Reconstructing a 55-million-year-old butterfly (Lepidoptera: Hesperiidae)

The oldest butterfl y fossil known, which was formed about 55 Ma in what is now Denmark, is described. The fragments of its forewing venation indicates it belongs to the Hesperiidae. Further reconstruction indicates that it fi ts in the Coeliadinae and is close to the extant genera Hasora and Burara. It is here described as Protocoeliades kristenseni gen. et sp. n. It is the fi rst butterfl y fossil found on a continent (Europe) where its closest relatives do not currently occur. Its position on the phylogenetic tree of the Coeliadinae and its importance in understanding the time dimension in the evolution of butterfl ies, and their ecological and biogeographic implications are discussed. ZooBank Article LSID: E0B5F7AD-5352-493B-9016-F16A9C596D70

6th, 2014.I feel it is appropriate to name the fossil after him.

MATERIAL AND METHODS
The fossil (Fig. 1) was found in the Fur formation on the island of Fur in the Limfjord in N Denmark.It is a 60 m thick marine deposit of diatoms and clay minerals with many layers of volcanic ash.It was formed shortly after the Paleocene to Eocene transition, about 55 Ma (Pedersen et al., 2004) and previously estimated at 57 Ma (Rust, 1998).The deposit is famous for its many fossils, not only marine animals but also terrestrial animals, which apparently were blown into the sea or actively fl ew over the water.Among the numerous insect fossils only the present fossil can be identifi ed as a butterfl y.
The fossil is of a relatively broad-winged insect with wings folded down enclosing the legs.Outline of hindwings diffi cult to make out and most of the head is missing, but the dark lines in front of it may be remnants of palpi.Apex and greater part of termen of forewings are absent.Because of its fi ne detail the specimen in Fig. 1a is thought to be the compressed insect (although little is left except coloration of the substrate), of which the imprint is preserved in the counterpart illustrated in Fig. 1b.Therefore, the uppermost forewing is supposed to be the right wing.In the rest of the fossil the veins of four wings are lying on top of each other, making it diffi cult to decide which vein belongs to which wing.Part of the underlying (left) forewing is free and, although in this wing the apex is missing, several veins are clearly visible and there are no structures lying under it and blurring the picture.The venation of the "free" part of the (supposed) left forewing (right part of the fossil in Fig. 1a) was drawn on a transparent sheet and slid over the other forewing.By moving the

INTRODUCTION
Butterfl y fossils are very rare.To date 48 fossil butterfl y species are named (de Jong, in prep.).In addition, a number of purported butterfl y fossils are mentioned in the literature.Sohn et al. (2015, Table 2) cites 142 fossil butterfl ies.This number, however, relates to specimens, not taxa (Sohn, pers.comm.), and does not include identifi ed taxa other than the 48 recorded so far (see also their additional fi le with literature sources).While modern estimates place the origin of the butterfl ies before the Cretaceous-Paleogene boundary (see chapter Importance of this fossil for determining the time dimension in the evolution of the butterfl ies), the oldest recorded fossils of butterfl ies date from the Early Lutetian, Middle Eocene, 48 Ma (Praepapilio; Durden & Rose, 1978).The present paper describes (and names) an older fossil, from near the Paleocene/Eocene boundary, initially dated at 57 Ma (Rust, 1998), but more recent estimates give a date of about 55 Ma (Pedersen et al., 2004), i.e. very early Eocene.This fossil was shown to me in 1996 by the late Niels Peder Kristensen, lepidopterist at the Zoologisk Museum, Copenhagen, with the request that I identify and describe it.Although I identifi ed it as a member of the Hesperiidae I did not describe it.A photograph of the fossil was published by Kristensen & Skalski (1999: 29), with the caption "An apparent hesperioid butterfl y…".The fossil was listed in Sohn et al. (2012: 62) under Hesperiidae, but with "genus incertae sedis".At long last I fulfi l my commitment to Niels with this paper, but regret that it appears after his untimely death on December The dark stripe approximately following the course of vein CuA 2 is much too broad to be this vein and is interpreted as a wing fold, which includes CuA 2 .Another dark stripe, still broader and further to the hind margin of the wing, is apparently overlying and obscuring vein 1A + 2A and is possibly part of a leg.
The nomenclature of the veins follows the convention of Nielsen & Common (1991), who broadly adopted the old Comstock system (Figs 4 and 5).originate from, their course clearly shows that they are unbranched (Figs 2 and 3; compare with Fig. 5) and do not branch off each other.Reconstruction of the possible complete forewing venation (Fig. 3; subcosta omitted, since no trace of it can be found in the fossil; it is not diagnostic) leads to the following observations.Cell length about 60% of wing length.The origins of CuA 1 and CuA 2 are not visible on the fossil, but if extrapolated (by following their course) towards their apparent (but approximate) origin from the cubitus, CuA 1 originates about halfway between wing base and origin of M 3 , while CuA 2 originates about halfway between wing base and origin of CuA 1 .
Etymology.The name Protocoeliades indicates that the fossil is an early member of the Coeliadinae (see discussion on taxonomic position), of which Coeliades is the type genus.Its gender is feminine.
Characters as for the genus, of which it is the sole representative.
Etymology.The new species is named after the late Niels Peder Kristensen (Zoologisk Museum, Copenhagen), who was the fi rst to recognize the fossil as something special and trusted the author with its description.

