Genome-wide association study of Arabidopsis thaliana leaf microbial community

Abstract

Identifying the factors that influence the outcome of host-microbial interactions is critical to protecting biodiversity, minimizing agricultural losses and improving human health. A few genes that determine symbiosis or resistance to infectious disease have been identified in model species, but a comprehensive examination of how a host genotype influences the structure of its microbial community is lacking. Here we report the results of a field experiment with the model plant Arabidopsis thaliana to identify the fungi and bacteria that colonize its leaves and the host loci that influence the microbe numbers. The composition of this community differs among accessions of A. thaliana. Genome-wide association studies (GWAS) suggest that plant loci responsible for defense and cell wall integrity affect variation in this community. Furthermore, species richness in the bacterial community is shaped by host genetic variation, notably at loci that also influence the reproduction of viruses, trichome branching and morphogenesis.

Figure 1. Genetic variation within A. thaliana shapes the composition of the best-sequenced members of the microbial community

(a) Using eigenvector techniques, inbred replicates of A. thaliana cluster together only when analyzing the most heavily sequenced bacteria. Nevertheless, the vast majority of the sequencing effort characterizes a small number (and %) of taxa in each community (b). Taken together, this implies that vagrant species and other poorly characterized/sequenced taxa (and occasionally, sequencing artifacts) obscure evidence that hosts shape their microbial communities. (c) Host-genetic variation within A. thaliana also affects the ability of fungi to colonize and proliferate on its leaves. All P values take into account technical confounders.

Figure 2. The most frequently observed genomic region in the results from GWAS of the 100 most heavily sequenced bacterial OTUs

The points illustrate the minimum P-value, per 10-kb region, from these separate analyses (i.e. separate GWAS of individual OTUs), and this region is shared in the extreme tail for 9 out of these 100 OTUs (100,000 permutations; P = 1 × 10⁻⁵). Notable a priori candidate genes include FAD2 and TBL1; as mentioned in the main text, the TBL gene family is involved in secondary cell wall synthesis and cellulose deposition. The association peaks, however, on TETRASPANIN 6 (TET6), a gene involved in metal ion transport.

Figure 3. Genes implicated in the community composition of the leaves

The ABC transporter C family members 7 and 8 (multidrug resistance-associated proteins 7 and 8) are associated (Chr 3, ~4.21 Mb) with the abundance of an OTU assigned to Mycosphaerella (a), while ABC transporter G family member 35 (pleiotropic drug resistance 7; Chr 1, ~5.23 Mb) and a pectinesterase (AT2G36710; Chr 2, ~15.392 Mb) are implicated in the abundance of an OTU assigned to Sphingomonas (b). Other pectin related enzymes include the pectate lyase (AT4G13210; Chr 4, ~7.67 Mb) associated with the abundance of Chryseobacterium (c) and the pectinesterase (AT5G26810; Chr 5 ~9.432 Mb) associated with the abundance of Xanthomonas (d). Notable a priori candidate genes also include TERPENE SYNTHASE 10 (TPS10; Chr 2, ~10.297 Mb) identified in (a), the resistance gene (R-gene) pinpointed (Chr 5, ~18.287 Mb) in (c), and the oxidoreductase (Chr 4, ~9.708 Mb) illustrated in (d). To assess genome-wide significance, a permutation approach was used that takes into account population structure (Methods).