Ovipositor and mouthparts in a fossil insect support a novel ecological role for early orthopterans in Pennsylvanian forests

Lobeattid insects represented a high portion of the earliest known, Pennsylvanian insect faunas. However, their systematic affinities and their role as foliage feeders which severely influenced their ecosystems remain debated. We investigated hundreds of samples of a new lobeattid species from the Xiaheyan locality using Reflectance Transforming Imaging combined with geometric morphometrics in order to assess its morphology, infer its ecological role, and phylogenetic position. Ctenoptilus frequens sp. nov. possessed a sword-shaped ovipositor whose valves interlocked by two ball-and-socket mechanisms. This unambiguously supports lobeattids as stem-relatives of all Orthoptera (crickets, grasshoppers, katydids). Given the herein presented and other remains, it follows that this group experienced an early diversification coupled with high levels of abundance. The ovipositor shape additionally indicates that ground was the preferred substrate for eggs. Visible mouthparts made it possible to assess the efficiency of the mandibular food uptake system in comparison to a wide array of recent species. The new species was omnivorous which explains the paucity of external damage on contemporaneous plant foliage.


47
The earliest known insect faunas in the Pennsylvanian, ca. 307 million years ago, 48 were populated by species displaying mixtures of inherited (plesiomorphic) and 49 derived (apomorphic) conditions, such as the griffenflies, but also by highly 50 specialized groups, such as the gracile and sap-feeding megasecopterans. A prominent 51 portion of this past insect fauna were the so-called 'lobeattid insects'. They have been 52 recovered from all major Pennsylvanian outcrops, where some species can abound 53 (1)(2)(3). Indeed, at the Xiaheyan locality, China, for which quantitative data are 54 available, they collectively account for more than half of all insect occurrences (4). 55 Additionally, another extinct group -the order Cnemidolestodea -composed of 56 derived relatives of lobeattid insects, was likewise ubiquitously distributed during the 57 Pennsylvanian until the onset of the Permian (5). 58 The systematic affinities of this abundant fraction of the earliest insect faunas are 59 debated. They have been regarded as stem-relatives of either Orthoptera (6) or of 60 other lineages within Polyneoptera (7, 8). A core point of the debate is the presumed 61 wing venation groundplan of insects, which, however, will remain elusive until 62 Mississipian or even earlier fossil wings are discovered. Ecological preferences of 63 lobeattid insects are also poorly known. They have been traditionally regarded as 64 foliage feeders (9) but, given their abundance, this stands in contrast to the paucity of 65 documented foliage external damage during that time. 66 The Xiaheyan locality is unique in several respects (4), including the amount of 67 insect material it contains. Over the past decade a collection of several thousand 68 specimens was unearthed, allowing for highly detailed analyses of e.g. ovipositor and 69 Ovipositor morphology 101 The external genitalia in insects consist primarily of a pair of mesal extensions, the 102 so-called gonapophyses, or ovipositor blades, and a pair of lateral projections, the 103 so-called gonostyli, or ovipositor sheaths on abdominal segments 8 and 9. These 104 sclerotized elements are collectively referred to as 'valves'. The studied fossils 105 possess three pairs of valves in their ovipositor, each strongly sclerotized ( Fig Phasmatodea showed comparatively high MAs with an almost linear curve 179 progression towards more distal parts of the mandibular incisivi whereas Plecoptera, 180 Zoraptera, and Grylloblattodea were located at the lower end of the MA range with a 181 gently exponential decrease towards the distal incisivi. The analysed extant 182 Orthoptera occupy a comparatively wide functional space, with lineages at the higher 183 and lower end of the MA range. The composite fossil mandible representation (CFMR) 184 of Ct. frequens (see Methods) is located in the centre of the observed range of MAs 185 for Orthoptera (Fig. 3). 186 A polynomial function of the fifth order resulted in the best relative fit on the MA 187 curves according to the AIC value (-661.3, see Methods). The five common 188 coefficients were subjected to a principal component analysis (PCA, Fig. 3E The CFMR of Ct. frequens is located at the centre of the first three PCs (Fig. 3E).

