The omission of critical data in the pursuit of ‘revolutionary’ methods to accelerate the description of species

Abstract

Recently, Meierotto et al. (2019) proposed a ‘revolutionary’ protocol for the description of understudied hyperdiverse taxa. The premise of their study was to champion exclusively DNA-barcode-based species descriptions (=diagnoses), which would dramatically increase the rate of description and provide a ‘human-readable record in the literature’ (unlike a Barcode Index Number, BIN; Ratnasingham & Hebert, 2013) that can later be supplemented with additional information. Species are always delimited against already known species (Linnaeus, 1753, 1758; Mayr, 1992; ICZN, 1999; Naciri & Linder, 2015; Renner, 2016). This was also recognized by Meierotto et al. (2019, p. 120): ‘Requirements for the publication of new species include (…) that they be accompanied by either a description or diagnosis which can separate them from any known species with which they are likely to be confused’. However, the latter authors failed to diagnose their 15 new Zelomorpha Ashmead, 1900 species from 51 out of 52 previously known species (only the type species was used in the analysis) and their three new Hemichoma Enderlein, 1920 species from any of the five previously known species. This is not the first case of its kind in zoology; Hebert et al. (2004) proposed to recognize ten species of skipper butterflies (genus Astraptes) based on DNA characters and ecology and some morphological characters, but the species were not formally named until Brower (2010) described them based exclusively on unique mutations in the DNA barcode region. Brower (2010) hailed this method as a flagship example of DNA barcoding's success in overcoming the ‘taxonomic impediment’ (Brower, 2010). However, it has received extensive criticism (DeSalle et al., 2005; Pons et al., 2006; Rubinoff et al., 2006; Elias et al., 2007; Dupérré, 2020). Meierotto et al. (2019) have taken this approach one step further by immediately assigning names to the lineages. This is, in our opinion, a step too far. Each description includes a lateral habitus image of a single specimen, a short diagnosis based solely on COI barcode nucleotide differences, brief notes on biology, and largely unannotated type specimen information except for brief mention of the locality and host caterpillar of the holotype. There are no morphological descriptions, and as mentioned, the ‘molecular diagnosis’ of their new Zelomorpha spp. are compared only with that of the type species, and not to the other 51 already known species of that genus, and in the case of Hemichoma, with none of the five species that were already described. We consider this poor taxonomic practice, and their approach to be fundamentally flawed. We urge innovators to pursue revolutionary new approaches that do not undermine the value of taxonomic expertise or produce sloppy results, but rather seek to draw on the latest methodological advancements to increase the rate of taxonomy without compromising on quality. Over the last 20 years, there have been many calls for an increase in the rate of taxonomic description (Mora et al., 2011). Few have delivered on that promise. Undoubtedly, the single steepest increase in this rate was made possible by DNA barcoding (Hebert et al., 2003). Yet, in the wake of the genetic revolution, there were already concerns that too much emphasis might be placed on DNA barcoding data alone, leading taxonomists to neglect the importance of other data—integrative approaches, taking the congruence of genetic signals with other datasets, would be required to keep describing biologically meaningful units (Dayrat, 2005; Ebach & Holdrege, 2005; Will et al., 2005). DNA barcoding is a rapid means to sort specimens into clusters, identify species and discover new ones (when a library of the relevant named species is already available), but does not overcome the bottleneck of the description process itself. Proposals for methods to speed up that process were dubbed ‘turbo’ or ‘fast-track’ taxonomy—an approach that does not differ fundamentally from previous species descriptions, but relies more heavily on formulaic descriptions of large numbers of new species (Butcher et al., 2012; Riedel et al., 2013). In essence, the approach of Meierotto et al. (2019) is simply another one of these turbo-taxonomic approaches except in three key aspects: Fundamentally, a diagnosis should identify features or combinations of features of a new species that are unique, that is, that allow it to be distinguished from all previously named taxa. It thereby gives a means not only to identify the new species but also to demonstrate that the new taxon is not a synonym of an existing one. The diagnoses of Meierotto et al. (2019) are based solely on DNA barcodes, but no barcodes are presented for 51 of the 52 existing species of Zelomorpha. Indeed, the existing taxa are summarily ignored, except for a statement that the notes of the second author (M.J. Sharkey) were used to verify that the new species are distinct from the existing taxa, without providing any evidence. As a result, it is impossible, based on the study of Meierotto et al. (2019), to assess whether or not their 15 new Zelomorpha names are distinct from 51 of the 52 species that were already described. In our opinion, Meierotto et al. (2019) have impeded, not enhanced, the taxonomy of these wasps. Ignoring almost all previously described species in a genus is indeed a way to speed up taxonomy—the process of comparison becomes very easy when you neglect practically all existing names—but it also creates chaos. Even the fastest approach to taxonomy will always require consideration of existing names before new ones can be established. Morphologically homogeneous (‘cryptic’) species are difficult to diagnose from one another, even when substantial differences exist in their DNA barcodes. This can delay taxonomy, because more effort must be invested per species to identify characters that do indeed differ. To overcome this problem, Renner (2016) recently called for more widespread inclusion of DNA sequence data in diagnoses. However, we do not believe that Renner (2016) envisioned the complete replacement of the diagnosis by single nucleotide changes, but rather expansion of concise but comprehensive diagnoses with such information. In some cases, restriction exclusively to genetic markers may be appropriate (e.g. where morphology is highly plastic, or where extremely distinctive genetic lineages are demonstrably cryptic in all other available lines of evidence), but such cases are likely to be the exception rather than the rule. Having complementary lines of evidence, such as morphology, is particularly important when, as is the case in Meierotto et al. (2019), only a tiny portion of the available names have DNA sequence