Shedding light on the enigmatic extinct insect order Glosselytrodea: new diversity and key morphological insights into non‐tegmenized Permian forms

Glosselytrodeans are enigmatic extinct (Permian–Jurassic) insects classified in their own order. The monophyly of Glosselytrodea is not universally accepted and its large‐scale relationships remain problematic, namely because its diversity is largely known from hardened (i.e. tegmenized/‘elytrized’) forewings of allegedly derived forms. Thus, new data on the group’s early diversity and/or complete specimens are critical. Herein we describe the oldest glosselytrodean known, Moscheloptera phantasma gen. et sp. nov. (Permoberothidae), based on a forewing from the Niedermoschel black shale (Germany; early Permian). The holotype of Permoberotha villosa Tillyard (Permoberothidae) (a forewing) and a previously described complete conspecific specimen, both from Elmo (USA; early Permian), are reassessed and photographically figured for the first time. A new, complete P. villosa specimen from Elmo preserves body structures hitherto not reported among glosselytrodeans, including complex pretarsal claws and detailed male genitalia. The holotypes of two further species, Sylvaelytron latipennatum Novokshonov (herein transferred to Permoberothidae) and Archoglossopterum shoricum Martynova (Archoglossopteridae), are also reassessed and figured. An emended diagnosis and an updated catalogue of the Glosselytrodea are provided. Glossopterum Sharov is returned to its own family: Glossopteridae Sharov. Our findings substantially expand the knowledge of the Permoberothidae, which is the earliest glosselytrodean lineage and lacks sclerotization‐related forewing specializations. Although the exact phylogenetic position of glosselytrodeans is still unresolved, the female and male terminalia described herein almost certainly support them as early holometabolan insects, possibly closer to the mecopterid rather than the neuropteroid lineage. Shedding light on Glosselytrodea improves our understanding of the early evolution of the Holometabola.

Sylvaelytron latipennatum Novokshonov (herein transferred to Permoberothidae) and Archoglossopterum shoricum Martynova (Archoglossopteridae), are also reassessed and figured. An emended diagnosis and an updated catalogue of the Glosselytrodea are provided. Glossopterum Sharov is returned to its own family: Glossopteridae Sharov. Our findings substantially expand the knowledge of the Permoberothidae, which is the earliest glosselytrodean lineage and lacks sclerotization-related forewing specializations. Although the exact phylogenetic position of glosselytrodeans is still unresolved, the female and male terminalia described herein almost certainly support them as early holometabolan insects, possibly closer to the mecopterid rather than the neuropteroid lineage. Shedding light on Glosselytrodea improves our understanding of the early evolution of the Holometabola. exclusively known from the Artinskian, lower Permian (Rasnitsyn & Aristov 2013). This taxonomic arrangement was deduced from a hypothesized process of forewing sclerotization that glosselytrodeans seemingly went through (Rasnitsyn & Aristov 2013). Recently, the family Polycytellidae Martynova, 1952 was restored based on a single diagnostic character, the area anterior to RA and opposite the end of ScP large (accounting for approximately one-quarter of the wing width); the group as currently understood is restricted to Gondwanan forms (B ethoux & Anderson 2021). The phylogenetic position of glosselytrodeans has always been and remains controversial. The group has been previously regarded as orthopteroid (Martynov 1938;B ethoux et al. 2001), belonging to the clade Eumetabola [= Acercaria + Holometabola] (B ethoux et al. 2007) or as holometabolans, with the current trend leaning towards deeming them as neuropterans (Tillyard 1932;Carpenter 1992) or, at least, neuropterids (Sharov 1966;Rasnitsyn & Aristov 2013). Glosselytrodea as a group has even been regarded as unnatural, with the Permoberothidae considered to be either stem-group or crown-group Neuropterida (as stem-group Eidoneuroptera), and the rest of the families, that is, the 'Glosselytrodea s.s.', viewed as an assemblage of contentious phylogenetic position, perhaps orthopteroid (Grimaldi & Engel 2005;Engel et al. 2018). The lack of data on the structures from their extreme wing base and the body structure, particularly the genitalia, has contributed to their disputed position.
In this work we describe the oldest glosselytrodean currently known and assign it to the family Permoberothidae. We also reassess the type and additional specimens from other species either previously or herein classified in Permoberothidae, as well as holotypes from other Glosselytrodea, figuring them as photographs for the first time. Overall, the data herein presented significantly advance the knowledge on the earliest lineage known of the Glosselytrodea, and provide key morphological information for reassessing matters such as the integrity of the group or their wide phylogenetic relationships.

GEOLOGICAL SETTING
The genus Permoberotha Tillyard, 1932, with its type and only species P. villosa Tillyard, 1932, is the type genus of the family Permoberothidae. The holotype of this species, as well as the two additional conspecific specimens figured and (re)described herein, are from the locality of Elmo, c. 5.5 km south-east from the unincorporated town of Elmo, Banner Township, Dickinson County, Kansas, USA (Dunbar 1924;Beckemeyer 2007). This locality, early Permian in age (Artinskian), belongs to the Wellington Formation (= Fm.), which ranges from north-central Kansas to north-central Oklahoma (Dunbar 1924;Beckemeyer & Hall 2007). The site represents one of the most prolific localities for Permian insects worldwide (Grimaldi & Engel 2005). The fossil insects are preserved in a light grey argillaceous limestone layer overlying a black carbonaceous clay shale level with in situ tree stumps and abundant plant remains (Dunbar 1924). The palaeoentomological specimens are preserved as impressions or, more rarely, compressions with a certain relief. The latter can show a high quality of morphological detail (RPF, pers. obs.), as shown by the new P. villosa specimen figured herein. The Elmo deposit has been reconstructed as corresponding to freshwater bodies set in a coastal environment (Dunbar 1924;Beckemeyer & Hall 2007).
