that models the differences in mean responses between individuals; it is assumed to be normally distributed with zero mean and variance, var i X , denoting the between-individual variance.Similarly, the residual error is assumed to be normally distributed, with zero mean and variance, var X , denoting the withinindividual, also called residual, variance.Note that this elementary mixed-effect model can be expanded to include individual-specific responses to a changing environmental cue, leading to the concept of random regression (Henderson Jr., 1982) popularized by studies of phenotypic plasticity (Arnold et al., 2019).Unlike constant coefficients in standard regression for a trait in terms of an environmental cue, the coefficients in random regressions are allowed to vary between individuals.From this model, the phenotypic variance, var z X , can be decomposed as var i X + var X , whereby the ratio , denotes the repeatability of the trait.We note that the the residual correlation (Figure 2).From Equation (4), it is clear that the phenotypic correlation depends on the repeatability, r X and r Y , of the traits.Studies often assume that the residual correlation, r (X, Y) , is negligible and, as a result, the phenotypic correlation is often hypothesized to be of same sign as the between-individual correlation.
However, Equation 4 can be used to derive the conditions under which the phenotypic correlations differ in sign from the between-individual correlation, and provide further masking of trade-offs.Therefore, since the negative between-individual correlations remove the effect of the residual correlations, they may provide a more reliable indicator of linear trade-offs than phenotypic correlations.Yet, it remains to be investigated whether negative residual correlation, caused by correlated measurements errors or correlated phenotypic plasticity, reflects a trade-off between the traits.However, it must be noted that between-individual phenotypic correlation is not fully genetically determined.To highlight this point, we consider the decomposition of the between-individual correlations.We note that that i Y,p can be expressed as the additive genetic value of the individual p and the common environment effects experienced by the examined individuals.As a result, the phenotypic correlation can be further decomposed as: with h X denoting the heritability of trait X and c X = r X − h X .The expression in Equation ( 5) includes the genetic correlation r A (X, Y) between the two traits that upon selection constrain their evolution (Gomulkiewicz & Kirkpatrick, 1992;Kingsolver et al., 2001;Roff & Fairbairn, 2007).Dingemanse and Dochtermann (2013) propose further decomposition of the residual correlation, which provides little biological relevant to understanding trade-offs, since it is fully determined by the environmental as well measurement error correlations.
From this decomposition it becomes apparent that negative genetic correlations do not always correspond to negative phenotypic correlations, since the latter are shaped by the remaining terms in Equation ( 5).For this reason, it has been argued that two traits should be considered in trade-off only if their genetic correlations are negative (Reznick, 1985;Roff & Fairbairn, 2007).Unlike Reznick (1985), our perspective on trade-offs does not diminish the importance of studies that focus on trade-offs inferred by negative phenotypic correlations or between-individual correlations.As in the literature on phenotypic plasticity, it can be argued that one should not exclude the possibility that artificial or natural selection may result in genetic assimilation whereby trade-offs can become genetically encoded (Pigliucci et al., 2006).A further argument for this perspective is that in the context of maximizing the rate of improvement of plastic traits (e.g., yield) via artificial selection, the selection response depends on both phenotypic and genetic correlations (Kirkpatrick & Bataillon, 1999).
Genetic correlations can be determined by using data from populations composed of parent-off-springs, full or half-sib families (Lynch & Walsh, 1998).They can be quantified by using linkage disequilibrium score regression and genomic restricted maximum likelihood (Roff & Fairbairn, 2007).In addition, recent advances have shown how summary statistics from genome-wide associations can provide estimates of genetic correlations using natural variability in a species (Zhang et al., 2021).However, the quantification of linear trade-offs by genetic correlations strongly depends on the population used and environments considered.Therefore, caution is warranted in applying the findings of genetic correlations to other populations or environments.To this end, we propose that as much as the experimental design allows, any investigation of tradeoffs should consider the three different types of quantification, based on phenotypic, between-individual, and genetic correlations (Figure 2).
Finally, we note that several primers for application of mixedeffect models have already detailed the effects of repeatability and sample sizes on the power and accuracy of inferring between-individual and genetic correlations (Dingemanse & Dochtermann, 2013;Houslay & Wilson, 2017).As a result, they can be readily employed to infer trade-offs by determining negative between-individual correlations and genetic correlations for pairs of traits.While the reporting of between-individual correlations have been advocated in ecology and behavioural biology research, we note that this is not yet applied in interdisciplinary studies that combine quantitative genetics with plant physiology or molecular biology (Careau & Wilson, 2017;Dingemanse & Dochtermann, 2013;Houslay & Wilson, 2017). (2)

| Inference of tasks in trade-offs from principal component analysis
When trade-offs between more than two traits are of interest, the mixed effect framework can be adapted to account for the larger number of traits (Dingemanse & Dochtermann, 2013;Wilson et al., 2011), requring careful experimental planning to gather sufficient data for parameter estimation.In addition, other, classical techniques from multivariate analysis, like principal component analysis (PCA), have been used to identify combinations of traits in trade-off.
