A new, large-bodied omnivorous bat (Noctilionoidea: Mystacinidae) reveals lost morphological and ecological diversity since the Miocene in New Zealand

A new genus and species of fossil bat is described from New Zealand’s only pre-Pleistocene Cenozoic terrestrial fauna, the early Miocene St Bathans Fauna of Central Otago, South Island. Bayesian total evidence phylogenetic analysis places this new Southern Hemisphere taxon among the burrowing bats (mystacinids) of New Zealand and Australia, although its lower dentition also resembles Africa’s endemic sucker-footed bats (myzopodids). As the first new bat genus to be added to New Zealand’s fauna in more than 150 years, it provides new insight into the original diversity of chiropterans in Australasia. It also underscores the significant decline in morphological diversity that has taken place in the highly distinctive, semi-terrestrial bat family Mystacinidae since the Miocene. This bat was relatively large, with an estimated body mass of ~40 g, and its dentition suggests it had an omnivorous diet. Its striking dental autapomorphies, including development of a large hypocone, signal a shift of diet compared with other mystacinids, and may provide evidence of an adaptive radiation in feeding strategy in this group of noctilionoid bats.

Differs from desmodontine, stenodermatine, carolliine and rhinophylline phyllostomids in having dilambdodont molars. Differs from most phyllostomines, macrotines, micronycterines, lonchophyllines, lonchorhynines and glossophagines in its myotodont rather than nyctalodont lower molars. Differs additionally from phyllostomines in M3 being large and with hypocone. Body mass. Using the equations of Gunnell et al. 38 and the proxies of upper first molar (M1) area, lower first molar (m1) area, and diameter of mid-shaft humerus, the body mass of eight of the ten known extinct and extant mystacinid taxa are given in Table 2. For the previously known mystacinids, these values range from ~8.5 g (Icarops paradox) to 39.3 g (Mystacina miocenalis). For Vulcanops jennyworthyae, the estimates are 42.6 g (based on M1 area) and 39.8 g (m1 area). This indicates a relatively large bat, compared with the median value of 13.8 g for 905 extant bat species (refs 38,39 ; see Discussion).

Phylogeny
The 50% majority rule consensus of post-burn-in trees from our Bayesian total evidence analysis is given in Fig. 3. Mystacinidae, Furipteridae + Noctilionidae, Thyropteridae, and Mormoopidae + Phyllostomidae formed clades, all with relatively high support (posterior probabilities shown in Fig. 3). Yangochiroptera had 100% support; Noctilionoidea and Vespertilionoidea were sister groups but with low support (50%). Myzopodidae and the emballonurid Saccopteryx bilineata grouped with relatively high support of 82%. Vulcanops fell within Mystacinidae, with a relatively high posterior probability of 81% but with relationships within the family less strongly supported (posterior probabilities 55-65%). Of the fossil taxa, Speonycteris aurantiadens grouped with phyllostomids rather than mormoopids 40,41 , but the others grouped in agreement with the results of previous studies, namely Phasmatonycteris spp. with Myzopoda spp. in Myzopodidae 22 , Australian Icarops spp. with New Zealand Mystacina spp. 27,30 and Notonycteris spp. with phyllostomine phyllostomids 42,43 .

