Phylogenetic study of the genus Sternolophus Solier (Coleoptera, Hydrophilidae) based on adult morphology

Abstract The phylogeny of the hydrophilid genus Sternolophus Solier, 1834 was examined in this study using 60 morphological adult characters, eight of them continuous and 52 discrete. The cladistic analysis resulted in a single most parsimonious tree with two major subclades corresponding, respectively, to species previously assigned to the subgenera Sternolophus s. str. Solier and Neosternolophus Zaitzev, although they are not re-instated. The species groups S. angolensis (Erichson, 1843) and S. solieri Castelnau, 1840 are recovered as monophyletic. The biogeography and diversification of the species of Sternolophus are briefly discussed.


Introduction
The genus Sternolophus Solier, 1834 is widely distributed in the tropics of the Old World, with only few species occurring in the temperate zones. In a recent taxonomic revision of the genus by Nasserzadeh and Komarek (2017), the number of species was increased from nine (Hansen 1999) to 17.
The phylogeny of Sternolophus has been poorly studied. Zaitzev (1909) split the genus into two subgenera, Sternolophus s. str. Solier, 1834 and Neosternolophus Zaitzev, 1909. His classification was based on the absence or presence of an emargination on the anterior clypeal margin. Although this subdivision was accepted by Orchymont (1919), this author considered the length of the spine on the metaventrite a more significant character. Smetana (1980) elevated Neosternolophus to generic rank based on the emargination of the anterior clypeal margin, but this change was later opposed by Hansen (1991). This subgeneric division was also rejected by Watts (1989) based on the wide inter-and intraspecific variation of the mentioned character within the Australian species. The phylogenetic relationships of Sternolophus species were also studied by Hansen (1991), Short (2010), Short and Fikáček (2013) and Toussaint et al. (2017), although these studies (with the exception of Short 2010) are mainly focused either on familyand tribe-level relationships (Hansen 1991;Short and Fikáček 2013) or had a biogeographic focus (Toussaint et al. 2017). Short (2010) included seven species of Sternolophus in his analysis of the subtribe Hydrophilina which resulted in the monophyly of the subgenus Sternolophus s. str. and the lack of resolution for species of Neosternolophus. Nasserzadeh and Komarek (2017) suggested changes to the subgeneric classification, and proposed two new species groups (the groups S. angolensis (Erichson, 1843) and S. solieri Castelnau, 1840) based on highest morphological similarity and without including a phylogenetic approach. These authors considered S. angolensis, S. inconspicuus (Nietner, 1856), S. mundus (Boheman, 1851) and S. solitarius Nasserzadeh and Komarek, 2017 as members of the angolensis group, and placed S. angustatus (Boheman, 1851), S. elongatus Schaufuss, 1883, S. mandelai Komarek, 2017, S. rufipes (Fabricius, 1792), and S. solieri in the solieri group. They left the remaining species (S. australis Watts, 1989, S. decens Zaitzev, 1909, S. immarginatus Orchymont, 1911, S. insulanus Nasserzadeh and Komarek, 2017, S. jaechi Nasserzadeh and Komarek, 2017, S. marginicollis (Hope, 1841, and S. prominolobus Nasserzadeh and Komarek, 2017) ungrouped.
Here the first comprehensive phylogenetic analysis of the genus Sternolophus is provided, based on a cladistics analysis of adult morphological characters. Considering the phylogenetic results, the biogeography and diversification of the species are briefly discussed.