The new glosselytrodean species described herein, Moscheloptera phantasma, is preserved in blue-grey, slightly contact-metamorphic, silty claystone from a roadcut east of Niedermoschel/Rhineland-Palatinate, the type locality of the Niedermoschel black shale (sensu Schindler 1997). This lithostratigraphical unit is restricted to the north-east of the Saar-Nahe Basin in south-west Germany. Stratigraphically, it belongs to the Jeckenbach Member, Meisenheim Fm., Lower Rotliegend Group, lower Permian, Asselian-?Sakmarian (cf. Boy et al. 2012;Schneider et al. 2020). The Niedermoschel black shale was deposited in the profundal zone of an ancient freshwater lake in an intramontane setting. It yielded an abundance of plant remains, aquatic arthropods such as diplostracans, syncaridans and xiphosurans, as well as fishes and a few branchiosaurs (Poschmann et al. 2021). Most remarkable are the abundance and the diversity of insects, which suggest that the Niedermoschel black shale may be one of the most important fossil insect-bearing units from the early Permian (Nel & Poschmann 2021;Poschmann & Nel 2021;Poschmann et al. 2021).

Specimens
The holotype of Permoberotha villosa Tillyard, 1932, YPM IP 005426p/cp, and the conspecific specimen YPM IP 015592p/cp were photographed at the YPM with a Leica M205 C stereo microscope and Leica DMC 4500 camera attachment running the Leica Application Suite X software, version 3. 0.7.19082, 2018. The new complete specimen of P. villosa, MCZ-ENT-PALE-5852a/b, was found among the Elmo MCZ's holdings lacking identification during the digitization of the museum's fossil insect collection (i.e. the Frank M. Carpenter Collection) back in 2015. This specimen was photographed with a Canon EOS 6D equipped with an MP-E 65 mm macro lens with a low focal depth setting to increase image sharpness; stacks of photographs were mounted using the software Helicon Focus Pro 6.0 (HeliconSoft Ltd). The holotype of Moscheloptera phantasma, NHMMZ PE 2020/ 5005-LS, was examined using a Leica MZ 7.5 stereomicroscope both dry and immersed in alcohol. Photographs were taken with the specimen under isopropanol using a Canon EOS 600D SLR camera equipped with a Canon MP-E 65 mm macro lens. Photographs were processed using the image-editing software Adobe Photoshop. Drawings were made from enlarged photographs. The holotypes PIN 966/1 of Archoglossopterum shoricum Martynova, 1958and PIN 1700/1701of Sylvaelytron latipennatum Novokshonov, 1998 were photographed at the PIN, Moscow using a Leica M165C stereomicroscope with a Leica DFC 420 camera.

Nomenclature
We generally follow the wing venation nomenclature proposed by B ethoux et al. (2007) B ethoux et al. (2007) showed that the vein between CuA and RP is concave in Eoglosselytrum perplexum (Riek, 1953) and corresponds to MP, while the first branch of RP is convex and corresponds to MA. Thus, we consider that MA is basally fused to RP into an RP + MA and separates again distally, while MP would appear to be independent of RP + MA in the Glosselytrodea. This wing venation pattern corresponds to that proposed by Shi et al. (2012) for the Neuropterida on the basis of the relative convexity or concavity of the veins; while Breitkreuz et al. (2017) proposed a different interpretation based on the wing tracheation pattern in which MA is completely separated from RP and has a long basal free stem with MP. Because the affinities of the Glosselytrodea are still contentious and tracheation patterns can be misleading (e.g. Schubnel et al. 2020), for now we prefer to use the basic wing venation nomenclature adopted by B ethoux et al. (2007). Nevertheless, after the recognition herein of a faint yet allegedly true vein interpreted as MP between the axial veins, at least in Permoberothidae, we differ from B ethoux et al. (2007) in considering the posterior axial vein as the anteriormost branch of CuA instead of MP, a stance coincident with other authors such as Martynova (1952)

SYSTEMATIC PALAEONTOLOGY
Order GLOSSELYTRODEA Martynov, 1938[= JURINIDA Zalessky, 1929 Revised diagnosis. (Only wing venation characters) (cf. Carpenter 1992;Rasnitsyn 2002;B ethoux et al. 2007). Fore and hind wings: RA very long, running parallel to the course of C until, at least, being close to the wing tip; RP and MA fused basally, separated distally; RA and RP closely running subparallel to each other for most of the wing's length; RP single or divided into a few pectinate, rather long branches; radial area with numerous cells that are variably dense; MA convex (+), straight or weakly zigzagged, very long; axial veins (i.e. anterior axial vein = MA, and posterior axial vein = CuAant) running parallel to each other for most of the wing's length, well-separated to closely spaced (sometimes seemingly fusing), roughly dividing the wing in two longitudinal halves, with no or few crossveins between them; a faint concave (À) vein (interpreted as MP) present or apparently absent (allegedly fused with either MA or CuAant), when present simple, emerging very close to base of RP + MA; stem of Cu divided into CuA and CuP near wing base; CuA convex, single or divided into a few pectinate branches; cubital area narrow basally and more or less broadened distally, with numerous cells that are variably dense; PCu and anal veins (up to three) straight, subparallel and simple; anal area independent of cubital veins.
Forewing: suboval to lanceolate in outline. Area between C and ScP moderately to greatly broadened, with simple veinlets (which can show a few crossveins between them) or a few to multiple rows of cells (= 'precostal' area); CuApost straight or variably extended to follow the wing margin, in the most extreme degree seemingly meeting an extended RA at the wing tip to form a continuous border (the so-called complete ambient vein).
Hind wing: suboval in outline, with anal lobe typically distinct. Slightly smaller than forewing. Area between C and ScP narrow, constant in width, bearing parallel, simple and short crossveins perpendicular to ScP (i.e. 'precostal' area always lacking); ScP very short to relatively long, ending on RA at or before the basal third of the wing's length; CuApost straight, not extending to follow the wing margin.
Remarks. Stephanastus (belonging to Skleroptera, putatively sister to the Coleoptera) has a superficial similarity to some representatives of Glosselytrodea. However, among other differences, Stephanastus lacks axial veins, with a concave M and a convex CuA that run alongside but in a curved fashion, not straight, and seemingly meeting basally, in contrast to the straight, convex MA and CuAant in Glosselytrodea that are fully separated down to the wing base (Nel et al. 2013;Kirejtshuk & Nel 2013. Family PERMOBEROTHIDAE Tillyard, 1932 Type genus. Permoberotha Tillyard, 1932.
