Establishing a genomic map of local adaptive cooperation in Arabidopsis thaliana

Get Complete Project Material File(s) Now! »

Towards unifying evolutionary ecology and genomics to understand genotype-by-genotype interactions within wild plant species


In a local environment, plant social networks include interactions between individuals of different species and among genotypes of the same species. While interspecific interactions have been highly recognized as a main driver of plant community patterns, intraspecific interactions have recently gained attention in explaining plant community dynamics. However, an overview of intraspecific genotype-by-genotype interaction patterns within wild plant species is still missing. The 77 experiments that we identified were mainly designed to test for the presence of positive interactions. Both the kin selection theory and elbow-room hypothesis were highly supported, despite their opposite predictions between the extent of genetic relatedness among neighbors and the level of positive interactions. In addition, we found that kin cooperation and overyielding were dominant in annual and perennial species, respectively. Nonetheless, we identified several shortcomings regardless of species considered, such as the lack of a reliable estimate of genetic relatedness among genotypes and ecological characterization of the natural habitats from which genotypes have been collected, which in turn impedes the identification of selective drivers of positive interactions. We therefore propose a framework combining evolutionary ecology and genomics in order to establish the eco-genetic landscape of positive GxG interactions in wild plant species.


During the course of its life cycle, a plant can interact directly or indirectly – consecutively and/or concurrently – with multiple neighboring plants. Plant social networks include interactions between individuals of different species (i.e. interspecific interactions) and among genotypes of the same species (i.e. intraspecific interactions) in a local environment. Plant–plant interactions play an important role in regulating the diversity and structure of plant communities and ultimately ecosystems functioning through their effects on resource availability and habitat structure (Brooker 2006, Martorell & Freckleton 2014). Studying the mechanisms underlying plant-plant interactions is therefore essential to understand the dynamics of plant communities, which may in turn help to predict the resilience of plant species in presence of anthropogenic-related global changes (Subrahmaniam et al. 2018). For instance, ongoing climate warming results in modifications of plant assemblages due to increase of plant biomass, reduced diversity (Baldwin et al. 2014) and shifts in the distribution areas of plant species (Gilman et al. 2010, Singer et al. 2013).
Plant-plant interactions can be divided into four main categories depending on the net benefit and cost associated with the interaction (Subrahmaniam et al. 2018). First, competitive interactions (-/-) come with a cost for both partners (benefit < 0, cost > 0 for both partners). Competition is characterized by reciprocal negative effects on plant growth or fitness caused by the presence of neighbors (Keddy 2015). Since all plants share a few basic requirements, limitations of resources such as the availability of nutrients, water or light could drive competition between plants (Turkington & Harper 1979, Chaney & Baucom 2014). Second, asymmetric interactions (+/-) yield benefit to one of the partner at the cost of the other interactor (benefit < 0 and cost > 0 for the helper; benefit > 0 and cost < 0 for the receiver). Parasitic plants are the prime example of this kind of behavior. In addition, plants releasing allelochemicals to negatively influence the physiology of their neighbors can be grouped under this category. Third, commensal interactions (+/0) are those that are beneficial for at least one of the partners, but there is no cost associated with providing such aid (benefit = 0 and cost = 0 for helper; benefit > 0 and cost = 0 for the receiver of the help). Many examples of such interactions exist at the interspecific level, like nurse plant effects in deserts or climbing plants that use the stems of other plants to avoid shade (Padilla & Pugnaire 2006, Gianoli 2015). Fourth, individuals can also reciprocally benefit (+/+) from being associated with a partner (benefit > 0 and cost < 0 for both plant partners). Many examples of such a reciprocal help have been described at the interspecific level. Plant-mycorrhizal associations that help nutrient sharing and transfer between different plant species are one example of such an association (Teste et al. 2014).
Estimating the relative importance of these broad categories in explaining patterns of plant communities is still under debate and mainly focused on interactions at the interspecific level. Interspecific competitive interactions have been traditionally recognized as the major factor responsible for the structure (Goldberg & Barton 1992), diversity (Chesson 2000) and dynamics of plant communities (Tilman 1985). However, more recently, the role of positive interactions among species (including both commensal interactions and reciprocal help) in regulating the composition of communities, has particularly gained attention. (Bertness & Callaway 1994, Callaway 1995, Brooker & Callaghan 1998, Bruno et al. 2003, Dormann & Brooker 2002, Kotowska et al. 2010, Wendling et al. 2017). In particular, positive interactions among species have been put forward to explain overyielding, which corresponds to the increase in productivity of species when grown in mixture as opposed to monoculture (Harper 1977, Vandermeer 1981, Loreau 2004, Schmid et al. 2008). However, upon decomposing species interactions into interactions occurring between genotypes of species, it has recently been argued that the interaction outcome depends on the genotype identity, rather than species identity (Ehlers et al. 2016). Genotype-by-genotype (GxG) interactions at the interspecific level might ultimately govern community diversity, composition and structure (Brooker 2006, Ehlers et al. 2016). Similarly, it is increasingly being recognized that studying GxG interactions at the intraspecific level might be a prerequisite for understanding eco-evolutionary patterns of plant communities (Hughes et al. 2008, Lankau 2018). Indeed, a huge number of genotypes of varying levels of relatedness can co-exist within a local population, even in the case of highly selfing species. For instance, a recent study on Arabidopsis thaliana revealed that the genetic diversity observed within a local population represents almost one-sixth of the genetic diversity at the worldwide scale (Frachon et al. 2017). Therefore, the patterns of interactions between different genotypes within one population are bound to vary as well.
Several meta-analyses have been carried out to understand patterns of GxG interactions at the interspecific level in herbaceous wild plant species (Maestre et al. 2005) as well as in trees (Piotto 2008, Zhang et al. 2012). However, an overview of GxG interaction patterns at the intraspecific level within wild plant species is still missing from literature. This review therefore aims to make a synthesis on such interactions. More precisely, based on 66 articles, we aimed at establishing general patterns of intraspecific GxG interactions by addressing the following questions: (i) Why GxG interactions were studied for ?, (ii) What plant material was used to study GxG interactions ?, (iii) What were the growth conditions used to estimate GxG interactions ?, (iv) What traits were phenotyped to study GxG interactions ?, (v) Can interactions between genotypes be indirect ?, and
(vi) What major conclusions can be reached in GxG experimental studies ? We then introduce several avenues that deserve to be explored to obtain a thorough picture of GxG interaction patterns within wild species. In particular, we stress the need to integrate genomics and evolutionary ecology to fully understand the complexity of intraspecific genetic interactions in wild plant populations.


