The oldest known bat skeletons and their implications for Eocene chiropteran diversification

The Fossil Lake deposits of the Green River Formation of Wyoming, a remarkable early Eocene Lagerstätte (51.98 ±0.35 Ma), have produced nearly 30 bat fossils over the last 50 years. However, diversity has thus far been limited to only two bat species. Here, we describe a new species of Icaronycteris based on two articulated skeletons discovered in the American Fossil Quarry northwest of Kemmerer, Wyoming. The relative stratigraphic position of these fossils indicates that they are the oldest bat skeletons recovered to date anywhere in the world. Phylogenetic analysis of Eocene fossil bats and living taxa places the new species within the family Icaronycteridae as sister to Icaronycteris index, and additionally indicates that the two Green River archaic bat families (Icaronycteridae and Onychonycteridae) form a clade distinct from known Old World lineages of archaic bats. Our analyses found no evidence that Icaronycteris? menui (France) nor I. sigei (India) belong to this clade; accordingly, we therefore remove them from Icaronycteridae. Taken in sum, our results indicate that Green River bats represent a separate chiropteran radiation of basal bats, and provide additional support for the hypothesis of a rapid radiation of bats on multiple continents during the early Eocene.


Introduction
With over 1,400 living species, bats are the second most speciose group of mammals [1,2]. They are found all over the world with the exception of polar regions and a few remote islands, and are ecologically highly diverse, occupying a wide variety of habitats and ecological niches [1,3]. Bats play a vital role in healthy natural ecosystems and additionally provide many ecosystem services important for human economies (e.g. pest control, pollination, seed dispersal)

Material and methods
No permits were required for the described study, which complied with all relevant regulations. We examined multiple specimens of Eocene fossil bats (S1 Appendix), as well as literature descriptions of all Eocene bats known from skeletons, and evaluated osteological characters including, but not restricted to, those discussed by Jepsen [14], Simmons and Geisler [37], Simmons et al. [12] and Smith et al. (2012). Dental homology nomenclature for premolars follows that of Simmons and Geisler [37], considering the missing elements in the premolar row to represent the P2 and p2 in the upper and lower dentition, respectively.
CT scans of the holotype of the new species were made using a high-resolution Phoenix v|tome|x micro-CT scanner in the Microscopy and Imaging Facility at the American Museum of Natural History. Visualization (S1 Fig) and analyses were performed in VGStudio MAX 3.0.2 and Avizo 9.5.0. Measurements of the new fossil bat were obtained from micro-CT images. Each measurement was taken three times and an average was taken to minimize measuring error. All measurements reported herein are from adult individuals with closed epiphyses. Measurements (in mm) are in S1 Table. To estimate body mass, we followed the methods described by Giannini et al. [46]. Body mass M (in grams) was estimated from least mid-shaft diameter data of the humerus (D) for fossil specimens. As recommended by Giannini et al., we transformed the equation: log10 (M) = (log10(D) − log10(b0))/b1 into a general model: log10 D = log10 b0 + b1 � log10 M + error, where b0 is the y-intercept and b1 is the slope parameter of the corresponding equation from extant bats (S1 File). Giannini et al. [46] showed that the least mid-shaft diameter of the humerus best explained the model (r 2 = 0.991) across a wide range of bat taxa with the corresponding values: b0 = -0.273; b1 = 0.363.
We conducted a phylogenetic analysis in order to determine the relationships of the new species of Icaronycteris to other species currently placed in the family Icaronycteridae as well as other well-known Eocene bats. Our morphology-based parsimony analysis included six other Eocene bat species known from complete skeletons (Icaronycteris index, Onychonycteris finneyi, Archaeonycteris trigonodon, Palaeochiropteryx tupaiodon, Hassianycteris messelensis, and Tachypteron franzeni), and the two other named species of Icaronycteris known only from the dentition (I.? menui and I. sigei). Extant bats representing living superfamilies (Saccopteryx bilineata, Emballonuroidea; Macrotus waterhousii, Noctilionoidea; Myotis lucifugus, Vespertilionoidea; Rhinolophus ferrumequinum and Rhinopoma hardwickii, Rhinolophoidea) were included in order to determine the relationship of the fossil taxa to crown Chiroptera. Erinaceus europaeus and Sorex bendirii (Eulipotyphla) were included as outgroups representing non-Chiroptera Laurasiatheria.