PCA identifies linear combinations of variables describing the set of items that capture maximum variance of the data set.
These linear combinations are referred to as principal components and facilitate visualizing the items in a lower-dimensional space.
When the variables and objects correspond to traits measured in individuals of a population, the principal components can also be seen as a task performed by the individuals.Using this approach, Shoval et al. (2012) have shown that in the presence of trade-offs between tasks, the individuals align within simple polygons (e.g., lines, triangles, or tetrahedrons) determined by the number of tasks.The vertices of the polygons may be regarded as individuals that are specialists for a single task and are referred to as archetypes.As a result, the performance of each individual can be seen as a linear combination of archetypes, with individuals closer to an archetype being more specialized to the corresponding task.This insight has been used to derive a statistical test for assessing the significance of the tasks in trade-offs obtained by fitting convex hulls in the lower-dimensional space obtained by PCA (Figure 3; Shoval et al., 2012).
Despite these advances in the application of multivariate statistical techniques for identification of trade-offs, it remains to be shown whether and to what extent the findings of archetypes and tasks in trade-off are affected by using between-individual correlations rather than phenotypic correlations on which PCA is based.That said, we would like to note that caution is warranted in applying PCA with traits scored in an experimental design with multiple levels.If PCA is applied the assumption of equality of trait covariances across levels needs to be ensured to avoid erroneous conclusions (see meta-analysis for an example; Poirier et al., 2020;Sauce et al., 2018).mutations in the selected environment have deleterious effects in nonselected environments (Brown & Kelly, 2018;Caspari, 1950;Williams, 1957) or by accumulation of mutations that are neutral in a selected environment but deleterious or neutral in another environment (Fournier-Level et al., 2011;Fry, 1996;Hall et al., 2010).

| G ENE TIC MECHANIS MS E XPL AINING TR ADE-OFFS: E X APLE S FROM PL ANT S TUD IE S
Second, negative genetic correlations may be indicative of linkage disequilibrium, whereby genes that are inherited together affect the traits in trade-off (Roff & Fairbairn, 2007).While random mating may break trade-offs caused by linkage disequilibrium, it is not expected to influence trade-offs due to pleiotropy.We note that while the classical designs for calculating genetic correlations cannot pinpoint the genetic mechanisms underlying trade-offs, using the summary statistics from genome-wide associations can provide some level of insight (see Section 3.1), although distinguishing between the underlying cause is challenging.It is also plausible that negative genetic correlations can also occur due to indirect effects of the genes interacting with the causal genes.
These two mechanisms hold true when the trade-off is explained by limited resources allocated to the traits in trade-off.However, for relative trade-offs, that depend on the environment, a third genetic mechanism, the interaction between the genotype and the F I G U R E 3 Trade-offs inference from principal component analysis.Each point in the space determined by the first two principal components (axes) corresponds to an individual.The individuals that are vertices of the convex hull (here triangle) enclosing all points are referred to as archetypes (a1-a3).Each point inside the triangle can be represented as a linear combination of the archetypes -Seen to specialize for a particular task environment, can explain genetic correlation between the traits.
Advances in genotyping technologies, precision phenotyping, and computational approaches for integration of the resulting data have propelled the application of genome-wide association approaches in determining genes controlling diverse plant traits, including those involved in trade-offs (Grimm et al., 2017;Seren et al., 2012), see Table 1.One can then ask if genes controlling a single trait, involved in a trade-off, also control the trade-off itself.To this end, the concept of antagonistic pleiotropy as a genetic mechanism underpinning trade-offs is particularly appealing in the context of genome-wide associations, since it translates into identifying associated markers that have opposing effects on two respective traits.
In recent years, natural variation together with quantitative trait locus (QTL) mapping or genome-wide association (GWA) analysis have indeed revealed genes controlling trade-offs through antagonistic pleiotropy.For instance, in Arabidopsis thaliana disease resistance and growth (measured as leaf initiation rate) were both independently linked to the same locus on chromosome 4, containing accelerated cell death 6 (ACD6) gene (Todesco et al., 2010).
Yet, ACD6 was linked to increased resistance and reduced growth.
