X-ray microtomography and phylogenomics provide insights into the morphology and evolution of an enigmatic Mesozoic insect larva

. Fossils sometimes show unusual morphological features absent in living organisms, making it difficult to reconstruct both their affinity and their function. We describe here a new lacewing larva, Ankyloleon caudatus gen. et sp.n. (Neuroptera) from the Cretaceous amber of Myanmar, characterized by an abdomen unique among insects, with ‘tail-like’ terminal segments bearing a ventral pair of vesicles. Phase-contrast X-ray microtomography reveals that these structures were dense and equipped with a median duct, suggesting that they were likely pygopods used for locomotion, holding the position through adhesive secretions. Our phylogenetic analyses, combining genomic and morphological data from both living and fossil lacewings, proved critical to placing Ankyloleon gen.n. on the lacewing tree of life as an early representative of the antlion clade, Myrmeleontiformia. These results corroborate the view that derived myrmeleontiform lacewings ‘experimented’ with unusual combinations of features and specializations during their evolutionary history, some of which are now lost.


Introduction
There is general agreement among scientists that winged insects (Pterygota) spawned the most spectacular radiation of metazoans on Earth (Misof et al., 2014;Haug J.T. et al., 2015). So far, more than one million extant pterygotes have been larvae) are more recent, dating back to the Moscovian stage (Carboniferous 315-303 Ma) (Nel et al., 2013;Haug J.T. et al., 2015). Holometabolans evolved a modular separation between postembryonic developmental (larva) and reproductive (adult) stages through a dramatic change of larval anatomy that takes place during an intermediate pupal stage. The evolution of this strategy exposed larvae and adults to different selective pressures, allowing a differential, temporal and compartmentalized exploitation of resources and thus eliminating intraspecific competition between life stages (Truman & Riddiford, 1999;Yang, 2001;Beutel et al., 2014;Rainford et al., 2014;Rolff et al., 2019;Truman, 2019). Larvae evolved into voracious feeders as their body developed to store the energy needed for metamorphosis and an adult life. In particular, the abdomen-the tagma most involved in energy storage-increased in size and weight, often becoming involved in locomotion unlike that of adults. To our knowledge, such a novel function of the larval abdomen has never been exploited by nonholometabolans, nor by any nonpaedomorphic, adult pterygotes. Neuroptera (lacewings) are a monophyletic group of holometabolans displaying marked disparity in larval life strategies and morphology, despite their relatively poor extant species diversity (ca. 6000 described species) Oswald & Machado, 2018). Neuroptera, together with the allied groups Raphidioptera (snakeflies) and Megaloptera (alderflies and dobsonflies), form the clade Neuropterida, which represents the sister group to Coleopteroidea (Strepsiptera and Coleoptera) Vasilikopoulos et al., 2020). The Cretaceous witnessed an explosive phase of evolutionary experimentation in lacewings. In that period, now-extinct larval phenotypes coexisted with forms that were almost identical to those living today, already displaying complex behaviours like camouflaging and mimesis (Pérez de la Fuente et al., 2012Fuente et al., , 2016Fuente et al., , 2018aFuente et al., , b, 2020Liu et al., 2016Liu et al., , 2018Wang et al., 2016;Badano et al., 2018;Haug J.T. et al., 2018, Haug J.T. et al., 2019a, 2019bHaug G.T. et al., 2020;Haug J.T. et al., 2020). Lacewing larvae and adults appear especially diverse and abundant in the Late Cretaceous 'Burmese' amber from Myanmar (ca. 99 million years) hinting at a possible prominent ecological role of this group at that time (Wang et al., 2016;Badano et al., 2018;Lu & Liu, 2021). Recent findings have shown that much of now-extinct lacewing morphologies are likely yet to be discovered and, indeed, the full depth of their evolutionary experimentation is not yet understood. Here, we describe a novel larval body organization from the Myanmar amber deposit, which, to our knowledge, has no analogue among living insects. Examined specimens, although sharing all the autapomorphic features that support the monophyly of Neuroptera, such as the modification of mandible and maxilla in a sucking apparatus and lack of maxillary palp (Beutel et al., 2010), drastically depart from the typical organization of lacewing larvae by having the second half of the abdomen abruptly narrowing, elongated and ending in a spatula-like structure equipped with relatively large ventral vesicles. Overall, the abdomen is reminiscent of the division in meso-and metasoma characterizing scorpions, although in this case the function is completely unknown. These modifications are not only unique throughout all known fossil and living insects but also among all hexapods, suggesting highly specialized habits. We examined the morphology of this fossil through synchrotron X-ray phase-contrast microtomography (XPCT) to explore in detail fine, internal anatomical features and hypothesize which selective forces likely shaped this extreme shape within the broader context of lacewing evolution. XPCT is a nondestructive 3D imaging technique endowed with higher image contrast and spatial resolution than conventional tomography, which also detects minor electronic density variations due to sensitivity to phase shift. This technique appears particularly helpful when applied to fossils, yielding detailed information, especially if combined with conventional micro-CT (Tafforeau et al., 2006;Lak et al., 2008;Soriano et al., 2010;Perreau & Tafforeau, 2011;Bidola et al., 2015). Moreover, we tested the affinities of fossil larvae with living species by combining morphological and molecular data. Our findings suggest that (i) complex anatomical structures evolved as dead-end innovations during a phase of morphological diversification and (ii) the integration of genomic data with morphological information allows for the determination of the affinities of enigmatic stem-group phenotypes.