A total of 568 discrete characters (248 craniodental characters and 320 postcranial characters) were scored as much as possible for each taxon. Completeness of taxon scoring ranged from a high of 95% (in extant species) to a low of 14% (in the incomplete fossil taxon Icaronycteris sigei). Multistate characters were treated as ordered only when clear intermediate states existed (e.g., i1 crown proportions: width greater than length; length and width subequal; length greater than width). A molecular backbone was used to constrain the relationships of living taxa following the relationships recovered by Amador et al. [47]. The dataset was compiled in Morphobank [48] and is available there as Project P4016. When possible, characters were scored from direct observation of specimens (S1 Appendix), otherwise characters were scored from the literature. Phylogenetic analyses were conducted in PAUP � version 4.0a169 [49], using a non-heuristic, exact search (i.e., branch-and-bound) to produce the most parsimonious trees. Due to the unstable position of the most fragmentary fossil taxa (i.e., Icaronycteris? menui and I. sigei) in initial analyses, a second analysis was conducted using a backbone tree constraining the relationships of the extant bats and fossil taxa known from complete skeletons.

PLOS ONE
The oldest known bat skeletons and their implications for Eocene chiropteran diversification means to estimate ages of specimens. The paratype was found in the same quarry as the holotype by Robert Kronner in 1994 in the bottom 60 cm of the sandwich beds of the FBM, Green River Formation.

Etymology
This species name is in honor of Gregg Gunnell in recognition of his extensive contributions to the understanding of fossil bats and chiropteran evolution.

Diagnosis
Icaronycteris gunnelli is the smallest known chiropteran from the Green River Formation (S1 Table) and is distinguished from other Eocene bats by the following combination of traits: claw present on wing digit I and II; tiny ossified third phalanx present on wing digit III-V; relatively short forearm and broad wing; relatively short, robust hind limbs with a possibly sutured tibia and fibula; lower canine tall and lanceolate; upper canine with anteromedial groove; diastema present between C and P1; P3 with raised lingual cingulum and a crown length relatively shorter than that of P4; P4 with well-developed lingual cingular cusp and labial cingulum; P4 length and width subequal; upper molars with strongly developed ectocingulum.

PLOS ONE
The oldest known bat skeletons and their implications for Eocene chiropteran diversification

PLOS ONE
The oldest known bat skeletons and their implications for Eocene chiropteran diversification

Description and comparisons
The skull of the holotype of Icaronycteris gunnelli is dorsoventrally crushed, so many cranial features, including structures of the basicranium, cannot be distinguished. However, the shape of the skull and several important traits of the skull and postcranium can be evaluated. Morphological traits of I. gunnellii are similar to those seen in Icaronycteris index and Onychonycteris finneyi except if explicitly stated otherwise below. The proclivous premaxilla in Icaronycteris gunnelli extends anteriorly beyond the level of the canine root and the nasal process of the premaxilla is well developed. The zygomatic arch appears complete and is broader than the mastoid region. Parietals appear to have not been inflated based on their rounded shape. The angular process of the dentary in I. gunnelli projects below the level of the occlusal plane, and the coronoid process is approximately two times the height of the condylar process. The dental formula of I. gunnelli is I2/3, C1/1, P3/3, M3/3 = 38. The upper incisors are orthodont and orthoclivous, and in this they resemble those of Onychonycteris finneyi. In contrast, the upper incisors in I. index are proodont and orthoclivous. The crowns of the upper incisors are not clearly differentiated from the root shafts in I. gunnelli, a condition similar to I. index but different from that seen in O. finneyi, in which the crown and root shaft of the upper incisors are clearly differentiated. The occlusal margin of I1 includes a large main cusp that is offset mesially from the axis of the tooth and tapers to a blunt point. A distal accessory cusp is present lateral to the main cusp on I1, with the accessory cusp approximately ¼ the size of the main cusp. The two upper incisors are subequal in height. Crown morphology of I2 is well developed and lingual cingulum is absent. Lack of lingual cingulum of the I2 is a characteristic also seen in I. index, whereas a lingual cingulum is present on this tooth in O. finneyi. I1 and I2 are subequal in height in I. gunnelli, also a condition shared with I. index. In contrast, in O. finneyi the height of I1 is less than that of I2. The upper canine (C) is separated from I2 by a diastema in all of these taxa.