Emended diagnosis. (Only wing venation characters.) Fore and hind wings with: (1) veins MA and CuAant (= axial veins) widely separated (i.e. medially, about 2/3 of the space between CuAant and the immediately posterior CuA branch); (2) a faint vein running between MA and CuAant (preferentially interpreted as MP), emerging rather basally; (3) cells in radial and cubital areas comparatively large and low in number, equally developed in both wings. Forewing: (4) with 'precostal' area not differentiated; and (5) lacking a complete ambient vein, with CuApost not even partially developed, straight. (6) Hind wing with ScP relatively long, reaching beyond the basal third of the wing's length.
Remarks. Note that the character considered as characteristic of Permoberothidae by Rasnitsyn & Aristov (2013, p. 696) of 'a long Sc of ordinary appearance' in the forewing is not included in the emended diagnosis because a shorter ScP is present in Moscheloptera phantasma. Genus PERMOBEROTHA Tillyard, 1932 1937 Dictyobiella Tillyard, p. 104.
Emended diagnosis. Forewing with costal area relatively narrow; ScP long, ending beyond the wing midlength; RP with branches running parallel to each other; CuA with pectinate branch between CuAant and CuApost bifurcating once; anal lobe moderately developed, as wide as costal area. Forewing with dense setation on longitudinal veins and crossveins; setae with well-marked insertion sockets. Both wings with radial and cubital areas subequally developed (reaching the same maximum width).
Remarks. The genus Permoberotha, and its type and only species P. villosa, can be distinguished from the other two genera herein classified in the Permoberothidae (i.e. Sylvaelytron and Moscheloptera; both also monospecific) by having at least the forewing with radial and cubital areas reaching the same maximum width (the radial area is more developed than the cubital area in Sylvaelytron, particularly that of the hind wing, whereas the opposite occurs in Moscheloptera) and having the anal lobe moderately developed (vs poorly developed, posteriorly straight, in both Sylvaelytron and Moscheloptera Additional material examined. Two specimens from the same locality as the holotype. Specimen YPM IP 015592p/cp, complete specimen preserved in lateral view with overlapping fore and hind wing from one side of the body plus the hind wing from the opposite body side exposed, originally described as the holotype and only specimen of Dictyobiella nervosa by Tillyard (1937) and later regarded as a junior synonym of P. villosa by Carpenter (1943); specimen herein figured as photographs for the first time. The specimen possesses one prothoracic leg extended, whereas the meso-and metathoracic legs are folded upon their respective pair. Specimen MCZ-ENT-PALE-5852a/b (part and counterpart), new specimen virtually complete, remarkably well preserved as a compression with slight relief, and in probable ventral view (although dorsal view is not ruled out). The configuration of the anterior part of the body, including the head and prothorax, is challenging to ascertain in the specimen based on the fossilization position. Although eyes  appear to be overlapping, this is also likely to be due to positioning.
Emended diagnosis. As for the genus (vide supra).
Head. Sub-hypognathous ( Fig. 2C, D). Compound eyes large; ovoid, c. 0.60 mm long, 0.43 mm wide; well separated, probably by a distance at least as wide as their own width. Vertex moderately pronounced. Ocelli, if present, not visible. Clypeus wide. Antennae relatively long, c. 3.30 mm long as preserved; inserted between compound eyes (in YPM IP 015592p/cp); scape and pedicel structure unclear but apparently longer and thicker than flagellomeres (YPM IP 015592p/cp); visible antenna with 10 preserved flagellomeres (in MCZ-ENT-PALE-5852a/b), each flagellomere narrower proximally and distally expanded (Fig. 3E); flagellomeres decreasing in length distally along the antennal length, with basalmost visible flagellomere more than 39 as long as wide. Labrum an apparently triangular sclerite. Mandibles well developed, at least 0.40 mm long, with incisor edge apparently sharp, at least one tooth probably present, other details unclear.
Thorax. Prothorax subquadrate, about as long as compound eyes, covering most of the sub-hypognathous head dorsally; meso-and metathorax well developed. All legs setose, setae up to 0.12 mm. Coxae (at least procoxae) well developed. Metatibiae bearing two moderately developed ventral spurs (Fig. 3F), not distinct in other legs. Tarsi five-segmented; tarsomeres 1-4 with particularly strong setae along plantar surfaces, the distalmost ones distinctly larger. All pretarsal claws complex, equal, widely separated from one another; each distally subdivided into at least two (possibly three) teeth ( Fig. 3G-I) and probably bearing additional shorter teeth basally, teeth arrangement uncertain, that is, pretarsal claws either bifid or trifid (with teeth subequal in length) or pectinate (with teeth progressively decreasing in length proximally); arolium apparently present, at least in the probable prothoracic leg, moderately enlarged (in MCZ-ENT-PALE-5852a/b) (Fig. 3G).
Forewing. (Measurements taken from the holotype, with the two remaining specimens assessed falling in the same size ranges) (Figs 1, 2A, B, 3A, C). Elongate and relatively narrow, 7.8 mm long, 2.6 mm maximum width (reached by the wing's distal third); crossvenation not dense, all veins (including most crossveins) highly setose, setae up to c. 0.13 mm, with setal insertions well marked; setae preserved in the costal margin with a thick straight base and a distal region thin and abruptly bent forwards (Fig. 3D). Costal space moderately broadened, with at least 11 visible simple veinlets. ScP ending on C beyond the wing's midlength. Basal ScP-R crossvein present immediately before the origin of RP + MA. RA simple, straight until the wing's distal 1/5, then slightly bent downwards following the wing's tip, with seven visible veinlets, all simple; RP + MA arising at c. CuApost; CuP straight, parallel to CuApost; PCu arising from a pronouncedly curved bulla, straight, running closely parallel to CuP; anal lobe moderately developed; anal veins simple, with a few crossveins between them; A1 approximating PCu gradually, then abruptly downcurving to end in A2 distad to the wing margin; A2 well developed, mostly straight.