Taxonomic position
Within the butterfl ies (as defi ned by Heikkilä et al., 2011) the fi ve unbranched radial veins in the forewing are a universal autapomorphy of the Hesperiidae (Ackery et al., 1999), in all other butterfl ies branching of some of the radial veins (sometimes including one median vein as well) occurs and often the radial veins are also reduced.Unbranched radial veins are very rarely found also in other Lepidoptera, but always combined with a very different, much narrower wing shape (see, e.g., Kristensen & Skalski, 1999).Thus, this character identifi es the fossil as a member of the Hesperiidae.
Cell length relative to wing length is variable throughout the Hesperiidae and a relative length of 60% does not place this fossil in any particular extant subfamily as recognized by Warren et al. (2009).It cannot be decided whether M 2 approaches M 3 at their origins (indicating relationship with Hesperiinae), since the relevant part of the wing is too blurred, but it is unlikely that it approaches M 3 suddenly in the blurred basal part of the veins, since in the extant species where M 2 approaches M 3 , it does so very gradually.
The origin of CuA 2 at 15% to about 25% of the cubitus from the base of the cell is universal in the Coeliadinae (Fig. 4), and found in some other Hesperiidae, viz. in the single species of Euschemoninae, and in various genera of Pyrginae (e.g., Oxynetra Felder & Felder, 1862, Gerosis Mabille, 1903, and Tagiades Hübner, 1819) and Eudaminae (e.g., Phocides Hübner, 1819, Epargyreus Hübner, 1819).In the extensive phylogenetic analysis of the Hesperiidae by Warren et al. (2008Warren et al. ( , 2009) ) these subfamilies are successively basal in the phylogenetic tree (Fig. 7).Since the Hesperiidae are not basal in recent trees for the butterfl ies (e.g., Fig. 6, based on Heikkilä et al., 2011), and this character does not occur in other butterfl ies (Ackery et al., 1999), it appears to be a non-universal apomorphy of the Hesperiidae.It either originated once at the root of the Hesperiidae, was lost in many taxa in the evolution of Eudaminae and Pyrginae and the common ancestor of Heteropterinae, Trapezitinae and Hesperiinae, or it originated separately in Coeliadinae, Euschemonidae, Eudaminae and Pyrginae (Fig. 7).
The origin of CuA 1 about halfway along the cubitus is found only in Coeliadinae (although it is approached in Oxynetra, a genus of Pyrginae restricted to the Neotropics), and within this subfamily it is found in all species of Hasora Moore, 1881 (Fig. 4) (but sex-linked in some species) and some species of Burara Swinhoe, 1893.The phylogenetic tree of the Coeliadinae is still uncertain.The phylogenetic tree in Fig. 8 was obtained by analyzing 55 morphological characters and 26 species (representing all nine genera of Coeliadinae and fi ve outgroups) (de Jong, 2007).The only other tree available is a part of the large tree which includes 210 hesperiid taxa (Warren et al., 2009), of which only six are Coeliadinae (representing six of the nine genera currently recognized; Fig. 9).Not surprisingly these trees differ considerably, but agree in not placing Hasora and Burara as sister groups.A partial molecular tree for Coeliadinae (viz.for the species occurring in Japan) is provided by Dodo et al. (2008), which only includes the genera Burara, Choaspes, Badamia and Hasora.Also in this partial tree Burara and Hasora are not sister groups, the fi rst being a sister group to the other three combined.
If that position of the origin of CuA 1 originated only once and was lost several times, this fossil could be placed either at the root of the Coeliadinae (Fig. 8), or one node up (Fig. 9) (indicated by an X in both fi gures).If the character originated several times independently, the fossil could be placed at two independent places in any of the two trees, making it useless for calibration purposes (see below).A double origin (in Hasora and Burara) may be the most parsimonious solution for this character, but without a simultaneous analysis of other characters of this fossil, it is not wise to speculate.There is an easy character for separating Hasora and Burara: in the former there is a sharp bend in the basal part of vein 1A + 2A (Fig. 4), while in Burara 1A + 2A is more or less straight or gently curved as is usual in butterfl ies.Unfortunately the basal part of this vein is not visible in this fossil.Since the fossil cannot be assigned to either Hasora or Burara, the genus Protocoeliades was erected to accommodate this fossil.For further discussion on the phylogeny of the Coeliadinae see the next chapter.

