The enduring value of reciprocal illumination in the era of insect phylogenomics: a response to Cai <i>et al</i> . (2020)

Arguably no other group within Coleoptera has received as robust and sustained investigation into their phylogenetic relationships as aquatic beetles (Short, 2018). Among this ecological guild, evolutionary relationships of the families within Dytiscoidea, a clade comprising the charismatic diving beetles (Dytiscidae) and their close relatives, have received particular attention (Ribera et al., 2002; Balke et al., 2005; Balke et al., 2008; Alarie et al., 2011; Hawlitschek et al., 2012; Toussaint et al., 2016). Very recently, four different studies were published investigating the phylogeny of Dytiscoidea, three of which utilized phylogenomic data (Table 1), the most recent by Cai et al. (2020). Cai et al. (2020) (hereafter CEA) approached investigating the evolutionary relationships among dytiscoid families by reanalysing the transcriptomic dataset of Vasilikopoulos et al. (2019) using different evolutionary models and data trimming regimes. CEA's analyses recovered three different topologies for relationships amongst Dytiscoidea (Fig. 1), two of which (Fig. 1A, B), have been recovered in several previous studies (Table 1). The primary difference among these topologies is the placement of Hygrobiidae, either as sister to (Dytiscidae (Amphizoidae + Aspidytidae)) (Fig. 1A), sister to Amphizoidae + Aspidytidae (Fig. 1B), or as sister to Dytiscidae (Fig. 1C). In CEA, topologies shown in Fig. 1A, C both received maximal (e.g. bootstrap values of 100 and posterior probabilities of 100%) to strong support respectively via their preferred model of evolution. Whereas, CEA's recovery of Hygrobiidae sister to Amphizoidae + Aspidytidae (Fig. 1B) was not as strongly supported, Gustafson et al. (2020) recovered this topology primarily with strong to maximal support across all analyses with comprehensive taxon sampling of Dytiscoidea. Rather than treating the three topologies recovered both within their own study and elsewhere as equally viable hypotheses (Table 1), CEA dismissed the relationships shown in Fig. 1A, B as the result of phylogenetic methodological error, promoting Fig. 1C as their preferred tree because it is '…consistent with morphology-based views of dytiscoid relationships.' (Cai et al., 2020: 5). Here, we address (i) the manner in which CEA approached reconciling conflicting hypotheses about the evolution of Dytiscoidea; and (ii) the misconception that dytiscoid relationships shown in Fig. 1C are the most consistent with morphology-based views in relation to those of Fig. 1A, B. Phylogenomic datasets present a new challenge in that nodes in recovered trees are often maximally supported (e.g. bootstrap values of 100 or posterior probabilities of 100%). In certain cases, even incorrect species tree topologies can receive strong statistical support due to systematic biases, incomplete lineage sorting, gene tree conflict, biased taxon sampling, model misfit, etc. (Phillips et al., 2004; Philippe et al., 2005; Rodriguez-Ezpeleta et al., 2007; Philippe et al., 2011; Sharma et al., 2014; Prasanna et al., 2020). However, issues with choosing among competing topologies have plagued systematists long before the phylogenomics era. Hennig (1950, 1966) promoted the use of 'wechselseitige Erhellung' or reciprocal illumination, to re-evaluate homology assessments using all available sources (morphological, ecological, biogeographical) in order to understand and resolve evolutionary relationships (Mooi & Gill, 2016). With reciprocal illumination, investigators use the results of current analyses as evidence to correct errors in prior conclusions and to inform about potentially spurious phylogenetic relationships or dubious homology assessments. Thus, reciprocal illumination serves as a philosophical test of the explanatory power of a hypothesis (in this case a tree topology) in relation to broader evolutionary theory, making it testable and thus preferable to empirical observations (e.g. re-analysis of datasets under different assumptions), as the explanatory power of such empirical tests are limited by the narrower circumstances under which the results were extracted (Grant, 2002). CEA relied almost exclusively on this latter option, reanalysing datasets under different evolutionary models and trimming settings, thus limiting the broader explanatory power of their results to a single model under a particular trimming regime. Furthermore, the trimming regimes implemented by CEA are recommended