Remarks. Re-interpretation of the forewing venation of P. villosa is based on the holotype and its counterpart (Fig. 1), which show the base of the wing (including the vein bullae) remarkably well preserved. The preserved basal structures identified as bullae are interpreted as such because the longitudinal veins appear to emerge directly from them and a more proximal position for median plates or axillary sclerites would be expected. The basal bulla of PCu (interpreted as A1 by Tillyard 1932) is characterized by its pronounced curved shape (see Schubnel et al. 2020). The weaker vein immediately anterior to PCu has to correspond to CuP, and the vein anterior to CuP corresponds to CuA; the bulla of the cubital vein (BCu) is also visible (Fig. 1B), showing that the separation of CuA and CuP occurs very close to the wing base. CuA has a strong stem (the posteriormost) running straight to the wing margin and two anterior branches. Anterior to CuA and posterior to MA, emerging from an indeterminate area probably after the branching of R and M, there is a straight vein, clearly weaker than MA and CuA, that we interpret as MP ( Fig. 1; see Discussion). Attribution of the specimen YPM IP 015592p/cp to P. villosa and its synonymy with Dyctiobiella nervosa done by Carpenter (1943) are herein supported. The forewings of both YPM IP 015592p/cp and MCZ-ENT-PALE-5852a/b show the same characters as those from the holotype except for the branching of RP taking place at about the wing's midlength instead of beyond the wing's distal third. Additionally, the forewing's RP of MCZ-ENT-PALE-5852a/b has three principal branches instead of two, the anteriormost RP branch briefly bifurcating near the wing's tip. These differences are regarded as intraspecies variability. Carpenter (1943) also noted some variability in the forewing pattern in the material that he examined and assigned to P. villosa. Setational differences apparently existing between the assessed forewings could be related to poor preservation and are, therefore, not reliable. Aside from wing characters, the remaining features (including size) between specimens YPM IP 015592p/cp and MCZ-ENT-PALE-5852a/b, accounting for putative preservational biases and terminalic differences, are compatible with their conspecificity. The body characters visible in YPM IP 015592p/cp are generally consistent with those described by Tillyard (1937) (see Discussion), who provided only descriptive notes and a drawing of the putative cercus. Although the author mentioned that two forewings and a hind wing were visible from the specimen, only one forewing overlapping with the hind wing is evident, given that the other visible exposed wing corresponds to the hind wing ( Fig. 2A, B).
Emended diagnosis. Forewing with costal area relatively narrow; ScP long, ending beyond the wing midlength; anal lobe weakly developed, narrower than costal area. Both wings (particularly the hind wing) with the radial area more developed than the cubital area (reaching a greater maximum width).
Remarks. In spite of the faint preservation of the wing venation in this species, the genus Sylvaelytron, and its type and only species S. latipennatum, can be separated from Permoberotha and Moscheloptera by at least the forewing having the radial area more developed (i.e. reaching a greater maximum width) than the cubital area (the opposite in Moscheloptera, subequally developed in Permoberotha). Other visible differences regarding Permoberotha include the less developed anal lobe in the forewing, whereas Sylvaelytron further differs from Moscheloptera in the narrower costal area and the longer ScP.   Remarks. The specimen is herein identified as male due to the presence of enlarged, broad gonocoxites bearing narrower gonostyli, the latter probably movable, concave, incurved distally; both the distal end of the gonocoxites and the inner side of the gonostyli appear to bear serrations (see inset in Fig. 5B).
Derivation of name. Named after the Moschel, a small creek and fourth-order tributary to the Rhine river close to the type locality, and -optera, a suffix referring to winged insects. Gender feminine.
Diagnosis. Forewing with costal area broad; ScP short, ending before the wing midlength; RP with branches zigzagging distally; cubital area well developed, reaching a greater maximum width than radial area; CuA with pectinate branch between CuAant and CuApost bifurcated twice at least; anal lobe poorly developed, narrower than costal area, posteriorly straight; anal margin bearing a fringe of distinctive thick setae.
Remarks. Moscheloptera, and its type and only species M. phantasma, can be distinguished from Permoberotha and Sylvaelytron by the forewing with a broader costal area, a shorter ScP, a cubital area that is more developed than the radial one, and zigzagging RP branches. Differences only with Permoberotha include forewing CuA with the branch between CuAant and CuApost bifurcated twice at least (bifurcated only once in Permoberotha) and the absence of a developed anal lobe. Also, although both Permoberotha and Moscheloptera appear to share the specialized forewing setae, these were present in Moscheloptera along the posterior wing margin, in the anal area, whereas they appear to be absent in that region in Permoberotha.
Description. Forewing suboval, 9.9 mm long, 3.6 mm maximum width reached at 2/3 of wing length; wing membrane with apparent maculations, in the radial and cubital areas roughly adjusting to the area within most of the cells; C bearing a distinctive dense fringe of well-developed setae; costal area broad, abutting, 0.7 mm maximum width, with c. 13 costal veinlets, apparently all simple; ScP relatively short, ending before wing midlength, at 4.2 mm from wing base, mostly straight and gently upcurving distally; branching of RA and RP + MA not discernible but rather basal; RA mostly straight, parallel to anterior wing margin after meeting point between C and ScP, distally curved following C but not reaching the wing tip; RP and MA probably separating at c. 1.9 mm from the wing base; RP with three branches, distally zigzagged; at least seven visible crossveins between RA and RP; MA straight; base of vein interpreted as MP not visible, but possibly arising between branching of RP + MA and that of MA; vein between axial veins (interpreted as MP) very weak, barely visible, discernible subdistally running closely parallel to MA; CuAant and CuApost forking at c. 0.6 mm from wing base; cubital region reaching a greater maximum width than radial region, CuA well developed, with a branch between CuAant and CuApost bifurcated at least twice; CuAant running parallel to MA, subsequent CuA branches distally downcurving towards wing margin; CuApost very gently downcurved; CuP weakly downcurved and closely parallel to CuApost; one PCu vein and one anal vein visible, details not assessable; anal area relatively narrow, slightly narrower than costal area; anal margin with a series of thick setae with a straight base and a distal region abruptly bent forwards.
