Classifying Cockroaches According to Forewings: Pitfalls and Implications for Fossil Systematics

Abstract

A reliable character system is crucial to taxonomy and systematics, and it promises valid downstream inferences, e.g., estimates of diversity and disparity, reconstruction of evolutionary history, and even stratigraphic correlations. Modern taxonomy and systematics of extant cockroaches requires an integrative study involving multiple lines of evidence with emphasis on genital and reproductive characteristics and molecular data. In contrast, many fossil cockroach taxa published recently are based solely on forewings. Many studies have shown that forewing-based taxa are questionable. In order to find out how much of the phylogenetic signal we could ascertain from venational similarity, and how confident we could be, this study used forewing characters to reconstruct phylogenies of the genera of well-recognized family-group taxa. The intuitively reconstructed phylogeny of 75 extant genera failed to recover those taxa or their relationships. Parsimony analyses of various datasets all yielded strong polyphyly and chaotic relationships. In conclusion, the forewing of cockroaches is not a universally competent character system. The underlying causes are the complicated nature of veins and the limitations of current analytical techniques. The uncertainty in forewing-based taxonomy and systematics has been underestimated in the literature. Forewing-based fossil taxa warrant re-evaluation; some of them are herein deemed nomina dubia in their current state.

1. Introduction

A reliable character system is crucial for a reasonable classification and a robust phylogeny of organisms, which in turn facilitates a valid estimate of diversity and helps to elucidate evolutionary history. Since McKittrick [1] inaugurated the era of modern systematics of cockroaches, the terminalia—especially male genitalia—have become the focus of morphological character systems. This is also the case for other insects [2]. In practice, many characters in addition to terminalia are considered altogether to obtain a better picture of a cockroach. Oth er important characters are the head (configuration of organs, etc.), pronotum (shape, color pattern, etc.), legs (segment ratio, spination, etc.), wings (venation, folding, etc.), abdomen (tergal gland, etc.), and also internal characters such as the proventriculus [3,4,5,6,7,8,9,10,11,12,13,14]. Modern systematics now incorporates DNA and other molecular data alongside multiple lines of evidence for species delimitation and phylogenetic reconstruction.

In contrast, wing venation remains a dominant character system for fossil cockroach classification. A substantial number of fossil taxa were established using solely the forewing, be it complete, incomplete, or fragmentary [15,16,17,18]. This practice appears to be a direct result of taphonomic biases. On the other hand, classifications and phylogenies based solely on wing venation have been found to be misleading (briefly reviewed by [19]: Text S3). First, studies have conclusively shown that intraspecific variations in venation can be considerable, particularly in the forewing [20,21,22]. Second, the higher-level classification and phylogeny proposed in the seminal wing-based study by Rehn [23] are no longer congruent with current, well-supported results derived from integrative morphological and molecular data [1,3,24,25,26] (among others). Given these considerations, the taxonomic identity of many fossil taxa encompassing both species and higher ranks warrants re-evaluation. However, two questions should be addressed prior to a revision: (1) To what extent does the reliance on the forewing introduce uncertainty into taxonomy and systematics? (2) In what cases can we trust a wing-based taxon, or what characters might be useful in taxonomy and systematics?

To evaluate the phylogenetic signal recoverable from forewing venational similarity and the associated confidence level, this study reconstructed phylogenies of the genera of well-recognized family-group taxa using only forewing characters. The results have implications for insect paleontological studies as well as geological studies because forewing fossils have been used for stratigraphic correlations (e.g., [27,28,29,30]).

2. Materials and Methods

2.1. Taxonomy and Sampling

In the present study, ‘cockroach’ refers to the broader sense, which includes ‘roachoid’ fossils. The taxonomic frame covering extant and extinct cockroaches follows Li [31]. The higher classification of extant cockroaches follows recent phylogenomic studies [26,32] with revisions within each superfamily [33,34,35]. The reference phylogeny used in this study is a summary and consensus of those studies.

