Moss mimesis par excellence: integrating previous and new data on the life history and larval ecomorphology of long-bodied craneflies (Diptera: Cylindrotomidae: Cylindrotominae)

Different physical structures play a central role in animal camouflage. However, in evolutionary studies of mimicry, the ecological and evolutionary significance of such structures has been poorly investigated. Larvae of long-bodied craneflies, Cylindrotominae, are all obligate herbivores and resemble plants. They are distinctively characterized by possessing numerous elongated cuticular lobes on the integument. A comprehensive overview of the biology and morphology of cylindrotomids, particularly their larval stages, is laid out, providing original data on nine species. To explore the ecological background of moss resemblance, host-plants of most examined species are clarified, revealing that terrestrial moss-feeding species tend to use specific groups of mosses, either belonging to Bryales or Hypnales. However, the evolution of cryptic forms remains paradoxical, due to the apparent absence of visual predators. Based on histological examinations, extensive internal musculatures within the cuticular lobes on the lateral side are discovered, shedding new light on their function in locomotion. Traditional functional explanations for these lobes, particularly as devices for respiration, locomotion and attachment, are challenged. This study promotes our understanding of the ecomorphology of mimicry devices, which is an angle often dismissed in evolutionary studies of mimicry.


INTRODUCTION
Camouflage is a fascinating example of adaptive evolution (Darwin, 1859;Wallace, 1867). A wide range of animals camouflage in their surroundings to avoid detection or deceive keen-sighted natural enemies in their vicinity (Stevens & Merilaita, 2009;Ruxton et al., 2018). Many insects, typically of the orders Diptera, Hymenoptera, Lepidoptera, Mantodea, Orthoptera and Phasmatodea, disguise themselves as plants. The cryptic forms of insects that resemble plants have presumably evolved as a survival strategy for a long time, with evidence in the fossil record extending from the Middle Permian to the Recent (Scott et al., 1992;Wedmann et al., 2007;Wang et al., 2010Wang et al., , 2012Wedmann, 2010;Garrouste et al., 2016;Liu et al., 2018). Coloration and patterning play a key role in various modes of mimicry and have historically been studied from ecological and evolutionary perspectives (Poulton, 1890;Cott, 1940;Protas & Patel, 2008;Cuthill et al., 2017). Moreover, visual camouflage is also associated with physical structures (Robbins, 1980;Sabaj et al., 1999), behaviours (Kang et al., 2015;Suzuki & Sakurai, 2015) and the active uptake of materials from the environment by the camouflaged organisms (Ruxton & Stevens, 2015). One famous example of such morphological structures is that of some lycaenid and riodinid butterflies, which bear narrow tails mimicking antennae on their hind wings -a so-called 'false head', which acts by deflecting the attention of birds (Robbins, 1980(Robbins, , 1981. Less obviously, larvae of the erebid moth genus Homodes Guenée, 1852 possess an anterior end with abundant long, bent, tentacle-like, spatulate setae, which apparently work by mimicking aggressive ants (Kalshoven, 1961;Leong & D'Rozario, 2012). A myriad of alluring examples of camouflage, in which physical structures are combined with coloration, are familiar to most people (Wickler, 1968;Unno, 2015). Nevertheless, the function and evolutionary significance of such structures has, generally, not been tackled in the study of the evolutionary ecology of mimicry.
Cylindrotominae is a subfamily belonging to the smallest cranefly family, Cylindrotomidae. Larval morphology of Cylindrotominae, known for more than two centuries (De Geer, 1773), deviates greatly from typical cranefly larvae -they resemble plants to a remarkable degree. Their extraordinary appearance is likely associated with their biological features as freeliving, obligate herbivores. Whereas the majority of cranefly species live in organic mud along stream banks or in rotten wood, and feed on decaying plant materials (Alexander, 1915;De Jong et al., 2008), all known larvae of the cylindrotomines live exposed on angiosperms or within the tufts of mosses and feed on living tissue of plants. Notably, the habitats vary among major clades, ranging from aquatic (Phalacrocera Schiner, 1863) to semi-aquatic (Triogma Schiner, 1863) to fully terrestrial (Cylindrotoma Macquart, 1834, Diogma Edwards, 1938, Liogma Osten Sacken, 1869. Larval cylindrotomines apparently attain camouflage among plants by several means. Many species are tinged with green or brown hues and are dorsally decorated with dark markings. In addition, elongated cuticular appendages longitudinally aligned on the integument are particularly noteworthy in mossfeeding species; these species attach to, and remain motionless on, the plant substrate. In combination, these features of coloration, patterning, integumental structure and behaviour, make finding the larvae in the moss carpets that surround them notoriously difficult. The suit of morphological traits in larvae varies significantly among taxa and is seemingly linked with biological attributes (e.g. habitats and host-plants). Notably, the elongated cuticular appendages may also be physiological, mechanical (physical) or behavioural adaptations to habitats; specifically, they have been hypothesized to serve as tracheal gills in aquatic species (De Geer, 1773, 1776Miall & Shelford, 1897;Müggenburg, 1901;Haake, 1922), an osmoregulatory organ (Brinkmann, 1991) or as a device for locomotion in both terrestrial and aquatic species (Zeller, 1842;Osten Sacken, 1869;Miall & Shelford, 1897;Peus, 1952), although these hypotheses do not conflict with one another. The highly cryptic form of the cylindrotomids is apparently associated with their habitats and biological interactions therein.
A question thus arises as to how a device for mimicry can evolve simultaneously alongside physiological, mechanical and behavioural functions. Osten Sacken (1897) referred to the larvae of cylindrotomids, writing, 'These larvae must, of course, show corresponding adaptations for such a mode of life, and it will be the task of future investigators to describe these structural differences in detail'.
To examine the link between ecological background and phenotypic variations of these complex traits, I herein provide an overview of the biology and key structures relevant to crypsis. Included are new data and analyses derived from my own investigations, as well as from relevant literature. Furthermore, devices and strategies of mimicry are challenged, with an emphasis on their functional aspects.
In Japan, seven species in four genera of cylindrotomids are known (Takahashi, 1960;Nakamura, 2001; The Editorial Committee of Catalogue of the Insects of Japan, 2014): Cylindrotoma japonica Alexander, 1919; Diogma caudata Takahashi, 1960;D. glabrata (Meigen, 1818); Liogma brevipecten Alexander, 1932;L. mikado (Alexander, 1919); L. serraticornis Alexander, 1919;and Triogma kuwanai (Alexander, 1913). Triogma kuwanai is subdivided into two subspecies, T. kuwanai subsp. kuwanai (Alexander, 1913) and subsp. limbinervis Alexander, 1953; occasionally they are treated as species (Oosterbroek, 2020). A recent phylogenetic analysis of major clades of Cylindrotominae using molecular (partial COI gene sequences) and morphological data has revealed discordance between the current taxonomy and phylogenetic relationships at both the species and genus level, suggesting that the taxonomic status of some Japanese and North American species should be changed (Kolcsár et al., unpubl. data). However, herein, the above-mentioned species and subspecies are treated following their current taxonomic status, as this does not cause any problems in demarcating these species.

data collection
In this study, immature stages were investigated for nine species, which cover all the described genera of Cylindrotominae: C. japonica, D. glabrata, L. brevipecten, L. mikado, L. nodicornis (Osten Sacken, 1865), L. serraticornis, T. kuwanai, Phalacrocera replicata (Linnaeus, 1758) and P. tipulina Osten Sacken, 1865 (Supporting Information, Table S1). For the species that were not originally studied here, data on their biology and morphology were retrieved from the primary literature (Supporting Information, Table S2). The original source is always cited if the information was gleaned from previous studies; this clearly distinguishes it from data obtained in this study and avoids possible confusion and inconsistency. Larvae have previously been described for six species in five genera of Cylindrotominae: C. distinctissima (Meigen, 1818), D. glabrata, L. nodicornis, T. exsculpta Osten Sacken, 1865, T. trisulcata (Schummel, 1829 and P. replicata. Immature stages of Stibadocerinae have not been described to date.

sampling, rearing and oBservations
For exploring life histories of cylindrotomines, insect larvae were collected in Japan (Hokkaido, Honshu and Shikoku) and the United States (California, Maryland, Tennessee and Virginia), between 2014 and 2020. Behaviour of adults and larvae, host-plant species and natural enemies were investigated in the field or in the laboratory. Host-plant classification follows APG IV (Angiosperm Phylogeny Group, 2016) for flowering plants and Goffinet & Buck (2020) for bryophytes. Insects were searched for in clumps of moss on humid forest floors and stream banks, and from trickling waters, waterfalls and lakes, and were collected together with the host-plants occurring in their habitats. The larvae of these species were usually found in those moss patches in which moss shoots were frequently damaged with characteristic feeding marks (see 'Biology and life cycle'). In order to collect L. nodicornis from the Great Smoky Mountains National Park in Tennessee, a research permit (GRSM-2017-SCI-2389) was acquired.
In order to identify the species at the larval stage, larvae obtained were reared until they became adults, where possible. In the laboratory, larvae were reared in plastic cases, which were occasionally moistened by a spray and provided with fresh hostplants. They were kept under laboratory conditions (15-24 °C). Rearing and observations of cylindrotomid larvae were performed either at Kyoto University or Ehime University (Japan), or at the Department of Paleobiology, National Museum of Natural History, Smithsonian Institution (USA).
To validate some functional hypotheses for locomotion of some terrestrial larvae, short films of larvae in motion were recorded with a Leica MC170HD camera mounted on a Leica M125C stereo microscope. The movement of the larval head of Cylindrotoma japonica was recorded from the ventral side (Supporting Information, Video S1; https:// youtu.be/dMSjisOKzV8). The larva was placed on a glass Petri dish and then flipped over; conveniently, it was able to attach itself to the dish, using its own attachment mechanism (i.e. fluid secretion). A larva of Liogma brevipecten at late instar was also observed from the ventral side (Supporting Information, Video S2; https://youtu.be/S1gywcO3ces). Due to the translucent integument, movement of the internal structure was observable from outside the body. The larva of L. brevipecten was easily detached on dried substrates, and thus placed on a dish of which the bottom was moistened with water. For L. brevipecten, L. mikado and T. kuwanai, adult insects, particularly their mating and oviposition behaviour, were also observed in the field.

morphological and histological examinations
For examining the ultrastructure of the external morphology, larvae were preserved in 70-99% ethanol, observed and photographed with the camera mounted on a stereomicroscope (as above). Larvae of four species were examined: C. distinctissima, L. serraticornis, P. tipulina and T. kuwanai. A JSM-5600LV (JEOL Ltd.) scanning electron microscope (SEM) was used to examine the ultrastructure of the spiracular field and integumental surface. Specimens were completely dried via the critical point drying method using a Leica CPD 300 Critical Point Drier; later, sputter coating was provided with a JFC-1600 Auto Fine Coater (JEOL Ltd.).
For examining internal structures of the integument, histological methods were employed for L. brevipecten at the first instar (N = 4). The specimens used for this purpose were fixed either with Bouin's solution or 5% glutaraldehyde solution; they were then cut into pieces shorter than 5 mm in length and embedded in paraffin. Serial sections, with a cross-section of 6-10 µm, were placed on slides, then deparaffinized by sequential immersion in the following for 2 min each: ethanol, 95%, 90%, 70%, 50%, 30% (v/v in distilled water) and distilled water; the sections were stained with Mayer's haematoxylin and eosin and then observed under a light microscope.
For external morphology, the terminology generally follows Beutel et al. (2009) andNeugart et al. (2009). All examined specimens are housed at the Graduate School of Science and Engineering, Ehime University, Japan.