data available. The lack of overlap between morphological and genetic data will further delay the process of clarifying whether or not the new names are synonyms of existing species. It is also important to note that DNA barcoding relies wholly on mitochondrial markers (usually cytochrome oxidase-I). Mitochondrial trees often disagree with nuclear species trees, especially in taxa where Wolbachia may be altering mtDNA introgression (Klopfstein et al., 2016). In these cases, and especially when genetic data are the sole basis of species-level recognition, congruence between nuclear and mitochondrial signal should be tested to better reinforce the species units identified. Moreover, as explained by Dupérré (2020), purely DNA-based descriptions will not only make the identification of millions of historical specimens impossible, it will impair this science in developing countries which house most of the undiscovered portion of biodiversity, due to high costs and lack of staff and technology. Considering the status of taxonomy as a fundamental science, this would drastically affect other related fields of study and, importantly, conservation. Renner (2016) also called for more emphasis on diagnosis and not description. With highly descriptive taxonomy, a great deal of time is invested in description of features that are not informative for the distinction of species from one another, which is time that could be spent instead diagnosing substantially more species. Instead, she and others have emphasized the importance of high-resolution photographs as supplements to diagnoses. We agree that detailed high-resolution photographs of specimens can indeed be highly valuable, but we contend that (i) there must be several photographs available, not a single lateral photograph of a single specimen, as provided by Meierotto et al. (2019), and (ii) some text highlighting important diagnostic features is valuable to experts, and of paramount importance to nonexperts, who must instead play a game of ‘spot-the-difference’ when such information is lacking. Experts might know the difference between variable and nonvariable characters, whereas such features cannot be distinguished by nonexperts, and it is the purpose of the diagnosis, if not the description, to point such features out. Moreover, we note that the photographs of Meierotto et al. (2019) are sometimes blurry and almost all of them cut off the tips of the antennae! Finally, it has been shown that (e.g. in case of tropical parasitoid wasps) the most time-consuming part of species discovery is field sampling (Sääksjärvi et al., 2004; Hopkins et al., 2019), and the actual description of the species may be written within minutes when material and expertise are already available. To make up the gaps in the existing barcode database, which contains maybe 2% of currently named species worldwide (see http://www.boldsystems.org/), far more survey work needs to be undertaken. A comprehensive barcoding database for a given taxon is a prerequisite to contemplating a DNA-only approach akin to that of Meierotto et al. (2019), and one that will require substantial further work to assemble. In the face of the Holocene (=sixth) extinction, taxonomists are racing to describe the 8 million unnamed eukaryotes that lie between the 2.07 million species currently named (Frid & Caswell, 2016), and the estimated ten million extant species (Mora et al., 2011; but see also Larsen et al., 2017 for well-reasoned estimates orders of magnitude higher). Currently, the rate of description is around 18 000 species per year (IISE, 2011), but with species going extinct at a rate 1000 times higher than the natural background rate of extinction, the annual species loss is clearly within or even higher than the rate of new descriptions (Dirzo & Raven, 2003; Mora et al., 2011), and thousands of species will undoubtedly go extinct before they can be described to science (IPBES, 2019). As the current average shelf life of new species between discovery and description is about 21 years (Fontaine et al., 2012), we do indeed need revolutionary new approaches to the discovery and description of new species. BINs and candidate species numbers (Vieites et al., 2009) already serve a valuable purpose as alphanumeric placeholders to recognize potentially evolutionarily significant diversity before it is taxonomically described. Simply assigning all BINs taxonomic names as Meierotto et al. (2019) propose would indeed complete the inventory of life on Earth extremely quickly (at precisely the same pace as the rate of barcoding)—that we do not dispute. But it would also remove the quantitative and qualitative difference between these preliminary identifiers (based on a single DNA marker) and full taxonomic recognition (based on a more comprehensive diagnosis, ideally supported by multiple lines of evidence including genetic data) that lend taxonomy its value. It would supplant taxonomists with technicians, who need to know nothing of the biology of the units with which they are dealing. The purpose of inclusion of molecular data in species descriptions should be to produce more precise taxonomic framework. A species description can be thought of as a hypothesis that can be supported or rejected when more data are obtained. Other researchers must have an opportunity to scientifically evaluate the status of the species in question. In our eyes, methodological changes to the way species are delineated and described are an important component of increasing the rate of species description, but dismissing the existing literature, and producing ‘descriptions’ that contain almost no information on the morphology of species, its variation, their unique features, their biology, or other aspects, do not constitute a revolution, and cannot be adopted. We note that real revolutions are undoubtedly coming, especially from the fields of machine learning and integrative species delimitation (Solís-Lemus et al., 2015; Favret & Sieracki, 2016), and also that it is possible to produce massive, rapidly assembled taxonomic monographs without compromising on quality (Rakotoarison et al., 2017). But we also want to emphasize that there is no shortcut to nirvana, and a true paradigm shift in taxonomy will come only when there is a revolution in the level of financial investment in taxonomy and the natural history museums that house the described and undescribed reference material of life on Earth (Wheeler, 2020), and when legislature stops acting to prohibit the collecting work of dedicated taxonomists while turning a blind eye to the innumerable organisms destroyed with every hectare of habitat that is lost (Britz et al., 2020). The authors declare no conflicts of interest. The data that support the findings of this study are available from the corresponding author upon reasonable request.

Topics & Concepts