Although this study centers on fossils, fossils are inadequate for the intended tests. Confidence in the natural relationships of the sampled taxa is paramount; therefore, extant species constitute the majority of samples. The resulting inferences can be extrapolated to fossils. Rigorous illustrations of wing venation (including hindwings) were collected from the literature on living Dictyoptera and fossils in amber, covering 144 extant cockroach species, seven fossil cockroaches, three extant termites, and four extant mantises (Table S1). The sampled genera mainly represent the following six family-group taxa: Blattoidea, Nocticolidae, Corydiidae, Pseudophyllodromiidae, Blattellidae sensu stricto (Blattellinae in Ectobiidae sensu lato), and Blaberidae. These taxa are commonly recognized as monophyletic taxonomic identities in both morphological and molecular studies.

2.2. Phylogenetic Reconstruction

First, 75 extant cockroach genera that have fully developed forewings with comparable venation were selected to intuitively infer the phylogeny based on the similarities in forewing venation. Intuitive analyses are subjective; therefore, the results cannot be directly compared between authors. Nevertheless, this approach is logical in practice, and it is exemplified by male genitalia. Taxonomy based on male genitalia has withstood tests from other evidence, and it is largely shared among taxonomists. Before the phylogeny in this study was inferred, the author was unfamiliar with most of the genera, especially their taxonomic or phylogenetic placements. The phylogeny was reconstructed against a hypothesized ancestor (Figure 1), of which the character states were summarized from a survey of the literature but remain highly conjectural (Appendix A). When interpreting the phylogenies, the term ‘basal’ (as opposed to ‘terminal’) was used [36].

Second, phylogenies were reconstructed through a parsimony approach. In order to interpret the homology with confidence, species were ruled out if the wings are reduced or highly tegminized, too many or too few veins present (≥60 or ≤14), or the venation is too specialized or disordered (e.g., Ectobius, Laxta, Paratropes). Finally, 72 species were selected from the venation library (i.e., Table S1). Together with additional 21 fossil taxa (Table S2), the taxaset comprises 66 species of true cockroaches (Blattaria), 3 extant termites (Isoptera), 2 extant mantises (Mantodea), 20 species of fossil cockroaches outside Blattodea (basal Dictyoptera and Eoblattodea) or uncertain placement, and 2 species of Protoptera (basal Neoptera, see [37]; see also [38], sister to “Dictyoptera” in the broader sense, which is essentially synonymous with Holopandictyoptera). Meanwhile, 14 out of the 20 fossil cockroach species are securely identified to 12 currently recognized families (Tables S1 and S2): nine species are the types of the families (pink-shaded in Table S2), three species are of the type genera of the families and closely resemble the type species (type specimens of the type species are incomplete), and two species outside of type genera (Divocina noci Liang et al., 2012 and Parvifuzia peregrina Wei et al., 2012) were determined to corresponding families according to the integrative diagnostic characters of the families [39,40]. In the analyses of Blattaria and Blattodea, the outgroup is Artitocoblatta asiatica Vishnyakova, 1968, which is a fossil of basal Dictyoptera [41]. In the analysis of Dictyoptera (only Blattodea and Mantodea included; basal lineages not), the outgroup is Karatovoblatta longicaudata Vishnyakova, 1968, a fossil of Eoblattodea [41]. In the analysis of Holopandictyoptera, the outgroup is Protoptera.

Parsimony analyses were performed in TNT 1.5 [42] with the collapsing rule set as “min length = 0” and searching algorithm as ‘Traditional search’. The analyses utilized 36 forewing characters (Table S3), weighted 1, 2, 4, or 8 (n = 2, 10, 23, 1; % = 5.6, 27.8, 63.9, 2.8) according to conjectural relative conservatism (i.e., negative plasticity). Characters are essentially venational, but they also include traits that strongly associate with veins or venational traits in the broader sense, namely, wing shape (length–width ratio and jugal lobe), surface structure (basal suture and diagonal channel), and pigmentation (sclerotization of veins and stigma). Some characters might be useful but were not included, because it is difficult to define the states, e.g., the curvature of a certain vein, and the degree of development of intercalary veins and cross-veins. Table S4 shows the complete data-matrix. The venational terminology follows Li et al. [19].

Third, Mesquite 3.61 [43] was used to trace possible family-level apomorphies on a phylogeny of the above-mentioned 66 true cockroaches with the topology constrained by the reference phylogeny (see above).