for the analysis of closely related species (Vasilikopoulos et al., 2020), not higher-level taxa like families, whose ancestors likely diverged hundreds of millions of years ago (Hawlitschek et al., 2012). Thus, CEA's results were obtained under biologically unrealistic settings (Vasilikopoulos et al., 2020), further limiting their explanatory power to an unrealistic dimension. The kinds of evidence utilized by reciprocal illumination, on the other hand, like complex traits and features, can be used to defensibly choose among competing hypotheses of homology and tree topologies (Grant & Kluge, 2004; Mooi, 2016). This last aspect of reciprocal illumination is increasingly relevant in the phylogenomic era when competing trees are often maximally supported, with differing topologies offering conflicting hypotheses about the evolution of a particular group. 'The sister-group relationship between Hygrobiidae and Dytiscidae was proposed by Burmeister (1976) based on morphology of the ovipositor and by Ruhnau (1986) based on larval morphology. Both adult Dytiscidae and Hygrobiidae also share the presence of prothoracic glands, among other characters (Forsyth, 1970; Beutel, 1986, 1988).' (Cai et al., 2020: 5). We revisited the literature CEA cited among others, in order to re-examine the morphology in light of the three different topologies for Dytiscoidea (Fig. 1A, B, C) for the purposes of reciprocal illumination. Forsyth, in a series of papers (Forsyth, 1968; Forsyth, 1970; Forsyth, 1972), documented the anatomical structure of the defensive glands, both gross and cellular, across all currently recognized adephagan families, with the exception of the more recently discovered Aspidytidae (Ribera et al., 2002) and Meruidae (Spangler & Steiner, 2005). Adephagans have two general types of paired defence glands: those located towards the apex of the abdomen, the pygidial defence glands, and those situated within the prothorax, the prothoracic defence glands. Pygidial defence glands occur in all adephagan beetles (Forsyth, 1968, 1970, 1972). Prothoracic glands are only known to occur in the families Hygrobiidae and Dytiscidae (Forsyth, 1968, 1970, 1972; Beutel et al., 2006; Dettner, 2019). In general, prothoracic exocrine glands are present in various beetle families (e.g. Chrysomelidae, Erotylidae, Histeridae, Pyrochroidae, Staphylinidae) (Dettner, 1987), however, complex prothoracic glands like those found in Dytiscidae and Hygrobiidae (Forsyth, 1968, 1970), are rare and known to have evolved outside these two families only in Tenebrionidae (e.g. Tribolium Macleay, Diaperis Geoffroy, Zophobas Dejean) (Roth, 1943; Sokoloff, 1975; Tschinkel, 1975). However, within Tenebrionidae prothoracic glands have potentially evolved multiple times (Tschinkel, 1975). '… Hygrobiidae and Dytiscidae are unique within Caraboidea in that both have thoracic defense glands. These have probably been evolved independently in the two groups.' (Forsyth, 1970: 68). Subsequent researchers have since agreed that homology of these glands between the two taxa seems questionable (Dettner, 1985, 1987, 2014, 2019) and have even considered them nonhomologous (Miller, 2001) and the likely result of convergent evolution (Kavanaugh, 1986; Lawrence et al., 2011). Others have contested that the glands are homologous, such as Burmeister (1976) who thought it unlikely such a 'differentiated' organ could result from convergent evolution, and most notably, Beutel (1986) who similarly thought convergent evolution of this gland was unlikely and cited the glands' similar sieve plates as evidence of common ancestry. Although it is true that both prothoracic glands in Hygrobiidae and Dytiscidae do have the secretory cell duct openings clustered together into groupings called sieve plates, the arrangement of these sieve plates is different. In Hygrobia Latreille they are positioned primarily along the lateral margin of the gland reservoir where they are unobstructed by the muscles dorsally covering the gland (Forsyth, 1970). In Dytiscidae the sieve plates are distributed randomly over the basal half of the reservoir only, additionally in between each sieve plate are many inwardly directed spine-like invaginations that are not found in the prothoracic glands of Hygrobia (Forsyth, 1968, 1970). Thus, the homology of the prothoracic sieve plates also seems questionable given the positional and structural differences (i.e. having spine-like invaginations separating them in Dytiscidae). Subsequent cladistic analyses coding the prothoracic defence gland as