200
Omnivorous Orthoptera and all herbivore taxa, with the exception of Apotrechus, are 201 located along the width of PC1, while there is a tendency for the carnivorous taxa Ovipositor morphology and phylogeny 218 Our analysis of material of Ct. frequens provides unequivocal evidence that an 219 olis2 occurs in this species. Therefore, the new species was an orthopteran. Owing to 220 the lack of jump-related specialisation in the hind leg, the species can be excluded 221 9 of 19 from Saltatoria . Due to its wing morphology, with a CuPa vein lacking a fork basal to  222  its fusion with the CuA vein, the new species can also be excluded from the  223 Panorthoptera. It follows that Ct. frequens, and its various Pennsylvanian relatives 224 collectively referred to as 'lobeattid insects', are stem-relatives of Orthoptera (Fig. 225 2G). 226 The organizational diversity of elements which form the external ovipositor in 227 Orthoptera made it difficult to reconstruct its evolution based on extant species only 228 (12-16; Supplemental Text section 2.2). Comparison has traditionally been made 229 between Grylloblattodea (rock-crawlers) and Orthoptera (15) even though the two 230 groups are not closely related. In both groups the ovipositor displays an elongate 231 gonostylus IX (gs9) and a ball-and-socket locking mechanism, the so-called primary 232 olistheter (olis1), interlocking gonapophyses IX onto gonapophyses VIII (gp9 and gp8, 233 respectively; Fig. 2G). This primary olistheter occurs widely among insects (11). 234 Orthoptera possess a variety of additional olistheters, including one interlocking gs9 235 onto gp9 (royal blue in Fig. 2; olis2), as exemplified by Rhaphidophoridae (cave 236 crickets; Fig. 2E and F) and Gryllacrididae (raspy & king crickets). The occurrence of 237 an olis2 is diagnostic of ensiferan ('sword-bearing') Orthoptera (14; and see below). 238 Even though it is unclear how far posteriorly olis2 extends in Ct. frequens, the 239 asserted phylogenetic placement of this species provides new insights on the 240 evolution of ovipositor interlocking mechanisms in Orthoptera (14; Fig. 2G). The 241 ovipositor interlocking mechanisms is comparable to the one of Rhaphidophoridae, 242 the main differences concern the rachis ('ball' as in 'ball-and-socket'), which is 243 limited to a short protrusion in these insects, while the aulax ('socket' as in 244 'ball-and-socket') extends further posteriorly. In addition, gs9 extends more ventrally, 245 concealing gp8 for some distance. Compared to Gryllacrididae the only notable 246 difference in Ct. frequens is the ventral extension of gs9 in the former. In 247 Anostostomatidae the ventral margin of gs9 enters a socket in gp8, regarded as 248 composing the premises of a third olistheter (olis3 and hooks (27). In Oecanthinae, in which oviposition functioning was studied in most 306 detail, the alternate back and forth movements of gp8 induce apices of gs9 to 307 alternately approximate and diverge (31), and therefore act as a shoving tool. 308 The Rhaphidophoridae commonly lay eggs into the ground, or, alternatively, into 309 rotten leaves or wood (32). In the latter case the ovipositor is often curved. 310 Interestingly, Ceuthophilus spp. use the ovipositor tip, somewhat truncated, to rake 311 ground surface above oviposition holes (33), presumably to hide them. 312 Anostostomatidae lay eggs in the ground or on walls of subterranean chambers (34, 313 35). These preferences also apply to both Gryllacrididae (36, 37) and Stenopelmatidae 314 (38; not represented in Figure 2G)

1.2
Documentation of fossil material

General aspects
Handmade draft drawings were produced using a LEICA MZ12.5 dissecting microscope and illustrated with the aid of a drawing tube (Leica, Wetzlar, Germany). Photographs were taken using Canon EOS 550D or 5D Mark III digital cameras (Canon, Tokyo, Japan), coupled to a Canon 50 mm macro lens, a 100 mm macro lens, or a Canon MP-E 65 mm macro lens, all equipped with polarizing filters. Each specimen was photographed under dry condition and covered with a thin film of ethanol. When available, both imprints were photographed. These photographs were optimized using Adobe Photoshop CC 2015.5 (Adobe Systems, San Jose, CA, USA) and assembled, together with handmade drawings, into a single, multi-layered document. Reproduced photographs referred to as 'composites' are a combination of photographs of a dry specimen and the same under ethanol. In addition to traditional photographs, we computed Reflectance Transforming Imaging (RTI) files for details of several specimens. This corpus of data was used to produce illustrations using Adobe Illustrator CS6 (Adobe Systems, San Jose, CA, USA). Multi-layered documents (photographs only) and RTI files are provided in the associated Dryad dataset (58). Investigated specimens are listed in the section 2.1.2.
Measurements were based on complete specimens illustrated herein and are provided in the following format: minimum/average/maximum.

1.2.3
Head and mouthparts morphology The head and mouthpart morphology was investigated based on six specimens. Four of them (viz. CNU-NX1-326, -747, -754, -764) were investigated for the mechanical advantage (MA; see section 1.4) of their mandibles. The specimens CNU-NX1-749, and -756 were excluded from the analysis because their mandibles were preserved with a slight rotation in the frontal plane, this impeding an accurate measurement of the MA (see below).

1.3
Documentation of extant material

Ovipositor morphology
We complemented the available literature on the morphology of female terminalia which form the ovipositor in polyneopteran lineages and in Orthoptera in particular [12-14, 57; and see Klass (11) and references therein] by preparation of material belonging to various extant species (see section 2.2). External habitus was photographed under various angles. Terminalia, together with the ultimate abdominal segments, were then cut off and mounted in a polyester resin. Three to four sections were made at various levels and hand-polished. Direct observation and photographs (same equipment as above) were used to document them.

Mandible morphology
To allow for inferences about the potential feeding ecology of the fossils, the MA was studied on a phylogenetically diverse sample of extant species including several orders of polyneopteran insects. Twenty-nine recent taxa of Polyneoptera (tab. S2) were investigated using micro-computed tomography (µCT) carried out at several synchrotron facilities: Beamline BW2 and IBL P05 of the outstation of the Helmholtz Zentrum Geesthacht at the Deutsches Elektronen Synchrotron (DESY), the beamline TOMCAT at the Paul Scherrer institute (PSI), the TOPO-TOMO beamline of the Karlsruhe Institute of Technology (KIT), and beamline BL47XU of the Super Photon Ring 8GeV (SPring-8).