3. Results

The intuitively inferred phylogeny of extant genera is shown in Figure 2. Both Rehn’s [23] and this study are unable to completely recover the six well-recognized family-group taxa or clarify their relationships (Figure 2). As for the three superfamilies, Corydioidea was recovered in Rehn [23], not in this study; neither Blattoidea nor Blaberoidea were recovered in either study. Blattidae and Nocticolidae are the only two families recovered in Rehn [23]; however, this is likely a result of limited taxon sampling (five genera in Blattidae and two genera in Nocticolidae). In this study, more specialized genera are included; then, the monophyly of Blattidae and Nocticolidae is no longer supported. Similarly, the strong polyphyly of Blaberoidea is due to a larger sampling in both Rehn’s [23] and this study. Subtaxa of Blaberoidea (i.e., Pseudophyllodromiidae, Blattellidae s. s. and Blaberidae) were also reconstructed as polyphyletic in both studies with a somewhat unexpected result that Pseudophyllodromiidae is almost recovered in this study (only one species was placed elsewhere). In addition, most of the subfamilies of Blaberidae were not recovered. After the reconstruction, the author re-examined the venation illustrations and found 30 characters that have potential value for taxonomy and systematics (Figure 1 and Figure 2, Appendix B).

Parsimony analyses all yielded strong polyphyly and chaotic relationships of the six family-group taxa (Figure 3). In addition, the topology appears to be prone to outgroup selection. The phylogeny of 66 Blattaria cockroaches (65 extant, 1 fossil) is largely unresolved, with Corydiinae being the only major clade recovered (Figure 3 and Figure S1). To investigate whether or not the forewings with somewhat irregular venation affect the phylogeny, an additional analysis was performed with seven species removed (Blaptica dubia, Salganea sp., Phortioeca phoraspoides, Eucorydia dasytoides, Eupolyphaga sinensis, Homoeogamia mexicana, Polyphaga plancyi). The 59-species phylogeny is highly resolved, but all of the six family-group taxa and Corydiinae are strongly polyphyletic (Figure S2). The phylogeny of Blattodea (= Blattaria + Isoptera) closely resembles the phylogeny of 66 Blattaria cockroaches with the Isoptera sister to a blaberid genus (Figure 3 and Figure S3); removing the seven species does not help to resolve the phylogeny (Figure S4). Remarkably, the inclusion/exclusion of the character ‘basal suture’ (autapomorphy of termites) does not affect the topology (Figures S3 and S4) even though this character is the only one weighted 8, which is the highest weight. The phylogeny of Dictyoptera is highly resolved, with Isoptera and Mantodea sister to two blaberid genera, respectively (Figure 3 and Figure S5). The topology is distinct from the above ones: Corydiinae is paraphyletic with respect to the rest of Dictyoptera. The phylogeny of Holopandictyoptera is also highly resolved, but the topology has little compatibility (if any) with the above ones (Figure 3). Six of the fossil species are supposed to belong in basal Dictyoptera, i.e., outside of Blattodea; however, two of them (Elisama extenuata and Huablattula hui) are deeply nested in Blattodea.

Mapping the characters used in parsimony analyses onto the reference phylogeny did not recover any family-level apomorphy.

4. Discussion

4.1. How Much Can We Learn from Forewings?

The forewing character system did not perform well in this study, nor did it perform well in Rehn’s [23]. While improvements in analytical methods and character interpretation are likely necessary, the current findings strongly suggest that forewing-based systems lack the robustness and reliability required for practical use in cockroach systematics. The uncertainty introduced by forewing-based approaches impairs taxonomy and systematics, invalidating downstream inferences such as estimates of diversity and disparity.

On the other hand, unique or diagnostic characters may recover certain widely accepted clades that are supported by integrative evidence. For example, Rehn [23] recovered Corydioidea, but this study does not, because a crucial hindwing character—flat vannus (no fanwise folds)—was considered in Rehn [23] but not herein. Nocticolinae (= Nocticolidae prior to Han et al. [34]) is the only one extant family-group taxon that is virtually defined with the forewing [44,45], of which the venation is reduced to a specialized state. Latindiinae is recovered even when more genera are sampled (this study, cp. Rehn [23]), because the forewing alone is distinguishable from other cockroaches (see also Han et al. [34]). Compared with families and subfamilies, smaller taxa are more likely to be characterized by forewing venation, e.g., a characteristic posterior branch of R is present in a few closely related blattellid genera (Blattella, Episymploce, Hemithyrsocera, Symploce, etc.) [19,23], and a diagonal channel can distinguish Austropolyphaga and Compsodes from other Latindiinae [45]. In general cases, the shape of the forewing, the shape and proportion of the veinal area, and the quantity of terminal branches, are more reliable than other venational characters [19,21]. Occasionally, the color pattern and maculae are, instead of venation, as unique as to validate a taxon, e.g., some species of Allacta, Panchlora, Paranauphoeta and Sundablatta [9,46,47,48]. A robust phylogeny encompassing a sufficient number of genera and species is the prerequisite for testing the suitability of a character or a set of characters for indicating a taxon; however, this is beyond the scope of this study.