general account of life-history notes
Biological features, including behaviour of larvae (feeding, protection and defence), pupation and adult oviposition, greatly differ between terrestrial mossfeeders (Diogma, Liogma and Triogma), herbaceous plant-feeders (Cylindrotoma) and aquatic mossfeeders (Phalacrocera). This section is mainly devoted to terrestrial moss-feeders (Diogma, Liogma and Triogma). For the others (Cylindrotoma and Phalacrocera), see the section of each genus.
The number of instars is not known exactly for many species. Cylindrotoma distinctissima distinctissima moulted four times (Brinkmann, 1991), as did Triogma trisulcata (Haake, 1922; but see the species descriptions below). Phalacrocera might moult eight to ten times (Bengtsson, 1897); L. brevipecten, L. mikado and L. serraticornis moult more than seven times (Imada, Y & Shindo, H, pers. obs.). According to Bengtsson's (1897) observation of P. replicata, the mode of moulting was the same for each moulting event: the old larval skin ruptured in the midline of the elongated lobes through a dorsal, straight longitudinal slit, extending from abdominal segments I to IV (sometimes V).
Larvae firmly adhere to leaf epidermis most of the time. The feeding manoeuvres of first-instar larvae differ to those of the second or later instars. The firstinstar larvae actively move around and pierce moss leaves with their mandibles oriented horizontally before sucking out cell contents. This feeding behaviour results in damaged leaf tissues with areas of white spots in which cells are devoid of chlorophyll (Fig. 1A, black arrowheads). The white spots on moss leaves are especially distinct on species of Mniaceae (Bryales), and are thus a useful signal by which the presence of Liogma spp. are detectable in the field; moss-feeding insects, other than cylindrotomine larvae, do not inflict such feeding damage. The feeding marks on other moss taxa, with small leaves, such as Thuidiaceae, are much less apparent and almost undetectable. Secondor later-instar larvae cause apparent chewing marks along the edges of leaves and stems (Fig. 1B). The larvae overwinter and then, in the next spring, pupate in the lower decomposed layers of moss carpets (Diogma, Liogma and T. kuwanai). Pupae (Fig. 1C) wiggle and crawl to actively escape from danger when disturbed.
Adult flies are sluggish and tend to stay among the vegetation where the larvae occur; they often hang down on branches or tree trunks, but do not dangle over them. The adult males usually emerge earlier than females and mate immediately after females eclose, before females have completely developed their coloured, chitinised skeleton (C. distinctissima, D. glabrata and L. mikado) (Mik, 1886;Müggenburg, 1901; this study, as in Fig. 1D). This behaviour is also known amongst some tipulids (Mik, 1886;Alexander, 1920). The females of Cylindrotoma cut the leaf epidermis near the leaf margin, using their cerci to make a slit in which single eggs are inserted (Fig. 1E); whereas in other species, the eggs are laid singly, attached to the surface of plant leaves or stems (Diogma, Liogma, Triogma and Phalacrocera) (e.g. Fig. 1F). Female oviposition behaviour is similar among the moss-feeding species (L. brevipecten, L. mikado and T. kuwanai). When ovipositing, the female bends its abdomen backward and places the cerci on the lower surface of a leaf so that the dorsal side of the cerci face the lower of the leaf; she then moves the cerci rapidly alternately and parallel to one another; the cerci are then slightly removed from the hypovalves, between which the egg appeared. The egg is released behind the moving cerci resting against the leaf.
The eggs have the lateral portion covered with a secretion fluid with which they adhere to substrates.
Without exception, the oviposition substrates are always plants, but not necessarily the host-plants of the larvae.

larval locomotion
The motion of final-instar larvae of Liogma brevipecten was investigated (Supporting Information, Video S2). Copulating couple of Liogma mikado on moss tufts; the female, on the left, was dragged out of the pupal exuvium by the male and thus the body has not been completely sclerotised. (E) Leaf of Stellaria (Caryophyllaceae) with two characteristic incisions in the epidermis of the lower leaf surface (white arrowheads) due to endophytic oviposition; the eggs are absent. (F) Egg of L. mikado laid on a shoot of Hylocomnium splendens (Hylocomniaceae). (G) Larva of C. japonica, which is being preyed upon by what is presumably a third-instar nymph of Himacerus apterus (Hemiptera: Nabidae). Scale = 1 mm.
The larvae firmly attach to moss leaves or stems solely using the force of their mandibles. The larvae can easily detach from the substrate, as they do not have any special adhesive structure on the ventral side; nor do they use a sticky substance, unlike Cylindrotoma. The larvae move around by crawling. The crawling motion starts at the posterior end of the body and shifts to the anterior end; it is, therefore, anterograde locomotion (Hughes, 1965). Multiple waves of steps (i.e. the movement in which a part of the ventral surface of the body is raised from, and landed on to, the substrate) can arise simultaneously. The ventral side of the body is broad and flat. The lateral lobes are attached to the substrate in the stance phase, and they are neither overtly deformed nor retracted during locomotion. The gut in the mid-portion of the body moves from posterior to anterior, before the surrounding cuticle begins to move. Such gut movement is reported for some lepidopteran larvae (Simon et al., 2010;van Griethuijsen & Trimmer, 2014). During each wave, three or four adjacent body segments are lifted simultaneously. Throughout the movement, the ventral lobes are solely used as friction points, which probably provide grip.

natural enemies
Records of predators and parasites of cylindrotomids are scarce. De Rossi (1876) reported that an ichneumonid wasp emerged from an unidentified species of cylindrotomid larva, but this is the only accessible record of parasitoid wasps of the Cylindrotomidae. The species of the infested larva was poorly described, as it was provided from memory. Various interpretations of the identity of this species by later authors as Triogma (Osten Sacken, 1897;Haake, 1922) or Diogma (Müggenburg, 1901;Alexander, 1920;Peus, 1952) are, therefore, not conclusive. Cameron (1918) examined over 100 larvae of C. distinctissima, which bore no parasitoids. Peus (1952) failed to find any parasitoids of any cylindrotomine genera, but he once observed numerous water striders, Gerris Fabricius, 1794 (Limnotrechus Stål, 1868 in the paper), feeding on a dead individual of P. replicata. I have observed a larva of C. japonica as the prey of, probably, a third-instar nymph of Himacerus apterus (Fabricius, 1798) (Hemiptera: Nabidae) on the upper surface of a dicot leaf (Fig. 1G).
First-instar larva: Newly hatched larva hyaline except head capsules (Fig. 1A). Elongated cuticular lobes present on dorsal, lateral and ventral sides; number and shape of lobes do not coincide with finalinstar larvae.
Final-instar larva: Body elongate, subcylindrical (Liogma, Diogma, Triogma and Phalacrocera) or markedly dorsoventrally flattened (Cylindrotoma) (Fig. 3A-E). Body trunk homonomous, consisting of head, three thoracic and eight abdominal segments without distinct differentiation. Body colour light green with seven symmetrical pigmented marks along the sides of abdominal segments I-VII (or I-VI); five marks on abdominal segments II-VI large and conspicuous, on segment VII smallest and least distinct (see 'Body coloration'). Head capsule prognathous and hemicephalic, completely retractable into prothorax; only weakly sclerotized, possessing paired dorsolateral incisions and a deep ventromedian incision at the posterior end. Antenna two-segmented; distal segment with diameter less than that of elongated basal segment. Maxilla apically narrow and pale. Maxillary palp with short, broad palp, with chitinised basal segment. Mandible toothed at adoral side, operating horizontally in first instar; in second or later instars, mandibles moving in vertical plane and working approximately parallel to one another (Bengtsson, 1897;Alexander, 1920;Peus, 1952; this study); prostheca arise on separate sclerotized lobe (Oosterbroek & Theowald, 1991;this study). Hypostoma distinct, constituting well-sclerotized, multidentated plate; completely fused with subgenal bridge (Oosterbroek & Theowald, 1991); in later-instar larvae, due to vertical operation of mandibles, mandibles work against hypostoma. Labium has about seven teeth on either side; teeth near median line relatively larger. Labrum not strongly chitinised, bearing few scattered bristles. Prothorax frontally slopes from anterior end, comprising a pronotal ridge lined with some tubercles or conical outgrowths; anterior opening forms a liplike lobe (except Cylindrotoma). Integumental surface structure differs among species and body regions (see 'Integumental structure'). Elongated cuticular lobes occur on dorsal, lateral and ventral sides of thoracic and abdominal segments with various size and form (see 'Elongated cuticular lobes'). Ventral creeping welts absent. Abdominal segment VIII (hereafter, anal segment) with one pair of spiracles typically surrounded by four lobes on dorsal and ventral sides (see 'Spiracular field'); each of ventral lobes possesses a sclerotized apical hook; anal lobes present on ventral side with two or four anal papillae or without anal papilla.

Body coloration
The matured larvae exhibit two different components of body coloration: green or brown hues occurring over the whole body, and dark pigmentation on the integumental surface. These features are typically apparent in larvae in the second or later instars.
The body colour appears after the first moult and the instar in which it becomes more distinctive varies among species. The green hue is attained mainly by haemolymph, whereas the integument is nearly transparent with a range of yellowish tints, on which the patterning of dark pigmentation occurs (Fig. 4A, B). Fat bodies surrounding the simple, cylindrical digestive tract, are pale yellow and do not significantly contribute to the hue. In some laterinstar larval individuals, the hue changes from green to pale brown. This colour change through ontogeny is seen in C. distinctissima (Alexander, 1920), C. japonica, T. trisulcata (Bengtsson, 1897;Haake, 1922;Hemmingsen, 1968), L. mikado (Fig. 4C, D), L. serraticornis and P. replicata (Wesenberg-Lund, 1915;Brinkmann, 1997;Pujante et al., 2016); whereas a brown morph has not been found in D. glabrata (Müggenberg, 1901;Wesenberg-Lund, 1915;Alexander, 1920; this study), L. nodicornis (Alexander, 1915(Alexander, , 1920 or L. brevipecten. For T. kuwanai, only brown larvae were found.
The dark pigmentation on the dorsal integumental surface is found in larvae at various instars. Occurrence of the pigmentation shows a well-organized patterning consistent with each body segment, especially on the abdominal segments I-VII; this also occurs on the thoracic segments in some species, but in a less organized manner. The pigmentation represents ontogenetic changes: in species of Liogma and Diogma, it appears in the second or later instars and is most conspicuous in the final instar. However, in C. japonica this marking is present only in larvae in early instars and later disappears. The patterning varies significantly among species; the coloration ranges from grey to nearly black, and the size of the shaded area can vary even within species due to unknown mechanisms that affect the entire body coloration.
Dorsal lobes occur on thoracic and abdominal segments; they are particularly well-developed and markedly differentiated among species on the abdominal segments II-VII (Figs 3A-E, 5A-K). The dorsal lobes are longitudinally aligned in a double row (Diogma, Liogma, Phalacrocera and Triogma, as in Fig. 6B-H), or, at least partly, in a single row (Cylindrotoma, as in Fig. 6A), along the central axis of the body. They are grouped in an ordered, repetitive manner, corresponding with each horizontal section on each body segment. The lobes on the pronotum are either tuberculate or conical and arranged in two different ways: dorsal lobes constitute a crown on the pronotal ridge (pr), except Cylindrotoma; in Cylindrotoma, the frontal edge of the pronotal opening is surrounded with the anterolateral lobes, whereas in other genera, it often bears many tubercles and constitutes a smaller crown (L. serraticornis and T. kuwanai) or it can be simple without such tubercles (L. brevipecten). On meso-, metathorax and abdominal segment I, each segment carries two pairs of lobes -one at the anterior and one at the posterior of the segment. These pairs tend to be close to each other due to the short segment length; and the lobes tend to be simple and conical, but in some species carrying auxiliary branches. On abdominal segments II-VII, each segment essentially possesses four pairs of lobes (except P. replicata, possessing two pairs), and the abdominal-segment lobes successively increase in size and complexity of shape from anterior to posterior. These forms vary, ranging from short and tuberculate (Cylindrotoma; Fig. 6A), to conical with blunt or sharp apex, with a various number of auxiliary outgrowths at the frontal side (Diogma, Liogma and Triogma; Fig. 6B-G), to extremely long and filiform (Phalacrocera; Fig. 6H). On the anal segment, they project laterally (Cylindrotoma), are conical (Diogma, Liogma and Triogma) or are fully developed (Phalacrocera).
Lateral lobes are simple, finger-or rose-thornshaped, flat laterally and bent backwards with a broad base (Diogma, Liogma and Triogma). They can also be long and filiform (Phalacrocera) or appear gently undulating (Cylindrotoma). On the thoracic segments, two pairs of lateral lobes are present. On the abdominal segments, three to four pairs occur per segment or at the border between two serial segments; in species with four pairs of lobes, usually the first (anteriormost) pair is small and associated with the second pair.
Ventral lobes can be absent on the prothorax in many species, but are present on meso-, metathoracic and abdominal segments I-VIII. In Phalacrocera, some of these lobes can be nearly as long as the dorsal lobes. The lobes on the abdominal segments represent small, rounded protuberances (Cylindrotoma), backwardly bent cones (Diogma, Liogma and Triogma) or filiform, as in the case of dorsal lobes (Phalacrocera). On mesoand metathoracic segments, the number of ventral lobes varies among species. Abdominal segment I bears three pairs (except Cylindrotoma and P. replicata). On abdominal segments II-VII, each segment bears five pairs of ventral lobes (except Cylindrotoma and P. replicata) arranged in a similar order to the dorsal ones, corresponding to each horizontal section, increasing in size from anterior to posterior.
The anal segment possessses cuticular lobes with varied number and form (see 'Spiracular field'). In all examined species, the spiracular field is surrounded by one pair each of dorsal and ventral lobes. The dorsal and ventral lobes are traditionally considered to be homologous to elongated cuticular lobes in other body segments (Lenz, 1919;Haake, 1922;Peus, 1952). For these lobes, the affinity to the spiracular lobes of Tipulidae is unclear (e.g. Young, 2004). Anal lobes with anal papillae at the apices, which are often soft, fingerlike or filiform; they are retractable except in the cases of Cylindrotoma and Phalacrocera.