Remarks. The broad costal area and ScP meeting C (instead of RA) indicate that NHMMZ PE 2020/5005-LS represents a forewing. Note that Moscheloptera phantasma is not the first Glosselytrodea known from Germany, given that Bashkuev et al. (2012, p. 176)   below. The degree of separation between the forewing veins MA and CuAant (i.e. the axial veins, putatively synapomorphic for Glosselytrodea) is variable across the order, ranging from a relatively wide separation in Permoberothidae and Archoglossopterum Martynova, 1958 (family diagnostic character 1) (Figs 1-3, 5-7) to a narrow separation (e.g. Eoglosselytrum Martynova, 1952, Glossopterum Sharov, 1966, Mongolojurina Ponomarenko, 1988) (Martynova 1952, 1961Ponomarenko 1988Ponomarenko , 2000Novokshonov 1998) (2021). The character 'complete fusion of the alleged vein MP with either MA or CuAant' could constitute a putative synapomorphy for the species classified in these three genera but it could also constitute a synapomorphy for those classified in Karajurina.
A weak vein running between MA and CuAant in both the fore and hind wings (diagnostic character 2 for Permoberothidae) had originally been overlooked in Permoberotha villosa Tillyard, 1932, for which only a 'groove or furrow' was described in this position (Tillyard 1932(Tillyard , 1937Carpenter 1943, fig. 4). Later, Carpenter (1992, p. 537) noted that this groove was 'apparently bearing a spurious vein'. Permoberotha villosa clearly possesses a continuous yet weak vein between the axial veins, as the images from the holotype and additional conspecific specimens herein show (Figs 1-3). The same vein was drawn from Sylvaelytron latipennatum Novokshonov, 1998 but overlooked in the description (Novokshonov 1998); the reassessment of its holotype also shows the presence of this weak vein running between the axial veins (Fig. 5). Archoglossopterum shoricum Martynova, 1958 also appears to possess this character, although its course cannot be fully resolved due to preservation artefacts (Fig. 7). We interpret this structure quite confidently as a vein as opposed to a flexion line. Although the permoberothid wing might have been folded to a certain degree longitudinally along the space between the axial veins (including the vein in between) in some permoberothid taxa as already noted by Tillyard (1937), the possibility that this longitudinal vein was merely a flexion line in the wing membrane of some Glosselytrodea such as Permoberothidae is considered unlikely due to the presence of a few macrosetal insertions preserved close to the base of this vein in P. villosa (see inset in Fig. 1A), and because it seems implausible that flexion lines would fossilize in the same way that veins do in fossils at least partly preserved as impressions. In the Permoberotha (and Sylvaelytron, to a lesser degree) specimens preserved as impressions herein presented, the putative MP veins have a similar appearance to the remaining veins, although they are clearly weaker, both having left a dark carbonaceous residue of cuticle. Furthermore, although we prefer to tentatively interpret the weakened longitudinal vein between the axial veins in Permoberotha, Sylvaelytron and Archoglossopterum (barely discernible in Moscheloptera) as MP, the lack of an apparent connection of its base with RP + M in these taxa, if not regarded as preservational, also enables the alternative interpretation that it could represent an intercalary vein (a fact that would render the medial vein simple following our present venational approach). This hypothesis seems less plausible because intercalary veins typically appear in multiple numbers, not single, along contiguous sets of principal veins across diverse insect lineages, and would appear to be less prone to bearing macrosetae. It is important to note that Martynova (1958Martynova ( , 1961 interpreted the vein MA of Archoglossopterum as being MP basally fused to RP. Likewise, in both Glosselytridae and Jurinidae Zalessky, 1929 the anterior axial vein had been interpreted as MP (Martynov 1938;Martynova 1952Martynova , 1961Rasnitsyn & Aristov 2013). We propose that this vein can be homologized with that of Permoberothidae and Archoglossopterum, and therefore it actually corresponds to MA, with a putative MP absent due to reduction or complete loss (allegedly concomitant with the narrowing of the space between axial veins) during the process of forewing sclerotization in Glosselytrodea. Future findings should be able to shed light on these and other matters concerning the wing venation interpretation in glosselytrodeans.
Permoberothidae (and Archoglossopterum) have larger cells in the radial and cubital areas when compared with most of the remaining glosselytrodean diversity (family diagnostic character 3), particularly Glosselytridae and Polycytellidae. Moreover, their cells are of equal size and morphology between the fore and hind wings, something that would not appear to be the case outside the group, as shown by Eoglosselytrum and Jurina Zalessky, 1929, where radial and cubital cells tend to be larger in the hind wing than in the forewing (Martynov 1938;Martynova 1952;Vilesov & Novokshonov 1994). Furthermore, permoberothids lack the development in their forewing of the so-called 'precostal' area (family diagnostic character 4), whereas other glosselytrodeans (including Archoglossopterum) show a development of this region that is variable in shape (abutting to lanceolate) and venation, with the alleged development of ScA and abundant crossvenation between branches of ScA and ScP, which form multiple rows of cells.
In glosselytrodeans the term 'ambient' is used to describe an apparently continuous vein that creates a ledge along the forewing margin. This ambient vein is actually composed of two veins, RA anteriorly and CuApost (also interpreted as CuP by some authors) posteriorly (cf. Eoglosselytrum, Surijoka Martynova, 1958, Protojurina Martynova, 1958, which showed different degrees of development throughout evolution and met at the wing tip as an alleged derived scenario according to Rasnitsyn & Aristov (2013). Although virtually all glosselytrodean wings have an RA that follows the course of C until, at least, being close to the wing tip, only some taxa have an RA that is fully extended following the course of C and which reaches the wing tip to apparently meet with CuApost. This condition occurs in all Glosselytridae and Polycytellidae; in the Jurinidae a fully ambient vein is present in, at least, Mesojurina Martynova, 1943, Shaanxiglosselytron Hong, 2007and Sinoglosselytron Hong, 2007(Martynov 1938Martynova 1943;Ponomarenko 1988;Vilesov & Novokshonov 1994;Martins-Neto & Gallego 2001;Hong 2007;Huang et al. 2007;B ethoux & Anderson 2021). The vein CuApost, however, is always straight, not even partially ambient, in the forewings of Permoberothidae (family diagnostic character 5) (Figs 1-6) and Archoglossopterum (see vein re-interpretation in Fig. 7), as well as those of Glossopterum. In the glosselytrodean hind wing (see below) RA and CuApost never form an ambient vein, their structure being very similar to those of the permoberothid forewing, that is, R following the course of C until being close to the wing tip but not reaching the median veins, and CuApost straight, not following the course of C (Tillyard 1937;Martynov 1938;Carpenter 1943;Martynova 1952;Vilesov & Novokshonov 1994;B ethoux et al. 2007). In contrast, the two species classified within the genus Glossopterum, G. sharovi Novokshonov, 1998 andG. martynovae Sharov, 1966, lack any development whatsoever of an ambient vein in the forewing (nor anteriorly or posteriorly, given that RA ends in the wing margin at a point similar to Permoberothidae and CuApost is fully straight as well), although an ambient vein is present in the Glosselytridae, Polycytellidae and Jurinidae. Therefore, there is no strong argument to maintain the genus Glossopterum in the Glosselytridae at present; it was transferred therein by Rasnitsyn & Aristov (2013). At the same time, Glossopterum cannot be accommodated into the Permoberothidae or the Archoglossopteridae due to the axial veins narrowly separated, the conspicuously denser cells in radial and cubital areas, and forewing with a distinct 'precostal' area with more than two cell rows. Thus, we transfer the genus Glossopterum and its two species back to its own family, Glossopteridae Sharov, 1966 stat. rest., acknowledging that future material should clarify the status of this taxon.