Materials and methods
Taxon sampling. More than 4000 specimens in all the 17 species of Sternolophus were studied as ingroup, and Hydrochara flavipes, belonging to the tribe Hydrophilini, was included as outgroup. A total of 271 specimens were measured. The specimens were obtained on loan from the following institutions and collections: The examined specimens are listed in Appendix 1. The specimens were selected according to: 1) geographical distribution, 2) morphological variation, and 3) status as type specimens.
Preparation for morphological studies. To study the male genitalia, the aedeagus was extracted and macerated in lactic acid for at least four days to become hydrated and cleared before examination. Bursa copulatrix, spermatheca, and spermathecal gland were also dissected (for details see Nasserzadeh et al. 2005) and mounted in DMHF or Euparal on transparent cards and pinned below the associated specimens. Morphological data for each species were obtained using a stereomicroscope (Zeiss Stemi SV11). Measurements were made through a micrometric eyepiece and presented in figures 1, 8, 14−15, 20−21. Line drawings of characters were adapted from Nasserzadeh and Komarek (2017). Photographs were taken using a 650D Canon digital camera.
Character selection and coding. Character selection and character state definition follow Smetana (1980), Nasserzadeh et al. (2005 and Nasserzadeh and Komarek (2017). A total of 60 characters (eight continuous and 52 discrete) was selected and scored from zero to 59 (see Table 1). Eight continuous characters involving ranges and ratios were treated as such, avoiding the use of ad hoc methods to establish ranges (Goloboff et al. 2008). Discrete characters contained 45 binary and seven multistate. Characters 0, 2−6, and 8−45 correspond to the external morphology, characters 1, 7 and 46−55 were derived from the aedeagus, and characters 56−59 were coded from the female genital membranous tube. Characters and character state compositions approach the logic of neomorphic and transformational pattern as indicated by Sereno (2007). There are no missing characters in the data matrix, and the inapplicable characters were coded as '?' (Appendix 2).
Phylogenetic analysis. Cladistic analyses were performed on all characters in 'Tree Analysis using New Technologies' (TNT) (Goloboff et al. 2008) with 'traditional' search based on 5000 replicates, through 'tree bisection reconnection' (TBR) branch swapping holding 100 trees by collapsing rule 'min. length=0'. Discrete characters were treated as unordered, and multistate characters were treated as polymorphic (e.g. [0 1]). The same analysis was performed only on the discrete characters and the consensus tree was obtained using strict and majority-rule methods. An analysis including Table 1. List of morphological characters, character states, and codes.

Codes List of characters and character states Continuous characters 0
Average length of body in millimeters. 1 Average length of aedeagus in millimeters (Fig. 15a).
2 Ratio width of head (from outer lateral margin of eyes) / width of clypeus in anterior margin (connecting with labrum) in males.
3 Ratio width of head in outer margin of eyes / length of clypeus (from the centre of frontoclypeal suture (Fig. 3a) to anterior margin of clypeus). 4 Ratio average length of body / average length of aedeagus. 5 Length of hind femur (Fig. 13a) / widest part (Fig. 13b).

6
Ratio distance of bare area between the apical angle of the pubescent part of submentum to the base of mentum (Figs 5c, 6c) / width of anterior margin of submentum (connecting to the mentum) (Figs 5d, 6d). 7 Ratio length of aedeagus ( Fig. 15a)/width (widest part of the parameres) (Fig. 15b).

38
If length of spine on metaventrite long, spine: (0) straightly elongated almost in parallel to the ventral side; (1) slightly and gradually bend upward distally toward posterior end.

59
Longitudinal rows of small tooth-like spines on the membranous wall of the bursa: (0) absent; (1) present.
all continuous and discrete characters was also conducted by retaining suboptimal trees 0.5 steps longer than the most parsimonious tree; the resulting trees were summarized by strict and majority-rule consensus methods. The synapomorphic characters and character states are mapped on the single most parsimonious cladogram (analysis A). Branch support was calculated by bootstrap (Felsenstein 1985), jack-knife (Farris et al. 1996), and symmetric resampling (Goloboff et al. 2003), with 2000 replicates. Different numbers of replicates (up to 5000) did not affect the results. In resampling analysis, the results of the absolute frequency summarize method was used, which were slightly higher than the analysis using frequency difference.
The consistency and retention indices (Kluge and Farris 1969;Farris 1989) of discrete characters were calculated using PAUP version 4.0b10 (Swofford 2002) (analysis D). All 52 discrete characters were equally weighted, and multistate characters were treated as unordered. Heuristic searches were selected with 20000 random additions followed by branch swapping using TBR and holding a single tree (NCHUCK = 1, CHUCKSCORE = 1) (Alipanah et al. 2010).