1.4
Analysis of the mandibular mechanical advantage

Introduction
The mechanical advantage (MA) is a straightforward biomechanical metric which was first introduced for vertebrates (59) and was used since in studies on vertebrate and arthropod jaw mechanics (52, 60, 61). The MA is defined as the inlever to outlever ratio. For dicondylic insect mandibles the inlever is the distance between the application of the input force and the joint axis, while the outlever arm is the distance from the biting point to the joint axis ( fig.  S1).
The MA thus indicates the percentage of force transmitted to the food item (i.e. the effectivity of the lever system). Although more detailed investigations concerning muscular insertion angles, muscle volumes, spatial arrangements and muscle characteristics would be needed to quantify the absolute forces applied to a given food item, the MA is a useful mechanical performance index: It allows a size independent comparison of the relative efficiencies within the mandibular lever system and it can be readily measured in a wide array of dried museum specimens as well as freshly collected ones. Here, we used it to assess the efficiency of the mandibular lever system of insect fossils for the first time.
Automatic segmentations of the mandibles were performed using the software ITK-snap (62) after which STL files were imported into the software Blender (www.blender.org) for further processing (fig. S1). The gnathal edge was defined sensu Richter et al. (63) as the area from the pars molaris (proximal to the mouth opening) to the pars incisivus (distalmost tooth). Since the homology of subparts of the gnathal area is debated (63-65), the gnathal outline, as seen when orienting the mandible in line with the rotation axis ( fig. S1), was scaled as a percentage of tooth row length. For this, ~800 points for each specimen were wrapped against the gnathal outline in Blender and the distance between each point orthogonal to the mandibular rotation axis (= outlever) was measured. Similarily, one point was placed at the insertion point of M. craniomandibularis internus on the mandible and the distance between this point orthogonal to the rotation axis was measured (i.e. inlever). MA measurements were carried out on the segmentations of the left mandible for each specimen. All measurements and calculations were carried out in the R software environment (v. 1.1.383) using custom scripting. Separate MAs for each studied fossil were computed and combined to a composite fossil mandible representation (CFMR) using a Procrustes superimposition as implemented in geomorph v.3.0.5. in order to account for uncertainties in MA extraction due to potential distortion artefacts. From this superimposition the mean MA shape was extracted and used together with the MAs of recent species for the further analysis steps. Polynomial functions of the 1 st -20 th order were fitted against all MA profiles. The Akaike information criterion (AIC) was used to determine the polynomial function with the best relative fit whose coefficients were then used for further analysis.

1.4.2
Phylogenetic signal Phylogenetic signal was assessed using the most recent phylogenetic estimate of the 1kite consortium (www.1kite.org) as a basis (66). The phylogeny was pruned in order to contain only the taxa analysed here. The fossils were fitted into the phylogenetic estimate based on inference derived from wing venation and leg and ovipositor morphology (Fig. 2). Specifically, Ct. frequens possesses a fusion of CuA (emerging from M+CuA) with CuPa, which is the defining character state of the taxon Archaeorthoptera (67), and olis2, the defining character state of the taxon Neoclavifera tax. n. (see below). Both taxa include crown-Orthoptera and some stem-relatives. The lack of a branching of CuPa (which would indicate a panorthopteran), and of jumping hind-legs (which would indicate a saltatorian) both indicate that Ct. frequens is not a crown-orthopteran.
Phylogenetic signal was assessed using the K statistic as implemented in geomorph v.3.0.5 (54) with 10,000 random permutations. This test statistic was found to be the most efficient approach to test for phylogenetic signal (68). Since significant phylogenetic signal was detected, a principal component analysis (PCA) as well as phylogenetic PCA as implemented in the phytools package v.0.6-44 (55) were carried out in order to compare the analysed specimens in MA shape space.

