Metabarcoding of soil environmental DNA to estimate plant diversity globally

Abstract

Introduction: Traditional approaches to collecting large-scale biodiversity data pose huge logistical and technical challenges. We aimed to assess how a comparatively simple method based on sequencing environmental DNA (eDNA) characterises global variation in plant diversity and community composition compared with data derived from traditional plant inventory methods.

Methods: We sequenced a short fragment (P6 loop) of the chloroplast trnL intron from from 325 globally distributed soil samples and compared estimates of diversity and composition with those derived from traditional sources based on empirical (GBIF) or extrapolated plant distribution and diversity data.

Results: Large-scale plant diversity and community composition patterns revealed by sequencing eDNA were broadly in accordance with those derived from traditional sources. The success of the eDNA taxonomy assignment, and the overlap of taxon lists between eDNA and GBIF, was greatest at moderate to high latitudes of the northern hemisphere. On average, around half (mean: 51.5% SD 17.6) of local GBIF records were represented in eDNA databases at the species level, depending on the geographic region.

Discussion: eDNA trnL gene sequencing data accurately represent global patterns in plant diversity and composition and thus can provide a basis for large-scale vegetation studies. Important experimental considerations for plant eDNA studies include using a sampling volume and design to maximise the number of taxa detected and optimising the sequencing depth. However, increasing the coverage of reference sequence databases would yield the most significant improvements in the accuracy of taxonomic assignments made using the P6 loop of the trnL region.

Keywords: TRNL; distribution; diversity; environmental DNA; molecular methods; plant; soil.

Figure 1

Correlations between eDNA, GBIF, Kreft & Jetz and Cai datasets. Numbers show correlation strength and direction, red asterisks show significance (*p < 0.05; **p < 0.01; ***p < 0.001).

Figure 2

Global distribution of (A) eDNA family richness, (B) GBIF family richness, (C) Kreft & Jetz estimate of richness and (D) Cai estimates of richness predicted using generalised additive models. Red points indicate sampling locations. The Sahara region was excluded from interpolations because of insufficient sampling.

Figure 3

dbRDA plot (distance-based redundancy analysis) showing effects climatic variables (CHELSA) and biogeographic realm on (A) eDNA and (B) GBIF family level composition (following variance stabilising transformation). Ellipses indicate 1 standard deviation around the centroids for different biogeographic realms.

Figure 4

Relative success of taxonomic assignment of eDNA reads (trnL P6 loop). Values are the logarithm of the ratio between the number of reads (A, B) or OTUs (C, D) getting a match against trnL intron sequence data in INSDC at family (A, C) or species (B, D) level and the number not getting a match. Higher values indicate greater proportions of reads or OTUs getting a match. Red points indicate sampling locations. Global predictions are the result of a generalised additive model. The Sahara region was excluded from interpolations because of insufficient sampling.