Discussion
Bayesian total evidence analysis (mitochondrial and nuclear genes plus dental characters) places the New Zealand Miocene bat Vulcanops jennyworthyae among Australasia's living and fossil mystacinids. The overall results of our phylogenetic analysis are broadly congruent with recent large-scale molecular studies of bats 17,19,20,24,25 . Like some of these studies (e.g. 20 ), our analysis raises questions about the inclusion of Africa's Myzopodidae within Southern Hemisphere Noctilionoidea, suggesting instead that myzopodids may be more closely related to cosmopolitan emballonurids. Our analysis finds a sister-group relationship between Madagascar's extant Myzopoda species and North Africa's Phasmatonycteris species, supporting referral of those fossil taxa to the family Myzopodidae 22 . These fossil taxa were described by Gunnell et al. 22 from the Eocene Birket Qarun (~37 Ma) and Oligocene Upper Jebel Qatrani (~30 Ma) Formations of the Fayum in Egypt and referred to Myzopodidae on the basis of their lower dentitions (upper teeth are unknown). Although there are similarities between Vulcanops and myzopodids in the morphology of the lower dentition (e.g. m1-2 paraconid buccally displaced, not aligned with metaconid and entoconid, with talonid conspicuously wider than trigonid; m1-3 cristid obliqua curved rather than straight, with inflection near trigonid, and contacting trigonid buccal to rather than at midpoint between protoconid and metaconid; m1-3 with only shallow hypoflexid; m3 reduced in length and width with respect to m1-2; see Differential diagnosis), our phylogenetic analysis indicates these similarities are likely homoplastic.
Unequivocal noctilionoid families, from the Americas and Australasia, first appear in the fossil record slightly later: mormoopids 32-30 Ma in Florida 44   With respect to Southern Hemisphere palaeogeography, these divergence times long postdate estimated dates for the separation of India-Madagascar and Africa from Gondwana (>100 Ma), with Madagascar isolated in the Indian Ocean for more than 80 Ma 50 . The divergence dates, however, span those estimated for the breakup of the Australia-Antarctica-South America landmass, with Australia and Antarctica separating ~45 Ma and South America and Antarctica ~41 Ma 5,6 . New Zealand has been isolated in the South Pacific from ~52 Ma 3,4 , possibly before the divergence of the mystacinid lineage from other noctilionoids.
Based on phylogenetic inference and tectonic events, a number of biogeographic hypotheses have been proposed to explain the modern distribution of noctilionoids in the Southern Hemisphere. These include: a trans-Atlantic dispersal of stem noctilionoids from Africa to North or South America in the Eocene (e.g. 18,23 ); a North American origin (or transit) of stem noctilionoids, with dispersal to South America via an Eocene proto-Caribbean archipelago (e.g. 21 ); or an American origin or transit with subsequent dispersal of ancestral mystacinids to Australasia (e.g. 51 ). Gunnell et al. 22 proposed that noctilionoids originated and initially diversified in Africa (giving rise there to myzopodids) with a subsequent dispersal to Australia (producing mystacinids) and then to South America via Antarctica (this lineage leading to the five neotropical noctilionoid families).
Even if myzopodids are not noctilionoids, as suggested by some recent molecular data and by our total evidence analysis, one of these scenarios may still be valid. The modern bat crown-clade is thought to have originated in either Africa [52][53][54] or Eurasia 55 , with estimates for the age of the base of the extant bat radiation ranging from 62.6 Ma 20 to 50.3 Ma 25 . Potential living sister-groups of Noctilionoidea (sensu 20, i.e. excluding Myzopodidae) are vespertilionoids and emballonuroids. These two speciose groups have cosmopolitan distributions, occurring on all continents except Antarctica today, but molecular data suggest their roots were in Africa (stem and crown) and their oldest fossils are from North Africa 54 . These data, and an estimated divergence time of ~50 Ma to 37 Ma for Noctilionoidea 20,24,25 , are not inconsistent with the many previous biogeographical hypotheses for the distribution of superfamily Noctilionoidea outlined above.
The data are also potentially consistent with a vicariant origin of Mystacinidae (e.g. 56 ). In the early Paleogene, global temperatures were up to 12 °C higher than today, mainland Antarctica supported a frost-free, paratropical flora until 50 Ma and Nothofagus forests until at least 15 Ma, and intercontinental distances in the Southern Hemisphere were generally less than now 57,58 . The Paleogene remnants of Gondwana may have supported a broadly distributed noctilionoid fauna. If so, final fragmentation of the supercontinent may have led to the extinction of noctilionoids in Neogene Antarctica as ice-sheets grew 59 , with Mystacinidae vicariantly isolated in the Australian region. However, fossil bats have yet to be found in Antarctica, and a divergence date for mystacinids from other noctilionoids of ~50 to 37 Ma, after isolation of New Zealand in the Pacific ~52 Ma 3,4 , suggests that their presence in at least New Zealand probably reflects one or more post-Gondwanan dispersals.
