Species delimitation of the Hyphydrus ovatus complex in western Palaearctic with an update of species distributions (Coleoptera, Dytiscidae)

Abstract The species status of Hyphydrus anatolicus Guignot, 1957 and H. sanctus Sharp, 1882, previously often confused with the widespread H. ovatus (Linnaeus, 1760), are tested with molecular and morphological characters. Cytochrome c oxidase subunit 1 (CO1) was sequenced for 32 specimens of all three species. Gene-trees were inferred with parsimony, time-free bayesian and strict clock bayesian analyses. The GMYC model was used to estimate species limits. All three species were reciprocally monophyletic with CO1 and highly supported. The GMYC species delimitation analysis unequivocally delimited the three species with no other than the three species solution included in the confidence interval. A likelihood ratio test rejected the one-species null model. Important morphological characters distinguishing the species are provided and illustrated. New distributional data are given for the following species: Hyphydrus anatolicus from Slovakia and Ukraine, and H. aubei Ganglbauer, 1891, and H. sanctus from Turkey.


History of classification
The genus Hyphydrus Illiger, 1802 represents a well-defined group of medium sized, globular shaped Dytiscidae. Altogether 139 species occur in all regions of the Old World, with most species distributed in tropical Africa (Miller and Bergsten 2016;Nilsson and Hájek 2017a). A taxonomic revision of the genus was published by Biström (1982).
Only three Hyphydrus species occur in Europe (cf. Nilsson and Hájek 2017b). While the Mediterranean H. aubei Ganglbauer, 1891 can be easily identified based on black markings on ferrugineous dorsal surface, the uniformly dark-ferrugineously coloured H. anatolicus Guignot, 1957 is very similar to the widespread western Palaearctic H. ovatus (Linnaeus, 1760) and it was not recognised until 1957. Hyphydrus anatolicus was described originally from Angora [= Ankara], Turkey (Guignot 1957). Subsequently Sanfilippo (1963) described the same species under the name H. carrarai Sanfilippo, 1963 from Italy. The synonymy of both species was established by Pederzani (1976). The species was later included in the revision of Biström (1982), who synonymized H. anatolicus with the older name H. sanctus Sharp, 1882, known previously only from the Levant region. Biström (1982) also argued that H. sanctus and H. ovatus should possibly be regarded as subspecies, but that more work was needed. Although Wewalka (1984) described the differences between H. anatolicus and H. sanctus, and a habitus photo of H. anatolicus was published by Hájek (2009), both mentioned species remain enigmatic, predominantly because of their similarity with H. ovatus, and because their distribution is not satisfactorily known.

Molecular data from museum specimens
With the advance of DNA Barcoding, extraction and amplification techniques have moved forwards in two directions. First towards high-throughput low-cost facilities racing from specimens to barcodes (Ivanova et al. 2006) and boosted by next-generation sequencing techniques (Shokralla et al. 2014). Second towards non-destructively generating DNA sequence data from older museum material with degenerated DNA (Gilbert et al. 2007). The latter will get ever more important as local and global extinction of species due to human activities means that getting fresh material of many species will be impossible or increasingly difficult. Therefore the only resort is to old, often dry-pinned or dry-mounted museum material, with the DNA degraded to various degrees. Little is known about exactly how fast DNA degrades under various conditions (but see Allentoft et al. 2012), but any probability model will have longer half-time the shorter the fragment. Thus, aiming for shorter amplicon size has been the preferred method, not least seen in the field of ancient DNA (Thomsen et al. 2009).
In this study, one of the three focal species is very rarely collected hence we attempt to amplify a >800bp segment of cytochrome c oxidase subunit 1 (CO1), from 19-25 years old dry-mounted specimens. We do this by using additives to standard DNA extraction lysis solutions and designing a number of internal primers to amplify the target segment in six short but overlapping fragments. Extractions are done on whole body but completely non-destructive, an important requirement for invaluable museum specimens.
We also use the general mixed Yule coalescence model (Pons et al. 2006) and a likelihood ratio test to explicitly test whether the H. ovatus-complex is better seen as one species (null hypothesis) or several species (alternative hypothesis) in a statistical likelihood framework. The GMYC model was developed as a tool for exploring and delimiting poorly known faunas based on DNA sequences. However here we use it in the context of testing questioned taxa of unsettled taxonomic status in an integrated toolbox where both DNA sequence data, speciation/coalescence models and morphological data bear evidence on the hypothesis.
To clarify the status and distribution of Hyphydrus anatolicus and H. sanctus, we provide a basal differential diagnosis of both species and related H. ovatus. We confirm the specific status of all taxa with molecular analysis. In addition, we review published records and add new faunistic data for H. anatolicus and H. sanctus, as well as the first record of H. aubei from Turkey.