Origin of specimens and depository
The examined fossil materials were extracted from the amber outcrops of northern Myanmar, located in Kachin Province, ca. 100 km west of the town of Myitkyina. These deposits are dated at the very beginning of the Late Cretaceous near the Aptian/Cenomanian border (Shi et al., 2012). However, the discovery of other outcrops in the region, dating back from the same stage but slightly younger (Zheng et al., 2018), makes it difficult to pinpoint with certainty the fossil provenance. As far as the authors can ascertain, the amber pieces were obtained legally and before the upheaval and humanitarian crisis in the area. The holotype of Ankyloleon caudatus sp.n. (MZURPAL00111) and Ankyloleon specimen 1 (MZURPAL00112) are deposited in the public collection of the Zoological Museum of Sapienza University of Rome, Italy; Ankyloleon specimen 2 (BUB3) is deposited in T. Weiterschan private collection, Höchst Odw., Germany; Ankyloleon specimen 3 (PED_0118) and Ankyloleon specimen 4 (PED_0454) are deposited in the public collection of the Ludwig-Maximilians-Universität, Munich, Germany.

Imaging
Specimens were examined and measured with a Leica© (Wetzlar, Germany) MZ 9.5 stereomicroscope equipped with an ocular micrometre. Photos were taken with a Canon© (Tokyo, Japan) EOS 600D digital camera with Canon© lens MP-E 65 mm. The resulting images were focus-stacked with the software Zerene© (Richland, WA, U.S.A.).

X-ray phase-contrast microtomography
The experiments were carried out at the TOMCAT beamline of the Swiss Light Source (Villigen, Switzerland). The incident monochromatic X-ray energy was of 20 keV. A PCO edge 5.5 camera coupled with optics resulting in a pixel size of 1.625 × 1.625 μm 2 and 0.32 × 0.32 μm 2 was set at a distance from the sample of 3 and 5 cm, respectively. The tomographic images were acquired with an exposure time of 90 ms for 1.625 × 1.625 μm 2 and exposure time of 220 ms for 0.32 × 0.32 μm 2 covering a total angle range of 360 ∘ , using the so-called half-acquisition mode, which allows to almost double the image field of view. Then, the tomographic projections were reconstructed on site by means of ad hoc software based on the Paganin's phase retrieval algorithm. Image processing and 3D rendering were made with the software ImageJ (https://imagej .net/Fiji) and 3D slicer (https://www.slicer.org/) and Amira.

Scanning electron microscopy
Extant lacewing larvae preserved in 70% ethanol were dehydrated in a graded ethanol series to 100% ethanol, critical point dried in a CPD 030 unit (BalTec©, Balzers, Liechtenstein) and secured to aluminium stubs using conductive double adhesive carbon discs. Subsequently, the samples were sputter-coated with a thin layer (30 nm) of gold, using a K550 sputter coater (Emithech©, Kent, UK) and examined with a Zeiss© Gemini Sigma 300 (Jena, Germany) field-emission scanning electron microscope, with an Everhart-Thornley© secondary electron detector and an acceleration tension of 5 kV, in the Laboratorio Interdipartimentale di Microscopia Elettronica, Roma Tre University, Rome, Italy.