Other bats were present in the Australian region in the early Paleogene, as demonstrated by the archaic Australonycteris clarkae from the 55 Ma Tingamarra fauna of southeastern Queensland, Australia 60,61 . The likely route taken by the first bats to reach Australia is unknown (the relationship of Australonycteris to other early chiropterans from Northern and Southern Hemispheres is unclear 62 ;). Between 55 and 26 Ma, there is long gap in the Australian mammal record 63,64 but when it resumes in the late Oligocene mystacinids were widespread, occurring in deposits in both central and northern Australia 27 . In New Zealand's oldest terrestrial mammal-bearing deposit, in 19-16 Ma sediments of the lower Bannockburn Formation near St Bathans, mystacinids are present and there is evidence that long-term ecological associations between Mystacina and its arthropod prey and roost trees and food plants were already established 30 .
If Vulcanops is a mystacinid, as we suggest here, it brings the number of representatives of this bat family in the Miocene St Bathans fauna to four 28,30 . In Australia, at least another four mystacinid species, all in the genus Icarops, are recorded from Oligocene to Miocene deposits in South Australia, Queensland and the Northern Territory, with two species co-occurring in some Queensland deposits 27 . In our total evidence analysis (Fig. 3), Vulcanops forms a clade with Mystacina species, with Icarops species paraphyletic relative to Vulcanops + Mystacina; this arrangement is congruent with a single origin of New Zealand mystacinids from an Australian source, but the topology receives only weak support.
A striking feature distinguishing the dentition of Vulcanops from previously known mystacinids (Mystacina spp. and Icarops spp.) is the presence of a large hypocone on its upper molars (Fig. 2). This structure is similar to that found in neotropical noctilionoids (phyllostomids and mormoopids). In that speciose group, it appears to have evolved multiple times 43 , but it is otherwise uncommon (and particularly rare on M3) in bats with a dilambdodont dentition. Outside Noctilionoidea, a large bulbous hypocone also occurs in the late Eocene Egyptian bat Aegyptonycteris knightae Simmons, Seiffert & Gunnell, 2016 65 . The latter is known only from its dilambdodont M2-3 and is the only member of its family whose relationships to other bats are unknown 65 . This large fossil bat differs significantly from Vulcanops in that its M2-3 also have a large conule at the base of the metacone and an ectostyle on the buccal margin, two features unknown in other bat families, living or extinct 65 . Among mammals, a hypocone increases occlusal area, effectively doubling the tooth surface devoted to processing food 66 . It is strongly correlated with a less strictly carnivorous diet, often involving an increase in plant consumption 66 . In Vulcanops, a long, broad, deep talonid on m1-2, low curved postcristid (=posthypocristid), cristid obliqua lacking carnassial notches, and long broad protocone on M1-2 are also horizontal shearing adaptations associated with a relatively more herbivorous diet. At the same time, elongation of the molar crests as also seen in Vulcanops (postmetacrista on M1-2 twice length of preparacrista, shallow ectoloph, open angle of m1-3 trigonids, cristid obliqua meeting trigonid buccal to centre of crown) are adaptations for vertical shearing, possibly indicating relatively more flesh eating.