For this review, we only focused on studies looking at intraspecific interactions within wild herbaceous species. We made this choice because the number of generations of wild herbaceous species is clearly smaller than the one of trees, therefore fitness proxies can be better estimated during their life cycle. Several keywords were used to gather these studies: GxG interactions, intraspecific interactions, intraspecific variation, intra/inter-population variation, group selection. The websites inspected included Google Scholar, Web of Science, Sci-hub, Researcher. Although we tried to do a comprehensive analysis to include a maximum number of studies reporting intraspecific GxG interactions, the list is certainly not exhaustive and some studies may have been gone overlooked. We gathered a list of 66 articles including 77 experiments (Supplementary Table
1) published in the last 35 years. Interestingly, we observed a sharp increase in cumulative number of experimental papers published over the years (Figure 1A), thereby illustrating the rising interest in examining intraspecific GxG interactions in wild plant species. The list includes 43 species belonging to 18 botanical families (Figure 1B). The most commonly studied botanical families comprise of Brassicaceae (40%), Asteraceae (12%), Fabaceae (10%) and Poaceae (9%). However, there is a significant bias in the family Brassicaceae towards A. thaliana as it constitutes about 90% of the studies from this family. Upon removing this species, the relative proportion of botanical families studied is consistent with the amount of species within each family (Figure 1B, Supplementary Table 1). The list of 43 species is divided roughly equally between annuals (~46%) and perennials (~54%) (Supplementary Table 1) and is dominated by selfing species that make up about 42% of the dataset. The remaining species comprise of mixed breeding system (28%), outcrossing (21%) and clonals (9%). Allochory (assisted seed dispersal) seems to be predominant in this list as about 65% of the listed species demonstrated this mode of seed dispersal, while only 35% of species listed have an autochorous (self) mode of seed dispersal (Supplementary Table 1).


Why GxG interactions were studied for?