The upper canine in I. gunnelli is elongated anteroposteriorly but does not project anteriorly. This tooth is not labially swollen and it lacks a raised cingulum or accessory cusp. No lingual cingulum is present, which is another trait shared with I. index. By contrast, O. finneyi exhibits a faint lingual cingulum on the upper C. The distal cingulum of the upper canine is small in I. gunnelli. A labial cingulum on the canine was not visible on the X-ray, but we could not determine if this indicated true absence or a lack of resolution in our images. An anteromedial groove is present on the upper canine, a structure that is not present in I. index or O. finneyi. A posteromedial ridge is present and the posterolingual and posterolabial surfaces of the canine are flattened. The upper canine has a tiny posterolateral accessory cusp, a structure that is absent in O. finneyi, and present but slightly smaller in I. index. The shape of this cusp is conical and is not clearly separated from the main cusp; it originates on the posterior face of the main cusp. A diastema is present between upper C and P1.
The upper premolars lie in line with the molar toothrow and no teeth are offset medially or laterally. The P1 is single rooted. This tooth has a well-developed crown which lacks a labial cingulum, similar to I. index but differing from O. finneyi, which has a labial cingulum on P1. The P3 in I. gunnelli also lacks a labial cingulum, again as in I. index, with both Icaronycteris species differing from O. finneyi, which exhibits a weakly developed labial cingulum on P3. The lingual cingulum of P3 is weakly developed and lacks a cusp, but the edge of the cingulum is slightly raised. This differs from the condition in I. index and O. finneyi, both of which have a lingual cusp on P3. The crown length of P1 is shorter than that of P3. This condition is shared with I. index, while P1 and P3 are subequal in crown length in O. finneyi. The postparacrista of P3 extends as a single crest to the distal edge of the tooth and lacks a distal accessory cusp. Both P3 and P4 have three roots, but the crown length of P3 is shorter than that of P4. The latter trait contrasts with the condition in I. index and O. finneyi, both of which have a P3 that is longer than P4. The height of P3 is subequal to that of P4 in I. gunnelli. This differs from both I. index, which has a P3 that is taller than P4, and from O. finneyi, in which P3 is shorter in height than the P4. No diastema present between the P3 and P4. The lingual cingulum of the P4 is very large and resembles that of I. index. In both of these taxa, the lingual cingulum forms a distinct lobe that extends posteriorly as far as the protocone of the first molar. This condition differs from that seen in O. finneyi, where the lobe of the lingual cingulum does not extend as far as the protocone of the first molar. The labial cingulum of the P4 is well developed in I. gunnelli and O. finneyi, but absent in I. index. The P4 in I. gunnelli has three roots. The lingual cingulum of the P4 is very large and resembles that of I. index. In both of these taxa, the lingual cingulum forms a distinct lobe that extends posteriorly as far as the protocone of the first molar. This condition differs from that seen in O. finneyi, where the lobe of the lingual cingulum does not extend as far as the protocone of the first molar. The labial cingulum of the P4 is well developed in I. gunnelli and O. finneyi, but absent in I. index. A lingual cingular cusp is well developed on P4 in I. gunnelli. This differs from the condition in I. index, in which the lingual cingular cusp is weakly developed. The condition of a lingual cingular cusp in O. finneyi cannot be determined from the specimens available. The P4 postparacrista is a single crest that extends as a continuous crest to the distal aspect of the tooth. The length and width of the P4 are subequal in I. gunnelli, in contrast to the conditions seen in I. index (P4 wider than long) and O. finneyi (P4 longer than wide).