Systematic Palaeontology
In this section the systematics at the family-group level and below conforms to the ICZN to ensure that the new species name is valid under this Code, while that above the family-group, left ungoverned by the corresponding code, conforms to the principles of cladotypic nomenclature (69; and subsequent accounts), itself compliant with the PhyloCode (70). Specifically, a cladotypic definition corresponds to an apomorphy-based definition using two species as internal specifiers (each being anchored to a specimen designated as type). There are minor discrepancies between cladotypic nomenclature practice on the one hand and recommendations of the PhyloCode on the other. Notably, the first author to have associated the selected defining character state and a taxon is to be acknowledged under the former procedure. Qualifying clauses. Several qualifying clauses are explicit when using a cladotypic definition, but they need to be specified for a PhyloCode usage. The name Neoclavifera shall be considered as invalid as that of a taxon if it occurs that (i) the defining character state was acquired by the cladotypes / specifiers convergently, (ii) the defining character state is a plesiomorphy, (iii) the cladotypes / specifiers belong to a single species and/or (iv) the defining character state does not occur in the specifiers (unless it is secondarily lost). There is no known evidence that one of these clauses might be challenged in our case.
Discussion. At the first glance, the name Ensifera Chopard, 1920, appears as a suitable name to convert. However, it is an explicit reference to the sword-like shape of the ovipositor valves in the corresponding insects, which composes a pre-occupation under cladotypic nomenclature (conversely, the taxon name Caelifera Ander, 1936, is an explicit reference to chisel-like shape of the valves). In other words, the name etymologically refers to a character state different from that used to define the new taxon, which makes it unavailable for the aimed purpose. The same applies to the taxon name Dolichocera Bei-Bienko, 1964 ('long horned'; and, conversely, 'Brachycera Bei-Bienko, 1964' for 'short horned'), favoured by Kluge (14). Moreover, current classificatory schemes customarily regard Ensifera and Caelifera as sister-groups, while our results predict that Caelifera is to be nested within Ensifera. Prolonged ambiguity on the conversion of 'Ensifera' as a defined taxon is then to be expected, not mentioning the fact that Ensifera Lesson, 1843 is a genus name for sword-billed hummingbirds, and Ensifera ensifera (Boissonneau, 1839) its type species. Given this situation, and the absence of a name composing a direct reference to the occurrence of olis2, we propose to coin a new one accordingly. Based on our literature survey, Kluge (14) is the first author to have discriminated a taxon on the basis of the defining character state only. This author stated that an olis2 is the autapomorphy of 'Dolichocera', but the name being a direct reference to another character state (see above), it follows that a new one is needed, hence Neoclavifera.
The meaning of the terms 'rhachis' and 'aulax' is critical to the proposed definition. Modest elevation and groove likely made the transition from adjoined smooth surfaces to ones bearing a proper rhachis and aulax. It is therefore necessary to define rhachis and aulax, as follows: a rhachis is a projection whose base is narrower than its projected part at its widest (best assessed in cross-section), and an aulax is its counterpart.
As defined, and based on species currently known, the composition of the taxa Archaeorthoptera and Neoclavifera overlap. We hypothesize that the defining character state of Archaeorthoptera was acquired in a hypothetical ancestral species distinct from the one of Neoclavifera, but the order of acquisition of their respective defining character states remains unknown.  (Brongniart, 1893), its most closely related species (SI Appendix, section 2.1), smaller size (deduced from forewing length) and prothorax longer than wide (as opposed to quadrangular).