a simple binary present-or-absent character have recovered it either as the only unambiguous synapomorphy uniting Hygrobiidae + Dytiscidae (Beutel & Haas, 1996), or in combination with the larval trochanteral annulus, position of the larval cerebrum (Beutel et al., 2006; Beutel et al., 2020) and most recently with the elongate larval antennomere 1 (Beutel et al., 2020). However, Baehr's (1979) detailed cladistic study of the prothoracic musculature of Adephaga and phylogenetic analyses utilizing molecular data from multiple genes other than the primary use of mitochondrial genes (Table 1), have failed to provide support for the synapomorphy of the prothoracic gland (de Pinna, 1991), instead corroborating the likely convergent evolutionary origins of these structures. '[The pygidial glands of Trachypachidae] show greater similarity to [Carabidae] than do those of Cicindelidae, (Forsyth, 1970) which Crowson prefers to include as a tribe of Carabidae.' (Forsyth, 1972: 267). Therefore, a pygidial gland with a secretory lobe composed of acini is likely a potential synapomorphy uniting Trachypachidae and Carabidae, possibly to the exclusion of Cicindelinae. This is consistent with the phylogenetic relationships recovered by Gustafson et al. (2020) and could potentially provide further morphological support for Cicindelinae as a family distinct from Carabidae (as hinted at by Forsyth in the quote above), pending study of the secretory lobe of Platychilini, which was recently recovered as sister to all other cicindelines (Gough et al., 2020). In Haliplidae, the sister group to Dytiscoidea, most genera have the plesiomorphic simple-elongate-tube-type of secretory lobe (Fig. 3A), for example Haliplus Latreille (Forsyth, 1968; Dettner & Böhner, 200) and Brychius Thomson (Dettner & Böhner, 2009). Peltodytes Régimbart on the other hand appears to have the acini-type secretory lobe (Fig. 3B), supporting its placement within Haliplidae as sister to the reamining genera (Dettner & Böhner, 2009; Gustafson et al., 2020). Within Dytiscoidea, Noteridae + Meruidae are consistently supported as being the sister lineage to the remaining families (Fig. 1, Table 1). The pygidial glands of Meruidae remain undescribed, but within Noteridae, Noterus Clairville, is known to the have the plesiomorphic simple-elongate-tube-type of secretory lobe (Fig. 3A) (Forsyth, 1968; Dettner, 1985). In nearly all Dytiscidae studied, they similarly possess the plesiomorphic simple-elongate-tube-type of secretory lobe (Fig. 3A). Within the subfamily Hydroporinae which was recovered in a clade with Hydrodytinae as being reciprocally monophyletic to the other subfamilies, Nebrioporus Régimbart (Dettner, 2014), Stictotarsus Zimmermann (Forsyth, 1968) and Hyphydrus Illiger (Forsyth, 1968; Dettner, 1985) exhibit this type of secretory lobe. In Laccophilinae (e.g. Laccophilus Leach (Fig. 3A) (Forsyth, 1968; Dettner, 1985) and Copelatinae (e.g. Copelatus Erichson [Dettner, 1985]) which are representative of the clades that are sequential sister lineages to the Agabinae, Colymbetinae, Dytiscinae and Cybistrinae, the same type of secretory lobe is found. In Agabinae [e.g. Ilybius Erichson (Forsyth, 1968)] and Cybistrinae (e.g. Cybister) an unmodified secretory lobe similar to the other dytiscid subfamilies is also encountered. Colymbetinae (e.g. Colymbetes Clairville) and Dytiscinae (e.g. Dytiscus) (Dettner, 1985; Dettner, 2014; Dettner, 2019) are among the most derived subfamilies (Gustafson et al., 2020) and show a slight modification to the secretory lobe, not seen elsewhere in Dytiscidae. The secretory lobe still consists primarily of a single elongate tube-like structure, however, it is branched apically (Dettner, 1985; Dettner, 2014). These branched secretory lobes could potentially be associated with increased pygidial gland secretion in Colymbetes (Dettner, 2019) and Dytiscus (2014). 