Integumental structure
The integument is unsclerotized and moderately elastic; the surface bears bumps, tubercles and macrosetae. The ultrastructure of the dorsal integument greatly differs between Cylindrotoma japonica, Liogma brevipecten and Phalacrocera tipulina. In these three species, the ultrastructure of the elongated cuticular lobes is conspicuously differentiated from that of the general dorsal integument. In C. japonica, evenly spaced, button-shaped microtubercles are present on the dorsal integument (Fig. 7A). These microtubercles become conspicuously swollen on elongated dorsal Figure 5. Schematic diagram illustrating diverse arrangements and numbers of the elongated cuticular lobes in 11 cylindrotomine species. Each diagram represents a simplified larval body, longitudinally grouped into six, based on the patterned arrangements of the lobes: (1) prothorax, (2) meso-and metathorax, (3) abdominal segment I, (4) abdominal segment II, (5) abdominal segments III-VII, and (6) abdominal segment VIII (anal segment). Each square partitioned by double lines in a row denotes the integument on the dorsal (dl), lateral (la), ventral (vl) sides of the corresponding segment(s). Each black dot symbolizes a cuticular lobe, of which position/size in a given space express the relative position/size of the lobe. Anal papillae are not shown. Lobe forms, auxiliary outgrowths, and relative length of segments are not shown here (see each species' own description). Each taxon name is abbreviated. For G and I, diagrams are based on the description in Brodo lobes (Fig. 7B). In P. tipulina, the dorsal integument bears similar microtubercles on a smoothly undulated, nearly glabrous, cuticular surface (Fig. 7G); each microtubercle possesses a minute micropapilla near the centre (Fig. 7H). The lobes in Phalacrocera are entirely armoured with conical micropapillae (Fig. 7I), as previously illustrated for P. replicata (Bengtsson, 1897: fig. 21). In L. brevipecten, the dorsal integument is markedly different from those of Cylindrotoma and Phalacrocera; it bears extensive wrinkles and becomes reticulated at the base of elongated lobes (Fig. 7D); the surface of the elongated lobes is smoothly undulated (Fig. 7E), and has nearly transparent, coarse, lateral spinules (Fig. 7F). The spinules in L. brevipecten and the micropapillae in P. tipulina are similar, in that both are unicellular, but each spinule of L. brevipecten has a base and is seemingly separate from the cuticular surface, whereas the micropapillae in P. tipulina lack (1967) and Peus (1952), respectively; caution is needed when these are compared with the species examined herein (A-F, I-K), because the lobe characteristics could be treated differently in the previous studies. For K, a single lobe is present at the location of asterisk (*) only in the abdominal segment VII but not in the other segments. such differentiation (Fig. 7F). The microtubercles and micropapillae do not show an internal structure such as air canals. The internal space (haemocoel) is compartmentalized into three regions in cross-section: dorsal sinus, held by a dorsal diaphragm (ddph); ventral sinus, defined by ventral diaphragms (vdph); and the space in-between -the perivisceral sinus (Fig. 8E). The dorsal sinus contains the dorsal blood vessel (dv), surrounded by thick fibres of dorsal longitudinal muscles. The ventral sinus contains the ventral nerve cord (ne) with segmented ganglia, extensively surrounded by ventral, longitudinal muscles (e.g. Fig. 8A). The longitudinal muscles contained in the dorsal and ventral sinuses can also be extended into the cavities of the dorsal and ventral cuticular lobes, respectively.

Internal structures
The integument is internally striated; two cuticular layers consisting of a thin, hyaline, exocuticle layer and an unsclerotized, achromatic, endocuticle layer and an underlying epidermal cell sheet are recognized (Fig. 8C). The thickness of the cuticle is heterogenous across different body regions, culminating in dorsal cuticular lobes (Fig. 8D).
The cuticle forms lobes on the dorsal, lateral and ventral sides, which do not have a basal articulation  and posterior (E) parts of abdominal segments, representing anterior hindgut; and anal segment (F), illustrating posterior hindgut. Note that integumental thickness is heterogeneous across body parts. Black arrowheads and arrows denote intrinsic and extrinsic muscles, respectively. Abbreviations: cm, circular muscle; dcl, dorsal cuticular lobe; ddph, dorsal diaphragm; dv, dorsal blood vessel; ed, endocuticle; epd, epidermis; ept, gut epithelium; ex, exocuticle; fb, food bolus; ft, fat body; lcl, lateral cuticular lobe; lm, longitudinal muscle; lu, gut lumen; mdm, midgut microvilli; mlp, malpighian tubule; mtb, microtubercle; ne, ventral nerve cord; pr, peritrophic membrane; pro, proventriculus; re, rectum; sal, salivary duct; stm, stomodeal valve; tr, trachea; vdph, ventral diaphragm; vcl, ventral cuticular lobe. Scale = 50 µm. with the body wall. Dorsal and ventral cuticular lobes are characterized by extremely thickened endocuticles (Fig. 8C); they do not receive intrinsic muscles, but contain longitudinal muscles, sometimes with attachment points (Fig. 9A). In contrast, muscles fully contained within the narrow cavity sheltered by lateral cuticular lobes are found, of which both ends are attached to the lobes (i.e. intrinsic muscles), throughout thoracic and abdominal segments; several short fibres are positioned closely from the apical to basal part of the lobes (black arrowheads, Fig. 8A-F). Although the intrinsic muscles near the distal end of the lobes are dorsoventrally oriented, some fibres near the base of the lobes are inserted in oblique directions (i.e. extrinsic muscles; black arrows in Fig. 8B, E, F). The intrinsic and extrinsic muscles in the lateral lobes are thus clearly distinguished from the longitudinal and circular muscles based on their positions and orientation. The internal structures of the dorsal, lateral and ventral lobes between those in the thoracic (Fig. 9A, B) and abdominal ( Fig. 9C-F) segments are similar to each other, except in the anal segment. In the anal segment, the dorsal sinus contains muscle fibres oriented dorsolaterally, as well as a pair of dorsoventral muscle fibres surrounding the dorsal vessel, which are attached to the dorsal diaphragm (Fig. 9A).
The digestive tract is nestled in the perivisceral sinus as an elongated tube with little differentiation in external appearance. The foregut is composed of the pharynx, crop and proventriculus. Salivary glands are strongly developed behind the head (Fig. 8A). The stomodeal valve (stm) (Fig. 8B) connects the posterior end of the proventriculus and anterior end of the midgut. The midgut (Fig. 8C) is simple without conspicuous diverticula. The epithelium bears microvilli (mdm) in the lumen (lu). The food bolus (fb), which is composed of fine particles, is enveloped within the peritrophic membrane (pr). The hindgut appears to consist of two parts -the ileum to the anterior (Fig. 8D) and the rectum to the posterior (Fig. 8E). The wall of the hindgut bears Malpighian tubules (mlp), which loop and diverge posteriorly (Fig. 8D). The digestive tract is surrounded by numerous fat bodies (ft) and tracheae (tr) (Fig. 8D). At the posterior end, four anal papillae (anp) are present near the anus (Fig. 9A); the tissue is histologically distinguished by a hyaline cuticle and ramifying, fine muscle fibres.

Spiracular field
Larvae are metapneustic. A pair of spiracles (sp) is present at the dorsodistal end of the anal segment. In Cylindrotoma, spiracles are located dorsally without a spiracular disc (Fig. 10A). In most species (Diogma, Liogma, Phalacrocera and Triogma), the spiracles lie in the retractable cavity of the spiracular field (i.e. flattened posterior face of the anal segment) ( Fig. 10C-E). When the spiracular dics are closed, the spiracles face each other (Fig. 10B, E). Typically, each pair of dorsal (dl) and ventral lobes (vl) are present on the spiracular field, but additionally, dorsomedial lobes (dm) are present in some species (D. glabrata, L. mikado, L. nodicornis, T. exsculpta and T. trisulcata), of which size and position various. Each of the ventral lobes possesses a sclerotized apical hook; the posterior side of the ventral lobe are pigmented as a dark line. In Phalacrocera, each of the dorsal lobes bears a small accessory lobe (accessory dorsal lobe; dac) near its base.
The external structure of the spiracles is similar to that of other cranefly larvae. The spiracular field of Cylindrotoma and P. replicata is glabrous (Fig. 10A), whereas those of other species bear a hair fringe (ha) (Fig. 10C, E). A central scar plug lies at the centre of each spiracle as a uniformly sclerotized circular disc, which is surrounded by aeropyles. Scanning electron microscope observations reveal that the central plug has an aperture and the aeropyles have slit-like openings in three species: Cylindrotoma japonica (Fig. 11A, B), Liogma brevipecten (Fig. 11C, D) and Phalacrocera tipulina (Fig. 11E, F).
The stigmatic ring internally has a developed spiracular atrium (at) (Fig. 9A-C). The atrium is lined with numerous tracheae (Fig. 9C, D), constituting a 'felt chamber' (Keilin, 1944). The spiracular structure differs between aquatic (P. replicata) and semi-aquatic (T. trisulcata) species (Haake, 1922). In Triogma trisulcata, the central scar plug is thin and heavily sclerotized; internally, the atrium chamber was reported to be wide and relatively short (Haake, 1922). The spiracles of Phalacrocera replicata are characterized externally by an extremely thick, weakly sclerotized, central scar plug surrounded by a stigmatic ring; the atrium chamber was relatively narrow and long by Haake's (1922) observation. Notes: Two species (including two subspecies), Cylindrotoma distinctissima subsp. americana Osten Sacken, 1865 and C. d. subsp. distinctissima, and C. japonica have been studied so far. These species are closely related to each other, and the life history and the behaviour of adults and larvae do not significantly differ among these taxa, and thus are given collectively below.

Life history
The typical habitat of Cylindrotoma is shaded, humid forest floor in marshy woodlands. The larvae are found on leaves and stems of herbaceous plants. Cylindrotoma distinctissima and C. japonica are polyphagous and they use a wide range of dicots and some monocots. The larvae of these species occur densely in patchy, small populations, where they feed on leaf tissues of various herbaceous plants with soft, green, unlignified stems; some larvae even feed on young shoots of woody plants with short heights. Adult flies appear in spring; for some populations of C. d. distinctissima, two broods occur per year (Brinkmann, 1991): one in spring (Schellenberg, 1803;Zeller, 1842) and the other in autumn (Boie, 1838). Larvae moult four times (Brinkmann, 1991). At the first instar, the larvae feed on upper leaf tissues, leaving the lower epidermis. At the second instar or later, the larvae skeletonize the leaf tissues by opening holes and leaving only the hyaline membrane and leaf veins. The larvae live on a flat, smooth leaf substrate, which provides no supporting structures to hold on to, and so they ensure their attachment via the suction-cup principle and sticky gland secretions (as noted in 'Larval behaviour'). The late-instar larvae become apathetic, do not eat and do not noticeably move. The larvae cease feeding after the final moult and overwinter among fallen leaves (Brodo, 1967), which presumably occur at the third instar (Brinkmann, 1991). The larvae pupate on a leaf or a stem, holding on via the sheath of the adhered larval exuvia (Zeller, 1842;Alexander, 1920;Peus, 1952).

Adult behaviour and mechanism of endophytic oviposition
Females exclusively insert eggs under the leaf epidermis (Fig. 1E); endophytic oviposition is only known so far from this genus among other genera of Cylindrotomidae. The female makes use of its ovipositor, which consists of two types of valves: an upper set of double-bladed cutting valves bearing a serrated blade; and a lower set of hypogynial valves that deliver eggs with an unserrated blade. The female fly places the upper side of her cerci against the lower leaf surface, near to the edge. She then slits the leaf epidermis using the cutting valves, whilst the lower valve guides the egg into the resulting slit. The bifurcated valve on the lower side apparently serves to keep the leaf in position during the act of ovipositing, since the leaf margin was held securely between this valve and the cutting valves (Cameron, 1918). When cutting a slit in the leaf epidermis, the cerci are kept close to one another and are moved alternately parallel to one another, which moves the abdomen from side to side (Peus, 1952). The eggs are partly concealed beneath the leaf epidermis and dorsally exposed through the slit; they were generally laid in close, parallel series (Cameron, 1918). The eggs were covered with adhesive material when oviposited (Peus, 1952) and became solid after drying, but unlike other genera, any adhesive substance was not observable (Hemmingsen, 1960).