The upper molars of I. gunnelli are characterized by a strongly developed ectocingulum. This is similar to the condition seen in O. finneyi, but differs from that in I. index, in which the ectocingulum is present but weak. A lingual cingulum and stylar shelf are present on both the M1 and M2, and the protocone and paracone are subequal in height. The protocone on both M1 and M2 is sharp and inclined anteriorly. The angle between the postparacrista and premetacrista is acute. The postparacrista contacts the premetacrista on the labial aspect of the tooth and closes the trigon basin labially on both M1 and M2. The trigon basin on both teeth is wide and shallow, and a mesostyle and mesostylar crest are absent. The contact between the postparacrista and premetacrista is more lingually displaced than the condition observed in I. index and similar to that of O. finneyi, and both M1 and M2 have a cuspule on the premetacrista as well as a small paraconule. The maximum mesiodistal length of M1 and M2 is less than half the distance between the paracone and metacone of the respective teeth.
The M1 has a long, straight, crestlike, parastyle that is not separated from the preparacrista. A single deep ectoflexus is present labial to the mesostyle. The preparacrista of M1 is subequal in length to the postparacrista, and the postmetacrista is about 1.5 times longer than the premetacrista. The trigon basin does not show any striations or enamel folds, instead being relatively smooth. The postprotocrista is oriented distolabially towards the metacone, and it extends labially to become confluent with the metacingulum and the edge of the hypocone shelf. A metaconule is present on the M1, but an endoloph is absent. The hypocone shelf is small and narrow but clearly distinct from the trigon basin. The M2 parastyle is crestlike and curved. Like M1, the M2 has a single deep exoflexus labial to the mesostyle. The metacone is located directly posterior to the paracone. The M2 postmetacrista, which is oriented distolabially towards the metacone, is 1.75 times longer than the premetacrista. An endoloph is present on M1, but is lacking on M2. The hypocone shelf on the M2 is very small and appears only as a narrow cingulum; no hypocone cusp is present.
The M3 is moderately reduced with a surface area <75% that of M2. The anterior segment of the M3 is nearly as large as that of the M2, and a long, crested curved parastyle and a paraconule are present. The premetacrista and postparacrista of M3 are subequal in length, and the metacone is present as a distinct cusp. A metastylar fovea is lacking on the M3, but the protocone and small hypocone shelf are present.
The three lower incisors in I. gunnelli are trilobed and somewhat procumbent; they are all similar in size and shape. The alveoli are evenly spaced, and there is no diastema between the last incisor and lower canine. The lower canine is tall, lanceolate, and lacks an anterior cuspule. The posterior cingulid on the canine is broad and a posterobasal cusp is strongly developed. The labial cingulum and cingular structures of the canine are not clearly visible given preservation of the holotype of I. gunnelli and hence cannot be evaluated.
As in other Eocene bats, I. gunnelli has three lower premolars. The p1 is large, single-rooted, cuspidate and has a distal cuspule. The crown length of p1 is less than that of p3. This is similar to the condition seen in I. index, but different from that of O. finneyi, in which p1 crown length is greater than p3 length. The p3 is premolariform, i.e., the cusps are not arranged in a tribosphenic pattern. Anterolingual and distal cuspules are present, but there is no lingual cuspule. Both p3 and p4 have two roots that are oriented longitudinally with respect to the long axis of the tooth row. The crowns of p3 and p4 are subequal in height and length but differ in morphology; in contrast to p3, the p4 is tribosphenic with a distinct trigonid and talonid. The paraconid of p4 consists of a single large cusp supporting a paracristid; this tooth also has a large metaconid, a long protocristid, and a long talonid. The protoconid is tall, sharp and slender. Labial structures on the premolars are not visible and therefore cannot be described in detail.