Phylogenetic analyses
The genomic dataset is an excerpt of the nucleotide alignment of Winterton et al. (2018) including 57 taxa-i.e., pruned of those with unavailable information on larval morphology-and 57546 bp. The morphological datasets were assembled in Mesquite integrating recently published matrices (Beutel et al., 2010;Badano et al., 2017;Jandausch et al., 2019) for extant species and (Badano et al., 2018) for fossils. The two morphological matrices included 137 characters and are identical except in the number of taxa. The matrix of extant species included 57 taxa, whereas the set encompassing fossils comprised 76 taxa.
When information about larval morphology was not available for terminal taxa, morphological characters were scored from congeners.
The partitioned maximum likelihood (ML) analyses, combining the genomic and the molecular datasets, were performed in IQ-TREE (v.2.1.1) (Minh et al., 2020) on Extreme Science and Engineering Discovery Environment (XSEDE) at Cyberinfrastructure for Phylogenetic Research (CIPRES) (Miller et al., 2010). Models were computed separately for the two datasets as different partitions, enforcing the Model Finder option implemented in IQ-TREE (Chernomor et al., 2016;Kalyaanamoorthy et al., 2017). The mixed partitions had different speeds and applied 50 independent tree searches. Node supports were estimated using ultrafast bootstrap (Hoang et al., 2018) performing 10 000 replicates (Fig. S4).
The Evolutionary Placement Algorithm (EPA) was developed (Berger & Stamatakis, 2010) and implemented in RAxML (Stamatakis, 2014). EPA integrates molecular and morphological datasets, placing taxa for which only morphological data are available in a molecular tree in a two-step process. We ran all the analyses using RAxML-HPC2, version 8, on the XSEDE at CIPRES Science Gateway (Miller et al., 2010). First, we generated a phylogenomic tree enforcing a GTR+I+Γ substitution model and included a rapid bootstrap search of 100 replicates starting at a random seed number. Then, we supplied the obtained molecular tree and the morphological dataset of extant taxa to calibrate weights for each morphological character, based on congruence between the molecular and morphological inputs. In the following step, we provided a morphological dataset inclusive of fossils, allowing EPA to estimate the likelihoods of fossil placement in the genomic tree through the morphological weights ( Fig. S5).
Maximum parsimony analyses of the morphological dataset were performed with TNT, version 1.5 (Goloboff & Catalano, 2016), with the "traditional search option" enforcing the following settings: general RAM of 1 GB, memory set to hold 1 000 000 trees, setting 1000 replicates with tree bisection reconnection and saving 1000 trees per replicate; multistate characters were treated as unordered. Character state changes (apomorphies) were optimized in WinClada, version 1.00.08 (Nixon, 2002) (Figs S6, S7).