Recently growth traits and colour in leaves were found to be colocalized to same QTL in A. thaliana, and allelic variation in flowering locus M (FLM) gene was found to mediate a trade-off between resource acquisition and resource conservation (Hanemian et al., 2020).In another case, GWA analysis revealed that grain weight and grain number in wheat were associated to the same loci posing opposite effects on the traits (Guo et al., 2018).In rice, grain number and grain size trade-off has been shown to be determined by the grain size and number 1 (GSN1) gene (Guo et al., 2018).In this study, the gsn1 knock-out mutants and transgenic lines with reduced GSN1 expression showed both increased grain size and reduced number, indicating that GSN1 mediates a trade-off between the grain size and number through negative regulation of grain size and positive regulation of grain number (Guo et al., 2018).Finally, antagonistic pleiotropy controlling hypoxia and drought in A. thaliana was identified using GWA analysis.These examples demonstrate that genes controlling trade-offs can be identified based on colocalization of significant associations of the individual traits involved in trade-offs.
Interestingly, however, several recent studies have provided evidence that (1) focal traits and trade-offs between them are controlled by different mechanisms, and (2) trade-offs depend on conditions or developmental stage of the plant.For example, a wellknown trade-off between defence and growth in plants could be overcome by conditionally inducing the defence response genes at a certain stage of development (Karasov et al., 2017).Using drought sensitive mutants, it has also been shown that the growth in these mutants can be improved by expressing genes that are known to enhance growth but are not responsive to drought (Kudo et al., 2019).
Moreover, the trade-off between the seed number and seed size observed in many plant species has a genetic mechanism that differs from those explaining the individual traits (Calderini et al., 2021;Guo et al., 2018;Sadras, 2007).Expressing expansin in seeds has been shown to overcome the trade-off between grain weight and number in wheat by increasing the weight without affecting the number of grains (Calderini et al., 2021).This evidence demonstrates that the pleiotropy of the genes controlling the traits does not alone explain the genetic mechanisms of trade-offs, but they can also arise independently due to GxE interactions, which affect genetic correlations.
Further, the increasing evidence for independent genetic control of the traits and entailed trade-offs suggests that the implicated genes could be under different evolutionary pressure; however, further work is needed to clarify these questions.

| PER S PEC TIVE S
Trade-offs between traits are inevitable because organisms function under diverse genetic and physicochemical constraints.In addition, expression of phenotypes is conditionally constrained by the environment (e.g., nutrient availability, organ/tissue or developmental stage).Yet, several studies have shown that individuals differ in their trade-offs, suggesting that the degree of trade-offs may not be under the same genetic control as the individual focal traits.This point is supported by the recent observations that trade-offs are conditional and influenced by external factors (Calderini et al., 2021;Karasov et al., 2017;Kudo et al., 2019).In this sense, we propose that trade-offs, like plasticity, should be viewed as traits themselves, provided there is sufficient evidence for their genetic control.This in turn would imply that trade-offs themselves can also be subject to First, dissecting the genetic architecture of trade-offs can be achieved by identifying genes underlying antagonistic pleiotropy for two or more traits.This problem can be readily addressed by mining the findings from genome-wide associations (Grimm et al., 2017;Seren et al., 2012).Another direction for future research consists of expanding the concept of relationship quantitative trait loci (rQTL) to detect genes controlling trade-offs.rQTL capture the association of a locus with a trait after controlling for the effect of another (Pavlicev et al., 2013).Such partialling out of trait effects can facilitate the identification of trade-offs between more than two traits at the cost of increasing the computational demand, because different combinations need to be considered.
Second, the discoveries related to the genetic architecture of trade-offs must go hand-in-hand with efforts directed at understanding the extent to which trade-offs are affected by the environment.This is particularly relevant if we aim to breed crop lines for future climate scenarios that are superior in multiple traits thought to be in trade-off, like size and nutritional value (Dwivedi et al., 2021).To this end, particular emphasis should be placed on identifying whether traits with contrasting plasticities over different environments tend to overcome or strengthen trade-offs involving these traits.For instance, a recent study found a prominent tradeoff between plant biomass and soil organic carbon under elevated carbon dioxide, suggesting that trade-offs may emerge due to climate change (Terrer et al., 2021).Importantly, future efforts should also involve experiments towards understanding the trade-offs in variable conditions, ultimately with more than one changing cue relevant in future climate scenarios.
In particular, identifying the genes that control plant trade-offs can have far-reaching implications for breeding efforts aimed at overcoming trade-offs involving agronomically relevant traits.In this direction of research, it is critical to understand the extent to which overcoming one trade-off may lead to new, previously unseen trade-off situations.To address this point, however, substantial research is required with respect to the identification and characterization of trade-offs in multiple environments using mutants and transgenic lines.