As body size increases in bats, species with dilambdodont molars often include small vertebrates in their diets [65][66][67][68][69][70] . The presence of a well-developed hypocone in the ~40 g Vulcanops, however, argues against a strictly carnivorous diet. A tall, rounded hypocone is absent in flesh-eating bats (e.g. nycterids, megadermatids and phyllostomines Vampyrum and Trachops; 71 ), although a crestiform hypocone is present in fish-eating noctilionids SCientifiC REPORTS | (2018) 8:235 | DOI:10.1038/s41598-017-18403-w (Noctilio spp.) and is similar to the condition seen in some specimens of Vulcanops (e.g. CM 2013.18.916; Fig. 2a). Other aspects of Noctilio teeth that are possibly adaptations for piscivory (e.g. the discontinuous centrocrista of M1-2, in which the central blades reach the buccal margin of the crown, and the cristid obliqua of m1-2, which extends to the lingual margin of the crown) are very different from those of Vulcanops. The latter's dentition, and diet, was perhaps most similar to some phyllostomines that consume invertebrates, nectar, fruit, flowers, as well as small vertebrates (e.g. the large-bodied omnivorous Phyllostomus hastatus; 65 ).
New Zealand's Recent Mystacina species also have very broad, omnivorous diets consisting of nectar, pollen, fruit, flowers, and flying and terrestrial arthropods, but are not known to hunt small vertebrates 72,73 . However, Vulcanops exhibits several dental apomorphies, such as a large, blunt hypocone and long, broad, deep talonid, that are lacking in Mystacina species (as well as in Australia's extinct Icarops species) and suggest additional feeding capabilities in this extinct bat. No other extant or extinct bat known from the Australasian region has similar dental features. If a large blunt hypocone is indicative of increased herbivory in bats, as argued above (see also 65 ), this may provide evidence for the wider adoption, both geographically and taxonomically, of phytophagy in noctilionoid bats by the early Miocene 74 . It may also have relevance to phylogenetic reconstructions of the ancestral diet in Noctilionoidea and its constituent families 17,43,[74][75][76][77] .
There is some evidence from dental remains that Australia's extinct Icarops species were more insectivorous than New Zealand's omnivorous extant and extinct Mystacina species 27,30 . The derived features present in the dentition of Vulcanops that are absent in other mystacinids signal a further shift in diet. This could reflect exploitation of new, abundant and/or underutilized food resources in New Zealand compared with Australia where omnivorous peramelemorphian (bandicoot) and phalangeridan (possum and kangaroo) marsupials were morphologically diverse, speciose and abundant in forest ecosystems shared with mystacinids 27,64 . Baker et al. 76 have argued that the adaptive radiation of feeding strategies seen in phyllostomid noctilionoids -the most radical derived from a common ancestor for any monophyletic group of mammals -was triggered by the dietary inclusion of plant material in addition to insects, in concert with new environmental opportunities in Oligo-Miocene South America.
The large body size (~40 g) estimated for the early Miocene New Zealand mystacinids Vulcanops jennyworthyae and Mystacina miocenalis Hand, Lee, Worthy & Archer, 2015 30 is notable compared with other extant and extinct mystacinids (Table 2), and especially given that the ancestral body mass for noctilionoids and the Mystacina lineage has been estimated at ~10-14 g 78 . The evolution of relatively large size in certain bat lineages has been associated with ecological release from the biophysical constraints imposed by flight and echolocation during aerial insectivory, and occurs in lineages exhibiting divergent dietary and behavioural specializations such as frugivory (e.g. pteropodids) or gleaning and perch-hunting behaviour in extreme insectivory and animalivory (e.g. megadermatids, noctilionids) 78 . That some mystacinids have reached notably large sizes may be another example of this evolutionary trend. Mystacinids are renowned for their peculiar walking habits which enable them to exploit an exceptionally broad range of plant and animal resources 72 , including ground-flowering plants and large invertebrate prey that they can pursue on foot.