'This situation [of bifurcate pygidial glands] can possibly be regarded as a derived basic plan characteristic of the Amphizoidae + Hygrobiidae + Dytiscidae, whereby the unbranched pygidial gland of most Dytiscidae is interpreted as secondary.' (Beutel, 1986: 47). However, given that an unbranched secretory lobe is found in Noteridae, the sister group to all other Dytiscoidea except Meruidae (whose secretory lobe form is undescribed), Haliplidae, the sister group to Dytiscoidea, Cicindelinae (recently recovered as sister to Carabidae + Trachypachidae (Gustafson et al., 2020) and Gyrinidae (including Spanglerogyrus) the sister group to all other adephagans, it seems most appropriate to regard this state as plesiomorphic, rather than secondarily derived. Furthermore, even if we were to assume the bifurcate structure of the secretory lobe was lost in Dytiscidae, one would not necessary expect an associated change in the microstructure; so if the single elongate tube of Dytiscidae were in fact secondarily derived, we would expect some evidence of the ancestral sieve plates to be present, which is not the case. The chemistry of pygidial gland secretions has also been demonstrated to serve as excellent phylogenetic markers in chemically defended beetles, for example the low molecular alkaloids in Stenus Latreille staphylinid beetles (Betz et al., 2018). In adephagan beetles, pygidial gland chemistry has frequently been used in the past to infer phylogenetic relationships or explore evolutionary trends (Dettner, 1985, 1987, 1990; Dettner & Böhner, 2009). Dettner (1990) quantified the average number of steps in the biosynthetic pathways necessary for synthesizing the main constituents of pygidial gland secretions in dytiscoid beetles. This was done as a way to identify which components should be regarded as derived characters, as indicated through comparatively more biogenetic steps necessary for their synthesis, allowing evolutionary relationships among dytiscoids to be inferred through presence of shared derived features. In general, Dettner (1990) identified Amphizoida, Dytiscidae and Hygrobiidae, as having shared derived biosynthetic pathways. The pygidial gland secretions of Hygrobiidae have traces of benzoic acid and 4-hydroxybenzaldehyde, that in Dytiscidae make up some of the main gland constituents (Dettner, 1985, 1987, 2019). Hygrobiidae also share biosynthetic precursors in common with Amphizoidae, such as the amino acid methionine, which in the former is used to produce a-hydroxy acids and lactides, and dimethyl disulphide in the latter (Dettner, 1990). However, Amphizoidae and Dytiscidae, particularly Hydroporinae, were found to share the most derived pygidial gland chemical components (Dettner, 1990). Notably, both Dytiscidae and Amphizoidae are able to produce the compound marginalin, which was found to be the most derived feature in terms of number of biogenetic steps necessary for synthesis (Dettner, 1990). Marginalin gives the pygidial gland secretions a yellow colour (Dettner, 2014, 2019). However, marginalin is only known from several members of Agabinae (e.g. Agabus labiatus (Brahm), A. undulatus (Schrank) and A. serricornis (Paykull) (Dettner, 1985, 2014, 2019) and Dytiscinae (e.g. Dytiscus marginalis L. (Dettner, 2014, 2019), which are both relatively derived members of Dytiscidae (Gustafson et al., 2020) and even within these genera, some congeners are not known to produce marginalin (Dettner, 2014, 2019). The pygidial gland secretions of Aspidytidae remain unknown, as do those of Meruidae. In general, the biosynthetic pathways and number of biogenetic steps necessary to produce the chemical constituents of the pygidial gland secretions seem to most strongly support the evolutionary relationships in Fig. 1A rather than those promoted by CEA (Fig. 1C). For larval morphology supporting a sister group relationship of Hygrobiidae + Dytiscidae, CEA cite Ruhnau (1986) who proposed at least four different features uniting these two taxa. Two of these (characters 36: tarsal claw with spinulae and character 37: presence of a vertical line behind eyes) are either homoplasious across a broad sampling of taxa [spinulae occur on the tarsal claws of numerous other adephagan larvae (Nilsson, 1988) or of questionable homology and have not been treated as viable synapomorphies since that publication. Indeed, Ruhnau (1986) questioned the homology of the vertical line behind the eyes as indicated by introducing character 37 with a '?'. The two characters that have persisted as valid synapomorphies from Ruhnau's (1986) work on larvae are the presence of a trochanteral annulus and an elongate larval antennomere 1. The trochanteral annulus is a line occurring on the trochanter of the larva in all Hygrobia species (Alarie et al., 2004; Michat et al., 2014a) and all Dytiscidae, but which is potentially absent in all other adephagan larvae (Meinert, 1901; Bertrand, 1972; Bousquet & Goulet, 1984; Nilsson, 1988; Alarie et al., 2011; Michat et al., 2017). The exact function of the annulus, whether a sulcus allowing increased flexibility, or an internal ridge providing support, is also currently unclear and its presence may affect the position of muscles originating from