Another direction in dissecting the genetic architecture of tradeoffs can focus on deriving indices that are indicative of trade-offs at the level of an individual which can then be employed with the existing approaches for genome-wide association analysis.For instance, the concept of the dependence reaction norm, applicable for two traits of an individual, can be used to address this point.In this direction, it would also be of interest to develop approaches that can identify the genetic architecture of trade-offs between more than two traits.All of these suggestions stress the possibility of using natural variation of crops and plant models to dissect the genetic basis of trade-offs.This perspective is feasible and supported by recent studies demonstrating that the growth-flavour trade-off in different crops (Gnan et al., 2014;Su et al., 2021).
Lastly, it is important to investigate how the knowledge about the genetic architecture of trade-offs and their dependence on environments can be combined with approaches for genomic prediction to identify individuals of desired performance (Tong & Nikoloski, 2021).To this end, prediction of traits in trade-off can capitalize on recent advances in computational approaches for prediction of multiple traits under different environments (Runcie et al., 2021).We envision that these directions for future research that rely on interdisciplinary approaches will enable us to obtain a better understanding of the role that trade-offs play in rapid adaptation as well as their more prominent or attenuating effects under future climate scenarios.
trade-offs can be calculated based on different measures and data from various experimental designs, resulting in several definitions of trade-offs.Here, we present a classification of the definitions of trade-offs based on: (1) the measures used for their quantification, (2) the dependence of trade-offs on environment, and (3) experimental design used (i.e., a single individual across different environments or a population of individuals in one or multiple environments).
instance, trade-offs can be identified in a single environment over a population of individuals or in a single individual over different environments.Yet, these approaches and experimental designs do not provide insights in the genetic basis of trade-offs, since such tradeoffs can be seen as population-level characteristic.To determine the genetic architecture of trade-offs, one must rely on a population of individuals exposed to different environments.This experimental design allows scoring of trade-offs in individuals by relying on intraindividual correlations.Here, we provide a succinct summary of approaches that have been proposed for inferences of trade-offs from data, and stress the distinction between trade-offs as individual or population-level characteristics.First, we deal with the problem of inferring trade-offs between two traits, and then cautiously consider the multivariate extensions.
calculated as with cov( • ) denoting the covariance and var( • ) representing the variance of random variables.If the correlations are calculated for focal traits scored in the same genotype over multiple environments, we can used them to quantify intraindividual trade-offs.
pq , F I G U R E 2 Classification of trade-offs.The definitions of trade-offs can be classified based on the measure used for their quantification, dependence on the environment, and level they apply to.a.Based on the measures for quantification, trade-offs can be classified into linear, quantified by Pearson's correlation coefficient, or nonlinear (convex or concave), quantified by the spearman or Kendall correlation coefficient.Negative correlations are indicative of trade-offs, irrespective of the coefficients used.(b) Trade-offs are referred to as absolute if they persist over all considered environments (e.g., black and red line show negative correlations; they are denoted as relative if they are manifested over a subset of environments.For instance, environments in the dependence reaction norm depicted in red can either lead to zero correlation (full line) or even positive correlation (dotted line).(c) Trade-offs can be revealed by using data from a single individual over multiple environments or from populations, either within or between generations.(d) Three types of trade-offs have been distinguished based on the biological process involved, namely: Resource allocation, resource acquisition, and specialization in a biological function.In majority of applications, trade-offs are identified based on phenotypic, between-individual, or genetic correlations, which depend on each other as depicted by the rectangles surrounding the concept of trade-offs repeatability represents an upper bound to the heritability, h X , of a trait, since var i X can be further decomposed into variance due to genotype and other factors.For another trait, Y, the model in Equation (1) takes the form of Equation (2), below Determining the covariance of the two traits, X and Y, one can easily derive Equation (3) with cov i X , i Y denoting the between-individual covariance and cov X , Y the residual covariance.The between-individual covariance can be understood as the correlation between the trait means across the different individuals (Figure 2c, right panel); the residual correlations correspond to the correlation between the residuals around the means of the respective traits in a single individual.From the decomposition of phenotypic covariance, one can readily obtain the following expression for decomposition of phenotypic correlation: with r i (X, Y) denoting the between-individual correlation and r (X, Y) Figure1c, measurements in two different environments can result in selection and play a role in steering evolution.It also has important implications when designing strategies aiming to engineer superior crop plants that can overcome trade-offs.We propose that the future research of plant trade-offs should focus on understanding the factors involved in uncoupling of traits in trade-offs.Hence, it is of great importance to understand: (1) the genetic architecture of trade-offs, particularly those involving agronomically important traits, (2) the extent to which trade-offs depend on environmental factors, (3) effects that trade-offs have on selecting lines with desired performance.Addressing these questions is particularly important, since based on basic research in model plants several genes and other biochemical components have emerged as possible leads to overcome trade-offs in crops(Dwivedi et al., 2021).