Larval behaviour
Debris carrying: Larvae of Cylindrotoma carry debris on the dorsal integument of the abdominal segments (Fig. 12F). This is done by newly hatched larvae, the bodies of which are nearly transparent and only the fresh gut content has a green tint; in later larval stages, as their haemolymph becomes green, this behaviour becomes less common. The larvae deposit faecal pellets on the dorsal integument by bending the distal abdominal segments forwards. The anal segment can only reach over the front of the metathorax, which, therefore, receives the densest accumulation of faeces. Peus (1952) described the debris-carrying behaviour in detail, as below. The pronotum was not covered with faeces because the anus cannot reach over the two high processes of the mesonotum (Peus, 1952). The posterior segments indirectly supplied with faeces as these pellets already deposited on the dorsal segments rolled over (Peus, 1952). The faecal pellets were dull black in colour, cylindrical to roundish in shape and adhesive (Peus, 1952). Larvae sticked firmly to the substrate with an anal secretion, which covered the entire ventral surface of abdominal segments VI-VIII (Peus, 1952).
Defensive behaviour: The larvae of Cylindrotoma are sensitive to the motion of surrounding objects and they show defensive behaviour when threatened. When disturbed, in response they bend the anterior part of the body backwards tightly, making an acute angle with the caudal end, and then demonstrate katalepsis; namely, they hold a frozen pose for ten seconds to a minute. In cases where the danger is within touching distance, the larvae on leaves display thanatosis (i.e. death feigning) by tumbling down from the substrate to the ground.
Larval locomotion of Cylindrotoma: The larvae live on the leaf surface making use of two, especially developed, attachment mechanisms that are used in combination: a suction cup and a sticky secretion. The head capsule is retracted into the thorax and thereby a cavity is created; this acts as a suction cup when it is tightly pressed against the flat leaf surface. The head in Cylindrotoma is the bead-like, soft, front edge of the prothoracic integument, and the rim is pressed against the substrate (Supporting Information, Video S1). The interior cavity is filled with a thin film of salivary gland secretion, which covers the larval head and prothorax (Fig. 12D), but with an opening (white arrowhead, Fig. 13A). When the secretion completely dries, it becomes a solid, reticulated aggregation, which is composed of delicate, thread-like strands (Fig. 13B), presumably silk. The fluid is drawn by a capillary between the integumental ring and the substrate, thus producing a tight seal. Although the silk-like threads are widespread on the ventral side of the prothorax, those occurring on the body elsewhere than the head capsule are coarse and have a greater diameter (Fig. 13C). When feeding or resting, the larvae normally keep themselves adhered to the substrate in such way that the head capsule remains entirely within the thorax but the mandibles are kept in a position to access the leaf surface.
The suction-cup mechanism also has a role in locomotion. The larvae move around by inching; when they transit from one leaf to another, they fix themselves with their caudal end, straighten up sharply and then move forwards the fore-body often accompany with a slow back and forth motion ( Fig. 12I-K). The ventral cuticular lobes tend to be small, rounded protubercles and they also are involved in larval movement. The ventral lobes of the abdominal segment VII (Fig. 13D) have a stamp-like broad surface end, and they work in cooperation with the ventral lobes on the anal segment (Osten Sacken, 1869;Peus, 1952); these lobes aid the movement of larvae up and down the leaves and stems of plants. The secretion from the salivary glands causes the larvae on the leaf surface to leave a slight, silvery, creeping trail, which fades after a few hours as a result of drying (Cameron, 1918;this study). Egg: Spindle-shaped, circular in cross-section, with a slightly tapered front pole. Chorionic sculpture smooth-edged. Egg covered with adhesive fluid right after oviposition and becoming solid after drying (Peus, 1952); adhesive area absent (Hemmingsen, 1960). Micropyle usually hidden under leaf epidermis; region around micropyle not differentiated well from surrounding area; instead, showing fine granulation (Peus, 1952).

Final-instar larva:
Length 20 mm (Peus, 1952). Light green, darker dorsally; brown individuals unknown (Peus, 1952). Larval trunk dorsolaterally flattened, dorsally with a single row of small, conical lobes along its body. Prothorax with eight tubercles lining at lateral margin and with one pair of small, simple, conical lobes on dorsal side (Peus, 1952). Spiracular field without obvious spiracular disc without hair fringe; spiracles present dorsally (Peus, 1952). Elongated cuticular Figure 13. Ultrastructure of Cylindrotoma japonica final-instar larva, SEM views. (A) larval head loaded with a mass of secretion (presumably silk material); secretion is reticulated and entirely conceals the head capsule but a small, circular opening is present near the centre (white arrowhead), which leads to the mouth underneath. (B) Close-up view of a secretion mass, which provides a cap for the head capsule, revealing delicate reticulated threads. Bright flaring is due to sample charge. (C) Isolated threads of the ventral thoracic segment; note that the diameter is greater than that of head part threads in image B. (D) Ventral lobe on abdominal segment VII, in which the cuticular surface is worn out due to abrasion. Scales as shown in each image.

Host-plants:
The known host-plants encompass nine angiospermous families: two monocot families and diverse eudicots consisting of seven families. Following genera and species were recorded as host-plants (Lenz, 1921;Peus, 1952;Tarasova, 1981;Stubbs, 2006;Uffen & Chandler, 2010;Paramonov, 2013) Final-instar larva: (N = 3) Larvae similar to C. distinctissima distinctissima, with few minor characters not described for other taxa. Body apparently green (Fig. 12G) or brown hue (Fig. 12H) when moderately dried, but nearly transparent with silvery shining ventral region when wet (Figs 4B, 12F). Dorsal integument with tadpole-shaped, dark pigment in young larvae (presumably at second instar; Fig. 12G); pigment absent in late instars (Fig. 12H). Larval trunk dorsolaterally flattened. Prothorax with eight tubercles lining at lateral margin and with one pair of small, conical, dorsal lobes (Fig. 12C); integument at frontal margin often wrapped with larval secretion (presumably silk) trapping debris (Figs 12D, 13A). Dorsal elongated lobes on abdominal segments small, simple and tuberculate, arranged longitudinally in a single row (Figs 5A, 12A). Lateral lobes on meso-, metathoracic and abdominal segments corrugated at the margins (Fig. 12A). Ventral lobes on meso-, metathoracic and abdominal segments I-VII as one pair of rounded tubercles, lined near posterior margin (Fig. 12B). Anal segment possesses one pair of dorsal lobes, oriented laterally (Fig. 12E); two pairs of lateral lobes, with bases near the anterior end of the segment; one pair of ventral lobes, papilla-like, with apices projecting downward; anal papilla absent. Posterior spiracles present dorsally, without forming an obvious spiracular disc (Fig. 10A) and devoid of hair fringe (Fig. 11A, B). Elongated cuticular lobes internally, without containing tracheal branches.  (Müggenburg, 1901); in Japan, the adult flying period is in summer (July to August), which is much later than the other co-occurring species (Liogma mikado and Liogma serraticornis). Larvae are found in terrestrial mosses growing on stones in limestone woodlands, and also in moss mats growing on soil. In Europe, they were found in forests in grassy, humid areas with Rhytidiadelphus squarrosus (Hedw.) Warnst. (Hypnales: Hylocomniaceae) (Zeller, 1842;Müggenburg, 1901). The number of moults is unknown. Final-instar larvae apparently do not have a preference for pupation sites and pupae were found anywhere in the lower damp layers of moss carpets (Alexander, 1920).
Oviposition: A female of this species laid about 60 eggs in total (Müggenburg, 1901), depositing them singly on the lower surface of leaves or stems of mosses (Hemmingsen, 1960).
First-instar larva: Newly hatched larvae ashy grey, but become moss green in colour at later stages (Müggenburg, 1901).

Final-instar larva:
Length 17-18 mm (N = 3). Body colour with dark green hue; dark pigmentation appears as small blotches on lateral sides, with various forms ranging from rhombus to trapezoid; also, a single, small, isolated blotch near the distal corner of the anterior end is present in the abdominal segment; size of pigmented area increases from abdominal segments I to VI; pigmentation small and pale on abdominal segment VII (Fig. 14A, B). Brown individual unknown (Müggenberg, 1901;Wesenberg-Lund, 1915;Alexander, 1920). Prothorax rugulose; dorsal lobes conical with sharply pointed apices, two pairs along pronotal ridge, anterior pair longer than posterior pair (Fig. 5B); lateral lobes one pair; ventral lobes two pairs. Meso-and metathoracic segments have two pairs of dorsal lobes; lateral lobes two pairs; ventral lobes two pairs, papilla-like, anterior pair closer to each other than posterior pair (Fig. 5B). Dorsal lobes on abdominal segments with bluntly pointed apices, third and fourth pairs bearing two, rounded auxiliary outgrowths (Fig. 6B); segment I with two pairs of lobes, each with two auxiliary outgrowths at the front; segments II-VII with four pairs, each of the posterior two pairs carry two teeth-like auxiliary outgrowths at the front, fourth pair longest and stout; segment VIII with a pair of small, simple, conical lobes without auxiliary outgrowth. Lateral lobes on abdominal segments simple, finger-shaped; segments I and II-VI with three and four pairs, respectively. Ventral lobes on abdominal segments rounded. Anal segment with one pair of dorsal lobes long; dorsomedial lobes (dm) one pair, short and conical (Fig. 14C); lateral lobes one pair, papilla-like; ventral lobes two pairs, anterior pair papilla-like. Spiracular field with hair fringe. Life history: This species is univoltine; adults emerge during May-July, in central Japan. Larvae of this species occur in well-moistened Mniaceae moss tufts in wet spots of forest or woodland, such as stream banks, trickling waters, rocks near waterfalls and lake margins. In Japan, they frequently co-occur with some other terrestrial moss-feeding species, such as L. serraticornis, L. mikado and T. kuwanai. In winter, the larva buries itself beneath the moss patches and sometimes its body is even embedded in the soft ground.
Final-instar larva: Length 14 mm (N = 47). Body colour dark green dorsally, due to oblique dark pigmentation on lateral side (Fig. 15A); colour generally uniform among individuals, brown individual unknown. Integumental surface rugulose due to undulated or reticulated ridges (Fig. 7D). Prothorax with two pairs of dorsal lobes along pronotal ridge, small and conical, with needlelike apices; anterior opening relatively smooth, without distinct tubercles (Fig. 15B); one pair of lateral lobes near the posterior end of the segment, small. Meso-and metathoracic segments with one pair of dorsal lobes; lateral lobes two pairs; ventral lobes two pairs, small and conical. Dorsal lobes on abdominal segments with translucent, sharply pointed apices; lobe covered with numerous sharp, translucent spinules on anterior and posterior sides of the lobes (Figs 6C, 7F, 15C); abdominal segments I and II-VII with two and four pairs, respectively. Lateral lobes on abdominal segments with sharp apices; abdominal segment I with three pairs; abdominal segments II-VII with four pairs (Fig. 5C); lobes bearing spinules as of dorsal lobes, but with lower density. Ventral lobes on abdominal segments small and conical. Anal segment with one pair of dorsal lobes bearing spinules; one pair of ventral lobes, devoid of spinules; dorsomedial lobe absent; lateral lobe absent; two pairs of ventral lobes, anterior pair papilla-like; four anal papillae present. Spiracular field with hair fringe (Fig. 10C). Elongated cuticular lobes internally containing tracheal branches (Fig. 15D).
Oviposition: Behaviour of four female individuals was observed in the field (   Life history: Liogma mikado is univoltine, having the adult stage in spring. This species markedly differs in habitat preference from other Liogma species, although it frequently co-occurs with L. serraticornis on a small geographic scale (within a range of a few metres). The larvae are most frequently found on arid moss tufts on rocks or on the ground, but also on decomposed tree trunks. The larvae are exposed on the surface of moss patches, unlike other species of Liogma (e.g. L. brevipecten and L. serraticornis), which sometimes bury deeply beneath the moss patches.
Mating and oviposition: Two adult females of L. mikado were observed during 13:00-16:10 on 19 May 2016, at Ikawa-touge (Shizuoka, Japan). The site was a forest floor with gentle slope facing a forest path toward a mountain pass. The males emerged before the females and waited beside the female pupae for the emergence of their mates. Mating took place mainly resting on the branches and leaves of the understory vegetation (Fig. 1D). When pairs are in copula, they readily take flight. The oviposition was observed once on the same day. One female flit about the tufts of mosses and lay an egg on a branch leaf of Hylocomnium splendens (Hedw.) Schimp. (Hylocomniaceae) (Fig. 1F).
Final-instar larva: Length 8-10 mm (N = 3). Body colour polymorphic, with light green (Fig. 4C) or brown hue (Fig. 4D); dark pigmentation making a narrow band, appearing as an oblique line in lateral view, from mesothoracic through abdominal segment VII; also, tips of small lobes tinged, including lobes on thoracic segments, first and second pairs and auxiliary outgrowths (particularly, second and third teeth) of third and fourth pairs of lobes on abdominal segments II-VII (Fig. 16A). Integument slightly tinged yellow (Fig. 4A), surface smooth. Prothorax dorsally with two, conical pairs of lobes on pronotal ridge with sharply pointed apices, posterior pairs longer than anterior pairs; two pairs of lateral lobes, papilla-like. Meso-and metathoracic segments with two pairs of dorsal lobes, each with two conical auxilliary outgrowths at the front; two pairs of lateral lobes, anterior pair longer than posterior pair; one pair of ventral lobes, papillalike (Fig. 5D). Dorsal elongated lobes on abdominal segments generally short, longest lobe shorter than body width in ventrolateral view (Fig. 6D); segment I with two pairs: anterior pair conical, with two tiny teeth-shaped auxiliary outgrowths at the front; posterior pair long, with three teeth-shaped outgrowths at the front; segments II-VII with four pairs of lobes, anterior two pairs conical and simple, third pair with two teeth at the front and fourth pair with three teeth near the base at the front; apices of teeth-like auxiliary outgrowths obtuse and tuberculate when dried, sharply pointed when wet. Lateral lobes on abdominal segments with sharply pointed apices; segment I with three pairs, posterior two pairs longer than anteriormost pair; segments II-VII with four pairs, middle two pairs longer than others. Ventral lobes on abdominal segments small, rounded or conical. Anal segment with one pair of dorsal lobes, long; one pair of dorsomedial lobes, short and conical; lateral lobes papilla-like; two pairs of ventral lobes, anterior pair papilla-like. Spiracular field with hair fringe.