In early Miocene New Zealand, V. jennyworthyae was part of a diverse faunal community living in semitropical to warm-temperate Gondwanan rainforest on the shores of the vast 5000 sq km Manuherikia palaeolake 30,79,80 . A number of distinctive vertebrate taxa present in the early Miocene St Bathans assemblage, like Vulcanops, disappeared sometime before the late Pleistocene. These include crocodilians, terrestrial turtles, flamingo-like palaelodids, swiftlets, several pigeon, parrot and shorebird lineages and non-volant mammals (e.g. 8,9,[31][32][33]36,37 ). Most of these were probably warm-adapted species 8,9,81 . After the middle Miocene, global climate change 59 brought colder and drier conditions to New Zealand, with significant changes to vegetation and palaeoenvironments 80,82 . It is possible that this general cooling and drying trend also drove extinction of the Vulcanops lineage, and overall loss in mystacinid diversity over time. In Australia, the Icarops lineage also went extinct, sometime after the late middle Miocene, with Mystacinidae being the only one of eight crown bat families known to have become extinct on that continent 62 . The reasons for this remain unclear, in part because the later Miocene and Pliocene Australian mammal record is too poor to pinpoint the time of their disappearance 62,64 .

Methods
Stratigraphic nomenclature for the St Bathans region follows Schwarzhans et al. 35 . Dental terminology follows Hand et al. 27  To assess its likely phylogenetic affinities, Vulcanops was added to a large morphological character matrix (MorphoBank Project 2737; http://morphobank.org/permalink/?P2737) comprising 292 dental characters scored for 45 yangochiropterans (35 extant and 10 fossil species) plus 2 yinpterochiropteran outgroup taxa. Vulcanops could be scored for 143 of 292 characters, rendering it 49% complete. 112 characters representing plausible morphoclines were specified as ordered. We also created a total evidence matrix by combining the morphological dataset with the molecular dataset of Amador et al. 20 . This comprises DNA sequence data from five nuclear genes [dentin matrix protein 1 (DMP1), recombination activating protein 1 (RAG1), recombination activating protein 2 (RAG2), exon 11 of the breast cancer susceptibility protein 1 (BRCA1) and exon 28  The total evidence matrix was analysed using an undated Bayesian approach in MrBayes 3.2.6 84 . First, PartitionFinder 2.1.1 85 was used to select an appropriate partitioning scheme and set of models for the molecular data, assuming linked branch lengths, and using the "greedy" algorithm and AICc for model selection; only models implemented by MrBayes were tested. The morphological data was assigned the Mk model of Lewis 86 , assuming that variable characters had been scored, and with a gamma distribution with four rate categories to model rate heterogeneity among the morphological characters. The MrBayes analysis comprised four runs of four chains (three "heated, " one "cold"), sampling trees every 5000 generations. The analysis was run for 5 × 10 6 generations, with the first 25% of sampled trees discarded as burn-in; the post-burn-in trees were summarised using 50% majority rule consensus, with Bayesian posterior probabilities as support values.
To estimate body mass in extinct bats, Gunnell et al. 38 developed a set of algorithms based on dental, skeletal and weight measurements in 1,160 extant bats from eight families. We used these equations, and the proxies of upper first molar (M1) area, lower first molar (m1) area, and diameter of mid-shaft humerus, to estimate the body mass of eight of the ten known extinct and extant mystacinid taxa ( Table 2).
The morphological datasets generated or analysed during this study are included in this published article's tables or are available in the MorphoBank repository as Project 2737 (http://morphobank.org/permalink/?P2737).
Nomenclatural Act. This published work and the nomenclatural acts it contains have been registered in ZooBank, the proposed online registration system for the International Code of Zoological Nomenclature. The ZooBank life science identifiers can be resolved and the associated information viewed by appending the life science identifiers to the prefix http://zoobank.org/. The life science identifier for this publication is 13BDAB9F-4BC3-4711-A331-4E883DE52DC2, for Vulcanops is 498FA8AA-7DAF-4703-94F3-02931CE7F85F, and for V. jennyworthyae is 3A625804-F490-46F6-BDA0-C93A6523EE6D.