Material and methods
Hyphydrus ovatus was sampled throughout Europe. We acquired fresh material of H. anatolicus from Russia and dry-mounted specimens from Turkey, Greece and Slovakia. Hyphydrus sanctus was available only as dry-mounted specimens from Israel and Turkey for molecular analysis; H. aubei was used as an outgroup in the parsimony and non-clock analyses. The specimens included in this study are deposited in the following institutional collections; for specimens included in molecular analysis, see Table 1. H. sanctus. The former was extracted in 96-well Wizard SV plates following the manufacturers instructions (Promega). The 3' end of cytochrome c oxidase subunit 1 (CO1) was amplified with the primers PatDyt or RonDyt (Isambert et al. 2011) and Jerry (Simon et al. 1994) using 1ul of DNA, Bioline Taq and the following cycling conditions: 94° for 2min, 35 to 40 cycles of 94° for 30s, 51-53° for 60s and 70° for 90-120s, and a final extension of 70° for 10 min. PCR products were cleaned with a 96-well Millipore multiscreen plate, sequenced in both directions using a Big Dye 2.1 terminator reaction, and analysed on an ABI 3730 automated sequencer. PatDyt and Jerry were used as sequencing primers. The older dry-mounted specimens were extracted using the QIAamp® DNA Micro Kit (QIAGEN®), following the tissue protocol with the addition of 20ul of DTT (Dithiothreitol)(Sigma-Aldrich). PCR was done with a set of 6 newly designed primer pairs (Table 2) amplifying the complete 825bp CO1 segment in shorter overlapping segments between 147 and 228bp long. We used Ready-ToGo™ PCR beads (Amersham Biosciences) together with 1ul of 10uM of each primer, 2ul of DNA and 21ul water in a 25ul reaction. Cycling conditions started with a 5 min denaturation step at 95°C followed by two cycles of 30 s at 95°C, 30 s at 45°C (first, second and fourth fragments) or 50°C (third, fifth and sixth fragments), and 40 s at 72°C, then two cycles of 30 s at 95°C, 30 s at 43°C or 48°C and 40 s at 72°C, and 39 cycles of 40 s at 95°C, 40 s at 41°C or 46°C, 50 s at 72°C, then a final extension step of 8 min at 72°C. PCR reactions were purified with Exonuclease I and FastAP (Fermentas) in the proportion 1:4, and sequenced with a BigDye™ Terminator ver. 1.1 Cycle Sequencing Kit (Applied Biosystems), cleaned with a DyeEx 96 kit (QIAGEN) and run on an ABI Prism 3100 Genetic Analyzer (Applied Biosystems). Sequences are submitted to Genbank under accession codes FN998871-FN998899 and JX221701-JX221703.