While the forewing likely harbors valuable phylogenetic signals, current methodologies may not be sophisticated enough to fully capture it. The intuitive reconstruction is subject to the understanding and experience of researchers, and the parsimony analyses are compromised by unverified character correlations and polarities. Additionally, the interpretation (quantification) of characters is unlikely as objective as expected. To unlock the full potential of wing venation, future studies would likely benefit from robust evolutionary models, which could be parameterized and also could utilize artificial intelligence techniques. Of course, even the most advanced methods rely on a foundation of meticulous documentation of venation and a sound interpretation of characters. However, even the most sophisticated methods cannot guarantee success (see below).

4.2. Why Is Forewing Not a Good Character System?

Owing to the formidable nature of the forewing—a carrier of an enormous amount of divergence and convergence, the complicated evolutionary history of venation is hard to trace. The forewing is a tangled web, woven with plesiomorphies and homoplasies, and decorated with few apomorphies; these characters are per se too variable (see also the discussions in Evangelista et al. [49]). This study highlights two key challenges in utilizing forewing venation for phylogenetic reconstruction: (1) discriminating phylogenetic signal from noise, and (2) the expectable depletion of characters following a rigorous screening for dubious characters. This depletion could result in an insufficient number of characters for robust phylogenetic inference.

Forewings are sometimes very conservative (plesiomorphies) or easily becoming similar (parallel evolution, convergence or overlap of variation). An obvious example of plesiomorphy of Blattodea is the claval veins that are parallel to the claval furrow or CuP and that reach the hind margin of the forewing. This trait is frequently found in cockroach fossils, especially those prior to Cretaceous (e.g., many distantly related species from Jurassic described by Vishnyakova [41]). In comparison, only a minority of extant taxa retain this trait, e.g., Ctenoneura and Latindiinae [19,45,50,51]. Convergence can be found between superfamilies. Salganea of Blaberidae, Blaberoidea and Catara of Blattidae, Blattoidea are similar in the unique modification of R veins that is absent from other cockroaches. Their R veins consist of penniform branches and a strongly developed pseudostem that points the costal margin immediately prior to the apical angle [19]. Such resemblance might be the result of similar habitat and habit, namely, feeding on and living in rotten woods in tropical forests. The forewings of some genera closely resemble each other and can be readily interpreted as closely related as siblings, but they actually belong to two families, even distinct families. For instance, Blattella of Blattellidae s. s. and Euphyllodromia of Pseudophyllodromiidae are similar in simple ScP, pectinate proximal R, multi-branched distal R, longitudinal parallel M and CuA, more developed M than CuA, and diagonal to longitudinal claval veins [4,19,23]; these genera were recovered as siblings in parsimony analyses (Figure 3 and Figures S1–S5), but their most recent common ancestor dates back to about 130–170 Mya [52]. Ctenoneura of Nocticolidae and some species of Pseudophyllodromiidae are similar in simple ScP, nearly pectinate proximal R, multi-branched distal R, more or less dichotomous M and CuA, and M being dominant in mediocubital veins [19,23]; these genera are recovered as a clade in the intuitively reconstructed phylogeny (Figure 2) and also some of the parsimony phylogenies (Figure 3 and Figure S5). Ocelloblattula and Huablattula are fossil genera found from amber [53,54]; they might be regarded as synonyms if we only know their forewings. Thanks to other characters (female terminalia in particular) well preserved in the amber, we readily know that these genera are not closely related. Accordingly, fossil taxa established on similarities in forewing venation can be polyphyletic, but they have little chance of being tested.