The reported experiments can be categorized into two main rationales that hypothesize opposite relationships between the extent of genetic relatedness among neighbors and the level of positive interactions (File et al. 2012). Rooted in evolutionary biology concepts, the first rationale is based on the kin selection theory that advocates that individuals increase their inclusive fitness by modifying their behavior to help a relative (Hamilton 1964). The ‘kin/non-kin recognition’ category concerns 52% of the experiments where the differential response of a genotype was tested in the presence/absence of a relative genotype (kin) vs a stranger genotype in pairwise experiments (Figure 2A, Supplementary Table 1). The second rationale is based on the elbow-room ecological hypothesis that assumes that intraspecific resource partitioning occurs and increases as the genetic distance between neighbors increases (Argyres & Schmitt 1992). This positive diversity-productivity relationship corresponds to overyielding at the intraspecific level. The ‘genotypic diversity- productivity relationship’ category concerns 32% of the experiments where fitness proxies were compared between monocultures using multiple kin individuals (compound intra-genotypic interactions) and mixtures of different genotypes (compound inter-genotypic interactions) (Figure 2A, Supplementary Table 1).
The remaining experiments (16%) that were grouped under the category ‘Others’ included experiments that aimed at (i) characterizing the genetic architecture underlying GxG interactions (Mutic & Wolf 2007, Botto & Callucio 2007), (ii) studying the effects of GxG interactions on intra-individual traits such as genome size variation (Smarda et al. 2010) and transcriptomic profiles (Bowsher et al. 2007), (iii) studying extended phenotypes such as root exudate profiles (Badri et al. 2012) and soil microbial communities (Burghardt et al. 2019, Fitzpatrick et al. 2019), (iv) testing the effect of adding a neighbor plant on genotype-by-environment interactions (i.e. GxExG instead of GxGxE) (Cahill et al.. 2010), (v) looking at local adaptation of genotypes (Linhart 1988, Espeland & Rice 2007), and (vi) investigating individual vs group selection in wild plant populations (Goodnight et al.. 1985, Donohue 2003). For the latter, we need to stress that the existence of group selection is still controversial (Nowak 2006, Nowak et al. 2010, Rousset & Lion 2011, Queller et al. 2015, Kramer & Meunier 2016) and it will not be addressed in this review.

READ  The Software Architecture Perspective: Understanding Interoperability 

What plant material was used to study GxG interactions?

Around 84% of the experiments listed were based on genotypes collected in natural populations (Supplementary Table 1). With the main goal of dissecting the underlying genetic and molecular mechanisms of GxG interactions, the remaining experiments were based on (i) experimental populations such as F2 populations or families of Recombinant Inbred Lines (RILS, 9%) (Goodnight et al. 1985, Griffing 1989, Mutic & Wolf 2007, Botto & Collucio 2007, Willis et al. 2010, Latzel et al. 2013, Wuest & Niklaus 2018), and (ii) mutant lines (~5%) (Cahill et al.. 2005, Crepy & Casal 2014, Wagg et al.. 2014, Zhang & Tielborger 2018) (Supplementary Table 1). Unsurprisingly, given the great amount of genetic resources publicly available, all these experiments dedicated to the study of genetic and molecular mechanisms concern A.thaliana (with the exception of Trifolium pratense, Wagg et al. 2014).
For experiments based on genotypes collected in natural populations, we observed a clear tradeoff between the number of genotypes used and the number of populations sampled (Figure 2B). The number of natural genetic lines used to evaluate GxG interactions is highly variable among experiments, ranging from 2 to 72 (mean ~12) (Figure 2B). On average, these lines have been collected from about four populations (min=1, max=60) (Figure 2B). In wild plant populations, intraspecific neighbors share common space over generations and this increases their probability for repeated interactions. Consequently, positive interactions are also likely to evolve between members of a single population rather than between members of different populations (Nowak 2006). Accordingly, most experiments that do not imply A. thaliana are based on natural genetic lines from a single population (30%) or sampled at a regional scale (between 2-14 populations, 37.6%). On the other hand, an opposite trend is observed in A. thaliana, which represents 35% of the dataset. Almost 90% of the experiments on this species utilized genotypes coming from worldwide collections. The main hypothesis to explain this bias in using worldwide genotypes in A. thaliana is related to its predominantly selfing breeding system, which initially suggested that most populations were monomorphic (Platt et al.. 2010). Therefore, the large public collections of genotypes that are available for A. thaliana mostly correspond to one representative genotype per population. However, more and more studies challenged this view by revealing extensive genetic diversity within populations (Le Corre 2005, Jorgensen & Emerson 2008, Bomblies et al. 2010, Platt et al. 2010, Kronholm et al. 2012, Brachi et al. 2012, Roux & Bergelson 2016, Frachon et al. 2017, Fulgione et al. 2017, Frachon et al. 2018), thereby giving an opportunity of studying more relevant GxG interactions in A. thaliana at the local scale.
Based on all the experimental experiments listed in this survey, we nonetheless identified two major shortcomings of the plant material used to study GxG interactions, regardless of species considered. Firstly, as previously mentioned, testing both the kin selection and elbow-room hypotheses requests estimation of the degree of genetic relatedness among interacting genotypes. Kin selection theory predicts partisan help given to close relatives. By contrast, according to the elbow-room hypothesis, genetically close relatives will compete for the same resources and increasing genetic distance between genotypes can translate into increasing niche partitioning. To test for these contrasting predictions would require integrating information about the extent of genetic relatedness among interacting genotypes. However, this crucial information has been poorly considered in these experiments. In our survey, only two experiments estimated the degree of genetic relatedness among interacting genotypes (Crutsinger et al. 2006, Crutsinger et al. 2008) (Supplementary Table 1). Secondly, positive plant-plant interactions at the interspecific level are expected to evolve in natural settings including a certain level of abiotic and/or biotic stress (Bertness & Callaway 1994, Brooker & Callaghan 1998, Bruno et al. 2003). Whether this stress gradient hypothesis is also relevant at the intraspecific level remains an open question. Still, in our survey, only 19% of the experiments have loosely described the ecology of the populations used in the experiments (Supplementary Table 1). At most, only a rough description of habitats from which the genotypes have been collected was given.