Differential diagnosis. Compared with Ctenoptilus elongatus
General description. Body length (excluding antennae, including ovipositor) about 42-52 mm (based on female individuals only). Head: prognathous, head capsule heart-shaped in dorsal view; md with strongly sclerotized and prominent incisivi and a well-sclerotized molar area; la with a strong apical tooth and a smaller sub-apical one; mp well-developed, with 5 observed segments; tentorium composed of well-developed ata, ct and pta, dorsal arms not visible; co located in the midline along the dorsal side of the head capsule, then branching into two diverging fc; ant long, filiform. Thorax: prothorax longer than wide, longer than head; boundary between mesothorax and metathorax not visible. Wings: ScP reaching RA distal to the two thirds of wing length; RA with few or no anterior veinlets; RA and RP strong, parallel for a long distance; RA-RP area narrow in its basal half; at the wing base, R and M+CuA distinct; MA and MP simple for a long distance, with similar numbers of terminal branches, usually 1-3, rarely more than 4; CuA diverging from M+CuA and fusing with CuPa; CuA+CuPa posteriorly pectinate. Forewing: length 31.5/36.1/41.2 mm, largest width 6.9/8.3/10.7 mm, membranous; ScP with anterior veinlets; RA-RP fork slightly distal to the point of divergence of M and CuA (from M+CuA); RP branched distally, near the second third of wing length, usually with 11-17 branches reaching apex, and occasionally 1-2 veinlets reaching RA; first split of M+CuA (into M and CuA) near the first fourth of wing length; between the origin of CuA (from M+CuA) and the first fork of RP, M very weak; first fork of M near wing mid-length; MA distinct from RP, connected to it by a short cross-vein, or occasionally fused with it for a short distance; median furrow located along M and then MP; CuA+CuPa with most of its main branches further branched, with a total of 16-26 terminal branches; in basal part, CuA+CuPa emitting strong posterior veinlets, vanishing before they reach the claval furrow; CuPb concave, weak and simple; AA1 with 3-4 branches; AA2 with about 10 branches; cross-veins mostly not reticulated, except along the apical and postero-apical section of the wing margin, and in the ScP/ScP+RA-RP area (where they are particularly strong); longitudinal pigmented areas located (i) along R, (ii) along CuA, and then the main stem of CuA+CuPa and (iii) along the posterior wing margin, distal to the endings of the first branches of CuA+CuPa; these three areas merge distally; additional pigmented area along AA1. Hind wing: as in forewing, except for the following: slightly shorter than forewing; RA-RP fork opposite the point of divergence of M and CuA (from M+CuA); RP usually with 11-16 branches reaching apex; M forked at the first quarter of the wing; M with 5-8 branches reaching posterior wing margin; CuA+CuPa with 5-8 branches; pigmented area forming an arc covering the apex, beginning along RA and ending close to the end of CuPb; plicatum well developed, with plica prima anterior reaching the posterior wing margin opposite the end of ScP (on RA). Legs: Fore-leg femur 4.9-6.3 mm long, 1.0-1.3 mm wide, tibia 5.2-6.3 mm long; mid-leg femur 5.2-6.4 mm long, tibiae 5.9-7.3 mm long; hind-leg femur 7.5-11.5 mm long, tibia 9.8-12.0 mm long; spines, probably in two rows, present along the ventral side of tibia of all legs, concentrated near the apex (fore-leg, at least 12 spines; mid-leg, at least 8 spines; hind-leg, at least 15 spines); tarsus 5-segmented, 2 nd , 3 rd and 4 th segments shorter, terminal tarsal segment with paired claws and arolium (deduced from well-preserved fore-legs). Abdomen: abdomen about 17-23 mm long (based on female individuals only); female with a prominent sword-like ovipositor (see more detailed interpretation below and specimens description). , CNU-NX1-326 (Fig. 1): Positive and negative imprints of an almost complete female individual, viewed dorsally, very well preserved, with head, thorax, leg remains (including well exposed fore-legs) and complete right forewing; right hind apex missing, left wings incomplete, left hind wing very incomplete, ovipositor apex concealed under right forewing. Head: about 6.6 mm long, 4.3 mm wide, prognathous; mandibles about 2.0 mm long, with prominent teeth at their apex; gnathal edge of right md clearly visible, heavily sclerotized, with the distal incisivus shorter than the subdistal ones; mp strong, but segments not visible; f large and separated to the vertex by a U-shaped line, laterally delimited to the well-developed genal area by a line; frontal and coronal sutures well-developed, located at the closest distance of the eyes to each other; Eyes large, laterally protruding from the head capsule covering about half of the lateral head profile; ant incomplete, 6.3 mm long as preserved. Thorax: prothorax about 5.5 mm long, 3.7 mm wide. Left forewing: preserved length 22.2 mm, best width 8.1 mm; M with its 2 main branches preserved, CuA+CuPa with 22 terminal branches preserved. Right forewing: length 32.6 mm, width 9.6 mm; RP simple for 14.3 mm, with 16 branches reaching wing apex and 1 reaching ScP+RA; MA connected to RP by a short cross-vein, with 3 branches, MP with 4 branches; CuA+CuPa with 26 terminal branches preserved; CuPa partly preserved. Right hind wing: preserved length: 30.4 mm, best width 8.6 mm; plicatum creased. Legs: fore-leg femur about 4.9 mm long and 1.2 mm wide, tibia 6.3 mm long and 0.7 mm wide, tarsus about 5.0 mm long, tarsal segments (5), paired claws and arolium visible; mid-and hind-legs incomplete and/or not well exposed. Legs: spines well exposed on foreleg tibiae and distal part of a mid-leg tibia. Abdomen: bent (probably a consequence of decay), about 17 mm long, ovipositor viewed laterally, possibly slightly obliquely; bases of gp8 strongly sclerotized, well visible. Fig. 2A and B, and fig. S7A-E): Positive and negative imprints of an almost complete female individual, wings incomplete and overlapping, body about 45 mm long. Head: about 6.4 mm long, 3.5 mm wide. Thorax: prothorax about 5.6 mm long, 3.7 mm wide. Legs: fore-leg femur 4.9 mm long, 1.2 mm broad, tibia 5.8 mm long, 0.8 mm broad, tarsus about 3.8 mm long; mid-leg femur 5.9 mm long, 1.0 mm broad, tibiae 7.3 mm long, 0.8 mm broad, tarsus about 4.9 mm long; hind-leg femur 6.1 mm long, 1.1 mm broad, tibia 10.1 mm long, 0.7 mm broad; spines visible, or even well-exposed, on each exposed tibiae. Abdomen: about 17 mm long (excluding ovipositor); sword-like ovipositor viewed laterally, about 8.4 mm long; antero-basal apophyses of gs9, gp9 and gp8 distinct, well delineated; near the ovipositor base, dorsal and ventral edges of gs9 and gp8, and ventral edge of gp9 well delineated; dorsal edge of gp9 visible in the distal half of the ovipositor; olis1 and olis2 visible near the ovipositor base, strongly sclerotized; olis1 located along the ventral edge of gp9 and dorsal edge of gp8; olis2 located close to (or along) the ventral edge of gs9, and laterally on gp9; olis1 and olis2 converging; ventral edge of gp8 with teeth more prominent and densely distributed near the apex.

