The thorax morphology of Epiophlebia (Insecta: Odonata) nymphs – including remarks on ontogenesis and evolution

The species of Epiophlebia are unique among the recent Odonata in showing a mixture of morphological characters of dragonflies (Anisoptera) and damselflies (Zygoptera). The status of the four described extant species of Epiophlebia is disputable from a genetic as well as from a morphological point of view. Here we present an analysis of the thoracic musculature of different nymphal instars of Epiophlebia laidlawi and Epiophlebia superstes to elucidate their morphology and ontogenetic development. In total, 75 muscles have been identified in the thorax of Epiophlebia. This represents the highest number of thoracic muscles ever found in any odonate. It includes six muscles that are reported for the first time for Odonata, and three of these are even new for Pterygota. In total, our results indicate that Epiophlebia has the most ancestral thoracic morphology among Odonata.

Insect flight and thorax morphology. The evolutionary success of pterygote insects can be attributed largely to the evolution of the ability to fly 17 . Despite of this evolutionary importance, the origin and evolutionary development of the insect flight apparatus are still only partially understood [17][18][19][20][21] . Within Pterygota, the Odonata are among those groups that show the most impressive flight skills 22 .
The mechanism of wing movement is realized in different ways among Pterygota: The Odonata have an exclusively direct flight mechanism where dorso-ventral muscles are attached to elements of the wing base, actuating the wings directly. Dorsal longitudinal muscles, which are a crucial part of the flight musculature of all other pterygote insects, are either extremely small or missing in Odonata 18,[23][24][25] . The mechanism that drives the wings in all other Pterygota works largely indirectly through deformation of the winged thoracic segments while the usually weaker direct flight muscles are mainly responsible for steering and flight control actions, e.g. adjusting the wing's angle of attack, etc.
As exclusively aerial predators adult Odonata depend even more on the performance of their flight apparatus than many other Pterygota 23 . Therefore, understanding its complex morphology is necessary for a better understanding of behaviour, phylogeny and evolution of this group.
The thoracic musculature of adult Epiprocta 1,23-25 as well as the pterothorax 18 , and to some degree the entire thorax 1,19 of adult Zygoptera have been comprehensively investigated. The morphology and development of the nymphal thorax of Odonata, however, have only been studied superficially 1,20,26,27 . The present investigation of the ontogenesis of the thoracic musculature of Epiophlebia nymphs will substantially supplement the hitherto available information, leading to a better understanding of the evolution and development of the odonate thorax.

Results
To compare the thoracic muscles of E. laidlawi and E. superstes with Anisoptera, Zygoptera as well as Neoptera, each muscle is identified according to the homologization proposed by Büsse et al. 18 and the muscle nomenclature proposed by Friedrich & Beutel 28 .
In total 75 muscles are identified in the thorax of Epiophlebia nymphs, 20 in the prothorax, 26 in the mesothorax and 29 in the metathorax (Table 1). This represents the highest number of thoracic muscles ever found in a species of Odonata 1,18,20,27 .
Detailed descriptions of the muscles together with information on the interpretation of their identity or homology are given in Supplementary data 1. Figures 1-3). For the skeletal elements of the thorax we use the nomenclature of Asahina 1 . Where necessary, this is supplemented with terms from Snodgrass 29 Figures 1& 2: pro-(F1), meso-(F2) and metafurca (F3)) are attachment points for several muscles (see Supplementary data 1).

Cuticle (Supplementary
On the postero-lateral surface of each coxa a short process is present, which serves as attachment point for the large pronator muscle of each leg. The tergites of the three thoracic segments (Supplementary Figure 3) are substantially different from each other. The prothoracic tergite is a broad plate covering the entire dorsum of the prothorax. In the centre of the prothoracic tergite a small spur, the first tergal apophysis, is present (Supplementary Figure 3: TEa1). The second, third and fourth tergal apophyses are represented by the segmental borders between pro-and mesothorax, meso-and metathorax and mesothorax and first abdominal segment, respectively (Supplementary Figure 3: TEa2, TEa3, TEa4). In the cranial half of the dorsal part of the mesothorax both pleurites arch towards each other and displace the mesothoracic tergite to a more caudal position. The meso-and metathoracic tergites are covered by the wing buds and each bears a stubby lateral process, which serves as attachment point for muscles (see Supplementary data 1). The pleurites (Supplementary Figure 1) are divided into the episternum and the epimeron. Whereas the prothoracic pleurite has no distinct apodemes or extensions, the meso-and metathoracic pleurites have prominent arched interpleural ridges. The metathoracic preepisternal apodemes arise just behind the intersegmental border. The spoon-shaped structures on both sides of the body extend towards the median axis and are connected by a transverse muscle (see Supplementary data 1), above the nervous system. Asahina 1 described a mesothoracic preepisternal apodeme in Epiophlebia superstes that serves as an attachment point for transverse muscles. Although these muscles could be identified in both species of Epiophlebia, determining the exact outline of the mesothoracic preepisternal apodemes was not possible in the specimens examined.