Host-plants:
The host-plant range of this species varies considerably among populations. Typically, the larvae are found and reared on some species of Thuidium  Life history: This species is univoltine (Alexander, 1920). It commonly occurs in damp, swampy woods, in gorges of creeks and on moist, shaded hillsides. Larvae are found on small patches of Mniaceae mosses along the slope of forest paths, dead tree trunks and detritus underneath.
Oviposition: Females deposited eggs singly on the leaves or branches; or lightly attached eggs to the axils of the leaves of Rhytidiadelphus squarrosus (Alexander, 1915).  (Brodo, 1967;Byers, 2002). This study confirms that larvae can also be reared on Rhizomnium appalachianum T.J.Kop. and Plagiomnium ciliare (Müller Hal Final-instar larva: Length 20 mm (N = 10). Body colour various with light green (Fig. 16E) or dark brown hue (Fig. 16F). Dorsal pigmentation pattern well organized: prothorax lacks pigmentation; meso-and metathorax nearly entirely pigmented; in abdominal segments I-VII, dark blotches appear as inverted Y-shape, broadly diffused anteriorly (Fig. 16G); in some individuals, small blotches also occur ventrally. Integument translucent, except for pigmented parts, surface rugulose (Fig. 6F). Prothorax dorsally with six lobes along pronotal ridge having blunt apices; middle two longer and conical, outer two shorter and tuberculate (Fig. 16H). Meso-and metathorax with two pairs of dorsal lobes, both bearing two button-shaped auxiliary outgrowths at the front; two pairs of lateral lobes, anterior pair larger; two pairs of ventral lobes. Dorsal lobes on abdominal segments slender and each pair clinging to each other with bluntly pointed apices ( Fig. 6F): segment I with three pairs, posteriormost pair carrying two auxiliary outgrowths; segments II-VII with four pairs, posterior two pairs bearing two auxiliary outgrowths (Fig. 6F). Lateral lobes on abdominal segments relatively long, apices obtuse and bent backwards; segments I and II-VII with three and four pairs, respectively. Ventral lobes on abdominal segments relatively long, apices significantly oriented backwards. Anal segment with one pair of dorsal lobes; dorsomedial lobe absent; two pairs of ventral lobes, anterior pair papilla-like (black arrowheads, Fig. 16I); four anal papillae (anp) present (Fig. 16I). Spiracular field with hair fringe. Life history: The habitat of this species is 'seepage areas and along short marshy spring rills and are often found in wet mossy meadows where there is no woody vegetation' (Brodo, 1967).

Egg: Unknown.
Final-instar larva: Following description based on Brodo (1967). Length 16 mm. Body colour undescribed. Prothorax glabrous; two and one pair(s) of dorsal and lateral lobes present. Meso-and metathoracic segments with two pairs of dorsal lobes, lobes of posterior pair with one small protuberance at the base; two pairs of lateral lobes; three pairs of ventral lobes, the middle pair most widely separated to each other. Dorsal lobes on abdominal segments as follows: segment I with two pairs of lobes, each with one anterior protuberance; segments II-VII with four pairs, first pair small, second pair slightly larger, third pair with two frontal tooth-like protuberances, fourth pair longest with one frontal tooth. Lateral lobes on abdominal segments simple; two and four pairs on segments I and II-VII, respectively. Ventral lobes simple, conical. Anal segment with three pairs of lobes, inclined toward each other, one small dorsomedial pair and one pair of dorsal lobes; one pair of dorsomedial lobes; two pairs of ventral lobes, anterior pair shorter. Ventral surface of anal segment possesses protuberances.

Host-plants:
Larvae were collected among the tufts of the aquatic moss Fontinalis antipyretica Hedw.
Life history: In Japan, adults often fly together with species of L. mikado, L. serraticornis and L. brevipecten at the same locality and about the same period of time (May-June, in Honshu). Larvae are found in moss patches on the forest floor. The cuticular surface is heavily armoured with debris and epibionts, which seems to be related to the cryptic habitat. All reared larvae died before reaching the second instar.
Egg: Length 1.3 mm (N = 1). Spindle-shaped, circular in cross-section, with a tapered front pole. Pale grey due to exochorionic sculpture, with yellowish tint of egg content (Fig. 2C), yet apparently lighter than L. brevipecten. Chorionic sculpture elongated and narrow, arranged as irregularly jagged, dotted dash lines; reticulated near micropyle.
Final-instar larva: Body with dark brown hue (Fig. 17A). Integument entirely rugulose with velvet-like texture, retaining high opacity in alcohol; dark pigmentation diffused on dorsal side, densest near dorsal lobes, making stripes in parallel with longitudinal body axis; small patch with purple lustre at bases of dorsal lobes (Fig. 17E). Prothorax dorsally with six tubercles along pronotal ridge; one pair of lateral lobes; ventral lobe absent. Meso-and metathorax with two pairs of dorsal lobes, posterior pair bearing two auxiliary outgrowths at the front; two pairs of lateral lobes, anterior pair longer; two pairs of ventral lobes, anterior pair more widely separated from each other than the posterior pair. Dorsal lobes on abdominal segments slender and apices acutely bent backwards (Fig. 6G); four pairs on abdominal segments I-VII, two anterior pairs smallest, and fourth pair longest with three auxiliary outgrowths at front. Lateral lobes relatively long and broad, apices strongly bent backwards. Ventral lobes on abdominal segments relatively long and finger-like (Fig. 17E). Anal segment (Fig. 17C) with one pair of dorsal lobes; dorsomedial lobes absent; two pairs of ventral lobes,  Life history: Larvae of this species occur in semiaquatic mosses on moorlands, such as those of Hypnum cuspidatum (Brindle, 1967), in stagnant waters on bogs (Brinkmann, 1997). They have also been found in slowly running water and springs (Steinmann, 1907-08;Hemmingsen, 1968), and in small alpine streams (Steinmann, 1907-08;Müller, 1908;Alexander, 1915). According to Haake (1922), the larvae underwent moulting four times until pupation (Haake, 1922); however, his explanation had a contradictory statement. As Haake (1922) noted, stages I, II and III have one, two and no moulting, respectively, which does not make the total of four instars that the author recorded. First-instar larvae were whitish without distinct colour (Haake, 1922). The larva lives among the stems of the aquatic moss Fontinalis antipyretica, to which it clings firmly with two strong chitinised hooks at the caudal end of the body (Haake, 1922). On pupating, it comes to the water surface and remains immobile for the first few days of pupation; after 8-10 days, it becomes more mobile (Haake, 1922).
Egg: Spindle-shaped, circular in cross-section, with a tapered front pole. Pale yellow. Chorionic sculpture elongated and narrow, arranged as irregularly jagged, dotted dash lines; reticulated near micropyle.
Final-instar larva: Length 19 mm (Alexander, 1920). Body colour light green with dark blotches (Alexander, 1920). Prothorax with three pairs of dorsal lobes along pronotal ridge; one pair of lateral lobes; three pairs of ventral lobes as tubercles (Peus, 1952). Meso-and metathoracic segments with each two pairs of dorsal and lateral lobes; lateral and ventral lobes two and three pairs, respectively (Alexander, 1920). Dorsal elongated lobes on abdominal segments shorter than body width in ventrolateral view, c. 1.5 mm in length (Alexander, 1920): abdominal segments II-VII with four pairs of lobes, posteriormost pair longest, bearing three or four teeth-like auxiliary outgrowths at the front (Alexander, 1915(Alexander, , 1920. Lateral lobes on abdominal segments conical; on segments I and II-VII, three and four pairs, respectively, anteriormost pair smallest (Alexander, 1920). Ventral lobes small, finger-shapd (Alexander, 1920). Anal segment with one pair each of dorsal, lateral and ventral lobes. Spiracular field with hair fringe (Alexander, 1920).

Larval behaviour
Defensive behaviour: Larvae of P. tipulina are usually passive and not responsive, even when poked. However, larvae of P. replicata curl up their body and clasp the moss shoot they are on when alarmed; that way, the larvae may avoid being washed away in flowing water (Haake, 1922;Peus, 1952). Larval cuticles of Phalacrocera are often armoured with ectobionts, such as algae, diatoms and other freshwater microorganisms (Fig. 6G), and even with gastropods (e.g. Planorbis O.F. Müller, 1774) (Miall & Shelford, 1897).
Locomotion and attachment mechanisms: Larvae live underwater and cling to aquatic mosses, making use of their ventral lobes. When feeding or resting, the larvae normally keep themselves adhered to the substrate in such way that the head capsule remains entirely within the thorax, but making the mandibles accessible to the leaf surface. This attachment mechanism also plays a role in locomotion. Larvae fix themselves to the substrate at the caudal end, straighten up sharply and move the fore-body with a slow back and forth motion (Fig. 16D, E).
phalaCroCera repliCata (linnaeus, 1758) (figs 5J, 10d, 18a, B) Life history: Larvae were found to live in stagnant, dystrophic waters or acid pools of moors or woodlands where aquatic mosses, such as Hypnum, Fontinalis and Sphagnum, were abundant (Bengtsson, 1897;Brindle, 1967;Brinkmann, 1997;Stubbs, 2010). Phalacrocera replicata had a single brood in a year (Bengtsson, 1897;Müggenburg, 1901;Alexander, 1920); admittedly, they might have a second brood (Miall & Shelford, 1897). Bengtsson (1897) confirmed that P. replicata moult at least eight and possibly ten times during their development. The first moult occurred 4-6 days after leaving the eggs (Bengtsson, 1897), which did not significantly affect the external or internal appearance. The second moult occurred at about 18-20 days, and the larvae dramatically changed in appearance (Bengtsson, 1897). First-instar larvae started feeding on the host-plants immediately after hatching; they scraped only the upper layer of leaves (Peus, 1952). They overwintered as larvae and thus endured long and severe cold under the ice (De Geer, 1773;Miall & Shelford, 1897;Wesenberg-Lund, 1915). The larvae were also able to live a long period of time (c. 5 days) in anoxic conditions (Miall & Shelford, 1897) and they could sit down deep in water 1 m in depth for a while in autumn (Wesenberg-Lund, 1915). Pupae were much more active than larvae and when disturbed, wriggled about and bent their bodies almost into a circle by flexion of the abdomen (Miall & Shelford, 1897). Usually, the pupae rested vertically, with the mesothoracic horns reaching the surface of the water (Miall & Shelford, 1897;Hemmingsen, 1952;Peus, 1952), by firmly grasping floating plants with its dorsal abdominal hooks (Miall & Shelford, 1897). Although the pupae could survive asphyxiation for six hours (Miall & Shelford, 1897), they occasionally floated at the water surface through weeds by abdominal movement, and then recovered their original position for respiration (Miall & Shelford, 1897).
Oviposition behaviour: Larvae were found in ponds in which moss vegetation flourishes and currents keep the water in a constant motion (Hemmingsen, 1952). Females walked on the water surface with the posterior end of the abdomen pricking into the water or into submerged moss (Hemmingsen, 1952). Female adults deposited the eggs in the leaf axils of submerged moss, usually singly (Miall & Shelford, 1897), or four to six egg batchs in each pocket of concave leaves of Fontinalis moss (Wesenberg-Lund, 1943); the number of eggs laid in each leaf seemed to depend on the leaf size (Hemmingsen, 1952). Females laid c. 60 eggs in total (Miall & Shelford, 1897).
Final-instar larva: Length 25 mm (Brindle, 1967;Alexander, 1920). Body colour on dorsal side dark brown or lighter, showing dimorphism among individuals at final instar (Wesenberg-Lund, 1915;Brinkmann, 1997;Pujante et al., 2016); dorsally with pale, indistinct stripes (Miall & Shelford, 1897), and ventrally whitish. Body with extremely long cuticular lobes (Fig. 16A). Prothorax with each one pair of dorsal and lateral lobes; ventral lobe absent. Meso-and metathorax with dorsal and lateral lobes, filiform and long, two and one pair(s), respectively; one pair of ventral lobes, small and conical. Dorsal lobes on abdominal segments filiform and long, with length exceeding length of segment; segments I-VII with two pairs of lobes, anterior pair simple and posterior pair deeply bifurcated (Fig. 5H). Lateral lobes on abdominal segments simple without bifurcation, filiform and long; abdominal segment I with two pairs; abdominal segments II-VII with three pairs, anteriormost and posteriormost pairs shorter than middle pair, with longest pair approximate dorsal lobes in length. Ventral lobes on abdominal segments simple; segments I, II and III-VII with two, three, four pairs of lobes, respectively, posteriormost pair longest; additionally, on segments I-VII, single papilla-like lobe present near posterior end of segment. Anal segment with three pairs of lobes in total (Fig. 16B): one pair of dorsal lobes (dl), filiform and simple, bearing a short, sclerotized accessory lobe (dac) at the base with dark pigmentation; lateral lobe absent; one pair of ventral lobes (vl), blackened distally. Anal papillae soft and filiform, as single papilla (anp) on each side of the anus (Fig. 8D). Spiracular field without hair fringe (Fig. 10D). Elongated cuticular lobes internally containing tracheal branches (Bengtsson, 1897;Peus, 1952).