Taxonomy
Ankyloleon Badano, Haug & Cerretti gen.n. Zoobank registration: lsid:zoobank.org:act:DB189598-97C1-4684-AE3A-0548CCD80F03 Etymology. The generic epithet is a masculine name from Greek, composed of a prefix referring to the dinosaur genus Ankylosaurus Brown (Ornithischia, Ankylosauridae) based on the shape of abdominal segment 10 of the larva, which resembles the caudal club of the famous dinosaur, but also on ankylos 'bent, crooked' after the apical bending of jaws; the Greek suffix -leon means 'lion' and is commonly used in antlion genera. Diagnosis. Campodeiform, elongate lacewing larva; head capsule sclerotized; stemmata on distinct tubercle; antenna short and three segmented; mandibular-maxillary stylets elongate, straight, curved inwards only at apex, without teeth; prothorax elongated, tubular; abdomen subdivided in an anterior section with six short segments and a posterior narrower section (Figs 1,  2).
Etymology. The specific name is an adjective derived from the Latin cauda and meaning 'tailed', after the tail-like section of abdomen.
Description. Body length 3.28 mm. Head capsule longer (length 0.58 mm) than wide (width 0.44 mm), globose in lateral view but tapering anteriorly; ventral surface flat (Fig 1A, B). Ocular region raised on tubercle, bearing six stemmata (five dorsal, one ventral) and with apical trichoid sensillum, twice as long as tubercle height and borne on elevated base (Fig. 1A). Antenna short, 1/5 the length of mandible, tapering apically (Fig. 1A). Jaws (mandibular-maxillary stylets) two times longer than head capsule (mandible length 0.96 mm), straight, curved inwards at apex, almost in contact at insertion; mandible wider at base, internal margin with minute serrations, without teeth, external margin with long trichoid sensilla (Fig. 1A, B). Labial palp absent. Posterior section of head progressively tapering in a neck-like region (Fig. 1A). Dorsal side of head with long trichoid sensilla orderly arranged in rows and raised from circular alveoli. Cervix short, apparently membranous.
Abdominal segments 1-6 not differentiated from thorax, relatively broad, not sclerotized and covered with long setae (Fig. 1A, B). Segment 7 as a short truncate cone. Segments 8-10 strongly modified into a narrow, elongate caudal section almost as long as the rest of the body. Segment 8 cylindrical, two times longer than wide. Segment 10 spatula-shaped, proximally invaginated within the cuticle of segment 9 (Figs 1A, B, 3B). Ventral side of segment 10 with a pair of prominent, symmetrical vesicle-like structures with a central pore prolonged internally into a tube-like channel (Fig. 1B, 3C). Segments 8-10 covered with thin hair-like setae, longer on body outline; terminal "spatula" with long thin sensilla along lateral margin (Fig. 1A, B).
Comments. These four fossil larvae can be assigned to Ankyloleon, though their state of preservation or completeness does not allow for the study of several key morphological features (Figs 2, S1, S2, S3). Thus, it is not possible to fully compare them with A. caudatus sp.n., only allowing us to place them to genus. However, these specimens share the apomorphic characteristics of Ankyloleon such as long jaws only bent at apex, antenna short, tubular pronotum and abdominal segments 9 and 10 differentiated from the rest of the abdomen as a narrow posterior section. These larvae are characterized by shorter, less differentiated segments 9-10 with respect to A. caudatus and by the absence of a distinct spatula. Among them, the best-preserved specimen, that is, Ankyloleon specimen 1, is also provided with paired sclerotized pygopods on abdominal segment 10, demonstrating that this character was not exclusive of A. caudatus nor a fossilization artefact (Fig. S1). However, this specimen differs from A. caudatus in chaetotaxy, as the setae covering the body are relatively short and stout (hair-like in A. caudatus) (Fig. 2; File S1). The similar size of the specimens suggests that these differences were not ontogenetic, so these larvae likely belonged to different species.
Morphological remarks. Ankyloleon shares the following combination of character states with Myrmeleontiformia: (i) ocular tubercles, (ii) temple region, (iii) neck region, (iv) posterior tentorial grooves on the anterior portion of the head capsule, (v) maxilla half the size of mandible (Badano et al., 2018), but lack of dolichasterine setae. Interestingly, maximum parsimony analysis of the morphological dataset reconstructed Ankyloleon within Myrmeleontiformia (Figs S6, S7) based on a series of inferred character state changes (59:1, premental elements widely separated; 68:1, gular region with a small anterior triangular sclerite; 69:1, hypostomal bridge present; 72:2, posterior tentorial grooves on anterior portion of head capsule), which are not observable in our specimens, due to bad state of preservation or amber impurities (Figs 1, 2). The jaws of Ankyloleon recall those of Nevrorthidae, but its head capsule is greatly different. Ankyloleon lacks labial palpus like the larvae of spongillaflies (Sisyridae), and the long sensilla on its head resemble those of mantispoid larvae (Berothidae, Rhachiberothidae, Mantispidae). This combination of features, along with its outstanding autapomorphic abdominal shape, makes this taxon not readily assignable to any recognized family-ranked group. The homology assessment of the abdominal segmentation in Ankyloleon was determined by comparing A. caudatus with the other larvae through the arrangement of the abdominal spiracles (File S1). Boundaries of abdominal segments are difficult to identify in larvae of Myrmeleontiformia as their cuticle is soft, wrinkled and almost annulate, lacking tergal and sternal sclerites. Immature Myrmeleontiformia have a pair of spiracles on abdominal segments 1-8. Ankyloleon instead bears one pair of spiracles only on abdominal segments 1-6; this is likely a consequence of the great morphological changes segments 7 and 8 have undergone. There are no morphological cues help determine at which instar the examined larvae of Ankyloleon belong. Interestingly however, Ankyloleon specimen 3 is an exuvia, which means that it could more likely represent a first or second instar because third (last) instars encase themselves in a silky cocoon before pupation and leave their exuvial remains within it.

