Targeting environmental adaptation in the monocot model Brachypodium distachyon: a multi-faceted approach

Abstract

Background: The local environment plays a major role in the spatial distribution of plant populations. Natural plant populations have an extremely poor displacing capacity, so their continued survival in a given environment depends on how well they adapt to local pedoclimatic conditions. Genomic tools can be used to identify adaptive traits at a DNA level and to further our understanding of evolutionary processes. Here we report the use of genotyping-by-sequencing on local groups of the sequenced monocot model species Brachypodium distachyon. Exploiting population genetics, landscape genomics and genome wide association studies, we evaluate B. distachyon role as a natural probe for identifying genomic loci involved in environmental adaptation.

Results: Brachypodium distachyon individuals were sampled in nine locations with different ecologies and characterized with 16,697 SNPs. Variations in sequencing depth showed consistent patterns at 8,072 genomic bins, which were significantly enriched in transposable elements. We investigated the structuration and diversity of this collection, and exploited climatic data to identify loci with adaptive significance through i) two different approaches for genome wide association analyses considering climatic variation, ii) an outlier loci approach, and iii) a canonical correlation analysis on differentially sequenced bins. A linkage disequilibrium-corrected Bonferroni method was applied to filter associations. The two association methods jointly identified a set of 15 genes significantly related to environmental adaptation. The outlier loci approach revealed that 5.7% of the loci analysed were under selection. The canonical correlation analysis showed that the distribution of some differentially sequenced regions was associated to environmental variation.

Conclusions: We show that the multi-faceted approach used here targeted different components of B. distachyon adaptive variation, and may lead to the discovery of genes related to environmental adaptation in natural populations. Its application to a model species with a fully sequenced genome is a modular strategy that enables the stratification of biological material and thus improves our knowledge of the functional loci determining adaptation in near-crop species. When coupled with population genetics and measures of genomic structuration, methods coming from genome wide association studies may lead to the exploitation of model species as natural probes to identify loci related to environmental adaptation.

Figure 1

GIS survey of the sampling area. Depiction of the sampling transect as evaluated through DIVA GIS. False colours were generated to represent the most limiting factors for a typical annual grass species among the 19 BioClim variables through Ecocrop model. Letters A to H denote sampling locations and identify the 9 local groups analysed. Sampling locations were chosen to introduce the most possible environmental variation in relation to rainfall and temperature. BioClim variables are reported in legend. For full meaning, see Additional file 1. From west to east, altitude and aridity increase.

Figure 2

Phylogeny based on the full set of SNPs. Bootstrap network tree based on 1000 permutations with Uncorrected P distances. A-H correspond to the nine sampling locations listed in Table 1. All compatible splits are represented in a single branch; the more parallel branches there are, the more alternative splits were present in the bootstrapped dataset. The reference genome (Reference) overlaps with the Bd21 inbred line genotyped for control sakes (*), and clusters with the inbred lines (I). Local groups do not separate following a strict geographical criterion, yet within-group relationships are maintained. Circles encompass grouping of local groups A, B and H, local groups C and C2, and local groups D, F and G. Location E is intermediate, also geographically. The main split occurs between central Turkey groups and eastern and western sampling points.

Figure 3

Phylogeny based on P/A regions as group-wise markers. A bootstrap network tree based on Jaccard’s distances of binary markers based on regions with consistent within-group presence/absence of reads. The tree topology, though more unstable, entirely overlaps with that produced by the SNPs in Figure 2. This suggests that distances deriving from P/A regions are primarily based on elements with segregation patterns similar to those of genetic variation, probably transposable elements and regions of DNA methylation.

Figure 4

Analysis of spatial structuration of molecular diversity. The measure of conditional genetic distance is shown in panel A. Note that local groups are artificially set in a circle, so edge lengths are not proportional to conditional genetic distance. Node size is proportional to genetic diversity within sampling group. This graph confirms the detachment of sampling groups A, B, H from all of the others. In panel B, visual depiction of the spatial PCA. Positive PCs represent global structure, negative PCs local structure. Note the value of the first PC (out of scale). This global structure overlaps with the main split emerging from other analyses.

Figure 5

Kinship analysis. Kinship relationships among samples according to VanRaden method. Two main groups can be seen, the largest comprising mostly A, B, C, C2 and H individuals. The bimodal distribution of the kinship values confirm the results from diversity analyses, and probably would have biased the association analyses.

Figure 6

Manhattan plots of the association tests. Manhattan plots depicting association across the five B. distachyon chromosomes with environmental PCs 1 to 3, according to LFMM method. On the y axis, the significance of each association test; on the x axis, the SNP locations across the chromosome. The dashed line reports the significance for the LD-corrected Bonferroni method (p-value < 1.37 × 10⁻⁴). Black dots represent significant associations also detected with CMLM. The two methods identify clear peaks. Association peaks mostly have the skewed appearance given by linkage disequilibrium between cis elements nearby the strongest associations.

Figure 7

Triplot from CCA. The CCA analysis uses sites (sampling groups; letters) as fixed points to evaluate the relationship of individuals (P/A regions; dots), and environmental variation (PC1-3; vectors). Though many of the P/A regions are located near the centre of the graph, some appear highly related to PC1 and especially to PC2.