Host-plants:
This species feeds on submerged or aquatic mosses, including Sphagnum s p.
Larval behaviour: Larvae cling to moss stems using their large anal hooks to secure the body and sway it from side to side, as if to promote respiration (Miall & Shelford, 1897). They creep from stem to stem by grasping with the mandibles and the anal hook alternately; when alarmed, they curl up like a caterpillar (Miall & Shelford, 1897).
phalaCroCera tipulina osten sacken, 1865 (figs 3e, 5k, 7g-i, 10e, 11e, f, 18c-h) Life history: This species occur in shaded seepage areas surrounded by deep grasses and clumps of ferns; the larvae occur in stagnant pools with brown and murky water due to the acidification by Sphagnum mosses (Byers, 2002; this study). The number of moults is unknown. This species can be multivoltine, as larvae with greatly variable sizes are observed both in February and November. The larvae are motionless and hang on shoots of Sphagnum mosses; they pupate near water surface.

Egg: Unknown.
Final-instar larva: Length 25 mm (N = 3). Body pale brown to yellowish on dorsal side, without distinct markings (Fig. 18A, B); whitish on ventral side (Fig. 18C). Prothorax devoid of cuticular lobe. Mesoand metathorax with two pairs of dorsal and lateral lobes, filiform; one pair of ventral lobes, small and conical. Dorsal lobes on abdominal segments filiform and simple, longest one as long as body width; segments I and II-VII with two and four pairs, respectively. Lateral lobes simple; abdominal segment I with three pairs of lobes, second pair longest; segments II-VII with four pairs of lobes, third pair longest. Ventral lobes on abdominal segments simple; segments I and II-VII with three and five pairs of lobes, respectively, posteriormost pair longest; additionally, only on abdominal segment VII, single papilla-like lobe present near posterior end of segment. Anal segment possesses five pairs of lobes in total (Fig. 18F): one pair of dorsal lobes (dl) bear accessory lobes (dac) at the base, with sclerotized area at the inner part and with macrosetae at the tip (Fig. 18H), and also basally connects with dorsomedial lobe (dm) projecting laterad; lateral lobes one pair; ventral lobes one pair, conspicuously blackened on distal side, bearing one sharp, recurved apical hook with three macrosetae (Fig. 18G). Anal papilla (anp) single, short and stout, placed at proximal end of anus (Fig. 18F). Spiracular field with hair fringe (Fig. 10E). Elongated cuticular lobes internally containing tracheal branches.

Host-plants:
This species feeds on an unidentified species of Sphagnum (Sphagnales: Sphagnaceae) in a population at Mountain Lake, Virginia, USA. At the same locality, Fontinalis dalecarlica (Hypnales: Fontinalaceae) is common in a gently flowing stream nearby, but is not used by P. tipulina. The validity of moss mimesis in Cylindrotominae is revisited, which Thienemann (in Lenz, 1919) previously cast doubt on. He weighed up two possibilities for explaining why the cylindrotomine larvae visually resemble mosses: the resemblance to mosses 'is causally related to the life in the mosses', by which he meant 'adaptive resemblance' (Sykes, 1905;Starrett, 1993); the alternative is that the mossmimetic forms evolved coincidentally ('non-adaptive resemblance'). Thienemann consequently disputed the first possibility; namely, he alleged that the cryptic traits (i.e. body colour and special integument) are not the products of cryptic mimicry favoured by natural selection, but are the incidental result of developmental inertia. Such an idea was based on two main reasons related to larval structures and ecological background. First, similar cuticle structures are seen in some distantly related dipteran groups that are not mossfeeders. Second, Thienemann pointed out the lack of evidence for predation by visually oriented predators. Birds are apparently absent in cylindrotomine habitats (i.e. moss tufts), which is inconsistent with an evolutionary scenario of anti-predator adaptation in such environments. Although Thienemann (1919) was the first to explore the link between ecological background and evolutionary consequences, the antipredator argument has long been dismissed by many authors subsequently (Alexander, 1920(Alexander, , 1927Haake, 1922;Brindle, 1967;Hemmingsen, 1968;Brinkmann, 1997; but see: Peus, 1952). The 'non-adaptive resemblance' of Thienemann is revisited, focusing on two pertinent issues. First, moss resemblance is not achieved by just the elongated cuticular lobes, but by complex character sets: coloured haemolymph, cuticular appendages and dark patterning, as well as protective and defensive behaviour. The cuticular appendages are among the most unique characters of Cylindrotominae, which are not seen in bryophyte-feeders in Tipulidae and Limoniidae, such as Dolichopeza, Tipula (Tipulidae), Limonia, Erioptera and Gonomyia (Limoniidae) (Byers, 1961;Coulson, 1962;Coulson & Whittaker, 1978;Caspers, 1980;Leffler, 1988;Todd, 1993;Smith et al., 2001;Arroyo-Rodríguez et al., 2007;Stubbs, 2010;Glime, 2017a). A parallel dipteran example is in the Syrphidae (Cyclorrhapha), the larvae of which use certain colours and shapes to blend into particular substrates, such as bark, pine needles and even bird droppings (Rotheray, 1986(Rotheray, , 2019. The evolution of complex traits implies that they facilitate the overall resemblance to plants as a product of natural selection under the threat of natural enemies. In fact, protective concealment is often achieved by complex traits, including coloration, patterning, texture, outline modification and shadow reduction/disruption (Waldbauer, 1988;Ruxton et al., 2018), which can be consistent with the case of cylindrotomines.

DISCUSSION
Second, adaptive appendages do not necessarily entail evolutionary novelties. Even those structures considered novelties can be generated through accumulated orchestrated changes in multiple pattern elements (Brigandt & Love, 2012;Almundí & Pescual-Anaya, 2019), as in the example of wing patterning in dead leaf butterflies (Suzuki, 2013;Suzuki et al., 2014;Suzuki, 2017). Conversely, ventral prolegs and similar structures are obtained, with different developmental mechanisms, in many pterygote orders (Nagy & Grbic, 1999;Suzuki & Palopoli, 2001;Bitsch, 2012). The lack of evidence for evolutionary novelty per se thus does not rule out the possibility that a structure is favoured by natural selection. In fact, the homology of the special cuticle modification of Cylindrotominae with similar structures in dipterans is highly questionable. Some findings on the structures of the cuticular lobes (see 'Morphological variation of cuticular lobes') imply that the cuticular lobes in Cylindrotominae can be a unique structure among other dipterans.
These aspects undermine 'non-adaptive resemblance' of cylindrotomines, although the effectiveness of moss mimicry has not been empirically tested. Development and maintenance of the cuticular outgrowths are bound to come at a cost to efficiency in both manoeuvrability and metabolism (Kaiser et al., 2007;Maddrell, 2018). The larvae, therefore, are likely to gain benefits from having these structures over possible costs, otherwise they would just be disadvantageous hinderances. Admittedly, protection against predators may not be the only driving force for the evolution of the elongated cuticular lobes, but physiological reasons may also exist. Peus (1952) suggested that these structures may primarily be used for physiological or behavioural purposes and may play only a secondary role in mimicry. The mechanical and functional explanations of mimicry-related traits are provided later.

host-plants
From the standpoint of the evolution of mimicry in Cylindrotominae, it is important to understand whether the moss-feeding species maximize their resemblance to host-plant species, as previously believed by many authors (Osten Sacken, 1897;Müggenberg, 1901;Alexander, 1915), or they only resemble mosses in general. This shall be viewed from two different perspectives: (1) body form matches with host-plants on an evolutionary timescale, or (2) plastic characters are influenced by host-plants on a developmental timescale.
First, the resemblance to a specific model requires the high degree of host-plant specificity in moss-feeding species. Whereas Cylindrotoma are associated with an extremely broad range of low-growing herbaceous angiosperms, each of the moss-feeding species in Diogma, Liogma and Triogma tend to be restricted to one specific family or order of mosses. Based on the extensive field survey and rigorous identification presented in this work, many species of Liogma feed on several Mniaceae genera (Plagiomnium, Mnium and Rhizomnium), which belong to Bryales; in contrast, L. mikado, Diogma glabrata and T. kuwanai are associated with various moss taxa belonging to Hypnales.
However, in examined terrestrial moss-feeding species, host-plants for feeding, and by adults for oviposition, are not necessarily coupled. In the case of T. kuwanai, the larvae did not even survive on the plants on which they are laid. The host-plant fidelity of moss-feeders thus are not expected to be high, unlike some herbivorous insect lineages that show the correlation between host-plants as oviposition substrates and larval foods (Jaenike, 1990;Thompson & Pellmyr, 1991;Gripenberg et al., 2010). Additionally, the appearances of some terrestrial moss-feeding species, which exhibit most elaborate patterns and forms, do not seem to be matched to those of their principal host-plants, i.e. Mniaceae. The 'evolutionary fit' hypothesis is, therefore, not supported by my observations.
Second, the larval body colour can be adjustable to that of their host-plants on a developmental timescale. Host-plant use is one of the key factors determining body coloration in Lepidoptera (Poulton, 1892;Fink, 1995;Canfield et al., 2009;Greene et al., 2009;Boege et al., 2019). Plasticity in body colour allows larvae to be polyphagous (Poulton, 1892;Skelhorn & Ruxton, 2011). In some cylindrotomine species, the body colour is, in fact, a plastic character. Host-plant attributes should be taken into account as a potential factor affecting body colours and patterns.

camouflage devices
The cylindrotomines may use several strategies for reducing the probability of detection by predators: the coloured haemolymph, the dark pigmentation patterning and the elongated lobes. The utility of these cuticular lobes in crypsis is not fully understood, but they may altogether break up the outline of the animal and facilitate disruptive coloration, in collaboration with coloration and pattern. Two modes of mimicry, countershading and crypsis (sensu Quicke, 2017), are likely involved as mechanism of the moss resemblance. Countershading helps to make an organism seem less three-dimensional or make its shadows less conspicuous (Ruxton et al., 2004;Rowland et al., 2008;Rowland, 2009). A substantial level of variation can be found in these camouflage-related traits, such as the pigmentation patterns, and both arrangement and form of elongated cuticular lobes.