Phase-contrast XPCT
XPCT images (Fig. 3, Movie S1, Movie S2) show that the state of preservation of the cuticle of the anterior portion of the holotype, including head, part of jaws, thorax and appendages did not provide enough contrast to reconstruct an accurate 3D outline. The tracheal respiratory system is rendered in some detail, showing the longitudinal branches of the anterior tracheal system extending from thorax to head (Fig. 3A). Abdominal segment 10 is rendered as a wide, relatively thick, dorsoventrally depressed, structure; it is roughly drop-shaped in dorsal view, narrowing at the insertion with segment 9 (Fig. 3B, D, E). The paired ventral vesicles are dense, thick, ovoid structures, clearly emerging from the body outline in lateral view (Fig. 3B-D). The XPCT cross-sections of vesicles show a perpendicular duct (Fig. 3C), which is externally visible as a median pore (Fig. 3B-E).

Phylogenetic analysis
We explored the phylogenetic placement of Ankyloleon and a selection of larvae of key extinct taxa on the lacewing tree of life, using two maximum likelihood-based optimizations (ML): (i) running a combined (or total evidence) phylogenetic analysis of genomic and morphological data (Figs 4, S4) and (ii) employing the EPA on a ML tree obtained with genomic data only (Berger & Stamatakis, 2010;Berger et al., 2011) (Figs 4, S5). Both approaches yielded trees that were largely consistent with that of Winterton et al. (2018) and largely agreed on the placement of fossil taxa (Fig. 4). Cretaceous taxa Pedanoptera Liu et al. and Hallucinochrysa Pérez de la Fuente et al. were recovered as sister to all extant chrysopids in the combined analysis. EPA instead recovered Hallucinochrysa as a crown-group Chrysopidae. Macleodiella Badano & Engel was consistently reconstructed sister to all the remaining Myrmeleontiformia. The position of Ankyloleon changed depending on the analysis but its affinities with Myrmeleontiformia were well supported under both methods (see also Figs S6, S7). The combined ML analysis recovered Ankyloleon nested within Myrmeleontiformia, as a member of the monophyletic group including Nymphidae (split-footed lacewings) and Ithonidae (moth lacewings) (Fig. 4). EPA instead found Ankyloleon as sister to Myrmeleontiformia, except Macleodiella (Fig. 4)

Discussion
The unique shape and puzzling features of Ankyloleon reinvigorate the question of how far lacewings explored the multidimensional ecological spaces of the Mesozoic. These features are so unique that make these specimens enigmatic from both phylogenetic and behavioural point of view.