The genera Sylvaelytron and Moscheloptera are herein added to the Permoberothidae, a family that up to now was composed only of Permoberotha and its type and only species, P. villosa. Sylvaelytron latipennatum was originally placed in the family Glossopteridae Sharov, 1966 but was transferred into the Glosselytridae by Rasnitsyn & Aristov (2013). This species is here tentatively transferred to the Permoberothidae based on all of the diagnostic characters of the family except character 4 (i.e. the lack of a 'precostal' area in the forewing), which is not clear due to poor preservation (Fig. 5). The assignment of Sylvaelytron to Permoberothidae was already suggested by B ethoux et al. (2001), who preferred to wait until re-examination of the holotype was possible. In contrast, Archoglossopterum shoricum fulfils at least half of the diagnostic forewing characters of Permoberothidae (i.e. characters 1, 3 and 5 for certain). A faint vein between the axial veins (character 2) is probably present, although following its course is hampered by poor preservation (Fig. 7). However, this species presents a distinctly developed 'precostal' area, albeit consisting only of two rows of cells. For that reason, in line with Rasnitsyn & Aristov (2013), we prefer to keep this species in its own family for now, Archoglossopteridae Martynova, 1958, acknowledging that this taxon probably forms a grade with the taxa classified in the Permoberothidae. Note that, although Martynova (1958Martynova ( , 1961 depicted the vein CuApost (CuP for that author) as partly ambient (i.e. following the course of the wing margin) in A. shoricum, our interpretation shows that the alleged distal part of that vein probably represents a deformation of the rock, and that CuApost actually ends on the wing margin much more basally than originally recognized (Fig. 7).

The body structure of Glosselytrodea
Information on the glosselytrodean body structure is scarce and has hindered the understanding of the relationships outside of and within the order. Body structure has been reported only in the families Jurinidae and (to a lesser extent) Permoberothidae. For the Jurinidae, body structures have been described and figured from Eoglosselytrum kondomense, E. perfectum Martynova, 1961, Jurina marginata Martynov, 1938, J. reducta Martynov, 1938and J. sojanensis (Martynov 1938Martynova 1961;Sharov 1966). An additional, undescribed, complete jurinid was figured by Rasnitsyn (2002). Moreover, Glossopterum martynovae and G. sharovi were described based on complete specimens (Sharov 1966;Novokshonov 1998;Rasnitsyn 2002;Zhuzhgova et al. 2015). Based at least on this evidence, non-permoberothid glosselytrodeans had a hypognathous to opisthognathous head, compound eyes well separated, antennae moniliform; thorax with movable prothorax, pronotum well developed, meta-and metathorax of subequal size and structure; legs cursorial, with well-developed coxae and five-segmented tarsi; wing position at rest roof-like; and abdomen narrow, much shorter than the wings (abdomen ending at about 2/3 of forewing length), bearing apparently one-segmented cerci (adapted from Rasnitsyn 2002). A reconstruction of Glossopterum by A.G. Ponomarenko was provided by Rohdendorf & Rasnitsyn (1980) (re-figured by Rasnitsyn 2002. Data on the body parts of permoberothid glosselytrodeans are even scarcer. It was hitherto based on only two specimens: the specimen YPM IP 015592p/cp originally described as the type of Dictyobiella nervosa Tillyard, 1937; and the holotype of Sylvaelytron latipennatum (Tillyard 1937;Novokshonov 1998). A new complete individual is herein added: the specimen MCZ-ENT-PALE-5852a/b. Details of these three specimens are provided below.
YPM IP 015592p/cp. The photographs of this specimen, provided herein for the first time, enable us to reassess several structures. Tillyard (1937, p. 105) described the specimen's head as having 'large and apparently strongly projecting compound eyes separated from one another by a space about equal to their own width', which are visible in the photographs (Fig. 2C, D). The clypeus was deemed as 'a clear, somewhat projecting plate', and he added that 'only the scape of one antenna can be seen'. The clypeus, indeed, appears to be well developed, but the scapes of both antennae, the pedicels, and a significant part of each flagellum (full antennal length not evident) are also visible (Fig. 2C, D). Additionally, the head is clearly subhypognathous, and at least the mandibles appear to be discernible as well (Fig. 2C, D). Tillyard (1937, p. 105) described the thorax as 'short and stout, the prothorax apparently small, the meso-and metathorax well developed; the femur and tibia of one foreleg are visible, each about as long as the head is wide', characters that seem to be present based on the photographs, and he added 'the femur not very stout, fusiform, and carrying many stiff hairs', which is visible in Figure 2C, D. All the remaining legs are preserved, but retracted and folded upon the body. Last, Tillyard (1937, p. 106) described the abdomen of YPM IP 015592p/cp as 'short, only about half as long as the forewing', this also being evident from our photographs. The author added 'the last segment shows, in the obverse only, a terminal triangular notch or incision, suggestive of a division into paired male genitalia of the hemerobioid type, closely appressed to the last segment on either side'. Indeed, this 'notch' is visible only in the part, probably due to the arrangement of the two structures interpreted herein as the gonocoxites + gonostyli 9 (Fig. 2E, F). Contrary to Tillyard (1937), we have interpreted the genitalia of this specimen as female. Tillyard (1937, p. 106) described and drew from the last abdominal segment a 'short, very pointed process, black at apex and ending in a stiff bristle, which I take to be one of a pair of reduced cerci '. Carpenter (1943) stated that Tillyard's interpretation was wrong due to the fossil not being completely exposed, and that both cerci were longer than originally interpreted (having four or five segments), and lacked a terminal bristle. Based on the photographs provided herein, five segments appear to be visible for the only visible structure interpreted as a cercus (Figs 2E,F,4A).