within the trochanter (Verhoeff, 1903; Ruhnau, 1986). A similar feature (i.e. trochanteral annulus) is found in larvae of the trichopteran genus Limnephilus Leach and has been suggested as providing elasticity to the trochanter (Tindall, 1963). The larva of the hydrophiloid Spercheus Kugelann also appears to have a membranous division of the trochanter (Fikáček, 2019: shown in fig. 19.3K) possibly increasing elasticity as well. Alarie et al. (2011) proposed the trochanteral annulus of Hygrobiidae and Dytiscidae functions similarly, increasing flexibility for the purpose of improving swimming capability in combination with of the Michat et al. the of on the as the likely plesiomorphic state for dytiscid with numerous occurring in several dytiscid Therefore, the presence of similar in larval Hygrobiidae is also possibly a result of convergent evolution, rather than shared with Dytiscidae et al., 2017). the trochanteral annulus is related to swimming as suggested by Alarie et al. this like the could also be homoplasious in Dytiscidae and Hygrobiidae, rather than appears to be for the convergent evolution of a trochanteral annulus outside that of Limnephilus (Verhoeff, and Spercheus (Fikáček, 2019). Ruhnau a certain Haliplus show like a line of in the posterior of their recently, Michat et al. (2020) a detailed of the larvae of two species of where they also recognized the trochanter was by an investigation into the trochanter of larval Haliplus species is to if this lineage also of a trochanteral of antennomere not than The larval antennomere 1 is strongly in Hygrobiidae and Dytiscidae but not in other the other dytiscoid families (Beutel et al., 2020: Although it is true that most have an antennomere 1 that is than this character state within Dytiscidae (Alarie et al., 2011). For an antennomere 1 that is not than in larvae of Copelatinae Copelatus & Agabinae (Alarie et al., Alarie et al., Laccophilinae & Michat & and Hydroporinae and & (Alarie & 2007; Michat et al., amongst Furthermore, even some Hygrobia have an elongate antennomere 1, like et al., like do not have antennomere 1 particularly than Alarie et al., Thus, in this feature is present within both Hygrobiidae and Dytiscidae. This character not been in morphological analyses (Beutel & Haas, Beutel et al., 2006; et al., 2011; Beutel et al., potentially due to such The position of the larval cerebrum in the of the was proposed as a synapomorphy for Hygrobiidae + Dytiscidae by Alarie et al. and recovered as a synapomorphy in cladistic analyses (Beutel et al., 2006; et al., 2011; Beutel et al., 2020). However, this character from both homology and character With to the position of the cerebrum could be a result of modification of other features common For larvae of Hygrobiidae have numerous morphological that associated with their on and larvae Alarie et al., 2004; Michat et al., Among these is a (Alarie et al., fig. likely used to in 1972; Alarie et al., Thus, of the could have in an of the cerebrum of not only these two characters, but the position of the cerebrum in Hygrobia and to be a result of Although it was suggested all larval have a cerebrum situated (Beutel et al., of and in which the of members of Hydroporinae, the cerebrum not situated but instead This could be a shared with other it is that the cerebrum of Hyphydrus fig. and other is not in a different position to that of other dytiscid but appears due to of the of the into the structure for et al., & 2018). Furthermore, this could also be a result of the posterior of the due to a for of the given an in certain is known to result in in other anatomical features & & those located the These homology of the position of the larval cerebrum both within Dytiscidae, and among and a character coding Alarie et al. the larval cerebrum of is also if fig. the larva of from Beutel to and in (1979) and the position of the cerebrum appears the position of the cerebrum of larval Beutel, Balke & shown in Fig. of Balke et al. is similar to that of the larvae as well. Thus, if this character is to to be used as a binary position of posterior of of as in Beutel et al. and Beutel, Balke & should also be as 1 along with Dytiscidae and However, given the issues with the homology of this character and thus on type of homology it may be appropriate to this character in morphological the of Hygrobia and Dytiscidae has no derived features in the of the ovipositor which do not also the of this gland is considered a synapomorphy for Hygrobia and Dytiscidae and in the this feature is Burmeister numerous