Received: 14 January 2019 | Revised: 20 March 2019 | Accepted: 10 April 2019 DOI: 10.1111/1755-0998.13027 RESOURCE ARTICLE A population genetic assessment of taxonomic species: The case of Lake Malawi cichlid fishes Catarina Pinho1 | Vera Cardoso1 | Jody Hey2,3 1CIBIO/InBIO, Centro de Investigação em Biodiversidade e Recursos Abstract Genéticos, Universidade do Porto, Vairão, Organisms sampled for population-level research are typically assigned to species Portugal by morphological criteria. However, if those criteria are limited to one sex or life 2Rutgers, the State University of New Jersey, Piscataway, New Jersey stage, or the organisms come from a complex of closely related forms, the species 3CCGG, Center for Computational assignments may misdirect analyses. The impact of such sampling can be assessed Genetics and Genomics, Department of Biology, Temple University, Philadelphia, from the correspondence of genetic clusters, identified only from patterns of genetic Pennsylvania variation, to the species identified using only phenotypic criteria. We undertook this Correspondence protocol with the rock‐dwelling mbuna cichlids of Lake Malawi, for which species Jody Hey, CCGG, Department of Biology, within genera are usually identified using adult male coloration patterns. Given high Temple University, Philadelphia, PA. Email: [email protected] local endemism of male colour patterns, and considerable allele sharing among spe- cies, there persists considerable taxonomic uncertainty in these fishes. Over 700 Funding information National Institutes of Health, Grant/Award individuals from a single transect were photographed, genotyped and separately as- Number: R01GM078204; Fundação para signed: (a) to morphospecies using photographs; and (b) to genetic clusters using five a Ciência e a Tecnologia, Grant/Award Number: PTDC/BIA‐BDE/66210/2006 widely used methods. Overall, the correspondence between clustering methods was and IF/01597/2014; European Social Fund; strong for larger clusters, but methods varied widely in estimated number of clusters. Ministério da Educação e Ciência The correspondence between morphospecies and genetic clusters was also strong for larger clusters, as well as some smaller clusters for some methods. These analy- ses generally affirm (a) adult male-limited sampling and (b) the taxonomic status of Lake Malawi mbuna, as the species in our study largely appear to be well‐demarcated genetic entities. More generally, our analyses highlight the challenges for clustering methods when the number of populations is unknown, especially in cases of highly uneven sample sizes. KEYWORDS cichlids, hybridization, population genetics, speciation 1 | INTRODUCTION frequently depend upon the use of diagnostic characters (Cracraft, 1983), but diagnostic characters per se play no role in the theory The processes of discovering new species and of assigning individuals of how species come into being or how they persist. Species do of to previously described species sometimes come into conflict with course evolve derived fixed traits that sometimes serve well as diag- the concept of a species as a sexual evolving lineage. For example, nostic characters, but cryptic species are also common (Bickford et species diagnosis and species identification, by human investigators, al., 2007). For species that are highly visual, and for which coloration This is an open access article under the terms of the Creat ive Commo ns Attri butio n‐NonCo mmerc ial‐NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. © 2019 The Authors. Molecular Ecology Resources Published by John Wiley & Sons Ltd. 1164 | wileyonlinelibrary.com/journal/men Mol Ecol Resour. 2019;19:1164–1180. PINHO ET AL. | 1165 patterns are cues, both for reproduction and for species diagnosis assumption underlying this type of sampling is that male nuptial by human investigators, we might expect that species diagnosis by coloration and other distinctive morphological features are indeed human investigators will accurately capture real species. good characters to delimit species. There are several reasons why We undertook a close examination of the process of species this might not be generally so. On the one hand, the diversity of co- identification in a classic species group where male colour variation lour could be driven in part by phenotypic plasticity. Also, intraspe- plays a large role in both reproduction and species identification by cific colour polymorphism in sympatry has been well documented in investigators. Specifically, we asked how well species assignment several species (Sinervo & Lively, 1996), including cichlids (Maan & based on morphological criteria (primarily coloration) corresponded