An unusual abdomen
Our analyses show that Ankyloleon is phylogenetically related to Myrmeleontiformia, although none of the extant and extinct representatives of this large and diverse group share such a highly derived abdomen. However, larvae of lacewings and allies are often provided with pygopods or anal prolegs (Snodgrass, 1935); that is, soft and deformable protuberances at the end of the abdomen (Fig. 5) that improve contact and attachment to the substrate through a complex combination of van der Waals forces, capillary forces and friction (Zurek et al., 2015). Pygopods often contribute to locomotion by acting as pivot points or grasping structures. Aquatic larvae of dobsonflies and fishflies (Megaloptera: Corydalidae), living in fast flowing rivers and creeks, evolved pygopods with terminal hooks or suckers to anchor themselves to the substrate. Conversely, terrestrial larvae of snakeflies (Raphidioptera) and of most lacewings-e.g., brown and green lacewings (Hemerobiidae and Chrysopidae), mantis flies (Mantispidae) and silky lacewings (Psychopsidae)-are provided with adhesive pygopods (Fig 5A, C-E). Lance lacewing larvae (Osmylidae), which might be terrestrial or amphibious, have paired soft, eversible structures with series of ventral hooks, forming a complex grasping apparatus (Fig. 5B). Finally, the larvae of antlions and owlflies (Myrmeleontidae) and thread-and spoon-winged lacewings (Nemopteridae) use abdominal segments 8 and 9 for digging and anchoring, usually by means of highly modified setae ( Fig. 5F). Although no other lacewings share a similar morphology, beetles may offer insights into the role of this combination of features. Peculiar configurations of the tip of the abdomen in beetle larvae can be associated with predatory habits or with defensive purposes. The elongated abdominal segments 8-10 and the presence of apparently glandular structures on segment 10 would suggest that Ankyloleon was able to raise up and curve the 'tail' over the body, perhaps secreting repulsive or attractive/appeasing substances, as do several rove beetles (Coleoptera: Staphylinidae) and flanged bombardier beetles (Coleoptera: Carabidae, Paussinae) (Di Giulio & Vigna Taglianti, 2001;Parker, 2016). Texas beetle larvae (Coleoptera: Brachypsectridae), which resemble owlfly larvae in shape and behavior, instead use the spine-like abdomen tip to stab prey (Crowson, 1973;Costa et al., 2006), suggesting that the abdomen of Ankyloleon might be potentially involved in predation, bringing the prey in contact with the mouthparts. The shape of abdominal segment 10 of Ankyloleon also recalls the condition of false flower beetle larvae (Coleoptera: Scraptiidae) (Švácha, 1995;. However, the caudal structure of scraptiids is accessory, as it can be shed off and then regenerated, whereas in Ankyloleon, segment 10 is an integral part of the abdomen, as evidenced by the presence of vesicles (Figs 1B, 3B-F). XPCT reconstructions suggest that the ventral vesicles were pygopod-like structures, perhaps associated with glands as found in some beetles where pygopods produce adhesive substances (Zurek et al., 2015) (Fig 3D-F). A secretive function is supported by the presence of a median duct and external pore (Fig 3C). A flexible abdomen, provided with prominent pygopods, might have proven useful to move on smooth surfaces, at the same time maintaining a grip through adhesive secretions. However, as often happens, a complex anatomical structure may have more than one role, as well as hidden functions that cannot be readily inferred from morphology without extant analogues.
What does the head say about the natural history of Ankyloleon?
As with bird beaks, jaws of lacewing larvae mirror alimentary specializations and life habits. Nearly all extant lacewing larvae (perhaps with few exceptions) are predators (Oswald & Machado, 2018) and Ankyloleon, being characterized by elongated, straight and apically hooked jaws was not an exception (Figs 1, 2, 3A). The articulation of the mandibular-maxillary stylets with the head shows that these larvae were not able to widely open their jaws as owlflies or antlions do in a trap-like fashion (Badano et al., 2017). This seeming disadvantage was likely counterbalanced by a highly mobile head-thorax articulation and elongate prothorax (Figs 1, 2), like the extinct Macleodiella and extant nevrorthids and thread lacewing (Crocinae) larvae (Beutel et al., 2010;Badano et al., 2018). Sense organs of Ankyloleon can also provide information on its life history. Both eyes (stemmata) and antennae were reduced, whereas labial palps were absent, suggesting that visual and chemical stimuli were likely not relevant for these larvae (Figs 1, 2, 3A). Ankyloleon was equipped with a diverse array of trichoid sensilla, especially on the head and segment 10 ( Fig. 1), suggesting these mechanoreceptors were the main sense organs, as in living antlions (Devetak et al., 2007;Podlesnik et al., 2019). The thin hair-like sensilla on the terminal club had likely a similar tactile function, allowing larvae to sense the surroundings and to navigate within dark and narrow spaces.

Palaeoecology
Arthropod assemblages preserved in Cenozoic ambers originating from angiosperms are known to be biased towards species living or climbing on the resin-producing trees (Solórzano Kraemer et al., 2015Kraemer et al., , 2018. In contrast, it is still debated whether arthropod amber inclusions from the Mesozoic provide an accurate rendition of a whole forest insect paleofauna (Peris et al., 2016). Nonetheless, bark dwelling neuropteroid larvae appear particularly well represented in Burmese amber, including snakeflies, silky lacewings and Myrmeleontiformia (Badano et al., 2018;Haug G.T. et al., 2020;Haug J.T. et al., 2020), implying that most of them were entrapped in resin in situ. Phylogenetic and fossil evidence also suggest that the oldest representatives of Myrmeleontiformia were arboreal (Oswald & Machado, 2018;Badano et al., 2018). All evidence hints at Ankyloleon as an arboreal thigmotactic predator, using the terminal abdominal segments 7-10 as a pivot point. The shape of head and jaws suggest that Ankyloleon struck prey from afar, probably tiny insects with soft cuticle.