Results
The parsimony analysis of all characters (analysis A) resulted in a single most parsimonious tree of 146.130 steps (Fig. 22). When suboptimal trees 0.5 steps longer than the most parsimonious tree were retained (analysis C), six most parsimonious trees were obtained. The consensus of these trees, either using strict or majority-rule methods, was congruent with the single most parsimonious tree from analysis A, except for slight differences in the position of the species within clades C and M (Fig. 23a, b). The analysis  of discrete characters only (analysis B) resulted in 36 most parsimonious trees of 110 steps. The consensus trees using both strict and majority-rule methods were different from previous trees in the position of the species in clade B (Fig. 24a, b). Analysis using PAUP on the 52 discrete characters (analysis D) estimated 38 parsimony informative characters, with consistency index (CI) = 0.56 and retention index (RI) = 0.72.
As shown in the single most parsimonious tree obtained with analysis A (Fig. 22), the examined Sternolophus species are divided into two major monophyletic clades, B and G, with 6 and 11 species respectively. Clade B contains S. decens as sister to clade C that is composed of five species, S. solieri, S. rufipes, S. angustatus, S. mandelai, and S. elongatus. Clade B is supported by five characters (0: 10.65-10.70, 1: 1.70-1.75, 6: 0.20, 30: 1, 37: 1), although it is weakly supported statistically. Except for the elongated spine on the metaventrite (37: 1), the characters sustaining this clade were homoplastic. The topology of clade B was slightly different in analysis C (Fig. 23), and the clade was not maintained in analysis B, with the six species unresolved in the strict consensus (Fig. 24a), whereas in the majority-rule consensus tree (Fig. 24b) S. decens was resolved as sister to clade G in 64% of the cases (24 out of 36 trees). The monophyly of clade G was well supported in all analyses (Figs 22-24). Monophyly of this clade is supported by the following five synapomorphies: the rufous to testaceous coloration of the labrum exceeding one third of its length (16: 1); the semicircular arrangement of the paired antero-lateral group of punctures on clypeus (17: 0); the presence of an emargination on the anterior margin of clypeus (19: 1); the moderately long maxillary palpus (22: 2); and the slim sternal keel of metaventrite (42: 0). All analyses also agreed in the monophyly of clade I, although with weaker support (Figs 22-24). Five synapomorphies sustain this clade: the narrow distance between paired antero-lateral groups of punctures on the clypeus (narrower than one-sixth of the width of clypeus at anterior margin of eyes) (18: 0); the absence of infuscation on the apex of fourth maxillary palpomere (21: 0); the belly shape of the pubescent area of submentum (24: 2); the presence of an emargination on the apical margin of ventrite 5 (44: 1); and the weakly curved and short male claw on fore leg (45: 0). Based on the results of analysis A (Fig. 22), S. australis is sister to clade I, whereas S. immarginatus is sister to the clade formed by S. australis and clade I. In all analyses, clades K, L, M, and N were found to be monophyletic with the same configuration. These clades are supported by one, two, three, and three synapomorphies, respectively (Fig. 22); however, the position of the four species within clade M was unstable in all analyses.
The comparison of the trees obtained using all characters (Figs 22, 23) with those obtained using only discrete characters (Fig. 24) reveals the influence of continuous characters in the formation of clade B. The exclusion of continuous characters from the analysis causes the species within this clade to collapse in a polytomy (Fig. 24). Clade B is supported by three continuous and two discrete synapomorphies. Similarly, continu-ous synapomorphies outnumber discrete synapomorphies within clade B, except for clade C with one continuous and three discrete synapomorphies (Fig. 22). The importance of continuous characters in shaping clade B can be explained by the fact that this character set (0 to 7) provides diagnostic features for separating the morphologically very similar species of the solieri species group (clade C) (Nasserzadeh & Komarek 2017). In all analyses, the topology of clade G remained consistent except for slight changes in clade M and variable support for clades G and I (Figs 22-24). On the other hand, Sternolophus decens was recovered in clade B in five of the six most parsimonious trees obtained using both continuous and discrete characters combined (Fig. 23b), whereas it was sister to clade G in more than 60% of the 36 most parsimonious trees obtained using discrete characters only (e.g., Fig. 24b), showing that the position of this taxon is also highly influenced of continuous characters.

