Geometagenomics illuminates the impact of agriculture on the distribution and prevalence of plant viruses at the ecosystem scale

Abstract

Disease emergence events regularly result from human activities such as agriculture, which frequently brings large populations of genetically uniform hosts into contact with potential pathogens. Although viruses cause nearly 50% of emerging plant diseases, there is little systematic information about virus distribution across agro-ecological interfaces and large gaps in understanding of virus diversity in nature. Here we applied a novel landscape-scale geometagenomics approach to examine relationships between agricultural land use and distributions of plant-associated viruses in two Mediterranean-climate biodiversity hotspots (Western Cape region of South Africa and Rhône river delta region of France). In total, we analysed 1725 geo-referenced plant samples collected over two years from 4.5 × 4.5 km² grids spanning farmlands and adjacent uncultivated vegetation. We found substantial virus prevalence (25.8-35.7%) in all ecosystems, but prevalence and identified family-level virus diversity were greatest in cultivated areas, with some virus families displaying strong agricultural associations. Our survey revealed 94 previously unknown virus species, primarily from uncultivated plants. This is the first effort to systematically evaluate plant-associated viromes across broad agro-ecological interfaces. Our findings indicate that agriculture substantially influences plant virus distributions and highlight the extent of current ignorance about the diversity and roles of viruses in nature.

Figure 1

French and South African sampling designs. (a) French and (b) South African sampling sites. Both 4.5 × 4.5 km² sampling grids contained 100 GPS-nodes with 500 m spacing and were located across agro-ecological interfaces between cultivated and uncultivated areas. Spatial interpolation of degrees of human-mediated disturbance at the (c) French and (d) South African sampling sites. Empirical Bayesian Kriging was performed based on scores depicting the level of intensity of agriculture using ArcGIS to visualise interfaces between uncultivated and cultivated areas. Every sampling point was ranked as follows: (0) intact native communities; (1) native communities degraded by disturbance or invasion; (2) fallow and old fields; (3) low-intensity polyculture (woodlots and pasture); and (4) intensive crop monoculture. (e, f) Examples of two geo-nodes at the French sampling site. (e) exemplifies an uncultivated sampling point at which four plants (numbers 1–4), each within 2.5 m of the geo-node and with a biomass >10 g, were considered ‘dominant’ and sampled. (f) exemplifies a cultivated sampling point, in this case where the vegetation is dominated by alfalfa, at which we collected three separate 5- g alfalfa samples (numbers 1–3) and one (number 4) from a dominant (that is, >10 g biomass) weed.

Figure 2

Virus prevalence associated with cultivated and uncultivated plants. Plant virus and mycovirus prevalence within cultivated and uncultivated plants are indicated in blue and light green, respectively. Significant differences in virus prevalence between cultivated and uncultivated plants are indicated by **=p-value <0.01 (two-tailed Z test for two population proportions). In (a–c) sample infection prevalence is defined as the proportion of plant samples that contained at least one plant-associated virus read or contig (PLAVs). In (d, e) individual prevalence is defined as the proportion of samples taken from individual plants that contained at least one PLAVs. Note that this comparison could not be made at the South African site because all cultivated plants that were sampled had <5 g of biomass and, as a consequence of this, multiple plants had to be bulked to obtain enough biomass for analysis. In (f–h) bulked prevalence is defined as the proportion of samples consisting of bulked material from multiple individual plants that contained at least one PLAVs.

Figure 3

Spatial associations of virus family communities with degrees of land usage. Association of virus communities with land use. The colour gradient represents the Pearson correlation coefficient in the fourth-corner analyses (testing virus vs environment relationships). Significance is indicated by: p-value<0.1, *p-value<0.05 and **p-value<0.01.