Fossil phylogenetic signal
The inclusion of fossil taxa in phylogenetic reconstructions provides information on character polarity, trait evolution and valuable calibration points for clade divergence time estimations (Ware & Barden, 2016). Nevertheless, the phylogenetic relationships of fossil taxa are often hypothesized based on non-numerical, morphological approaches, which are only rarely tested using explicit, quantitative methods. In the case of lacewings, the long-lasting concept that larval morphology is generally a reliable indicator of higher-level phylogeny (Aspöck et al., 2001;Badano et al., 2017; has been recently challenged by phylogenomic analyses, which highlighted that the evolution of morphological traits is likely more complex than predicted by cladistic methods Vasilikopoulos et al., 2020). Therefore, we integrated morphological and genomic data to make the phylogenetic placement of fossil taxa explicit and replicable. Our ML analysis of the combined dataset reconstructed the phylogenetic position of fossil taxa largely consistent with Badano et al. (2018) (Fig. 4). The main difference from the cladistic reconstruction of Badano et al. (2018) involves the affinities of Cladofer, which was not retrieved sister to Macleodiella (see also Figs S6, S7) but deeply nested within Myrmeleontiformia (Fig. 4). Moreover, EPA recovered Hallucinochrysa and Pristinofossor as crown-group Chrysopidae and Myrmeleontidae (Fig. S5), respectively, whereas the combined analysis agreed with Badano et al. (2018) by yielding Hallucinochrysa sister to Pedanoptera and Pristinofossor as a stem-group myrmeleontid (Figs 4, S4). The position of Ankyloleon varies according to the method employed. Despite these differing results, both approaches agreed in placing Ankyloleon within Myrmeleotiformia, even though this taxon does not share all the apomorphies supporting monophyly of the crown group. The state of preservation, the amount of available information and the unusual combination of characters likely account for the variable phylogenetic reconstructions of Ankyloleon. Myrmeleontiformia diverged from other Neuroptera in the Late Triassic, whereas the major subgroups diversified in the Jurassic or Early Cretaceous Vasilikopoulos et al., 2020). Our analyses confirm that all major extant lineages of Myrmeleontiformia were already present at the beginning of the Late Cretaceous and coexisted with stem-group relicts from the first diversification wave of the clade. The fossil placement also corroborates that early representatives of Myrmeleontiformia, including Ankyloleon and Macleodiella, were characterized by small-sized elongated larvae that likely lived in arboreal settings. Our results show that integrating morphological and molecular evidence permits the reconstruction of the phylogenetic placement of extinct body organizations, enabling the testing of hypotheses about insect evolution.

Conclusions
The fossil record continues to deliver new information about insect evolution, revealing that our current understanding of their impressive diversity and disparity is largely distorted by extinction events. Fossils also show that the radiation of insects was much wider than present-day diversity would suggest. Our study supports the view that the inclusion of extinct forms into phylogenetic analysis often has a deep impact on the relationships among living organisms and is critical to our understanding of the origin of certain morphological and behavioural features (Mongiardino Koch & Parry, 2020). The picture rendered by phylogenetic trees including fossils suggests that extinction often acts as a tight sieve that only allows a few lineages through. In this evolutionary process, survivors further diversify, evolving new features that might eventually disappear in favour of new ones.

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article.       File S1. Data matrix for phylogenetic analysis.
Movie S2. Dorsoventral view 3D reconstruction of the head of Ankyloleon caudatus holotype. (Sapienza University of Rome) for the critical review of the manuscript. We acknowledge Andrea Basso (Istituto Zooprofilattico Sperimentale delle Venezie) for his suggestions on the analyses. We thank Graham Montgomery for his photograph of the osmylid larva. Finally, we thank two anonymous reviewers for helpful comments and valuable suggestions on the manuscript. Davide Badano was supported by a SAPIExcellence BE-FOR-ERC fellowship (Sapienza University of Rome), Project "Tempo and Mode of Lacewing Evolution". Joachim T. Haug was kindly funded by the Volkswagen Foundation via a Lichtenberg professorship. The study was supported by the German Research Foundation (DFG HA 6300/6-1). The authors declare no conflict of interest.

Author contributions
Davide Badano and Pierfilippo Cerretti conceived and designed the study. Joachim T. Haug, Thomas Weiterschan and Jürgen Velten provided the materials. Davide Badano, Pierfilippo Cerretti and Joachim T. Haug studied, photographed and described the material. Michela Fratini, Laura Maugeri, Francesca Palermo, Nicola Pieroni and Alessia Cedola performed the experiment using XPCT and designed digital reconstructions and animations. Andrea Di Giulio examined the materials at scanning electron microscope. Davide Badano and Pierfilippo Cerretti assembled the dataset and analysed the data. Maurizio Mei contributed to logistics and materials tools and made the line drawings. Davide Badano and Pierfilippo Cerretti wrote the paper and made the figures. All authors edited and checked the manuscript.

Data availability statement
The data that supports the findings of this study are available in the supplementary material of this article