The lower molars of I. gunnelli lack a lingual cingulid and the trigonid fovea is open lingually. The hypoconulid is located on the distolingual border of the talonid, well separated from the entoconid and closer to the midline than to the entoconid. The height of the lingual cusps appears relatively low compared to the labial cusps. The entocristid is disrupted (the entocristid is short so that talonid basin drains lingually) on all three lower molars. The m1 and m2 are subequal in size and have similar crown morphology including a well-developed protoconid, paraconid, and metaconid. I. gunnelli and I. index both have a large hypoconulid on the m1 and m2, but in O. finneyi the hypoconulids on these teeth are small and low. The protoconid on m1 is markedly higher than the hypoconid, but is subequal in height to the protoconids on the m2 and m3. The m3 has two roots and bears a large hypoconulid.
The axial skeleton of I. gunnelli generally resembles that of other Eocene bats. An accurate vertebral count in the thoracolumbar region could not be determined due to poor preservation. The tail has 11 free caudal vertebrae. The scapula has an acromion process that extends anterolaterally beyond the glenoid fossa, and the lateral and dorsal edges of the acromion are flat. The infraspinous fossa is narrow and triangular shaped, tapering anteriorly towards the glenoid. A thick lip is present along the axillary border of the scapula. The coracoid process is short, stout, and appears to curve ventrolaterally.
Like other known Eocene bats, I. gunnelli has a well-developed wing skeleton with elongated hand and finger bones. The relative length of the humerus compared to the radius is 76%, which is comparable to that of O. finneyi (~76%) while somewhat greater than that of I. index (~71%). The head of the humerus is spherical, and the distal articular surface of the humerus is laterally displaced. The distal end of the radius is wider than the mean width of the shaft. An ossified ulnar patella is present. The ulna and radius are fused distally starting at approximately the midpoint of the forearm. The metacarpal formula is IV � V = III > II > I.
The hindlimb of I. gunnelli slightly differs from other Green River bats. The greater trochanter does not extend proximally as far as the femoral head, which differs from the condition seen in I. index and O. finneyi, in which the greater trochanter extends as far proximally as the level of the femoral head. The lesser trochanter is large and forms an extended flange in I. gunnelli as in O. finneyi. By contrast, the lesser trochanter is present only as a small tubercle in I. index. A third trochanter is absent in all three Green River bat taxa. The femur shaft is straight in anterior view and a femoral neck is absent. A small articulation is present between the distal ends of the tibia and fibula. The fibula is fully ossified and the proximal ends of the fibula and tibia are possibly sutured.
The first digit of the foot is shorter than the remaining digits. There is no visible contact between metatarsals of adjacent digits although intermetatarsal contact and/or facets may have been present, but are not detectable in our specimens. The shape of the proximal margin of the proximal phalanx seems to be flat in lateral view, but it is possible that this is due to taphonomic alteration. Foot digits II-V each have three phalanges.

Phylogenetic relationships
Phylogenetic analyses including all taxa resulted in three most parsimonious trees (MPTs) of 1511 steps (Fig 6A). The strict consensus tree recovered a sister relationship between Icaronycteris gunnelli and Icaronycteris index with high bootstrap support (bootstrap value 100). Onychonycteris finneyi fell sister to the Icaronycteris clade with lower support (bootstrap value 52), while Archaeonycteris trigonodon was recovered sister to the clade of Green River bats with low support (bootstrap value 37). Icaronycteris menui fell sister to the above clade in the strict consensus tree, also with low bootstap support (bootstrap value 46). The only other clade recovered in the strict consensus tree is the sister relationship between Rhinolophus ferrumequinum and Rhinopoma hardwickii, which received relatively high support (bootstrap value 76). Tachypteron franzeni in this analysis occurred in a polytomy with the previously mentioned clades, the remaining extant bats, Palaeochiropteryx tupaiodon, Hassianycteris messelensis, Icaronycteris? menui, and I. sigei.