BMNH
Sequences were assembled and edited in Sequencher 4.8 (Gene Codes Corporation) and aligned in ClustalX 2.0 (Larkin et al. 2007) with default settings of 15 as gap opening penalty and 6.66 as gap extension penalty. The alignment contained no gaps. Table 2. Newly designed primers (apart from Jerry and PatDyt) used to amplify 825bp of CO1 in 6 overlapping fragments from 11-25 years old, dry-pinned, Hyphydrus specimens. Jerry  CAACATTTATTTTGATTTTTTGG  1  178bp  Hyp178rw  AATATGCTCGAGTATCAAC  1  Hyp161fw  GTTGTATGAGCTCATCATATA  2  189bp  Hyp349rw  TAGATGAATTTGCAAGGACTAC  2  Hyp276fw  AGCTACCCTTCACGGATCTC  3  125bp  Hyp400rw  CATAATGAAAGTGAGCCACTAC  3  Hyp371fw  GTAGTCCTTGCAAATTCATCT  4  228bp  Hyp598rw  CAGGATAGTCTGAGTAACG  4  Hyp507fw  TTACAGGACTATCATTAAATTCTA  5  147bp  Hyp653rw  CTCCAATAAATGATATAGTAGATC  5  Hyp616fw  CTCGACGTTATTCAGACTATCC  6  210bp  Patdyt  TCATTGCACTAATCTGCCATATTAG  6 Bayesian analysis was done with MrBayes 3.2.1 (Ronquist et al. 2012). We set up a partitioned model based on 3 rd resp. 1 st +2 nd codon positions and applied a HKY+G+I model to each partitions, unlinking statefrequencies, t-ratio, shape and proportion of invariable sites. Partitions were allowed separate rates with a variable rate prior. All other prior and proposal settings were left as default. We ran two separate runs each with four chains (one cold and three incrementally heated) 3 million generations sampled every 1000 th generation. First 25% was discarded as burn-in. For the first analysis we used a time-free model and rooted the tree with the outgroup Hyphydrus aubei. For the second analysis we excluded the outgroup and instead tested the placement of the root with a clockmodel. We used a Bayes Factor test to assess if the data was compatible with a strict molecular clock or if a relaxed clock should be used. A heuristic parsimony analysis was run in Nona (Goloboff 1999) (hold 10000, Mult*100, hold/10, mult*max*) spawned from Winclada (Nixon 1999(Nixon -2002. The parsimony analysis was followed by optimising the characters on the most parsimonious tree. This was done to show discrete character support for the three species. We performed a species-delimitation analysis using the general mixed yule coalescence model (GMYC) as implemented in R (R Development Core team 2005) with the package Splits (Ezard et al. 2009;Fujisawa and Barraclough 2013). We tested the nullhypothesis that the Hyphydrus ovatus-complex is a single species versus the alternative hypothesis that it consists of more than one species with a likelihood ratio test under the GMYC model. The GMYC method optimizes the likelihood of a single threshold across an ultrametric gene-tree. The threshold defines speciation branches towards the root from the threshold and within-species coalescence branches towards the tips from the threshold. The older branches are modelled with a Yule (speciation) model while the younger branches are delimited into n-groups where each group is modelled with a separate coalescent process model. The maximum likelihood solution of the GMYC model (the likelihood is calculated placing the threshold at each node across the tree) is compared against a model treating the entire gene-tree as a single coalescence (i.e. as a single species) in the likelihood ratio test. We used the ultrametric clock-tree generated above as input to the species delimitation test.

Morphological observations
The specimens were examined using an Olympus SZX12 stereomicroscope. Measurements were taken with an ocular graticule. Habitus photographs were taken using a Canon MP-E 65mm f/2.8 macro lens with 5:1 optical magnification on bellows attached to a Canon EOS 550D camera. Drawings were made based on photographs taken using an Olympus SZX12 microscope equipped with a Canon EOS 1100D digital camera. Images of the same specimen/structure at different focal planes were combined using Helicon Focus 5.1.19 software. To avoid artefacts due to desiccation of poorly sclerotised parts, the genitalia were illustrated mounted in dimethyl hydantoin formaldehyde resin (DMHF) on the same card as the beetle.