What were the growth conditions used to estimate GxG interactions?

Performing experiments in controlled and field conditions is complementary (Bergelson & Roux 2010, Brachi et al. 2010). Experiments conducted under controlled conditions drastically reduce environmental noise, thereby allowing establishing a direct link between phenotypic observations and genotype performance under a given set of stable environmental conditions. On the other hand, in the field, plants are exposed to a greater but more ecologically realistic range of abiotic and biotic fluctuations than typically encountered in controlled conditions. Nonetheless, encompassing all these environmental fluctuations request the field experiments to be repeated over several years.
In our survey, almost 79% of the experiments were conducted under laboratory conditions (Supplementary Table 1). Out of these, ~67% and ~21% of the experiments were conducted in greenhouse conditions and growth chambers (including root chambers and growth tunnels), respectively. The remaining experiments (~12%) have been performed under in vitro conditions. On the other hand, few experiments (~13%) have been conducted under field conditions, even less in the native habitats (only two reported experiments, Supplementary Table 1). Finally, four experiments (~5%) were conducted in both greenhouse and field conditions (Espeland & Rice 2007, Anderson 2014, Ehlers et al. 2016). The type of growth conditions used to study intraspecific GxG interactions is therefore strongly biased in favor of laboratory conditions, notably when compared to other types of biotic interactions. For instance, in a recent review on Genome-Wide Association studies (GWAS) performed on plant – pathogen interactions, 60% of the studies were conducted in controlled conditions (greenhouse/growth chambers) and 40% under field conditions (Bartoli & Roux 2017).
Noteworthy, around 66% of the experiments in our survey tested the effect of a particular environmental factor on GxG interactions, either in controlled or field conditions (Supplementary Table 1). Abiotic treatments concern light quality, nutrient status, CO2 concentration and drought, whereas biotic treatments mainly concern density and the effect of soil conditioning by one or a mixture of genotypes, in particular for the ‘diversity-productivity relationship’ rationale (Bukowski & Petermann 2014, Semchenko et al. 2017, Bukowski et al. 2018). However, since no thorough ecological characterization has been conducted on the habitats from which the plant material has been collected, the treatments applied may not be ecologically relevant.

What traits were phenotyped to study GxG interactions?