Due to the fragmentary nature of both Icaronycteris menui and I. sigei, we conducted a second analysis omitting those taxa. This analysis resulted in two MPTs of 1420 steps each ( Fig  6B).

Discussion
The morphology of Icaronycteris gunnelli unambiguously places this new taxon within Icaronycteris as the sister taxon of I. index. The family Icaroncyteridae as previously recognized (e.g., [27]) contained one genus and three species, with only one species (I. index) known from North America. The two other referred species are from the early Eocene of Europe and Asia: Icaronycteris? menui from France [16,38] and Icaronycteris sigei from India [27]. It has always been somewhat ambiguous as to whether these latter species actually belong to the genus Icaronycteris in the sense of forming a clade with the type species, I. index, or simply share plesiomorphic features with that taxon. If the Icaronycteris (and by extension, the family Icaroncyteridae) is defined as including I. sigei, both the genus and the family can seemingly be diagnosed only by lack of apomorphies seen in other Eocene bat genera [54]. It is therefore possible that Icaronycteris and Icaronycteridae as thus defined might not represent a single lineage [54].

PLOS ONE
The oldest known bat skeletons and their implications for Eocene chiropteran diversification Icaronycteris gunnelli was placed as sister to I. index in all of our analyses with high support (Fig 6A and 6B). Both taxa are known from the same locality in the early Eocene of western North America, indicating at least a modest radiation of Icaronycteris in the Green River fauna. The sister group of Icaronycteris was found to be Onychonycteris, also from the Green River Formation. Onychonycteris was never recovered as the most basal bat, but instead consistently formed a clade with I. index and I. gunnelli. The placement of Onychonycteris received high support in the analysis excluding the more fragmentary taxa (bootstrap value 89; Fig 6B). This relationship-a sister-group relationship between Icaronycteris and Onychonycteris-has been recovered previously [55], but Onychonycteris has more frequently been recovered as the earliest diverging stem bat [12,56,57].
Icaronycteris? menui from the early Eocene of France was never recovered as grouping with the type species of Icaronycteris (I. index) but instead was placed as sister to a clade consisting of Icaronycteris, Onychonycteris finneyi, and Archaeonycteris trigonodon in our strict consensus tree. Relatively low bootstrap values support the above clade (bootstrap value 46; Fig 6A), with still lower support suggesting that it belongs to a clade with the North American Icaronycteris (bootstrap value 36). These results suggest Icaronycteris? menui likely represents a stem bat, but based on our analyses this taxon cannot be confidently assigned to the genus Icaronycteris or the family Icaronycteridae. We therefore state that it will be treated as Chiroptera incertae sedis until additional fossils or evidence emerges to allow definitive placement. Similarly, I. sigei from the early Eocene of India never grouped with other species of Icaronycteris in our analyses, instead occurring in a basal polytomy in our strict consensus tree (Fig 6A). Bootstrap values supporting the position of I. sigei are low, but place I. sigei either in a clade with Saccopteryx bilineata, Macrotus waterhousii, and Tachypteron franzeni (bootstrap value 33), sister to M. waterhousii (bootstrap value 32), or in a clade with S. bilineata, M. waterhousii, T. franzeni, and Myotis lucifugus (bootstrap value 31). These results suggest that I. sigei likely does not belong within Icaronycteris or Icaronycteridae, and may in fact be a crown bat. Icaronycteris sigei is currently known only from parts of lower dentition [27], and it does not seem that this material is sufficient to allow unambiguous assessment of its phylogenetic and taxonomic affinities other than to show that it does not seem to be an icaronycterid. Recovery of an upper dentition of I. sigei might help to stabilize its position within Chiroptera. Upper molars tentatively assigned to Icaronycteridae are known from the I. sigei type locality [26], but these have not been formally associated with I. sigei and thus were not included in our analysis. Pending further study, we state that I. sigei will be treated as Chiroptera incertae sedis until additional fossils or evidence emerges to allow definitive placement. With neither I.? menui nor I. sigei referred to Icaronycteris, the definitive geographic distribution of this genus (and the family, which is currently monogeneric) is thus restricted to North America.