Forewings are sometimes too variable. Considerable intraspecific variation has been reported [20,21,22]; such degree of variation has been used to distinguish between genera in fossils (e.g., Nehevblattella Vršanský, 2004 [55] and Irreblatta Vršanský, 2008 [56]; Elisama Giebel, 1856 sensu Vršanský, 2004 [57], Svabula Vršanský, 2005 [58] and Vrtula Vršanský, 2008 [59]). A dedicated study on intrageneric variation in extant cockroaches is unavailable in the current literature; nonetheless, evidence from generic revisions (e.g., [50,60,61]) implies that some fossil genera or families might be the result of excessive split (e.g., genera mentioned above; Compsoblattidae Schneider, 1978 [62] and Phyloblattidae Schneider, 1983 [63], see the type specimen of Compsoblatta in Schneider [62] and that of Phyloblatta in Handlirsch [64]). Disparity of a subfamily or a family in living cockroaches can further undermine our confidence about classifying cockroaches using forewings alone. Blaberidae is among the good examples: it can be imagined that Panchlora, Paranauphoeta and Salganea would be regarded as representatives of three far distinct families or even superfamilies if they were known for forewing fossils only (see their wide differences in Li et al. [19]). The disordering of veins might contribute to the variability of venation in larger-sized cockroaches such as Blaberidae. In the forewing of many species of Blaberidae and Corydiidae, merging between adjacent veins, sudden turns, and free veins is frequent, and the distribution of veins is more or less irregular and uneven [19,23]. In their smaller-sized relatives (e.g., Blattellidae s. s., Pseudophyllodromiidae and Nocticolidae), the forewing venation tends to be more regular, conservative, and readily comparable with fossils. It is therefore hypothesized that an increase in size (hence an expansion of forewings) among certain cockroach lineages coincided with a disordering of veins; this disordering might be caused by the insertion of additional veins that disarrange the pre-existing veins.

4.3. Implications for Paleontological Studies

While forewing venation holds limited significance for extant cockroach systematics, it has traditionally played a central role in fossil cockroach classification. Consequently, misinterpretations of wing venation have minimal impact on studies of extant cockroaches, but they can have profound consequences for paleontological disciplines reliant on these characters. Logically, one might expect a more rigorous approach to interpreting venation in fossils compared to extant specimens. However, the current practice often falls short of this expectation. Interestingly, some studies (e.g., [65,66]) have reported vein variability of fossil species, whereas it is the extent of variation that was used to delimit those species.

When forewings are used as index fossils for stratigraphic correlations, the uncertainty in taxonomy becomes the uncertainty in time. The correlation requires a genus or even a species to minimize errors. Imagine that we have many forewings that are similar enough to include in one genus or species (from an average perspective). If we are lucky that the observed variation is limited within serval closely related species, then the correlation would be fine. If the observed variation is shared between sibling genera, e.g., Anaplectoidea and Malaccina (Anaplectoideini, see [19,67]), then the correlation could be accepted. If the observed variation is commonplace across many genera, e.g., Allacta, Balta, Imblattella and Latiblattella of Pseudophyllodromiidae [19,23,68,69], then the uncertainty can be up to 120–150 Ma (the age of Pseudophyllodromiidae as estimated in Li [52]). In extreme cases, the uncertainty can reach 130–170 Ma (see above).

Taxonomy and systematics of extant cockroaches have become firmly grounded in an integrative approach with emphasis on genital and reproductive characteristics and, more recently, molecular data. Remarkably, the ovipositor and male terminalia, bearing essential genital and reproductive characters, have long been considered as the basis of fossil cockroach taxonomy [31,70,71,72]. In spite of the prevalence of forewing-based taxonomy in cockroach fossils, a few families are good examples of integrative taxonomy: Raphidiomimidae and Fuziidae are defined with unique characters based on relatively complete fossils, i.e., the prognathous head of Raphidiomimidae [73] and the pincer-like male cerci of Fuziidae [74]. In view of potential variations, Raphidiomimidae is indistinguishable from Liberiblattinidae in the forewing venation; likewise, Fuziidae is indistinguishable from Blattulidae. More good examples are found in descriptions of genera and species, among which are Shelford [75], Vishnyakova [41], Gorokhov [76], Anisyutkin [77], Anisyutkin and Gorochov [53], Anisyutkin and Gröhn [78], Gao et al. [79], Li and Huang [45,80,81], and Qiu et al. [54].