PIN 1700/1701. Although Novokshonov (1998) recognized the composite eyes, the mandibles and the possible laciniae in his drawings of the holotype of S. latipennatum, these are photographically figured herein for the first time (Fig. 5). The right antennal insertion and the basal stretch of said antenna also appear to be preserved. Moreover, the presence of enlarged gonocoxites inserted ventrally and bearing distal teeth-like projections (as in Permoberotha villosa) and narrower gonostyli, the latter probably concave, incurved distally, and with visible inner serrations, is herein reported, enabling us to regard the specimen as male.
MCZ-ENT-PALE-5852a/b. The new Permoberotha villosa specimen from Elmo is remarkably well preserved, and provides additional data on the body parts of Permoberothidae, namely on the fine structure of the pretarsal claws and the male terminalia. The pretarsal claws, apparently in all legs, are complex, that is, they bear multiple teeth, although their specific arrangement is challenging to discern (Fig. 3G-I). Complex tarsal claws have evolved independently among a wide range of insects, also encompassing the majority of holometabolan lineages (e.g. McAlpine et al. 1981McAlpine et al. , 1987Goulet & Huber 1993;Sattler & Wojtusiak 1999;Leschen et al. 2010;Leschen & Beutel 2014;Beutel & Leschen 2016), including both Neuroptera (Lambkin 1986) and Mecoptera (Beutel & Gorb 2001;Grimaldi & Johnston 2014). Therefore, this character, or at least the present understanding of it, does not seem to be informative when aiming to elucidate the relationships of Glosselytrodea. Moreover, the terminalia of this specimen is consistent with those of S. latipennatum (see above), showing the development of massive, independent, and ventrally inserted gonocoxites 9 bearing irregular teeth-like projections distally, to which relatively smaller, narrower, incurved gonostyli apparently bearing spine-like setae are articulated (Figs 3J, K, 4B). The serrations and/or spinelike setae in the dorsal and inner surfaces of gonocoxites and gonostyli in male permoberothids (as evidenced by P. villosa and S. latipennatum) are likely to have increased the grasping ability of these structures during nuptial and/ or copulatory phases. Other structures visible in the new P. villosa male specimen from Elmo include apparently one-segmented cerci (Figs 3J, K, 4B). Rasnitsyn (2002, p. 191) stated that the glosselytrodean male genitalia was 'rather small, symmetrical, with ovoid gonocoxa and small articulated gonostylus', which was based on an interpretation of the holotype of Glossopterum martynovae (A.P. Rasnitsyn pers. comm. 2021). Although that assessment is compatible with our interpretation of the genitalia of P. villosa and S. latipennatum, these structures would appear to be relatively large in these species.

The contentious phylogenetic relationships of Glosselytrodea
The glosselytrodean wing venation patterns have not been able to shed light on the phylogenetic position of the group. Glosselytrodeans have a stem of Cu distinctly separated from that of R and M. That rules out affiliation to the polyneopteran superorder Archaeorthoptera sensu B ethoux & Nel (2002), given that in Archaeorthoptera the CuA is fused with M at the very base of the wing (B ethoux & Nel 2002;B ethoux et al. 2007). Similarly, attribution of this order to the superorder Acercaria is unlikely because CuA is not fused to R + M at the wing base (Nel et al. 2012). On the contrary, a stem of Cu independent from those of R and M is present in the Palaeodictyoptera (Prokop et al. 2018), the non-archaeorthopteran Polyneoptera and the Holometabola, this condition probably being symplesiomorphic. In any case, there is no clear putative synapomorphy for supporting the attribution of the Glosselytrodea to the Holometabola based on wing venation characters. However, Beutel & Gorb (2001) proposed to support the Holometabola on the basis of a five-segmented tarsus, a character present in the Glosselytrodea, but also present in several non-holometabolous groups (e.g. Mantophasmatodea, stem Odonatoptera).
Assignment of the Glosselytrodea to the Neuropterida based on venation characters is hardly justified despite having been the general trend. Veins ScP and R/RA running closely parallel are not only present in Neuroptera, but also in other groups such as the superorder Mecopterida (= Amphiesmenoptera + Antliophora). The shape of the anal veins in the Glosselytrodea and Neuropterida is different, with rather simple parallel veins in the former clade. Although the general pattern of venation in Permoberothidae is reminiscent of that of Neuroptera, especially in the veins MA and MP running closely parallel, the shape of the cells in the distal half of the wing, and that of the branches of RP, these characters could be the result of convergence. In Glosselytrodea it is still uncertain whether R and M have a common stem or separate ones running closely parallel to each other.
The terminalia characters described from the complete permoberothids figured herein offer new data to assist the elucidation of the broader relationships of Glosselytrodea. In the Polyneoptera, the female gonostyli are either lost or strongly reduced, rendering an affinity of glosselytrodeans to this group based on this character highly unlikely. The female terminalic conformation of the P. villosa specimen YPM IP 015592p/cp appears to differ from both the neuropteroid and the mecopterid ground plans, and is instead compatible with that closer to the hypothesized holometabolan ancestral structure in which the ovipositor has been reduced (therefore unable to insert eggs in the substrate) but gonapophyses 8 are still well developed, gonocoxites and gonostyli 9 are present, even if reduced, and multisegmented cerci have been retained (Mickoleit 1973;Grimaldi & Engel 2005;H€ unefeld et al. 2012). Moreover, according to H€ unefeld et al. (2012) the groundplan condition for the female postabdomen of the Mecopterida is telescoping, which is not apparent in the specimen. Note that the configuration of the female terminalia of YPM IP 015592p/cp is very similar to the structures sketched by Sharov (1966) from a specimen assigned to Eoglosselytrum kondomense Martynova, 1952 (Jurinidae) by Martynova (1961). In contrast, the male terminalia in the P. villosa specimen MCZ-ENT-PALE-5852a/b would appear to be closer to that of the antliophoran type, or at least plesiomorphic when compared with that of Neuropterida, due to the presence of massive, independent gonocoxites 9 inserted ventrally, which are primarily reduced and integrated with the remaining abdomen in neuropterids (Asp€ ock & Asp€ ock 2008).