CNU-NX1-749 (
CNU-NX1-742 ( Fig. 2C and D, and fig. S8A-C): Positive and negative imprints of an almost complete female individual, partly disarticulated, left forewing missing; body about 52 mm long. Head: detached from the rest of the body, mouthparts not discernible. Thorax: prothorax about 7.0 mm long, 3.6 mm width. Wings: a forewing and two hind wings visible, poorly preserved. Legs: fore-leg femur 5.7 mm long, 1.1 mm broad, tibia 6.2 mm long, 0.9 mm broad; spines well exposed on one hind-leg tibia, some visible on one fore-leg tibia. Abdomen: strongly bent, segments not discernible; ovipositor very well preserved, detached from the rest of the abdomen, about 9.5 mm long; antero-basal apophyses of gs9, gp9 and gp8 distinct, well delineated; near the ovipositor base, dorsal and ventral edges of gs9 and gp8, and ventral edge of gp9 well delineated; dorsal edge of gp9 visible at the extreme base and in the distal half of the ovipositor; olis1 and olis2 visible near the ovipositor base, strongly sclerotized; olis1 located along the ventral edge of gp9 and dorsal edge of gp8; olis2 located close to (or along) the ventral edge of gs9, and laterally on gp9; olis1 and olis2 converging; ventral edge of gp8 with teeth more prominent and densely distributed near the apex.
CNU-NX1-754 ( Fig. 3A and B, and fig. S7I-K): Positive and negative imprints of an almost complete individual, well-preserved, wings overlapping, incomplete and partly creased , end of abdomen missing. Head: about 6.8 mm long 4.5 mm wide; md with strongly sclerotized and prominent incisivi and a well-sclerotized molar area; terminal teeth of la visible; ca distinguishable; co located in the midline along the dorsal side of the head capsule, then branching into two diverging fc. Thorax: prothorax about 5.9 mm long, 4.4 mm wide. Legs: fore-leg femora 5.3 mm long and 1.1 mm broad, tibiae 5.2 mm long and 0.7 mm broad, tarsus about 4.0 mm long; mid-leg femur 5.4 mm long and 1.1 mm broad, tibia 5.9 mm long and 0.8 mm broad, tarsus about 4.5 mm long; fore-and mid-leg tarsi well preserved, 5-segmented with paired claws and arolium; 2 nd , 3 rd and 4 th segments shorter, ventral process (projecting forward) of 3 rd and 4 th segments visible; hind-leg femora 7.5 mm long; end of hind-leg tibiae missing, 7.1/5.9 mm long, 0.7 mm broad; spines well-exposed on one of the forelegs tibiae. Abdomen: about 14 mm as preserved, segments not discernible.
CNU-NX1-764 ( Fig. 3C and D): Positive and negative imprints of an almost complete, isolated head, posterior part possibly overlapping with prothorax; mouthparts well preserved; md in occlusion, 2.1 mm long, 1.1 mm wide at their base, provided with strongly sclerotized and prominent incisivi and a well-sclerotized molar area; distal part of la visible, provided with a strong apical teeth and a smaller sub-apical one; tentorium composed of well-developed ata, ct and pta, dorsal arms not visible; ct 1.2 mm long and 0.3 mm wide.
CNU-NX1-752 ( fig. S2A and B): Positive and negative imprints of a partly incomplete individual, head and prothorax well exposed, a single fore-leg preserved, wings partly spread, right hind wing creased, most of abdomen missing. Thorax: prothorax about 7.0 mm long, 4.0 mm wide; Right forewing: preserved length 35.1 mm, width 8.8 mm. RP simple for 14.9 mm, with 12 branches preserved; M poorly preserved, MA simple, MP with 3 branches reaching the posterior wing margin; CuA+CuPa incomplete, with 15 visible branches. Left forewing: apex missing, preserved length 32.3 mm, width 8.7 mm; M not visible in its median portion; a portion of CuPa basal to its fusion with CuA visible. Left hind wing: length 29.5 mm, width 10.0 mm; plicatum in resting position and creased; RP with 11 branches reaching apex; M with 8 terminal branches.
CNU-NX1-738 ( fig. S2C and D): Imprint of an individual with parts of prothorax and thorax preserved, a left forewing (as negative imprint) and a right hind wing (as positive imprint). Left forewing: length 31.3 mm, width 10.3 mm; RP simple for 15.2 mm, with 13 branches reaching wing apex; MA connected to RP by a very short cross-vein; M with a total 5 distal branches (MP simple); CuA+CuPa with 15 preserved terminal branches. Right hind wing: partly creased, plicatum not discernible/preserved, wing base not discernible; length 29.6 mm, width 7.2 mm; RP with 13 branches reaching wing apex.
CNU-NX1-759 ( fig. S3A and B): Imprint of a nearly complete individual, most of head missing, left forewing twisted, right hind wing concealed under right forewing, a mass of circular cavities probably indicates the location of abdominal remains. Thorax: prothorax about 6.4 mm long, 3.1 mm wide; Right forewing: apex missing, anal area not discernible; length 34.1 mm, best width 8.0 mm; RP simple for 13.7 mm, with 11 branches preserved; one reaching ScP+RA; MA fused with RP for 0.6 mm, with 2 terminal branches, MP with 2 branches; CuA+CuPa not fully discernible, with 16 terminal branches preserved. Right hind wing: anterior wing margin and plicatum not discernible; RP with 7 branches preserved; M with 5 branches (2,3), CuA+CuPa incomplete, with 4 branches. Legs: left legs almost missing, right fore-and mid-leg with femur and tibia preserved; right fore-leg, femur 5.7 mm long, tibia 5.3 mm long; right mid-leg, femur 6.4 mm long, tibia 8.4 mm long; right hind leg, femur 8.3 mm long, tibia 11.1 mm long, tarsus about 6.6 mm long, with 5 tarsal segments, claws and arolium visible; spines visible on one of the hind-leg tibiae.
