Multiple species delimitation approaches with COI barcodes poorly fit each other and morphospecies – An integrative taxonomy case of Sri Lankan Sericini chafers (Coleoptera: Scarabaeidae)

Abstract

DNA taxonomy including barcoding and metabarcoding is widely used to explore the diversity in biodiversity hotspots. In most of these hotspot areas, chafers are represented by a multitude of species, which are well defined by the complex shape of male genitalia. Here, we explore how well COI barcode data reflect morphological species entities and thus their usability for accelerated species inventorization. We conducted dedicated field surveys in Sri Lanka to collect the species-rich and highly endemic Sericini chafers (Coleoptera: Scarabaeidae). Congruence among results of a series of protocols for de novo species delimitation and with morphology-based species identifications was investigated. Different delimitation methods, such as the Poisson tree processes (PTP) model, Statistical Parsimony Analysis (TCS), Automatic Barcode Gap Discovery (ABGD), Assemble Species by Automatic Partitioning (ASAP), and Barcode Index Number (BIN) assignments, resulted in different numbers of molecular operational taxonomic units (MOTUs). All methods showed both over-splitting and lumping of morphologically identified species. Only 18 of the observed 45 morphospecies perfectly matched MOTUs from all methods. The congruence of delimitation between MOTUs and morphospecies expressed by the match ratio was low, ranging from 0.57 to 0.67. TCS and multirate PTP (mPTP) showed the highest match ratio, while (BIN) assignment resulted in the lowest match ratio and most splitting events. mPTP lumped more species than any other method. Principal coordinate analysis (PCoA) on a match ratio-based distance matrix revealed incongruent outcomes of multiple DNA delimitation methods, although applied to the same data. Our results confirm that COI barcode data alone are unlikely to correctly delimit all species, in particular, when using only a single delimitation approach. We encourage the integration of various approaches and data, particularly morphology, to validate species boundaries.

Local stochastics and ecoclimatic situation shape phytophagous chafer assemblage composition

“…However, the fact that we found no correlation between species composition (for total assemblage) and geographic distance (Figure 5), even for adjacent localities situated in the same forest type also in the same mountain range (e.g., L2, L4), may indicate either that species generally might tend to disperse also over moderate‐to‐longer distances or that species disperse very little. Limited dispersal is supported further by molecular evidence (Ranasinghe et al, in review; Ranasinghe, Eberle, Thormann, et al, 2022), since different, the same here investigated localities shared almost no haplotypes. This latter conclusion would be not surprising as previous studies have also shown high turnover rates of assemblage composition at higher elevations independently from geographical distance (García‐López et al, 2010).…”

“…Multiple traps from one locality have the same color and colors correspond to Figure S1. molecular evidence (Ranasinghe et al, in review;Ranasinghe, Eberle, Thormann, et al, 2022), since different, the same here investigated localities shared almost no haplotypes. This latter conclusion would be not surprising as previous studies have also shown high turnover rates of assemblage composition at higher elevations independently from geographical distance (García-López et al, 2010).…”

Contrasting results of multiple species delimitation approaches cause uncertainty in synecological studies

“…Sequences were assembled, edited and aligned using Geneious R7 (version 7.1.9, Biomatters Ltd.). All data are deposited in BOLD (project: SCOIB) and GenBank (accession numbers MW698204 -MW698469 (Sericini; Ranasinghe et al, 2022a) and XX-XX, (other taxa; see Table S1)).…”

“…Furthermore, the match ratios for all pairs of delimitation methods were calculated analogously as explained above and compared in a similarity matrix. Subsequently, the matrix was transformed into a distance matrix and a principal coordinate analysis (PCoA) was performed in PAST v.3.25 (Hammer et al, 2001) to visualize the similarity of outcome between the different methods (Ranasinghe et al, 2022a). The same was done for species inventories resulting from individual subclade analyses and cumulated inventories from individually performed species delimitation analyses on subclades, in which entities of each subclade delimitation were taken again together for the entire assemblage, to explore whether species delimitation on subclades alone affected the overall outcome of delimitation analyses.…”

“…Among the different species delimitation methods proposed so far, the choice of a particular method has a considerable effect on estimated species entities and thus also on species richness estimates (Ahrens et al, 2016;Eberle et al, 2019;Zhou et al, 2019). With the ongoing employment of integrative approaches using multiple species delimitation methods, it became clear that neither of these always perfectly matches the morphospecies entities (e.g., Ahrens et al, 2013Ahrens et al, , 2016Dalstein et al 2019;Eberle et al, 2019;Lukic et al, 2021) nor do they rarely ever match among each other (Ranasinghe et al, 2022a). Although metabarcoding approaches use mainly distance based clustering algorithms with predefined thresholds for species circumscription (Callahan et al, 2017;Piper et al, 2019), several different pipelines or "cluster parameter values" are in use that may reveal alternative outcome (e.g., Potter et al, 2017;Alberdi et al, 2018;Creedy et al 2019;Bailt et al, 2020) The threshold values used for sequence filtering and the number of reads for the identified MOTUs have an effect on the assessment accuracy of data (Edgar & Flyvbjerg, 2015;Potter et al, 2017;Meyer et al, 2021;Piper et al, 2019Piper et al, , 2022.…”