Discussion
In the following section the nymphal musculature of E. laidlawi and E. superstes is compared with that of zygopteran and anisopteran nymphs. Additional information is taken from the descriptions of Maloeuf 27 and Büsse & Hörnschemeyer 20 for Anisoptera, from Asahina's 1 work on E. superstes and from the analysis of the musculature of adult Zygoptera by Büsse et al. 18 . A comparison of muscle nomenclatures of different authors can be found in Supplementary table 1.
The muscle numbers used by Maloeuf 27 and Asahina 1 are given in parenthesis. An additional number in parentheses within the first set denotes the homologous muscle in the meso-or metathorax. Muscles not recognized by Maloeuf 27  Dorsal longitudinal muscles. The dorsal longitudinal muscle IIdlm1 (Fig. 1) is small and is missing in nymphal Zygoptera. However, it is present but very small, consisting of just a few fibres, in adult Zygoptera 18 . This might indicate that IIdlm1 develops only in the latest instars and was not present in the instars investigated. It is present in nymphs of Anisoptera 20 .
Dorsoventral muscles. Idvm18 (14) (Fig. 2) originates from a large area of the prothoracic tergum that also encompasses the origin of Itpm3. Idvm18 inserts on an apodeme on the posterior base of the procoxa. It is by far the largest prothoracic muscle. Maloeuf 27 described its lateral branch as a discrete muscle (15). Idvm18 does show a slight dichotomy, yet all of its fibres run from the tergum to the apodeme. E. superstes and E. laidlawi show the same characteristics. Neither in Anisoptera nor in Zygoptera Idvm18 shows any striking dichotomy that would suggest the presence of an independent muscle (15) 18,20 . A muscle homologous to II (III)dvm7 (-) is not directly described by Maloeuf 27 nor Asahina 1 , yet in Maloeuf 's table 8 muscle IIdvm4 is listed as being dichotomous. Given the position of IIdvm4, it is quite probable that one of its alleged portions in fact is equivalent to muscle IIdvm7. In the nymphs of Anisoptera 20 and in the adults 18 and nymphs of Zygoptera muscle II (III)dvm7 is missing. In the nymphs of E. laidlawi and E. superstes it originates anterior-median to II (III)dvm4, runs parallel to it and finally is attached through a tendon to the trochanter while II (III)dvm4 inserts laterally on the coxa. Consequently, both muscles II (III)dvm4 as well as II (III)dvm7 seem to be present in nymphs and in adults of Epiophlebia.
In Neoptera muscle II (III)dvm7 originates at the central region of the notum of its segment and inserts at the trochanter 28 . In the Epiophlebia nymphs II (III)dvm7 originates at the antero-ventral rim of the meso-and metathoracic wing bud, which is a part of the notum.
IIdvm8 (-) is an intersegmental muscle, stretching between the mesofurca and the metathoracic tergite. Its homologues are Idvm10 and IIIdvm8 as they connect similar structures. Muscle IIdvm8 was not found in Anisoptera 20 , or in the zygopteran thorax 18 but in Neoptera 28 and in both species of Epiophlebia. Therefore, IIdvm8 probably is a muscle of the pterygote ground pattern that was, among the Odonata, only preserved in Epiophlebia.
The four muscles II (III)dvm1 and II (III)dvm2 are not present in Epiophlebia but could be confirmed for Zygoptera nymphs. Muscle II (III)dvm1 is present in nymphs of Anisoptera and in adult Zygoptera 18,20 .