Molecular analyses
Amplification was highly successful with the short fragment PCRs of old dry-mounted material ( Table 3). The full-length 825bp segment was achieved for the two H. sanctus specimens from Israel, 665bp for one of the Turkish specimens, and a 147bp segment of the second Turkish specimen, with three ambiguous base calls. The last specimen also gave a 175bp sequence from primer pair 1 ( Table 2) that turned out to be contaminated DNA with closest BLAST hit on Genbank being saccharomycete fungi. This is always a risk when extracting DNA from the whole body of a specimen. All three dry-mounted H. anatolicus specimens yielded full-length CO1 sequences. Genetic distances between the three presumed species in the ovatus-complex turned out to be large (Table 4). The distance between H. ovatus and H. anatolicus or H. sanctus was 9.4-11.4% (K2P-model). The distance between H. sanctus and H. anatolicus was slightly less, 6.7-7.1%. These genetic distances strongly indicate that we are dealing with three valid and separate species in the ovatus-complex. Within-species variation was less than 1.4%. The time-free bayesian analysis as well as the parsimony analysis, both rooted with H. aubei as outgroup, confirmed that the three presumed species are reciprocally monophyletic and separated from each other with long branches (Figs 1-2). Posterior probability support values were 1.0-0.98 for all three species. H. sanctus and H. anatolicus are sister species according to this single-gene phylogeny both in the outgroup-rooted trees (Figs 1-2), and in the clock-rooted tree (Fig. 3). Parsimony analysis and character optimization confirmed the H. sanctus + H. anatolicus sister group relationship with 17 supporting unambiguous and non-homoplasious substitutions (Fig. 2). Also all three presumed species were supported with between 16 and  24 unambiguous and non-homoplasious substitutions (Fig. 2). The Bayes factor test strongly favoured the strict clock (LnL=-1701) over a time-free model (LnL=-1764) (2*LnBF=125), hence a strict, as oppose to a relaxed, clock model was used to generate an ultrametric tree (Fig. 3). The GMYC model delimited three clusters congruent with the three presumed species as the maximum likelihood solution (Fig. 3). An approximate confidence interval of 2log likelihood units from the maximum likelihood (3 clusters) did not include any other solution. The explicit likelihood ratio test of the null hypothesis of a single coalescing unit (species) was refuted in favour of the alternative hypothesis of three separately evolving and coalescing units (-Log L one species = 211.9965, -Log L three species = 218.2261, Likelihood ratio=12.4592, p=0.00596).

Published records. Bosnia and Hercegovina:
Male. Longer metatibial spur long, nearly as long as metatarsomere I-II combined (Fig. 7a); spur bisinuate with only indistinct serration basally (Fig. 7a). Male genitalia as in Fig. 8a-d, median lobe in ventral view slightly narrowing from base to apex.
Female. Both shiny and matt forms known of females of H. anatolicus. Shiny form agreeing well with male; matt form with whole surface densely reticulated, meshes somewhat elongate on elytra. Large punctures well visible, small punctures indistinct among reticulation. Longer tibial spur shorter than in male; broad and with serration  in basal two thirds, narrowed, slightly curved and without serration in apical third. Female genitalia as in Fig. 8e-g. Habitat. The species inhabits various types of standing water, predominantly densely vegetated pools, ditches and small ponds. H. anatolicus tolerates also saline habitats.
Distribution. The species is distributed in the Eastern Mediterranean and in south-eastern Europe. It occurs in Italy, southernmost Slovakia, Hungary, the Balkan Peninsula, Turkey, southern Ukraine and Russia up to latitude 55° and east to the Ural Mountains (Fig. 9). First record from Slovakia and Ukraine.