The less information is known, the less confidence there is in taxonomy. Forewing-based taxa should be considered nomina dubia unless the forewing bears autapomorphies or a distinctive set of characteristics. Undefined forewing-based taxa become valid only when their identity is established by corroborative evidence from well-preserved topotypes, new findings from the types, etc. First things to consider are the families that are virtually established on a single incomplete forewing (e.g., Archimylacridae, Caloblattinidae, Mancusoblattidae, Phyloblattidae, Spiloblattinidae, and Subioblattidae), which are followed by the families based on a nearly complete but indistinctive forewing (e.g., Argentinoblattidae, Blattulidae, Compsoblattidae, Delpuenteblattidae, Mesoblattinidae, Mutoviidae, Mylacridae and Necymylacridae). Of these families, Blattulidae, Caloblattinidae and Mesoblattinidae are common in the literature with extended taxonomic concepts. The name Mesoblattinidae is already deemed detrimental to taxonomy [82]; the following are comments on the remaining two families.

Blattulidae. The type genus of this family, Blattula Handlirsch, 1906, was established on Blattina (Mesoblattina) dobbertinensis Geinitz, 1884, of which the type specimen is a forewing [64,83]. Vishnyakova proposed Blattulidae, which had diagnostic characters from the genitalia of females and males [84]. The genital characters are from non-type species, namely, Blattula brevicaudata Vishnyakova, 1968 and Blattula rectinervosa Vishnyakova, 1971 [41,72]. However, these non-type species (from the Late Jurassic of Karatau, Kazakhstan) and the type species (from the Early Jurassic of Germany) were bridged only by the equivocal resemblance in the indistinctive forewing venation; their supposed affinity has not been justified. Recently, Vršanský and Ansorge revised the genus Blattula [66]. They synonymized the type species and many other species with Blattula langfeldti (Geinitz, 1880), and they complemented the type species description using topotypes and specimens from nearby localities [66], but the information of terminalia is still missing. In summary, the current concept of Blattulidae is largely from the species that are dubiously included in Blattula, while the characteristics found from the type species are inadequate for establishing a family.

Caloblattinidae. Vršanský established this family on Caloblattina Handlirsch, 1906 (assigned to “Geinitz, 1883” by Vršanský) [65]. The type species of Caloblattina, Blattina mathildae Geinitz, 1883, is based on an incomplete forewing, of which the original figure is upside down [85] (plate 6, figure 1). Vršanský and Ansorge completed the forewing venation of the type species with the help of topotypes and specimens from nearby localities [66]. The complete forewing exhibits a generalized venation pattern of common cockroaches from the late Mesozoic to the present. Although Vršanský originally claimed that Caloblattinidae consists of about 50 genera [65], Vršanský and Ansorge listed only 15 genera in addition to Caloblattina [66]. To date, none of the 15 genera has been justified to be closely related to Caloblattina. Therefore, Caloblattinidae seems to be merely another taxonomy trash bin.

5. Conclusions

The forewing is not a universally competent character system for cockroaches. The underlying causes are the complicated nature of veins and the limitations of current analytical techniques. The taxonomic value of forewings should be interpreted with caution, and the interpretations in the literature should be adopted with caution as well. The uncertainty in taxonomy and systematics led by the forewing character system has long been underestimated and likely introduces errors into downstream inferences, e.g., estimates of morphological disparity, character variability and taxon diversity, reconstruction of evolutionary history, and stratigraphic correlations. An assessment of the forewing characters used is the prerequisite for validating these inferences. Forewing-based taxa warrant re-evaluation, especially common families in the literature; e.g., Blattulidae, Caloblattinidae and Mesoblattinidae are considered nomina dubia in their current state.

An ideal taxonomic system should encompass both extant and extinct species, and an ideal, integrative character system should incorporate multiple lines of evidence. However, the character systems of extant cockroaches and fossils still need to be compatible, and the forewing-based system in fossil taxonomy falls short of contemporary standards. How similar are the forewings enough to be included in one genus, and how different are the forewings enough to be placed in different families? While neontologic studies no longer struggle with these questions, these questions persist as significant challenges in paleontology.