Muscle II (III)dvm3 seems to be unique for Zygoptera nymphs 18,20 . The muscles II (III)dvm2 were found in the nymphs of Zygoptera for the first time. They show the same points of origin as in Neoptera 28 , whereas the insertions lie at the anterior margins of the corresponding coxae and not on the trochantins as described by Friedrich and Beutel 28 . However, free trochantins are not present in Odonata 30  Sterno-coxal and pleuro-coxal muscles. The pleuro-coxal muscles (Fig. 3) II (III)pcm2 are among those that undergo the most extensive growth in the pterothoracic segments. They start out very slender in the early instars and grow to be among the largest muscles in the respective segment in the latest   instars. In contrast to the description by Asahina 1 , we found clearly separated origins of II (III)pcm2 and II (III)pcm1.
Among the sterno-coxal muscles (Fig. 4) Iscm4 (-), IIscm4 (-) & IIIscm4 (-) were not mentioned by Maloeuf 27 or Asahina 1 . All three muscles were found in the nymphs of E. laidlawi and E. superstes. A possible explanation could be that all three muscles only exist in juvenile stages. At least IIIscm4 is also present in nymphs of Anisoptera 20 and IIscm4 was found in Zygoptera nymphs. However, none of the three muscles have been found in adult Zygoptera 18 . The homology established for IIscm4 in Büsse & Hörnschemeyer 20 holds for Iscm4 and IIIscm4 as well.
IIscm1 (-) & IIIscm1 (-) are very thin muscles running close to the meso-and metasternum and inserting by means of an apodeme on the lateral side of the coxa. These two muscles were neither listed by Maloeuf 27 nor Asahina 1 , but are present in E. laidlawi, E. superstes, in the Zygoptera and Anisoptera 20 . Not being mentioned by Büsse et al. 18 for adult Zygoptera, IIscm1 and IIIscm1 seem to be exclusively nymphal muscles in Odonata.
IIscm8 (-) is a funnel shaped muscle connecting the mesothoracic coxa and the metathoracic preepisternal apodeme. It has no serial homologue in the pro-or metathorax. Its presence is confirmed for both species of Epiophlebia, the Zygoptera and the anisopteran nymphs 20 . Neither Maloeuf 27 nor Asahina 1 or Büsse et al. 18 mentioned IIscm8, indicating that it also is a muscle that is restricted to Odonata nymphs.
IIIscm2 (-) connects the base of the metafurca and the posterior base of the metacoxa. It is confirmed for E. laidlawi, E. superstes, the Zygoptera and the Anisoptera 20 . Since it was not found in investigations of adult Odonata 1,18,27 , it probably also is an exclusively nymphal muscle. In the nymphs of the Ansioptera the homologue muscle IIscm2 could be identified 20 .
Sterno-pleural muscles. The sterno-pleural muscle (Fig. 4) Ispm1 (-) originates from the lateral surface of the apex of the profurca and inserts in the anterolateral area of the prothoracic epimeron. It was found in E. laidlawi, E. superstes, in Zygoptera and in Anisoptera nymphs 20 . Ispm1 resembles IIspm2 and IIIspm2, which originate from the apex of the meso-and metafurca, but it inserts in a different area. Ispm1 is a nymphal muscle 1,18,27 . Tergo-pleural muscles. The origins and the insertions of the tergo-pleural muscles (Fig. 5)

IItpm3
(-) & IIItpm3 (-) lie inside the wing buds. They are the smallest muscles in the nymphs. Both were found in the nymph of the two species of Epiophlebia and in the nymphs of the Anisoptera 20 . They are Similarly, the muscles Itpm7, Itpm8 and Itpm9 are found exclusively in Epiprocta nymphs.
Transverse muscles. The transverse ventral muscle (Fig. 6) II (III)tvm1* (-) was described by Maleouf 27 as muscle (69). It is supposed to run transversely between intersegmental sterno-pleural processes. These processes, depicted by Maloeuf 27 in an adult dragonfly, have been named preepisternal apodemes by Asahina 1 , who also found IItvm1 in the nymphs of E. superstes. However, both authors never described a transverse muscle by name. Maki 30 indicates the presence of a transverse muscle but gives no description. This muscle is neither present in Zygoptera nor in Anisoptera 18,20 .

