Measuring species turnover in aquatic macroinvertebrate communities using barcoding and metabarcoding Alfried P. Vogler, Carmelo Andujar, Cesc Murria, Kat Bruce Imperial College London and Natural History Museum, London, UK NatureMetrics Ltd., Egham, UK
Roadblocks to biomonitoring of aquatic invertebrates High abundance, small body size, taxonomically challenging, multiple co-occurring life stages Complex communities WFD prescribes quick and cheap surveys required on the spot Simplistic measures of water quality, at high taxonomic levels, e.g. BMWP* or EPT vs. OCH** Misses species-level responses to disturbance (including functional traits affected) *Biological monitoring working party, **Ephemeroptera, Plecoptera, Trichoptera versus Odonata, Coleoptera, Hemiptera
How barcode and metabarcode data can improve our understanding of community composition Geographic turnover: mountain freshwater streams across Europe (Murria et al., Mol Ecol, submitted) Community turnover at genetic, species and higher levels Sanger barcoding (>3000 specimens of insects) Disturbance-induced turnover: insecticide spill in R. Kennet, southern England (Andujar et al., Meth Ecol Evol, in preparation) Known community rearrangement in impacted sites Illumina metabarcoding
The patterns of diversity: Capturing aquatic invertebrate diversity in the European continent 6 biogeographic regions, ~12 sites each Standard cox1 barcode for >3000 specimens of selected lineages Rif (Morocco) Betic (South Spain) Picos (North Spain) Jura (France) Carpathians (Slovakia) Jämtland (Sweden) Murria et al., Mol Ecol, submitted
Taking a lineage perspective: Phylogeny and distribution of species (example of Baetidae) Narrow clusters of haplogroups, corresponding to GMYC species (OTUs, BINs) Most GMYC species are narrowly distributed in one or two sites. Ten GMYC species morphologically identified as Baetis rhodani; nine GMYC species are identified as Baetis alpinus
=> Differentiation at taxonomic levels: Genus < Species < Haplotypes A community perspective: Turnover across Europe at different hierarchical levels Genus level Species level Haplotype level
Lessons from Europe-wide study 1. High turnover among biogeographic regions, and within each region 2. Differences not clearly evident when studied at higher taxonomic level 3. Linnaean binomials split into multiple mitochondrial clusters, each with small distribution range => studies of water quality etc. have to be seen in the context of the local community
River Kennet insecticide spill (Thompson et al. Freshwater Biology, 2016) Impact Control study using metabarcoding (Andujar et al.)
Testing the metabarcoding pipeline In any treatments the overlap of OTU recognition is always >90%. Standardised parameters applied uniformly for all samples
Comparing cox1 and 18S rrna Approx. 500 OTUs detected with cox1 (5 and 3 ) No great differences with various clustering algorithms (Usearch, Swarm, Crop) SSU Fewer OTUs detected with SSU, collapsing separate species Platyhelmithes and Nematoda detected only with SSU Rotifera only detected with cox1
Control Impact comparisons using all OTUs Post-spill communities clustered in small part of ordination space Similar pattern at all hierarchical levels Turnover, not nestedness: postspill communities not subset (e.g. insecticide resistant survivors) of pre-spill communities
Comparison with morphological analysis Above and below spill Impact (Below spill) Control (Above spill) Baetis rhodani [C/I] Baetis rhodani [C/I] Baetis scambus [C/I] Baetis vernus [C/I] Baetis fuscatus [I] Baetis phoebus [I] Baetis rhodani [C] Baetis sp. [C] Baetis vernus [C] Chironomidae: 57 OTUs 19 OTUs [C/I] 4 OTUs [C] 34 OTUs [I] Gammarus fossarum [C/I] Gammarus minus [C] Gammarus nekkensis [C] Gammarus pulex [C/I] Thompson et al. Freshwater Biology Volume 61, Issue 12, pages 2037-2050, 24
How to extract invertebrate samples Surber sample bulk Meso 183 80 103 24 Macro 127 207 OTUs of Arthropoda
Conclusions Metabarcoding provides high-quality species-level inventories, after appropriate marker choice and bioinformatics filtering We need: Good taxonomy (many Linnaean binomials correspond to multiple mitochondrial OTUs) Good inventories at local levels (high turnover among sites, even when closely adjacent) Good ecological data (assess the correlates of distribution and susceptibility for each species)
Turnover with geographic distance Distance decay only evident at high hierarchical levels, already reached maxium at lower levels
Turnover within biogeographic regions (<100 km distances): controlled by geographic distance, environment, stochastic factors Variance partitioning of contribution to beta diversity of distance, environment, environment plus distance, and residuals Genus level Species level Haplotype level
Total number of OTUs with cox1-5, cox1-3 and SSU Approx. 500 OTUs detected with cox1 (5 and 3 ) No great differences with various clustering algorithms (Usearch, Swarm, Crop)