Other relationships among stem bats recovered in our analyses differed somewhat from previously published phylogenies. Archaeonycteris trigonodon from Germany was recovered as sister to the Icaronycteris + Onychonycteris clade (Fig 6A) rather than sister to a clade of all bats except Icaronycteris and Onychonycteris, where it was placed in previous analyses (e.g., [12,57]). This placement of Archaeonycteris received only low to moderate support (bootstrap value 37 in analyses including I.? menui and I. sigei and 54 in analyses excluding the previous taxa). Future analyses including other species of Archaeonycteris, Palaeochiropteryx, and Hassianycteris might better resolve relationships among these taxa.
Palaeochiropteryx and Hassianycteris are typically recovered as the closest relatives to the chiropteran crown group among stem bat lineages, with most analyses indicating that Palaeochiropteryx is sister to crown bats (e.g., [12,37]). The position of both P. tupaiodon and H. messelensis in our analysis was unstable, with the pair forming part of a basal polytomy in the strict consensus including both I.? menui and I. sigei (Fig 6A). In the strict consensus tree excluding the two more fragmentary taxa, P. tupaiodon and H. messelensis were sister taxa and part of a larger clade also including I. index, I. gunnelli, O. finneyi, and A. trigonodon (Fig 6B). The most commonly recovered positions of P. tupaiodon and H. messelensis in the analysis including both fragmentary "Icaronycteris" taxa was a sister relationship between the two (bootstrap value 42), part of a clade including extant bats, Tachypteron  . Tachypteron franzeni in our analyses fell within crown Chiroptera as predicted in its description [58], but in our analysis was placed as sister to a clade consisting of Saccopteryx bilineata and Macrotus waterhousii (representing superfamilies Emballonuroidea and Noctilionoidea, respectively) rather than sister to S. bilineata as would be predicted for an emballonurid (bootstrap value 72 in analysis excluding more fragmentary taxa, 42 in analysis including all taxa). However, our placement might be the result of very limited sampling of extant taxa, so the familial relationship of Tachypteron with extant emballonurids as suggested by Storch et al. [58] cannot be considered refuted by our results and requires further testing with a larger taxonomic sample.
Results of our phylogenetic analyses suggest a number of potential synapomorphies in the dentition of Icaronycteris (= Icaronycteridae), consisting of I. index and I. gunnelli, including: (1) I1 crown undifferentiated from the root shaft, (2) upper canine lacking a lingual cingulum, (3) weak posterolateral accessory cusp on the upper canine, (4) a P4 postparacrista that does not extend to the distal margin of the tooth, (5) M1 and M2 with paraconule, (6) M1 postprotocrista becomes confluent with metacingulum distobuccally, (7) lower molars with relatively low lingual cusps, and (8) coronoid process of the dentary twice the height of the condyloid process. Character polarity in early bats is difficult to determine due to the lack of closely-related outgroups (especially outgroups with a relatively unmodified tribosphenic dentition). However, species of Icaronycteris have been suggested to be united solely on the basis of symplesiomorphies, clearly an undesirable situation [54]. The features listed above are not shared by most other early diverging taxa (e.g., Onychonycteridae, Archaeonycteridae) and may prove useful for identifying other members of Icaronycteridae, although we acknowledge that some of these characters may ultimately be shown to represent symplesiomorphies if more primitive bats, or close chiropteran outgroups, are discovered in the future.