Ivlm3 (11) &
Ivlm3' (11') have been found in adults of Epiprocta 1,27 . E. laidlawi, E. superstes and the nymphs of the Anisoptera 20 do not show a division of Ivlm3. It is also not reported for Neoptera 28 . The only argument for identifying Ivlm3' as a separate unit is the fact that it does not insert directly on the tentorial bar, like Ivlm3, but on a membrane right underneath it. It is very likely that Ivlm3' is only a separate strand of Ivlm3, which is missing in Zygoptera 18 . Muscle IIvlm3 is only present in zygopteran nymphs and IIIvlm3 is only present in Epiprocta 20 . IIvlm7 (42) is a longitudinal muscle connecting the profurca and the metathoracic preepisternal apodeme. According to Asahina 1 , it supposedly connects the profurca and the first abdominal segment through IIvlm6 (68). Büsse & Hörnschemeyer 20 stated that IIvlm7 connects the metafurca and the abdomen in Anisoptera, an interpretation confirmed here. In E. laidlawi and E. superstes, however, IIvlm7 does not continue through the metathorax. It inserts on the anterior margin of the metathoracic preepisternal apodeme. Muscle IIvlm7 is missing in the Zygoptera nymphs investigated.
IIvlm6 (68) originates from the posterior surface of the metathoracic preepisternal apodeme. It has exactly the same width as IIvlm7 and inserts on the anterior process of a thin structure arising from the ventral phragma of the first abdominal segment. This structure might either be a cap tendon or a very fragile apodeme, since it could not be properly identified from the datasets. It might be a nymphal muscle since it is present in immatures of Zygoptera and Anisoptera 20 , but not in adult Zygoptera 17 . Muscle Ivlm6 only occurs in Zygoptera nymphs. IIIvlm4* (64) has its origin on the anterior process of the thoracic-abdominal tendon mentioned above, just posterior to IIvlm6. It is the last muscle in a row of longitudinal muscles connecting the tentorial bar, the furcae and the abdomen. It is missing in nymphs of Anisoptera 20 , but is present in all examined nymphs of Epiophlebia. In adults of Zygoptera IIIvlm4 is replaced by a tendon, which connects the metafurca with the bar between the first and second abdominal segment 17 . Muscle IIIvlm4 is missing in nymphs of Zygoptera. According to Maloeuf 27 this muscle is absent in adult Odonata, but Asahina 1 identified it in adults of E. superstes. A possible explanation is that IIIvlm4 retracts at a certain stage into the abdomen, which gets elongated after the last ecdysis. Further examination might be necessary, to clarify the status of IIIvlm4.
Muscles Ivlm6 und IIvlm3 could be identified in nymphs of Zygoptera, the first evidence for the presence of theses muscles in Odonata. Muscles Ivlm6 and IIvlm3 show exactly the same attachment points as described for Neoptera 28 .
Muscle summary. All muscles described for the Odonata by Maloeuf 27 and for adults of E. superstes by Asahina 1 could be identified in the nymphs of E. superstes and E. laidlawi. Five muscles differ from the descriptions of both authors: Ivlm3, muscle, IIvlm7, IIvlm6 and IIIvlm4. Six muscles, IIIvlm4, II (III) tvm1, Iscm4, IIdvm7 and IIIdvm7 could be newly identified in the Odonata 1,18,20,27 .
In Anisoptera four muscles are present that are missing in Epiophlebia: II (III)dvm1, IIscm2 and IIscm7.
These results confirm that E. superstes and E. laidlawi are highly similar in almost all aspects of their thoracic morphology as well as on the genetic level, as stated by Büsse et al. 2 .
Poletaïev 31 reported that the wing buds of Odonata appear in the 3 rd or 4 th instar but that the corresponding musculature is still indiscernible. Maloeuf 27 noted that the flight muscles in these instars are still diminutive. Our investigation confirms Maloeuf 's observation: some flight muscles of the adult first appear in early instars as sets of very few muscle fibres. Some are scarcely traceable, like for instance Itpm10 and Itpm11, and then grow significantly during ontogenesis, like II (III)dvm3, II (III)pcm2, II (III)tpm7. Some other muscles, like IIIdlm1, are not necessary for flight in the adult but seem to be important for the movements of the nymph 1,27 . IIIdlm1 starts out as a broad muscle whose origin covers roughly a fourth of the posterior surface of the intersegmental ridge in the two earlier instars and shrinks to a few fibres in the last instar. Maloeuf 27 also stated that nymphs of Odonata have more and larger leg and cervical muscles than the adults 1,20 . We can confirm these findings: muscles Idlm1, Idvm10, Itpm10, Itpm11, IIscm7, II (III)spm2, IIvlm7, IIIdlm1, IIIvlm2, IIIvlm4 are present in the nymphs and absent in the adults of Epiophlebia 1 .