Importance of this fossil for determining the time dimension in the evolution of the butterfl ies
Exciting as it may be to have the oldest known butterfl y in hand, its importance for estimating divergence times in butterfl y or hesperiid phylogeny is moderate.Recent analyses indicate that several butterfl y families younger than the Hesperiidae (based on phylogenies reported in several recent studies, such as Heikkilä et al., 2011; see Fig. 6) originated before the Cretaceous-Paleogene (K-Pg) boundary.For the four younger families the estimates are as follows: Pieridae 81-112 Ma (Braby et al., 2006), Nymphalidae 90 Ma (Wahlberg et al., 2009), Riodinidae 81-96 Ma (Espeland et al., 2015), Lycaenidae, as sister to the Riodinidae, should be similar in age.In addition, the age of the older family Papilionidae was recently estimated at 68 (53-87) Ma (Simonsen et al., 2011).Clearly, this cannot be true at the same time, with the youngest family at the root.Differences and uncertainties are due to diffi culties with calibration (and possibly the trees can still be improved).For instance, for the calibration of the Pieridae tree four fossils were used, the identifi cation of which is doubtful if not incorrect (de Jong, 2007).But we are not concerned with the butterfl ies as a whole in this paper.Whatever the correct ages of the "younger" families, the Hesperiidae must be older and fi nding the universal hesperiid apomorphy of unbranched radial veins in a fossil of 55 million years old does not alter this fact.
Similarly the basal move in the origin of vein CuA 2 is of limited value for calibration.It occurred once and was lost many times, or many times independently.In both cases we need to determine the minimum age for its occurrence at the root of the Hesperiidae.9. Phylogenetic trees.6 -redefi ned butterfl ies (Heikkilä et al., 2012).7 -Hesperiidae (Warren et al., 2008(Warren et al., , 2009)).8 -Coeliadinae (de Jong, 2007).9 -Coeliadinae (Warren et al., 2008(Warren et al., , 2009)).X = allocation of Protocoeliades as possible calibration point, see text for further explanation.
Only the basal move in the origin of vein CuA 1 has some potential for calibration.It is found in only two genera.Depending on the phylogenetic tree, it fi rst arose at the root of the Coeliadinae (Fig. 8), or one node up (Fig. 9).Evidently, a new analysis of Coeliadinae, based on morphology and molecular markers, and including all genera, is needed for a fruitful further discussion.
Such a new analysis is not only crucial for estimating divergence times in Coeliadinae, but also for butterfl ies in general.In Fig. 8, Hasora is at the root of the Coeliadinae.It only feeds on Fabaceae, the age of which varies from 56 to 82 Ma depending on the authors (see next chapter), and thus, the fi rst split in the hesperiid lineage in Fig. 8 cannot be older than that.Since the Hesperiidae are rather low in the butterfl y tree (see Fig. 6), this could considerably constrain the estimates for the entire tree.If, on the other hand, Hasora is in the position in Fig. 9, or another non-basal position, the constraint would be less.
Apart from a Pleistocene copal hesperiid fossil assigned to an extant species, Andronymus neander (Plötz, 1884) (Skalski, 1976), there is just one other fossil hesperiid, Pamphilites abdita Scudder, 1875.It is assignable to the subfamily Hesperiinae (de Jong, in prep.) and of Oligocene/Miocene age (around 25 Ma).The Hesperiinae are predominantly grass feeders.Their expansion and radiation undoubtedly coincided with the evolution of the other large group of grass feeding butterfl ies, the Satyrinae (Nymphalidae) in the Oligocene (Peña & Wahlberg, 2008), long after the origin of the present fossil.Being less than half the age of the present fossil, it would be of interest to compare P. abdita with P. kristenseni as a calibration point (the fi rst at the root of the Hesperiinae) using a molecular clock analysis.Vanessa vetula von Heyden, thought to be a hesperiid by Scudder (1875) who erected the genus Thanatites for it, cannot be assigned to any particular butterfl y family (de Jong, in prep.).