Material examined.
We have examined more than 600 specimens from the Czech Republic, Finland, France, Germany, Great Britain, Russia, Slovakia, Sweden, and Ukraine, deposited in NHRS and NMPC. Diagnosis. Habitus as depicted in Figs 4b, 5b. Clypeus with anterior margin medially nearly straight (Fig. 6b). Reticulation of dorsal surface confined to head, more distinct and impressed anteriorly (Fig. 6b). Punctation of head fine, visible only in posterior half, punctures on clypeus imperceptible due to strong reticulation (Fig. 6b); punctures dense, distance between them smaller than their diameter (Fig. 6b). Punctation of pronotum double, coarse, distance between larger punctures smaller than their diameter. Punctation of elytra double, diameter of small puncture about half of diameter of large punctures; distance between large punctures, at least basally, smaller than their diameter. Epipleura smooth with fine punctures. Metatibia with outer margin nearly straight.
Male. Longer metatibial spur short, only slightly longer than metatarsomere I (Fig. 7b); spur nearly straight, broad with distinct serration (Fig. 7b). Male genitalia as in Fig. 8h-k, median lobe in ventral view parallel-sided in most of its length.
Female. Both shiny and matt forms are known for females of H. ovatus. Shiny form agreeing well with male; matt form with whole surface densely reticulated, meshes distinctly elongate on elytra. Large punctures well visible, small punctures indistinct among reticulation. Longer tibial spur similar to that of male. Female genitalia as in Fig. 8l-n. Habitat. The species inhabits various types of standing and slowly flowing water bodies with at least some vegetation. The typical habitats represent (frequently eutrophic) ponds, densely vegetated pools, ditches, oxbows or open swamps.
Distribution. Widely distributed Palaearctic species. With the exception of the Iberian Peninsula, it occurs in most of territory of Europe and temperate Asia east to the Baikal Lake (east Siberia).
Male. Longer metatibial spur long, nearly as long as metatarsomere I-II combined (Fig. 7c); spur broad and straight in basal two thirds with small but distinct serration, attenuated and curved apically (Fig. 7c). Male genitalia as in Fig. 8o-r, median lobe in ventral view slightly narrowing from base to apex.
Female. Only matt females of H. sanctus are known so far. Whole surface densely reticulated, meshes on elytra somewhat elongate. Large punctures well visible, small punctures indistinct among reticulation. Longer tibial spur similar to that of male, but almost straight in apical third. Female genitalia as in Fig. 8s-u.
Habitat. Similarly to the other two species, H. sanctus inhabits various types of standing and slowly flowing water bodies with at least some vegetation. Wewalka (1984) reported several specimens from a densely vegetated pool and single occurrences from an artificial pool with clear water, an irrigation ditch and from a stream.
Distribution. A species distributed in the Levant region of the Near East. So far recorded from several localities in Israel, Jordan and Syria (Fig. 9). First record from Turkey.

Hyphydrus aubei Ganglbauer, 1891
Note. Hyphydrus aubei is the fourth European species in the Hyphydrus ovatus species group sensu Biström (1982). It does not belong to the Hyphydrus ovatus complex as here defined and it is easily separated from the preceding three species based on colouration (Figures 4-5).

Key to species
Key to western Palearctic species of the Hyphydrus ovatus species group 1 Elytra with distinct black maculate colour pattern on elytra; head bicoloured, testaceous anteriorly but distinct black areas posteriorly (Figs 4a, 5a)  Punctation of pronotum and elytra (males and shiny females) very coarse; distance between larger punctures smaller than their diameter. Longer male metatibial spur only little longer than metatarsomere I; straight and with distinct serration (Fig. 7b)  Punctation of pronotum and elytra (males and shiny females) finer; distance between larger punctures larger than their diameter. Longer male metatibial spur almost as long as metatarsomeres I-II combined; spur not straight, bisinuate or apically curved; serration of spur small to indistinct (Fig. 7a, c) ...........3 3 Clypeus with anterior margin medially nearly straight; exterior side of metatibia almost straight; longer male metatibial spur straight basally but curved apically and with serration small but visible (Fig. 7c) ......Hyphydrus sanctus -Clypeus with anterior margin rounded; exterior side of metatibia somewhat sinuous; longer male metatibial spur bisinuate and with indistinct serration basally (Fig. 7a)  It is highly probable that many records may refer to the other two species, but whether H. ovatus is replaced by, or sympatric with, these remain to be investigated for many areas.