Discussion
Taxonomy. The species formerly included in the subgenera Sternolophus s. str. and Neosternolophus were recovered into two major subclades, B and G, respectively. However, due to the following considerations, subgeneric status was not re-instated: i) Unreliable topology of clade B in different analyses and absence of support for its monophyly as well as monophyly of the subclades. ii) Questionable position of S. decens within clade B. Sternolophus decens was included in the subgenus Sternolophus s. str. by Zaitzev (1909), and was found to be closely related to S. rufipes and S. solieri by Short (2010). However, it was recovered in a monophyletic clade together with S. marginicollis (and some unidentified Sternolophus species) by Toussaint et al. (2017), which was included in the subgenus Neosternolophus by Zaitzev (1909). In the trees obtained in analyses A and C (Figs 22-23), S. decens was recovered as sister to clade C. The species of this clade (S. solieri, S. rufipes, S. angustatus, S. mandelai and S. elongatus) (Fig. 22) were grouped in the solieri species group by Nasserzadeh and Komarek (2017) based on highest morphological similarity. iii) A nearly similar topology was obtained for clade G in the different analyses, all of them including S. marginicollis, with strong support. Based on the topology obtained here and those of Short (2010) and Toussaint et al. (2017), we believe that reinstating subgenera within Sternolophus is premature and would not reflect the evolutionary history of the genus. Further investigations including larval and molecular characters of as many species of the genus as possible, as well as other techniques such as scanning electron microscopy, are required to resolve its phylogenetic relationships. Short (2010), in his phylogenetic analysis of the subtribe Hydrophilina based on adult-morphological characters, found evidence for monophyly of the subgenus Sternolophus s. str., but the species formerly grouped in the subgenus Neosternolophus were unresolved and formed a basal polytomy within the genus. In our analysis, on the contrary, strong evidence was found for monophyly of Neosternolophus, whereas monophyly of Sternolophus s. str. is more questionable for the reasons mentioned above.
Finally, the four species (S. solitarius, S. mundus, S. inconspicuus and S. angolensis) grouped by Nasserzadeh and Komarek (2017) as the angolensis species group based on morphological similarities, are resolved here as clade M confirming their close relationship, although weakly supported (Fig. 22).
Biogeography and diversification. In Figure 22 (right table), clade C consists of the solieri species group distributed in the Afrotropical, Palaearctic and Oriental regions. Distribution of S. decens overlaps with those of clade D. On the other hand, most members of clade G have an Oriental-Australasian distribution. The exceptions are representatives of the angolensis species group, with S. solitarius, S. mundus, and S. angolensis restricted to the Afrotropical Region whereas S. inconspicuus is widely distributed in the Oriental Region to the eastern boarder of the Palaearctic Region. Sternolophus insulanus and S. jaechi are two sister species with insular distribution in the Malay Archipelago (see Appendix 1). Toussaint et al. (2017) postulated an Afrotropical origin for Sternolophus, dispersing toward Australia in the Oligocene/Miocene. There are many New Cenozoic fossil findings of taxa closely related to Sternolophus in Europe and North America (e.g. Fikáček et al. 2008Fikáček et al. , 2010aFikáček et al. , 2010b, whereas the only record of this genus is a dubious fossil likely belonging to S. rufipes from the Early Pliocene of the Tsubusagawa Formation in Japan (Hayashi et al. 2003). The current distribution of Sternolophus in the Old World, i.e. without protruding into northern Asia, Europe, Tasmania and New Zealand (Nasserzadeh and Komarek 2017), which were largely covered by ice, and its absence in the fossil records from Europe and America, suggest a sensitivity of this group to climate change and glacial periods as inhibitor factors for its distribution, and also highlight the effect of eustatic changes in accelerating its dispersal in the Old World towards Australia.

App1.
List of the specimens examined.