CNU-NX1-731 ( fig. S3E and F): Positive and negative imprints of an almost complete individual, very well-preserved, legs and left hind wing missing; abdomen broken. Head: preserved length 6.7 mm. Thorax: prothorax about 4.7 mm long, 3.5 mm wide. Left forewing: length 39.6 mm, best width 9.1 mm; RP simple for 13.4 mm, with 10 branches reaching apex and one branch reaching with ScP+RA; M well preserved, connect with RP by a long, oblique cross-vein; MA with 3 branches, MP with 1 branch; CuA+CuPa with 17 branches reaching the posterior wing margin, CuPb poorly preserved. Right forewing: preserved length 35.9 mm, best width 9.2 mm; a vein interpretable as ScA partly preserved; RP simple for 13.1 mm, with 10 branches reaching apex and one veinlet reaching with ScP+RA; M well preserved, connected with RP by a long, oblique cross-vein; MA with 2 branches, MP with 2 branches; CuA+CuPa with 19 branches reaching the posterior wing margin; AA area with serval branches preserved. Right hind wing: apex missing; RP with 10 branches directed towards apex and 2 veinlets reaching with ScP+RA; MA and MP simple for a long distance, with 3 and 2 branches reaching the posterior wing margin, respectively; CuA+CuPa with 6 terminal branches preserved; plicatum folded, creased.
CNU-NX1-747 ( fig. S4A-D): Positive and negative imprints of an almost complete individual, left forewing and right hind leg missing. Head: about 6.0 mm long, 4.4 mm wide; md about 1.7/2.0 mm long, 1.4 mm wide at their base; apical tip of la visible, ct 0.9 mm long 0.2 mm wide; compound eye oval; circumoccular ridge well developed. Thorax: prothorax about 5.5 mm long, 3.3 mm wide. Right forewing: preserved length about 29 mm, best width 8.2 mm; RP simple for 12.3 mm, with 10 branches, two of them reaching ScP+RA; MA and MP with two branches each; CuA+CuPa with 19 terminal branches visible. Hind wings: plicatum folded, with numerous anal veins, not clearly discernible. Left hind wing: length 30.1 mm, best width 7.8 mm; RP simple for 9.6 mm, with 11 branches reaching wing apex and a single veinlet reaching ScP+RA; MA and MP simple for a long distance, each with 3 branches; CuA+CuPa posteriorly pectinate, with 6 terminal branches. Right hind wing: overlapping with right forewing, only partly discernible; RP simple for 8.3 mm, with 9 branches preserved. Legs: right legs poorly preserved and/or incomplete; left fore-leg femur 6.3 mm long and 1.0 mm wide, tibia 5.2 mm long and 0.6 mm wide; mid-leg femur 5.0 mm long and 1.1 mm wide, tibia 6.8 mm long and 0.6 mm wide; hind-leg femur 8.6 mm and 1.0 mm wide, tibia 9.8 mm long and 0.6 mm wide; spines well exposed on both foreleg tibiae, and the preserved mid-leg and hind-leg tibiae.
CNU Thorax: prothorax about 6.6 mm long, 3.9 mm wide. Right forewing: length 32.1 mm, best width 8.8 mm; RP simple for 12.1 mm, with 9 branches preserved; M poorly preserved, MA and MP with 3 and 2 branches, respectively; CuA+CuPa incomplete, with 14 branches preserved. Right hind wing: length 28.7 mm, best width 14.6 mm; RP simple for 14.6 mm, with 8 branches preserved; MA and MP simple for a long distance, M with 7 branches reaching the posterior wing margin; fusion of CuA (emerging from M+CuA) with CuPa visible; CuA+CuPa with 7 terminal branches; CuPb partly preserved; plicatum almost fully deployed, large, probably with vannal folds; AA with 9 branches preserved.
CNU-NX1-740 ( fig. S6C and D): Positive and negative imprints of an incomplete individual, with forewings and right hind wing poorly preserved, abdomen not discernible. Head: 7.5 mm long, 4.7 mm wide. Thorax: prothorax about 5.5 mm long, 4.5 mm wide. Left hind wing: apex missing; fusion of CuA (emerging from M+CuA) with CuPa visible; plicatum well deployed, large, with several veins preserved (attributable to AA). Legs: fore-leg femur length 5.5 mm long and 1.2 mm wide; mid-leg femur 5.2 mm long and 1.2 mm wide; hind-leg femur 11.5 mm long and 1.2 mm wide, tibiae 12.0 mm long and 0.8 mm wide, tarsus about 6.2 mm long, paired claws and arolium preserved.
CNU-NX1-199 ( fig. S6E and F): Positive and negative imprints of isolated right hind wing; wing base not discernible, apex missing; preserved length 17.4 mm, best width 10.1 mm; RP simple for 8.9 mm, with 6 branches preserved; M with 5 branches reaching the posterior wing margin; CuA+CuPa with 8 terminal branches; plicatum well-deployed, with 17 branches preserved (attributable to AA).
CNU-NX1-753 ( fig. S6G and H): Negative imprint of an isolated left hind wing, plicatum not discernible/preserved; length 34.2 mm, best width 10.1 mm; at the wing base, M+CuA distinct from R; RP simple for 9.5 mm, with 16 branches reaching wing apex; MA and MP simple for a long distance, with 3 and 4 branches, respectively; CuA+CuPa with 5 terminal branches preserved; plicatum with several visible veins (attributable to AA).
CNU-NX1-756 ( fig. S7F-H): Positive and negative imprints of an almost complete female individual, wings poorly preserved and incomplete, total length (excluding ant) about 51 mm. Head: 7.1 mm long, 3.6 mm wide; md open; left md with well-discernible il and mo; left la with a strong apical tooth and a smaller sub-apical one; co located in the midline along the dorsal side of the head capsule, then branching into two diverging fc; ant long, filiform; ce 1.4 mm long and 0.8 mm wide. Thorax: prothorax about 5.8 mm long, 4.0 mm wide. Abdomen: length about 23 mm, segments not discernible; exposed portion of ovipositor about 5.0 mm long. Compared with known species, the new one is mostly similar to Ct. elongatus, Ne. mazonus and Lom. udovichenkovi owing to the elongate to very elongate shape of the forewing (presumed in the latter). A further similarity of the new species with Ct. elongatus and Lom. udovichenkovi is the occurrence of numerous posterior basal veinlets of CuA+CuPa vanishing before reaching CuPb. Strikingly, the new species and Ct. elongatus share a very particular forewing coloration pattern, with three longitudinally-orientated, pigmented bands. We therefore propose to assign the new species to the genus Ctenoptilus Lameere, 1917.
Note that Béthoux & Nel (1) identified, in one specimen of Ct. elongatus, a linear structure they interpreted as MP, that would indicate a basal position of the first fork of M. However, based on data on the new species and on the original descriptions of Ne. mazonus and Lom. udovichenkovi, we assume that the 'linear structure' is more likely the median furrow alone. If so, the first fork of M, in Ct. elongatus, might well be located closer to the middle of the forewing, as in the new species and in Ne. mazonus and Lom. udovichenkovi. Note that this fork is located more basally in Lom. udovichenkovi than in the new species.

