Metabarcoding a diverse arthropod mock community

Abstract

Although DNA metabarcoding is an attractive approach for monitoring biodiversity, it is often difficult to detect all the species present in a bulk sample. In particular, sequence recovery for a given species depends on its biomass and mitome copy number as well as the primer set employed for PCR. To examine these variables, we constructed a mock community of terrestrial arthropods comprised of 374 species. We used this community to examine how species recovery was impacted when amplicon pools were constructed in four ways. The first two protocols involved the construction of bulk DNA extracts from different body segments (Bulk Abdomen, Bulk Leg). The other protocols involved the production of DNA extracts from single legs which were then merged prior to PCR (Composite Leg) or PCR-amplified separately (Single Leg) and then pooled. The amplicons generated by these four treatments were then sequenced on three platforms (Illumina MiSeq, Ion Torrent PGM and Ion Torrent S5). The choice of sequencing platform did not substantially influence species recovery, although the Miseq delivered the highest sequence quality. As expected, species recovery was most efficient from the Single Leg treatment because amplicon abundance varied little among taxa. Among the three treatments where PCR occurred after pooling, the Bulk Abdomen treatment produced a more uniform read abundance than the Bulk Leg or Composite Leg treatment. Primer choice also influenced species recovery and evenness. Our results reveal how variation in protocols can have substantial impacts on perceived diversity unless sequencing coverage is sufficient to reach an asymptote.

Keywords: DNA barcoding; community ecology; ecological genetics; environmental DNA.

Figure 1

Protocol employed to examine species recovery from the mock community. Four amplicon pools were examined. Two derived from bulk DNA extracts (Bulk Abdomen and Bulk Leg). The others derived from DNA extracts from single legs that were either pooled (Composite Leg) or kept separate (Single Leg) prior to PCR. All four amplicon pools were sequenced on three platforms (Illumina MiSeq, Ion Torrent S5 and Ion Torrent PGM). There were three technical replicates for each treatment except Single Leg [Colour figure can be viewed at http://wileyonlinelibrary.com]

Figure 2

Rarefaction curves showing BIN recovery versus the number of sequences analysed for the four amplicon pools (Bulk Abdomen, Bulk Leg, Composite Leg and Single Leg) on the three sequencing platforms. Two amplicon lengths (407 and 463 bp) were analysed on the S5, but just one (407 bp) on the other platforms [Colour figure can be viewed at http://wileyonlinelibrary.com]

Figure 3

Heat map showing the relative log abundance of the 369 BINs in each treatment for the four amplicon pools. This heat map was created using the JAMP package (https://github.com/VascoElbrecht/JAMP). Technical replicates are indicated with numbers while in silico pooled results are designated by the letter P [Colour figure can be viewed at http://wileyonlinelibrary.com]

Figure 4

(a) Bray–Curtis dissimilarity dendrogram for the four amplicon pools (BA = Bulk Abdomen, BL = Bulk Leg, CL = Composite Leg and SL = Single Leg). Replicates are numbered 1–3 while P is the result from pooling the replicates. The 463 bp amplicon is indicated with an asterisk (*). (b) Nonmetric multidimensional scaling (NMDS) ordinations using Bray–Curtis dissimilarity for the four amplicon pools. Coloured ellipses represent 95% confidence intervals for the BIN composition of the different treatments using ordiellipse (Oksanen et al. 2012). The shapes within each ellipse represent replicates for the four combinations of sequencing platform–amplicon length for three treatments. No replicates were available for the Single Leg treatment, so it has just four points [Colour figure can be viewed at http://wileyonlinelibrary.com]