cryptic coloration and patterning
There was an inconsistency among previous authors in the views on which organs or tissues determine the body coloration. The green hue was previously thought to be the consequence of transparency of the food bolus in the alimentary canal (Bengtsson, 1897;Miall & Shelford, 1897;Alexander, 1920;Peus, 1952), fat bodies (Bengtsson, 1897) or the uppermost zone of the integumental epicuticle (Bengtsson, 1897;Müggenburg, 1901).
This study clearly demonstrates that body coloration of larvae fundamentally consists of two separable elements: (1) a green or brown hue and (2) dark pigmentation. The body colour is attained mainly by the haemolymph, whereas the integument is nearly transparent (Fig. 4A, B). Notably, some individuals change their body colour, from green to brown, at later stages. Even within local populations, the coloration can markedly vary (Figs 4C,D,12G,H,16E,F). The mechanisms underlying changes in body hue are currently unknown, but it has been hypothesized that these changes are triggered by the colour of the surrounding mosses (Müggenburg, 1901;Lenz, 1919;Haake, 1922) or by seasonal differences in temperature (Alexander, 1920).
Another element of coloration is the dark (grey-brown to nearly black) pigmentation on the dorsal surface. In most species, the cuticular pigmentation becomes visible after the second or later instars, in concert with increased coloration of the haemolymph. The pigmented line pattern is organized in moss-feeding species (Diogma, Liogma and Triogma) in the final instar (e.g. Fig. 4A) but is less evident in Phalacrocera (Fig. 18A, C) and Cylindrotoma (Fig. 12G). The dark markings are likely to be a countershading strategy (Kiltie, 1988;Ruxton et al., 2018). Dorsal darkening is commonly seen as an anti-predator defence, which improves background-matching camouflage when viewed from above against dark backgrounds under sunlight (Behrens, 2009;Rowland, 2009;Allen et al., 2012;Cuthill et al., 2016), although it may also pertain to other physiological functions, such as the protection from UV or abrasion and thermoregulation (Rowland, 2009), especially in some species with diffused pigmentation patterns (T. kuwanai and L. serraticornis).
The variation and polymorphism of colour elements between and within species presumably play an important role in concealment (Turner, 1961;Allen, 1988;Booth, 1990). These two elements of colourgreen (brown) hue and dark pigmentation patterning -may enhance each other to achieve disruptive camouflage, possibly in combination with the dorsal elongated lobes. Although most tipulid larvae lack such distinct coloration and markings, mature larvae of Dolichopeza spp. (Tipulidae: Dolichopezinae), which feed on terrestrial mosses, also possess a similar green hue and dark pigmentation on the dorsal integument (Byers, 1961). The co-occurrence of similar body coloration and patterning in distantly related clades of Tipuloidea suggests that these traits are effective as a survival strategy, especially for terrestrial moss-feeders.

morphological variation of cuticular loBes
The arrangement of cuticular lobes is, below, compared to similar structures, such as other integumentary outgrowths or appendages such as prolegs, in other dipteran groups and in endopterygotes. The cuticular lobes can be viewed in two ways: from longitudinal and horizontal aspects. The number, position and form of the lobes are longitudinally divided into six groups of consecutive body segments (Fig. 5). The lobes on prothoracic and anal segments often exhibit irregular patterns, presumably because they are associated with feeding and excretion, respectively. The lobes on thoracic and abdominal segments are not soundly differentiated. The dorsal lobes on the abdominal segment I tend to represent the intermediate state of the neighbouring body segments (i.e. metathoracic and abdominal segment II) in number and form. The lateral lobes also are arranged continuously in a row from the mesothorax through abdominal segment VII, without marked differences. The longitudinal configuration of the lobes in cylindrotomines does not follow the general pattern in Diptera. In most dipterans, the prolegs are absent on the meso-and metathoracic segments (Hinton, 1955a). From the horizontal aspect, each of the dorsal, lateral and ventral lobes in cylindrotomines have some distinct features: in particular, only the lateral lobes have both intrinsic and extrinsic muscles, whereas the dorsal and ventral lobes are basically thickened cuticles. Thus, the lateral lobes are essentially different from the dorsal and ventral lobes (see also 'Insight into locomotory and attachment mechanisms').
B a s e d o n t h e a b o v e -m e n t i o n e d u n u s u a l characteristics, it is suggested that the cuticular lobes in the cylindrotomines may represent a unique structure among Diptera. Such proleg-like structures repeatedly originated across many lineages in Panorpida (Hinton, 1955a;Bitsch, 2012), although the internal structures are poorly understood. The integumental outgrowths have been found in many species of Bibionidae (Plecia Wiedemann, 1828 and Penthetria Meigen, 1803), Phoridae and Fanniidae (Fannia Robineau-Desvoidy, 1830), which tend to live under layers of decomposed leaves or in leaf mould (Banks, 1912;Morris, 1921Morris, , 1922Hardy, 1945;Hartley, 1960;Lyneborg, 1970;Ferrar, 1987;Rotheray & Gilbert, 1999). The fleshy projections are also apparently similar to those in some mecopteran larvae, such as Panorpa Linnaeus, 1758 (Panorpidae) and Bittacus Latreille, 1805 (Bittacidae); these lobes on dorsal, lateral and ventral sides lacked intrinsic muscles but had extrinsic muscles connecting with endosternal apodeme (Suzuki, 1991;Bitsch, 2012).
It should be noted that, in two fully aquatic species, Phalacrocera replicata and P. tipulina, the lobe characteristics are remarkably different from each other. Phalacrocera replicata have fewer and longer lobes than P. tipulina on dorsal, lateral and ventral sides (Fig. 5J). Phalacrocera tipulina lacks prothoracic lobes and has a pair of well-developed additional lobes (dorsomedial lobes) in the anal segment. The lobes of P. tipulina tend to be more similar to terrestrial mossfeeders than to P. replicata, in terms of number and configuration (Fig. 5K). The morphological variation between these species implies that even fully aquatic species can be at different levels of dependency on the aquatic habitat, although it remains unclear. Both species have a pair of accessory dorsal lobes (dac) near the bases of the dorsal lobes on the anal segment, the presence of which can be regarded as an autapomorphy of Phalacrocera. The cuticular surface of the accessory dorsal lobes (Fig. 18H) are found to be different from those of the dorsal lobes in the anal segment. They were suspected to be a sensory organ (Bengtsson, 1897), but this has not been tested.

respiratory gill hypothesis revisited
The possible function of cuticular lobes as a device for respiration is discussed in the light of three common methods for gas exchange: tracheal system connected to the spiracles (Klowden, 2013); cuticular surface, including plastrons, which are water-repellent structures that allow arthropods to use their air-filled tracheal system underwater (Thorpe, 1950;Marx & Messner, 2012); and tracheal gills (Wigglesworth, 1945). In general, terrestrial species of tipulids have a pair of posterior spiracles on the anal segment, through which they acquire oxygen (Brown, 1910;Gerbig, 1913;Whitten, 1959Whitten, , 1960. However, the structure of spiracles in cranefly larvae tend to differ depending on the habitat (aquatic vs terrestrial), and spiracles can even be non-functional in some aquatic tipulids (Brown, 1910;Gerbig, 1913;Wardle, 1926;Keilin, 1944;Pritchard & Stewart, 1982).
The respiratory methods of cylindrotomids was, somewhat arbitrarily, believed to be various according to the wide range of habitats. Particularly, there was an extensive argument over whether the spiracles in Phalacrocera are functional or not (Bengtsson, 1897;Miall & Shelford, 1897;Haake, 1922). Previous authors claimed that P. replicata does not take in atmospheric oxygen (Miall & Shelford, 1897;Haake, 1922). The spiracles in Phalacrocera were once assumed to be completely closed (Miall & Shelford, 1897), but later authors found that they are open at the central plug (Bengtsson, 1897;Haake, 1922) and thus concluded that they are probably functional.
The SEM images of both terrestrial (Cylindrotoma, Liogma) and aquatic (Phalacrocera) species show that all examined species maintain the posterior spiracles opening both at the central scar plug and aeropyles (Fig. 11A-F). Importantly, the SEM images of the spiracles of P. tipulina are consistent with the idea that they are functional. This is also supported by the larval behaviour, as they occasionally surface (Fig. 18E). In P. tipulina, the spiracles lie in the retractable spiracular field (Fig. 10E) and are surrounded by a hair fringe (Fig. 11E); these structures are generally thought to prevent water getting into spiracles (Pritchard & Stewart, 1982). However, in this regard, P. replicata does not bear hair fringe near the spiracular field (Fig. 10D).
The ultrastructure of the integument (Fig. 7G-I) shows that the cuticular surface of P. tipulina does not bear bristles, hairs or microtrichia, which function for retaining air (Thorpe & Crisp, 1947;Thorpe, 1950). The integument apparently lacks any potential aerial chambers (e.g. a three-dimensional meshwork structure with minute pores) that are internally connected to the tracheal system (Crowe & Magnus, 1974;Fielden et al., 2011;Davis & Landry, 2012). Some characteristic ultrastructures of the integument, such as microtubercles and micropapillae, are hard to interpret, as they are not comparable to any orthodox plastron structures known in arthropods (Hinton, 1947(Hinton, , 1955b(Hinton, , 1957(Hinton, , 1967(Hinton, , 1968(Hinton, , 1976(Hinton, , 1981Marx & Messner, 2012). Hence, any convincing evidence for cutaneous respiration with plastrons cannot be found in Phalacrocera. The lack of cuticular plastrons may reflect their fairly unusual habitat. Phalacrocera inhabits low-oxygen, acidic water, including Sphagnumdominated peatlands (Brauns, 1954;Desrochers & Van Duinen, 2006;Stubbs, 2010). Their host-plants, Sphagnum mosses, acidify water during the process of peat production (Schofield, 1985) and these mosses favour stagnant pools with low levels of dissolved oxygen where their spores have higher germinability (Feng et al., 2018). The plastron of insects only operates in well-oxygenated water, because when insects are in water with low oxygen levels, oxygen diffuses from air bubbles trapped in the plastron into the water, which is the reverse of its function as a gill (Klowden, 2013).
The elongated cuticular appendages as devices for gas exchange have been substantially debated (De Geer, 1773;Zeller, 1842;Bengtsson, 1897;Miall & Shelford, 1897;Brinkmann, 1997). This hypothesis was proposed particularly for aquatic species of Phalacrocera, which were armoured with extremely long lobes (Fig. 18A). The lobes in Phalacrocera are morphologically comparable to tracheal gills in some truly aquatic moths (Crambidae: Acentropinae), of which the gills are filamentous, simple or branched processes (Welch, 1922;Wichard et al., 2002;Lancaster & Downes, 2013;Graça & Solis, 2018;De-Freitas et al., 2019), and each lobe contains a trachea (De Geer, 1773;Peus, 1952;Miall & Shelford, 1897; this study). However, structural aspects of the lobes cast doubt on their role as respiratory appendices. In general, cutaneous respiration occurs through thin-walled, delicate parts of the cuticle (Snodgrass, 1935). In Phalacrocera, the cuticle is thick and solid, so it is unlikely that it allows oxygen to pass through, as Miall & Shelford (1897) pointed out. Furthermore, the tracheae contained in the appendages are not ramified as much as those in the anal gills (i.e. anal papillae) of some craneflies (Alexander, 1920;Young, 2004). Another possibility is that they act as blood gills, as in some caddisflies (Morgan & O'Neil, 1931;Thorpe, 1933), although this has not been sufficiently examined either. It is thus suggested that both terrestrial and aquatic species can acquire oxygen through the spiracles, although the possibility of elongated lobes in Phalacrocera acting as respiratory device have not been tested.
insight into locomotory and attachment mechanisms Historically, much attention has been paid to the function of cuticular lobes in locomotion. Ventral lobes were previously regarded as a major attachment mechanism (Haake, 1922;Peus, 1952). In particular, the roles of the ventral lobes on the anal segment, often called 'pseudopodium' or 'fixing apparatus', were emphasized (Osten Sacken, 1869;Alexander, 1920;Haake 1922;Peus, 1952), as they are apparently used for attaching the body to plant substrates. Hinton (1955a) found that the ventral lobes of cylindrotomines lacked intrinsic muscles and thereby concluded that they were not prolegs; internal structure of the dorsal and lateral lobes was not mentioned, or perhaps it might not have been examined.
This study reveals that lateral lobes are distinguished from dorsal and ventral lobes by the presence of intrinsic and extrinsic muscles (see 'Internal structures'). The lateral lobes contain some fibres of intrinsic muscles throughout the thoracic and abdominal segments (Fig. 8A-D). The intrinsic muscles near the tip of the lobes are closely positioned in a dorsoventral direction, whereas those near the base are inserted in an oblique position (Fig. 7C). The extrinsic muscles are inserted into the lobes at one end at an oblique angle (black arrows in Fig. 8B, E, F). However, the dorsal and ventral lobes tend to be simply thickened cuticles (Fig. 8C, D), except that the ventral hooks on the anal segment contain a huge amount of intrinsic muscle internally (Fig. 9C).
The musculature is particularly relevant to locomotion, casting new light on the traditional functional scenarios of the lobes. Terrestrial mossfeeding species are crawlers (Supporting Information, Videos S1, S2), as is a typical locomotion strategy for many legless dipteran larvae (Roberts, 1971;Pritchard, 1985;Rotheray, 2019). The crawling can be explained by three phases of muscle contractions, in refenrece to the well-studied examples in caterpillars (e.g. Pryor, 1951;Hughes, 1965;Gillott, 1995). First, when the dorsal longitudinal muscle and transverse muscles contract, a body segment is shortened dorsally and the posterior end is lifted up. This results in the segment behind being lifted from the substrate. Second, intrinsic and extrinsic muscles within the lateral lobes can contract, which results in lifting the ventral side of the segment from the substrate. At this moment, the intrinsic muscles in cylindrotomines can be comparable to dorsoventral muscles in caterpillars' prolegs. The extrinsic muscles, comparable to retractor muscle of prolegs, can particularly be important in propelling the larva forward (Law et al., 2014). Finally, contraction of the ventral longitudinal muscles, in combination with the relaxation of intrinsic and extrinsic muscles, moves the segment forward and down to the substrate. As above, the lateral lobes can be a functional analogue of ventral prolegs in the caterpillar.
The larval locomotion indicates that the ventral lobes serve only as friction points (Fig. 13D) and the ventral hooks are not fully engaged, at least when the larva is crawling on a wet, flat surface (Supporting Information, Video S2). The use of the ventral hooks as fixing apparatus may be conditional, for example, when larvae ascend moss stems in arid conditions. The reason why intricate musculature does not occur in the ventral lobes, but does occur in the lateral lobes, is unclear. One possible explanation for this can be that the distribution of muscles on the lateral side may enable the larvae to stabilize their body when they lift it, as it may contribute to keeping the centre of mass relatively low. Such a design can be efficient in terms of balance, unlike lepidopteran larvae of which the prolegs have to support the body mass from the ventral side (Van Griethuijsen & Trimmer, 2014). Studies of comparative anatomy, cytology of musculatures and ontogeny are warranted to understand how the cuticular lobes evolved. The musculature of tipuloids is poorly studied and thus do not allow us to make a comparison yet (e.g. Eldridge, 1966).
In contrast, Cylindrotoma move by inching, which is a unique locomotory strategy for cranefly larvae (Fig. 11I-K). Cylindrotoma use the prothoracic integument ( Fig. 11C; Supporting Information, Video S1) and ventral anal lobe (Fig. 11E) as anterior and posterior anchors, respectively. The 'suction-cup' principle involving the prothoracic integument can be seen as a strategy for attachment and feeding on herbs. The internal structure of Cylindrotoma is not described herein.
Taken together, some lobes partly constitute the device for locomotion. Especially, the lateral lobes can assist locomotion in terrestrial moss-feeding species. However, the lobes lack any special, single-purpose attachment mechanisms, such as crochets, dense setae/hairs and welts. The attachment mechanisms are instead associated anteriorly with their feeding apparatus. The larvae use the mandibles effectively for attaching to a moss shoot, while the head is completely sheltered in the prothoracic integument (Fig. 2B).