During ontogenesis the thoracic muscles are, in part, newly formed, transformed or reduced 26,27,31 . The extent of these modifications seems to be exceptional in Odonata, compared to other non-holometabolous pterygote insects, which usually display a nearly complete set of muscles from the first instar 21,32 . Muscles missing in Epiophlebia. As mentioned above, muscles II (III)dvm1, IIscm2 and IIscm7 are present in Anisoptera but not in Epiophlebia. The muscles II (III)dvm1 and IIscm7 are present in Zygoptera and Neoptera 28 and muscle II (III)dvm1 is also known for Ephemeroptera 32 . This distribution through higher taxa indicates that the presence of these muscles is a plesiomorphic condition for Pterygota. Therefore the missing of these muscles in recent Epiophlebia is a derived state. This interpretation may also be true for muscle IIscm2, which, among Odonata, is only present in Anisoptera 20 . Assuming a new formation of this muscle in Anisoptera is not parsimonious, because it is also present in Neoptera 32,33 .
Zygoptera have four muscles that are missing in Epiophlebia and Anisoptera: II (III)dvm2, Ivlm6 and IIvlm3. These are nymphal muscles in Zygoptera 18 . The muscles Itpm3, Itpm7, Itpm8 and Itpm9 as well as II (III)tpm3 and IIvlm7 are only present in nymphs of Epiprocta. Since Itpm3, II (III)tpm3 and  IIvlm7 are missing in Zygoptera 18 but are present in Neoptera 28 and IIvlm7 also in Ephemeroptera 32 , it is most parsimonious to assume that their presence is a plesimorphic character for Epiprocta. The muscles II (III)scm1, IIscm8, IIIscm2 and Ispm1 have only been found in nymphs of Odonata. They seem to be generally missing in the adults 1,18,27 . The odonate ground pattern most likely encompassed IIvlm3 (present in Zygoptera) as well as IIvlm7 (present in Epiprocta), with IIvlm3 secondarily missing in Epiprocta, and IIvlm7 in Zygoptera, because both muscles are present outside of Odonata, i.e. in Neoptera and in Ephemeroptera 28,32 .
Asahina 1 depicted and labelled transverse muscles (II (III)tvm1) in the thorax of adult E. superstes. Likewise Maleouf 27 and Maki 30 mentioned ventral transverse muscles, and Barlet 34,35 and Matsuda 33 found them in Zygentoma and in Archaeognatha. None of them named these muscles. Chadwick 36 gave an overview of the occurrence of ventral transverse muscles in several insect orders. Together with these data, our results clearly indicate that a ventral transverse muscle belongs to the ground pattern of the pterygote thorax. Therefore, the absence of such a muscle in Zygoptera and in Anisoptera suggests that it was lost independently in the last common ancestors of each of these two monophyla, and that only Epiophlebia retained this plesiomorphic character.
Assuming the commonly accepted monophyly of Ptergygota, it is most likely that its last common ancestor was morphologically similar to the extant species of Zygentoma. This also indicates that the number of muscles in the thorax was quite high in the ground pattern, since in Zygentoma and Archaeognatha 33-37 many more thoracic muscles are present than in any extant pterygote insect investigated. This interpretation is supported by our results. We found several muscles in the thorax of Odonata that are not present in Neoptera 28 (cf. Fig. 7). Among Odonata, species of Epiophlebia are those with the highest number of thoracic muscles, indicating that this state is plesiomorphic and that the missing of muscles like Iscm4 & IIIscm4 in Zygoptera or IIscm4 in Anisoptera represent the apomorphic state.
Therefore, the statement of Blanke et al. 4 that Epiophlebia has preserved the most ancestral characters in Odoanta is supported.

Material and Methods
Three different instars (early, middle and last) of both, E. laidlawi and E. superstes, as well as nymphs of three species of Anisoptera and two species of Zygoptera were investigated ( Table 2). The specimens of E. laidlawi were fixed in 4% formaldehyde and stored in 80% ethanol. The other specimens were fixed and stored in 80% ethanol. Prior to scanning, the samples were critical point dried (Balzers CPD030) and mounted on facility specific specimen holders. All applicable regulations concerning the protection of free-living species were followed.  Table 2. Specimens investigated and voxel resolution of μCT data. Voxels are isometric in x-, y-and z-axis in all datasets.