Ovipositor comparative analysis
This section complements schematic reconstructions provided in Fig. 2. Schemes representative of Grylloidea, Gryllacrididae and Anostostomatidae were derived from previous accounts (12-14).

Grylloblatta chandleri Kamp, 1963 (schematized under 'Grylloblattodea' in Fig. 2G)
Our observations corroborate previous accounts (15, 75), in particular regarding the occurrence of a long olis1 connecting gp9 and gp8. Its rh is slightly dejected externally. We also noticed the occurrence of an olistheter interlocking left and right gp9 along their dorsal margins. A specimen we observed had an egg engaged in the ovipositor. Due to the large diameter of the egg olis1 unlocked, as well as the dorsal gp9-gp9 olistheter. It can then be assumed that olistheters are comparatively labile structures in the species. In resting position (i.e. without engaged egg), when viewed externally, the ventral part of gp9 is not concealed by gp9. Most of the area of gp9 concealed by gs9 is not as strongly sclerotized as its ventral part, except for the very base and its dorsal, ventral and apical margins.

2.2.2
Anacridium aegyptium (Linnaeus, 1764) (schematized under 'Caelifera' in Fig.  2G) Our observations corroborate previous accounts on other caeliferan species reporting the occurrence of an olis1 connecting gp9 and gp8 along the entire ventral edge of the former (14, 16). Unlike reported by Ander (57), we found no evidence of an olistheter interlocking the 'inner' (i.e. gp9) and 'posterior' (i.e. gs9) valves (i.e. olis2). The gp8 and Ander's (57) 'lateral basivalvular sclerite' are extensively fused: they share the same lumen, and the dorsal and ventral fusion points are conspicuous in cross-section, owing to a clear invagination, coupled to a substantial and well-delimited thickening, of their shared wall.

2.2.3
Ceuthophilus sp. (Fig. 2E and F; schematized under 'Rhaphidophoridae' in Fig.  2G) We concur with previous accounts reporting that olis2 occurs in this genus and in other Rhaphidophoridae (12, 14, 76). Unlike other orthopterans, the rh of olis2 is a short projection directed posteriorly, while its al covers a broader range (as it is, the antero-ventral half of gp9). Viewed laterally, the al of olis2 is slightly convex. This configuration possibly provides some degree of rotational freedom to gs9 vs. gp9 & gp8 (interlocked by olis1, which extends more posteriorly than olis2, including its rh), using the rh of olis2 as a slightly movable axis. This supposed ability would allow gp8 postero-ventral teeth to be exposed (instead of concealed by gs9) and then used by the insect to appreciate the adequacy of substrate for oviposition. The gp8 is only partially concealed by gs9.

2.2.4
Tettigonia viridissima (Linnaeus, 1758) (schematized under 'Tettigoniidae' in Fig.  2G) The observed configuration of the ovipositor valves conforms that described by Cappe de Baillon (12). Unlike assumed by Kluge (14; among others) we argue that the olistheter interlocking gs9 and gp8 (thereafter olis3) is not homologous with olis2. Firstly, a protrusion from gs9 and directed towards gp9 (viz., the characteristic features of olis2) occurs at various levels along the ovipositor. It is clearly distinct from another well-delimited olistheter (viz. olis3). Secondly, as stated by Kluge (14), the Anostostomatidae possibly represent an 'intermediate' stage is which a well-delimited olis2 co-occurs with the premises of an olis3, in the shape of a projection of the ventral margin of gs9 into gp8. If two olistheters occur (in addition to olis1), they cannot be homologous. It follows that there is an olis3 besides olis2.

2.3
Analysis of the mandibular mechanical advantage Progression of mechanical advantage curves for the studied taxa are represented in fig. S9. Results of the PCA are summarized in tab. S3 and represented in fig. S10, including the phylogenetic PCA. Animated versions of the PCA represented in Fig. 3E are provided in the associated Dryad dataset (58).  Figure S1. Workflow for the extraction of the mandibular mechanical advantage based on 3D models.