other possiBle functions of cuticular loBes
Other than the potential functions touched on above (i.e. respiratory gills, locomotory and attachment devices), some other functions for the modified integument have been deduced from similar structures in examples of insects. Thick cuticles can also be a useful aid for feeding (Nation, 2008), as mechanical protection against predators (Gross, 1993;Gorb, 2002;Boevé & Angeli, 2010) or abiotic stress (e.g. desiccation and abrasion) (Smith, 1989;Gorb, 2002;Nation, 2008). Integuments tend to be tougher and thicker in terrestrial species (Diogma, Liogma) than in aquatic species (Phalacrocera spp. and Triogma trisulcata). Although tremendous efforts have been made in pursuit of the parasitoids (Cameron, 1918;Peus, 1952;this study), there is only a single record known (De Rossi, 1876); the low infection loads may be a consequence of the extremely thick cuticle in terrestrial species, which can prevent parasitoids attacking the larvae. The integumental surface is distinctively warty in many species. In particular, Phalacrocera spp. and L. brevipecten have minute unicellular spicules (acanthae) on the surface of the dorsal lobes (Fig. 7F, I). Although the physical properties and function of these spicules are so far unknown, they may be related to water saturation because these species live in aquatic or frequently inundated habitats.
It is noteworthy that the anal papillae are hypothesized to be used for osmoregulation (Brinkmann, 1997) or respiration (Wardle, 1926). The osmoregulatory function was originally proposed for Tipulidae, mainly since the number and size of anal papillae seem correlated with aspects of their habitats (Chiswell, 1956;Brindle, 1957), although this has not been empirically tested (Pritchard, 1983). The anal papillae in Tipula greatly differ from those in non-biting midges and mosquitos, which are known to have an osmoregulatory function (Thorpe, 1933;Wigglesworth, 1933Wigglesworth, , 1938Koch, 1938;Lawson, 1951;Bradley, 1987;Te Velde et al., 1988). Recently, Krivosheina (2005) concluded that the structure and function of anal papillae could vary among dipteran groups.
In the case of Cylindrotominae, the anal papillae vary in form and in number between species and they may be partly retractile (Diogma, Liogma and Triogma) or not (Cylindrotoma and Phalacrocera). In L. brevipecten, the anal papillae tissue is stained differently from that of the integument (Fig. 10A), which may correspond to a hind gut origin of the anal papillae (Edwards & Harrison, 1983). Moreover, anal papillae are not extensively tracheated, nor is considerable circulation of haemolymph seen. The anal papillae in cylindrotomines, therefore, are not comparable to the 'anal gills' of Tipulidae (Brown, 1910;Chiswell, 1956;Pritchard, 1983). synthetic view on herBivory, camouflage and ecomorphology The biology and larval morphology of cylindrotomines give insight into two aspects of evolutionary ecology of mimicry: interactions with host-plants and enemies, and functional morphology of the camouflage devices.
Evolution of herbivory generally accompanies a series of adaptations, such as morphology, behaviour and physiology: devices and strategies for locomotion and attachment or efficient ingestive/digestive systems (Strong et al., 1984;Bernays, 1998). Some morphological features in larvae and adults are apparently linked with the host-plants. In the terrestrial moss-feeding larvae, the body form and crawling movement may be correlated with their habit of clinging to mosses and slowly crawling on the slippery moss mat within some small patches. The female adult ovipositor (cerci) of Cylindrotoma, used for endophytic oviposition (Hemmingsen, 1952(Hemmingsen, , 1960Peus, 1952), can also be seen as an effective strategy for angiosperm-feeding of the larvae. Endophytic oviposition is nearly impossible for moss-feeding taxa because moss leaves consist of one cell-layer thick and the stems are too small and weak to support the load of the female's terminalia. The differences in a series of characters -morphology, modes of attachment and locomotion in larvae, and that of oviposition in adults -underscore how colonization to some different plant groups (angiosperms and mosses) have an impact on insect evolution.
With regard to the evolutionary background of moss mimesis, it is hardly explicable, as ground moss patches are regarded as an enemy-free space. The lack of visual predators attacking cylindrotomines can be explained by two circumstances: such predators may have simply been overlooked because the ground moss patches have been more poorly studied than other vegetation types; otherwise, there may be virtually no visual predators that attack cylindrotomine larvae in the present, due to the successful camouflage. The larvae of Cylindrotoma show a stereotypical behaviour for protection and defence when disturbed, whereas the larvae of terrestrial, moss-feeding species are sluggish and do not respond to disturbances. The larvae of Cylindrotoma, living in open habitat, may be more prone to predation than moss-feeders (Shaw & Askew, 2010;Humphreys & Ruxton, 2018).
The cylindrotomines provide a unique opportunity to study the function of proleg-like appendages in an ecological context. It is surprising that, for soft, turgid appendages in many insect larvae, structures and functions are poorly known. Disentangling the link between ecological setting and functional structures, properly interlinked by mechanistic explanations, is a future task for understanding how these curious designs evolved.

CONCLUSION
The larvae of Cylindrotominae achieve complex crypsis via multiple elements, including coloured haemolymph, dark dorsal patterning, elongated cuticular lobes, as well as defensive and protective behaviour. The moss-mimetic forms of cylindrotomines are thus considered to offer a protective resemblance to their surroundings. However, the evolution of mimicry in cylindrotomines is paradoxical because knowledge about visual predators is lacking. The orthodoxy of ground moss patches as an enemy-free space should be challenged, in a broader context.
The morphological and behavioural traits vary among taxa in concert with their habitats. Species within Cylindrotoma constitutes the group of terrestrial herb-feeders that use a wide range of angiosperm hosts, altogether encompassing 15 orders, and exhibit evident stereotypical protective behaviour (i.e. debris-carrying, thanatosis and katalepsis). This study clarifies that two distantly related moss groups are used by the terrestrial moss-feeders: Mniaceae (Bryales) are the principal host-plant clade, which are associated with Liogma (except L. mikado); in constrast, most of the recorded host-plants of D. glabrata, L. mikado and T. kuwanai belong to Hypnales. Nevertheless, the random ways of female oviposition suggest that they do not show high fidelity to their host-plants. This study thus implies that the cylindrotomines only imitate the general appearance of mosses, but not some particular moss taxa. Evolution of moss resemblance may be operated by mechanisms on evolutionary and developmental timescales. Future studies should address the question as to how the structural elements (e.g. forms of cuticular lobes) and body coloration come into play. Some functional hypotheses for elongated cuticular lobes are challenged. Although the extremely long lobes in Phalacrocera were believed to be tracheal gills, they are not likely to be the primary respiratory method; however, it is still possible as a subordinary method, as functional spiracles are present and any apparent plastrons are not found within the cuticular lobes. The possible roles of the lateral lobes in locomotion are emphasized, based on observation of the musculature and movement. The integumental appendages may also play some other roles, such as avoidance of parasitoid attack. Overall, the modified integuments (elongated cuticular lobes) may not only facilitate the resemblance of cylindrotomines to host-plants but also simultaneously may be used for different purposes.
To address the evolution of cryptic mimicry in Cylindrotominae, future research should investigate the ecological network (i.e. interactions with predators and host-plants) they are part of, and infer the trait evolution built on robust phylogenetic framework. Moreover, the biology and morphology of Stibadocerinae, especially the immature stages, should be explored. ACKNOWLEDGEMENTS I express my gratitude to Dr Makoto Kato for igniting my interest in these fascinating bryophyte-feeding insects and for providing some materials with valuable biological information. This paper greatly benefited from the help of Dr Levente-Péter Kolcsár who shared unpublished data, identified adult craneflies in field photos and provided critical comments in many fruitful discussions. I also thank Takayuki Ohgue for identifying some bryophytes; Honoka Shindo for her dedication to collections and rearing of larval cylindrotomids; Noriyuki Muro and Esko Viitanen for providing some larval specimens; Yoshiko Yamane, Dr Conrad C. Labandeira, Yumiko Imada and Yasuharu Imada for assisting in fieldwork; Dr Takeshi Kaneta was instrumental in training histological techniques; Dr Kazutaka Yamada and Dr Tadashi Ishikawa for identifying the hemipteran nymph; Dr Makiko Fukui, Manabu Sawa and Manami Nitta for initial assistance of SEM operation and visualization; Dr Nikolai Paramonov for sharing with me a copy of a Russian paper. I thank two anonymous reviewers who made a tremendous effort to improve clarity and accuracy of the previous versions of the manuscript. I gratefully acknowledge the support of the interlibrary loan staff of Ehime University Library for collecting literature. I am also grateful to the staff of the Great Smoky Mountains National Park and of the biological station of Mountain Lake, Virginia, USA. I also thank Dr Robert Davis for improvement of the English of

SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article at the publisher's web-site. Table S1. Specimens used for morphological and/or biological assessment in this study, with host-plant records. Taxon name in host-plant list followed by asterisk (*) were detected as fragments in faeces of reared larvae; those without asterisks are confirmed to feed on in the field and by rearing experiments. Table S2. List of literature cited for describing biology, larval morphology, and behaviour of adults and larvae. Video S1. Movie showing presumable suction cup of Cylindrotoma japonica making use of circular rim of prothorax, ventral view. https://youtu.be/dMSjisOKzV8 Video S2. Movie showing peristaltic movement of Liogma brevipecten on a wet glass petri dish, ventral view. https://youtu.be/S1gywcO3ces