Sefc, 2013, and references therein). to species as genetic entities, as estimated using clustering methods It is also possible that sampling on the basis of one set of mor- that minimize linkage disequilibrium (LD) and inbreeding disequilib- phological criteria can fail to reveal the presence of cryptic species rium among multiple polymorphic loci. that are similar by those criteria. Cryptic species are common in The cichlid fishes of the great African lakes represent arguably many biological systems (Bickford et al., 2007), and they have been the single most striking known radiation of vertebrate species, with reported among Lake Malawi cichlids (Kidd et al., 2006). If cryptic each of the lakes harbouring hundreds of species, and with some species are frequent, our view on cichlid diversity might be signifi- parts of these radiations having occurred very recently, and thus ex- cantly underestimated. ceedingly rapidly (Kocher, 2004; Seehausen, 2006). This system con- In this study, we asked the following question: how do the spe- stitutes one of the most important natural laboratories for studies in cies identifications of individuals, based only on traditional mor- speciation, particularly because of the complex interplay between phological criteria, correspond to genetic entities (i.e., gene pools) ecological opportunity and sexual selection, which presumably discovered using a population genetics approach? To be clear when drove the formation of multiple species (Salzburger, 2009; Wagner, referring to instances of the two kinds of groups, we refer to the spe- Harmon, & Seehausen, 2012). However, the system also presents cies to which individuals were assigned using morphological criteria, significant challenges in species diagnosis and identification (Turner, as “morphospecies” and to the groups identified only by patterns of Seehausen, Knight, Allender, & Robinson, 2001). genetic variation as “genetic clusters” or “genetic populations”. The systematics of African cichlids has traditionally relied on fea- We focused mainly on a community of cichlids, the rock-dwell- tures such as body shape and morphology of the pharyngeal and oral ing “mbuna” species of Lake Malawi that live primarily in the shal- jaw apparatus (Rüber, Verheyen, & Meyer, 1999; Trewavas, 1984). low waters of the rocky portions of the coastline. Individuals were More recently, male secondary sexual characters such as bower collected without regard to coloration by sweep netting a transect shape (Kidd, Kidd, & Kocher, 2006) and especially male nuptial color- along a rocky shore area of Lake Malawi. A large collection of juve- ation have also been used to describe species (Genner et al., 2004). niles and adults of both sexes were then photographed and identi- A reliance on male breeding coloration to delimit species necessarily fied to morphospecies, and then later genotyped at a set of highly overlooks the identification of females and juveniles, which are com- polymorphic loci to identify genetic clusters. With these two funda- paratively drab in coloration (Kornfield & Smith, 2000). Because of mentally different types of identifications in hand we could then ask this difficulty, sampling schemes in evolutionary or ecological stud- how well the two different groupings corresponded to each other, ies usually rely on typical males that can indeed be readily identified, and we could potentially identify hybrid individuals. Finally, we could while overlooking other individuals, including less typical males, fe- also ask: had we sampled only males in breeding coloration, would males, juveniles and possibly intermediate phenotypes. we have a biased or misleading sample of the genetic populations? When species are studied by sampling only males in breeding coloration there are implicit and possibly erroneous assumptions 2 | MATERIALS AND METHODS about the nature of those species. One assumption is that such males are representative of the species, which will be in error if they 2.1 | Sampling and photographic record of cichlids somehow differ genetically from other life stages or from females. Secondly, a reliance on typical male individuals might skew investi- Samples were collected off the rocky shore of the western side of gators’ perception of the magnitude of species colour differences Domwe Island, in southern Lake Malawi. Our choice of sampling and cause individuals that are intermediate between investigators’ habitat reflects the intention of collecting especially the brightly diagnostic patterns for related species to be overlooked. Species hy- coloured rock-dwelling mbuna, which are popular objects of both re- brids or backcrossed individuals are likely to be overlooked by this