The inclusion of Permoberothidae within the Glosselytrodea does not appear to be universally accepted. Permoberotha remained separated from the other Glosselytrodea (Tillyard 1932(Tillyard , 1937Carpenter 1943) until Martynova (1961 (as corroborated by Sharov 1966) indicated a relationship of this taxon with the remaining glosselytrodeans. Later, Rohdendorf & Rasnitsyn (1980) and Carpenter (1992, p. 357) adopted that stance, the latter simply stating 'It now seems certain that the Permoberothidae are, in fact, members of the Glosselytrodea', without providing references. B ethoux et al. (2001) considered the Permoberothidae as a separate lineage from Glosselytrodea s.s. and classified within the Neuropterida. Although Rasnitsyn (2002) did not question the glosselytrodean status of permoberothids, Grimaldi & Engel (2005, p. 332) considered that glosselytrodeans 'may be an unnatural group, consisting on the one hand of the Early Permian Permoberothidae and on the other hand as an assemblage of Late Permian and Triassic-Jurassic families '. Later, B ethoux et al. (2007) assigned Permoberotha to Glosselytrodea, listing several potentially synapomorphic characters, and so did Rasnitsyn & Aristov (2013). However, more recently, Engel et al. (2018) included only the Permoberothidae in a phylogeny of Neuropterida. The data discussed herein support permoberothids and the remaining glosselytrodeans as having a very similar body structure, and reinforce the idea that they belong to the same lineage. At present there are no strong arguments to support the view that the two groups are not closely related phylogenetically. A difference between Permoberothidae and the remaining Glosselytrodea would include the development of the cerci, which are multisegmented in Permoberothidae but apparently onesegmented or at least less developed in other Glosselytrodea. Besides the similar body structure of permoberothids and the remaining glosselytrodeans, their significant differences in forewing structure can be interpreted as resulting from a sclerotization process, with taxa (Archoglossopterum or Glossopterum) showing morphologies apparently bridging the two (Rasnitsyn & Aristov 2013).

Forewing sclerotization in Glosselytrodea
The sclerotization of the forewings is a process that has taken place multiple times across neopterans to increase the protection offered to both the hind wings and the abdomen when the former are kept retracted upon the latter. Sclerotized forewings include diverse tegmina present across several polyneopteran groups (e.g. dermapterans, dictyopterans, orthopterans), the hemipteran hemelytra, and the coleopteran elytra (Grimaldi & Engel 2005). Other instances are known in a few Psocodea (Lienhard 2005) and the Mecoptera, that is, in the only extant species of the relictual Eomeropidae, Notiothauma reedi MacLachlan, 1877(Crampton 1930Mickoleit 1971).
It has been posited that the glosselytrodeans also underwent a general trend of forewing sclerotization (i.e. tegmenization or 'elytrization' in its widest sense) and an increase in the protective role of forewings throughout their evolutionary history (Rasnitsyn & Aristov 2013), although the mode and tempo of such a process still need to be elucidated. Interestingly, most (if not all) of the forewing diagnostic characters of Permoberothidae can be interpreted in the context of absence of forewing sclerotization and, thus, they are putatively plesiomorphic: axial veins well separated (far from being fused) with a faint vein retained in between, low vein/cell density, 'precostal' (and anal) areas not (or barely) differentiated, and ambient veins absent. Although most of these characters are also present in Archoglossopterum shoricum, the presence of a 'precostal' area in this species (Fig. 7) appears to be a step closer towards forewing sclerotization.

CONCLUSION
The oldest known Glosselytrodea, Moscheloptera phantasma gen. et sp. nov., is described from the lower Permian of Germany, and classified in the family Permoberothidae together with Permoberotha villosa and Sylvaelytron latipennatum. The genus Glossopterum is excluded from the Glosselytridae and classified in its own family, Glossopteridae Sharov, 1996 stat. rest. Glosselytrodean diversity is currently classified into six families (Appendix S1). The permoberothid body structure, the knowledge of which has been significantly expanded with the new data presented herein, is consistent with that known from other glosselytrodeans, supporting the integrity of the 'Glosselytrodea s.l.' (i.e. including Permoberothidae) as a natural group. Body structures herein figured and described with the greatest detail thus far in the group, such as the pretarsal claws and both the female and male terminalia, increase the characters able to assist the elucidation of the obscure phylogenetic relationships of glosselytrodeans. Although it would currently seem certain that glosselytrodeans are early holometabolan insects, their affiliations are still doubtful and require further morphological data. One option, given the seemingly plesiomorphic nature of the holometabolous male terminalia of P. villosa and S. latipennatum, with independent gonocoxites and movable gonostyli, and the very reduced ovipositor in female specimens, is that glosselytrodeans could be stem-aparaglossatans, that is, sister to the clade consisting of Holometabola excluding Hymenoptera (e.g. Song et al. 2016). The permoberothid male terminalia presented herein, however, appear to bear a greater resemblance to those of the antliophoran lineage (Mecoptera, Diptera) rather than to the neuropterid one (Raphidioptera, Megaloptera, Neuroptera), the lineage to which glosselytrodeans have been most frequently related to in the last decade. In any case, a phylogenetic analysis of the whole order Glosselytrodea is necessary to clarify both its large-scale relationships, the tegmenization/ 'elytrization' hypothesis that allegedly drove the evolution of the group, and the limits of the currently accepted families. For this goal, it is necessary to continue to increase the diversity of glosselytrodeans, particularly that of permoberothids due to their putative 'plesiomorphic' status, as well as to expand the knowledge on their body structures. By doing so not only the origins of this enigmatic extinct order, but probably also the early evolutionary history of the entire Holometabola, will be one step closer to resolution. Improving the known glosselytrodean diversity will also lead to a better understanding of their diversity trends through time, including assessing potential impacts of the Permo-Triassic mass extinction event.