synapomorphies associated with the but all of these only Hygrobiidae, Amphizoidae and Dytiscidae. Burmeister two characters uniting Hygrobiidae and Dytiscidae, these features in and to such as the for of and and and close together in position Thus, Burmeister relied the prothoracic glands and the morphology of the proposed by Ruhnau (1986) for uniting Hygrobiidae and Dytiscidae as a of the of these three taxa and similarly found synapomorphies uniting all three families but supporting Hygrobiidae + Dytiscidae on of the morphological features cited by CEA support a sister relationship between Hygrobiidae and Dytiscidae. the the most complex and unique morphological feature found in these two the prothoracic glands, strong evidence for the convergent evolutionary origins of these structures. the pygidial glands that are but have been present a potential synapomorphy uniting Hygrobiidae, Amphizoidae, and likely Aspidytidae as well. In & 2019) we the characters based on the literature cited and the the exception of the larval antennomere 1 form The of character coding can be found in Fig. as as in the for coding certain taxa. we these characters the dytiscoid clade from the preferred trees of Vasilikopoulos et al. (2019) Gustafson et al. (2020) the dytiscid taxa to the other and CEA were under three different (i) unambiguous characters with unambiguous character state (ii) and 1970; & each topology The CEA topology one synapomorphies uniting Hygrobiidae + the larval trochanteral annulus (Fig. in the CEA topology Dytiscidae is as having secondarily lost the bifurcate secretory lobe, an unlikely of both and anatomical as in the CEA topology (Fig. the bifurcate secretory lobe is convergent between Hygrobiidae and Amphizoidae, a similarly unlikely The topology of Vasilikopoulos et al. (2019) under as is particularly in line with evidence of evolutionary relationships as inferred through chemistry of the pygidial gland secretions (Fig. Here, the presence of acid is recovered as a synapomorphy uniting Amphizoidae (including Aspidytidae whose secretions remain and Dytiscidae. Although not supporting homology of the prothoracic gland as an synapomorphy under this topology also as the unlikely of bifurcate secretory lobes in Dytiscidae under (Fig. The topology by Gustafson et al. (2020) the bifurcate secretory lobe of the pygidial gland as an unambiguous synapomorphy uniting Hygrobiidae, Amphizoidae and Aspidytidae (Fig. This topology under is particularly (Fig. with the prothoracic glands an independently evolved synapomorphy of Dytiscidae and likely Hygrobiidae as of prothoracic glands outside of with the trochanteral annulus homoplasious in the two with on this topology (Fig. also prothoracic glands as synapomorphies in these two families but instead the trochanteral annulus as a synapomorphy for all dytiscoids except Noteridae and with occurring in Amphizoidae + which is not larvae of these two families primarily over in rather than Alarie & 2005; Michat et al., character to the that the results of CEA are not consistent with morphology-based views of dytiscoid relationships and not more so than the other two topologies (Fig. 1A, it seems that CEA's results are most consistent with that of Beutel et al. (2020) only, this topology has before been recovered by either or morphological analysis (Table 1). on this tree of Dytiscoidea, it be to address and test a series of hypotheses the evolution of many morphological in (Cai et al., 2020: a phylogeny where the prothoracic glands are recovered as homologous, like Fig. to address and test a series of hypotheses the convergent evolution of these morphological in Dytiscoidea. is that in the phylogenetic hypotheses are given through the use of reciprocal illumination. we more attention be to the pygidial defence glands for the morphological evolution of Adephaga as these are both and their homology is not in We are to and two whose and have this are no of The morphological character and results from all ancestral state are in the supporting used in character morphological character implemented for ancestral state with the Vasilikopoulos et al., taxa and the Gustafson et al., and results from all ancestral state using for the three different topologies are available in this topologies are available in this the preferred phylogeny of Vasilikopoulos et al., the preferred phylogeny of Gustafson et al., the preferred phylogeny of Cai et al., The is not for the or of supporting by the than should be directed to the for the

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.