All the traits measured across the 77 experiments are listed in Supplementary Table 2. To assess GxG interactions, an average of 3.4 traits per study have been measured (min=1, max=9). We divided the list of traits into four broad categories, each related to a distinct eco-function of the plant (i.e. root related traits, shoot related traits, life history traits and seed production related traits). While ~43% of the experiments scored life history related traits (e.g. germination and flowering timing), ~53% and ~79% of the experiments measured root (e.g. root length and biomass) and shoot (e.g. plant height and dry biomass) related traits, respectively. Seed production related traits (e.g. number of fruits and number of seeds per fruit) were measured in ~40% of the experiments. Interestingly, most experiments focused on collecting phenotypic information using either two (45%) or three (27.2%) categories. About 39% of the experiments looked at both root and shoot related phenotype while ~34% of experiments focused on both shoot related and life history related traits. Only four experiments (~5%) focused on all four categories (Wilson et al.. 1987, Linhart et al.. 1988, Masclaux et al.. 2009). The relative proportion of trait categories are similar between the different rationales (Supp. Fig 1).
Measuring individual specific root responses is often recognized to be very difficult in experiments on plant-plant interactions. More often than not, to measure specific root traits, the total root biomass contributed by all genotypes present in a pot has been considered as a dependable measure, which obviously impedes the estimation of the relative contribution of each individual genotype in the pot. However, developments in non-destructive phenotyping technology has added new directions to start teasing apart the respective underground behavior of each genotype in both laboratory and field/natural experimental setups. For example, the 3D root system architecture of plants within natural or field soils can be easily created using low invasive tools (minirhizotrons; Johnson et al. 2001). More recently, X-ray Computed Tomography has been described to be useful for studying in details the root system, such as lateral root growth and orientation patterns under laboratory conditions (Subramanian et al. 2015).

Table of contents :

I. Importance of plant-plant interactions
I.A. General overview
I.B. The genetics underlying natural variation of plant-plant interactions, a beloved but forgotten member of the family of biotic interactions
I.C Towards unifying evolutionary ecology and genomics to understand genotype-by-genotype
interactions within wild plant species
I.D. Prevailing questions in the study of intraspecific interactions?
II. Arabidopsis thaliana as a model species to unravel the adaptive genetic and molecular bases of plant-plant interactions?
II.A. General characteristics
II.B. Genetic and genomic resources and tools available for Arabidopsis thaliana
II.C. A model for studying natural variation of plant-plant interactions?
III. How to identify the genetic basis of adaptation?
III.A. Genome wide association (GWA) mapping
III.B. Genome-Environment Association (GEA)
IV. Outline of the thesis
V. References
Chapter 1 Establishing a genomic map of local adaptive cooperation in Arabidopsis thaliana
I. Introduction
II. Objective
III. Materials and methods
III.A. Biological material
III.B. Experimental design
III.C. Phenotypic trait measurement
III.D. Statistical analysis of natural variation of positive interactions
III.E. GWA mapping using a Bayesian hierarchical model
III.F. Testing for signatures of local adaptation
III.G. Estimating putative ecological drivers of positive interactions
IV. Results
IV.A. Natural genetic variation of positive GxG interactions
IV.B. Genetic bases of positive GxG interactions and their adaptive status
IV.C. Selective ecological drivers of positive GxG interactions
IV.D. Genetic bases of local adaptive cooperation in 52 local populations
V. Discussion and perspectives
V.A. Natural genetic variation of intraspecific positive interactions
V.B. Genetic bases of local adaptive positive interactions
V.C. Selective ecological drivers of intraspecific positive interactions
V.D. Candidate genes underlying the identified QTLs: pre-eminence of functions related to
VI. References
VII. Supplementary information
Chapter 2 Investigating positive GxG interactions at the genomic level in a local population of Arabidopsis thaliana
I. Introduction
II. Objective
III. Materials and methods
III.A. Biological material
III.B. Experimental design
III.C. Phenotypic trait measurement
III.D. Statistical analysis of natural variation of positive interactions
III.E. GWA mapping combined with a local score analysis
III.F. Estimating genetic relationships underlying the ‘super overyielding’ strategy
IV. Results
IV.A. Natural genetic variation of GxG interactions within a local population
IV.B. Genetic architecture of positive GxG interactions
IV.C. Investigating relationships between cooperators at the genomic level
IV.D. Genetic bases associated with the ‘super overyielding’ strategy
V. Discussion and perspectives
V.A. The occurrence of a ‘super overyielding’ strategy in the local TOU-A population
V.B. The ‘super overyielding’ strategy observed in the TOU-A population may be driven by compatibility genes
V.C. Candidate genes underlying the ‘super overyielding’ strategy
VI. References
VII. Supplementary information
General discussion and perspectives
I. Introduction
II. Natural genetic variation of intraspecific positive interactions at different geographic scales
III. Biotic factors as the main selective drivers of positive interactions?
IV. What are the molecular determinants of cooperative plant-plant interactions?
V. References
VI. Supplementary information


Related Posts