Ecological and biogeographic considerations
The extant species of the Coeliadinae feed on a wide range of food plants, belonging to some 30 families (Vane-Wright & de Jong, 2003;Larsen, 2005;Cock, 2010), the estimated age of which collectively range from 30 to 110 million years (Hedges & Kumar, 2009; estimates differ considerably between authors; here a recent publication was used).Almost all genera of Coeliadinae are polyphagous, and consequently their food plants are not the maximum constraint on their age (except the 110 million year boundary for the subfamily, just mentioned).Two genera are an exception.The West African endemic and monotypic genus Pyrrhiades Lindsey & Miller, 1965, is only known from Malpighiaceae (Larsen, 2005), a plant family (estimated age 65-71 Ma; Hedges & Kumar, 2009) also widely used by other Coeliadinae.Moreover, Chiba (2009) concludes that Pyrrhiades is not monotypic, but also contains a number of species previously included in Coeliades.This taxonomy needs additional supporting evidence, which does not concern us here.Pyrrhiades sensu Chiba lives on Malpighiaceae and Asclepiadaceae (also widely used by Coeliades).The relatively large Oriental genus Hasora (extending to Australia and into the Pacifi c as far as the Sandwich Islands) is confi ned to Fabaceae.The age of this family is given as 56-68 Ma by Wikström et al. (2001), 59.9 Ma for the entire order Fabales by Magallon &Sanderson (2001), and76-82 Ma by Forest &Chase (2009).This gives a maximum constraint for the age of Hasora, although the wide range is not very helpful.Protocoeliades is younger than the oldest food plants for Hasora and Burara (the food plants of which belong to Malpighiaceae, Combretaceae, Araliaceae, Myristicaceae and Myrsinaceae, according to Vane-Wright & de Jong, 2003), or in other words it falls within the maximum constraint for both genera.
The extant subfamily Coeliadinae occurs throughout the tropics from Africa to Australia and far into the Pacifi c, with a very slight extension in East Asia to Japan and the Amur region (Southeast Siberia) (Evans, 1937(Evans, , 1949)).It can be divided into eight or nine genera, three of which are confi ned to Africa, and fi ve or six to Asia and Australia.Hasora and Burara belong to the latter group.The fi rst genus covers almost the entire area of the subfamily in Asia/Australia.
The fossil described was formed at a time when tropical and subtropical conditions were prevalent in large parts of Europe.Worldwide the Early Eocene was the warmest period in the Cenozoic; with some fl uctuations it was a continuation of the warm climate in the Cretaceous (Crowley & North, 1991).Relationships of fossil taxa from the Cretaceous and Paleogene in the Northern Hemisphere with extant taxa nowadays restricted to tropical areas, are well known (Eskov, 2000;Larsson, 1978).P. kristenseni is the fi rst record of a butterfl y fossil found on a continent (i.e., Europe), where at present its closest relatives do not live, and it fi ts in the pattern of a once, in the early Paleogene, widespread tropical fauna that largely or totally became restricted to tropical areas around the equator by cooler and drier climates from the Eocene onwards.

Genus
Protocoeliades gen.n.ZooBank taxon LSID: E771863A-8D48-4480-8D1A-72B216B9F31D Type species: Protocoeliades kristenseni sp.n.Length of preserved part of forewing 23 mm, actual wing length estimated at about 25.6 mm.In the right forewing, fi ve radial veins (R 1 -R 5 ) are clearly visible.Although it is diffi cult to make out exactly what part of the cell they sheet to give the best fi t, several dark lines in the apical part of the right forewing and other structures were identifi ed.