Body mass
Our reconstructions suggest that Icaronycteris gunnelli weighed approximately 22.5-28.9 g., similar to the body mass estimates for Icaronycteris index (24-27 g.; [47]). This is remarkable because I. index appears to be a slightly larger bat in terms of forearm length (43.5mm in I. index vs. 40.6mm in I. gunnelli). The overlap in our mass estimates for these taxa can perhaps be explained by the way the body mass calculations are made. Although midshaft diameter of long bones has been shown to be a better indicator of body size in extant bats than other metrics [46], it is also a parameter that is highly affected by taphonomic processes. Due to deformation or compression of the bone during fossilization, the original shape may be altered. The left and right humerus of the holotype of I. gunnelli were measured with different diameters, 1.65 and 1.81mm respectively, explaining the wide range of body size estimates for this species. The differences in measurements of the right and left humeri are doubtless a result of taphonomy since in living animals such measurements are nearly identical. It remains possible that the actual body mass of I. gunnelli did not overlap with that of I. index or that the latter species was indeed significantly larger than the former species, but we cannot rule out considerable overlap given the wide range of the body mass estimation for the former taxon.

Stratigraphy
The published stratigraphic record of the Green River Formation regarding fossil bats is extremely limited. Often information on the exact location of where the fossil was collectedlet alone details of which stratigraphic layer it came from-is lacking in species descriptions. Details of the collection locality of the holotype of Icaronycteris index (YPM:VPPU:018150) are not clear, with different sources reporting different information. According to archival records of the Yale Peabody Museum, the holotype of I. index was purchased from Clarence Cushman (La Barge, WY) in 1940 and was collected in 1935 from the Holland brother's quarry, approximately 5 miles west of Kemmerer, Lincoln Co., Wyoming. However, according to Jepsen [14], the fossil was found by C. Cushman (Rangely, CO) around 1933 and presented to Princeton University in 1941, whereas Grande [41] wrote that this fossil was found at the Lewis Ranch site #2, also known as the Smith Hollow Quarry. While the exact location of the collection locality holotype of I. index thus remains confused, all three sources agree that it was discovered in the famous 18-inch layer (also referred to as "F-1" or "black-fish" layer; [41]) of the Green River Formation. While at least two other specimens of I. index have been found in this layer (UW:FV:21481a/b; FOBU:13777), other specimens come from the Sandwich Beds about one meter below the 18-inch layer.
Both known specimens of Onychonycteris finneyi (ROM:Palaeobiology-Vertebrate Fossils:55351A & AMNH:FM:142467) and the holotype of Icaronycteris gunnelli were found in the American Fossil Quarry. According to the archival records of the ROM and AMNH, the ROM specimen of O. finneyi was found 16.5 cm below the Tri-ash layers (= layers about 21 cm below the top ash of the lower sandwich beds). The AMNH specimen of O. finneyi was found below the "13-inch layers" about 91 cm above the "Red line ash" which forms the base of the sandwich beds. The holotype of I. gunnelli was discovered about 115 cm below the Tri-ash layers by Terry Rickords in 2017. The paratype of I. gunnelli was found in the bottom 60cm of the Sandwich Beds, just below the holotype of I. gunnelli. The two specimens of I. gunnelli are collected lower in the stratigraphic column than other bats mentioned above and thus represent the oldest two bat skeletons known. Interestingly, our stratigraphic analysis indicates no overlap in the temporal ranges of known Green River bat species, with the genus Icaronycteris including both the youngest and the oldest species (Fig 5).
Isolated teeth from Clarkforkian deposits in North America have also been referred to cf. Icaronycteris sp. by Gingerich [59,60], thus potentially extending the range of Icaronycteris deeper in time to the late Paleocene. Other fragmentary material suggests that Icaronycteris may have persisted in North America until the Gardenerbuttean, an early Bridgerian interval that ended approximately 50 Mya [61]. Icaronycteris sp. is also mentioned from early-middle Eocene locations from Wyoming [62]. However, whether or not these specimens actually represent members of the Icaronycteris clade identified here (consisting of I. index and I. gunnelli) remains to be determined. In this respect, we may only be at the brink of discovering the true diversity of the early Eocene bat fauna in North America.