Discussion

In the present paper we investigated the DNA taxonomy of a megadiverse assemblage of chafer beetles (Sericini) with particular focus on the performance of commonly used species delimitation methods. The setup of examining barcodes of a single locality was chosen to investigate molecular species delimitation performance using data without geographic bias. While we know that match ratios strongly vary in tropical taxa (e.g., from 0.14 to 1.00; Ahrens et al., 2016), we theoretically expected that match ratios would go against one due to the exclusion of geography-induced variance. Instead, for different standard species delimitation methods, we could also not report match ratios higher than 0.77. Interestingly, the 3% threshold clustering that is commonly used in metabarcoding approaches did not perform worse than more sophisticated approaches (like PTP, or TCS), however, an accuracy of only less than 80% is not really what one could call a reliable taxonomy assessment.
DNA-based species delimitation approaches may oversplit morphological entities (Ahrens et al., 2016), while at the same time the opposite may be also the case (Dalstein et al., 2019), even in the same taxon (as demonstrated here for the tribe Sericini). This particularly proved to be true in presence of incomplete lineage sorting and hybridisation and if geographic bias is not excluded (match ratio < 0.5; Dalstein et al., 2019). Extreme over-splitting has been reported for both mtDNA and nDNA, when sex-biased dispersal occurs (Eberle et al., 2019) and were the general dispersal is in consequence also very limited.
Over-splitting in our data is caused by the relatively deep coalescence in 21% of the species, which widely corresponds with the missing match to the morphospecies, which is also reflected by the lack of a classical barcoding gap (Fig. 4). The impact is high with only 31 out of the 56 morphospecies matching perfectly the boundaries of inferred MOTUs (Fig. 2). The nature of maternal inheritance of mtDNA and its very low recombination rate is probably the major reason for these patterns of deep coalescence. Historically acquired genetic differentiation, for example in previously isolated populations, is maintained in secondarily mixing populations. The more often such isolated populations occur in time and space, for example due to climatic fluctuation during the Pleistocene in geographically highly structured areas such as Southeast Asia, the more often we encounter such ”paleogeographically induced” infraspecific variation which leads to the same result as current geographic variation. This effect consequently impedes species delimitation methods in the same way, particularly in a single marker system (e.g., cox1 ).
Similarly, in our data we could also report cases of incomplete lineage sorting and/or hybridisation. In three cases, morphospecies were not monophyletic (Microserica sp 11/ Microserica sp 13 vs.Microserica varians ; Neoserica sp 29 vs. Neoserica martinui ; Maladera sp 27 vs. Maladera sp 9.), while another three morphospecies shared identical haplotypes (Maladerasp 3a, sp 3b; sp 4). In all cases, we may exclude cross contamination based on the position of the single samples on the DNA-extraction microtiter-plates. These cases do occur in only rather closely related species, which might show similar life traits (e.g., daytime activity inMicroserica ), chemical communication, or mating behaviour (which is however, unknown for all species). In those instances, lumping of morphospecies in DNA based species delimitation seems to be more likely; however, also over-splitting was observed (e.g. Microserica ). Despite strong divergence in male genital morphology, hybridization between closely related species of Sericini have been reported (e.g., . The rather divergent structure of the aedeagus of the different species might function with females by mechanical isolation (lock-and-key hypothesis) that prevents mating between different species . However, although there have been some recent work on the morphology of female genitalia, our knowledge on copulation functionality and mechanics is still not sufficient to tell if morphological structures of males and female genitalia actually function as a barrier, if only through tactile recognition by cryptic female choice .
Again, the present study demonstrates the necessity of an integrative taxonomy in the sense of Yeates et al. (2011) (see also . We showed that the use of different clustering- and tree-based delimitation methods (Carstens et al., 2013) with the same single maker reproduces the same erroneous signal in slightly different variations. It is thus critical to corroborate results with data from other sources (e.g., genital or larval morphology, feeding traits, behaviour, etc.; e.g. Janzen et al., 2009) to allow for independent testing of species boundaries.
Sericini chafers proved to be a valuable model system, because of robust morphospecies assignments that were facilitated by the highly dissimilar and morphologically complex male genitalia that perfectly serve as a species diagnostic trait .
Overall, the initial hypothesis of impeccable DNA-based species boundaries in syntopically co-occurring species assemblages clearly had to be rejected. This was rather unexpected, especially since there was no additional evidence from other sources that these over-splittings could relate to cryptic diversity (Janzen et al., 2009, 2017; Janzen & Hallwachs, 2011).
Given the highly simplified parameters of DNA based species delimitation in this one-site species assemblage, it becomes clear how complex species delimitation with DNA-based methods is. Performance with mean error rates of more than 30% are under the expectations for proper use for applied sciences and conservation management. Even more sophisticated methods did not perform better than over-simplified threshold clustering methods as used for example in metabarcoding. Once more, we highlight the necessity of morphology for the verification ofde novo species delimitation results and the constant need of integrative taxonomic approaches.