The emerging vertebrate model species for neurophysiological studies is Danionella cerebrum, new species (Teleostei: Cyprinidae)

The four described species of Danionella are tiny, transparent fishes that mature at sizes between 10–15 mm, and represent some of the most extreme cases of vertebrate progenesis known to date. The miniature adult size and larval appearance of Danionella, combined with a diverse behavioral repertoire linked to sound production by males, have established Danionella as an important model for neurophysiological studies. The external similarity between the different species of Danionella has offered an important challenge to taxonomic identification using traditional external characters, leading to confusion over the identity of the model species. Using combined morphological and molecular taxonomic approaches, we show here that the most extensively studied species of Danionella is not D. translucida, but represents an undescribed species, D. cerebrum n. sp. that is externally almost identical to D. translucida, but differs trenchantly in several internal characters. Molecular analyses confirm the distinctiveness of D. cerebrum and D. translucida and suggest that the two species are not even sister taxa. Analysis of the evolution of sexual dimorphisms associated with the Weberian apparatus reveals significant increases in complexity from the simpler condition found in D. dracula, to most complex conditions in D. cerebrum, D. mirifica and D. translucida.

With the two species of Danionella that are used in these studies, D. dracula and D. translucida, at the start of their career as vertebrate models, we began to take a comparative taxonomic and anatomical look at these fishes. The skeletal anatomy of D. dracula was already the subject of a monographic study 9 , but information on D. translucida is so far limited to the scant original description 3 . When comparing cleared and double stained specimens of D. translucida from the type locality and from additional localities around the southern tip of the Rakhine Yoma in Myanmar, we noticed a striking difference in the chondrocranium between individuals from different localities. This led us to compare in detail the skeletal anatomy of these two forms, and the other species of Danionella, and to complement our anatomical study with molecular data. Here we show that the organism that has been introduced as a promising model organism for neurophysiological research is not D. translucida, but an undescribed species, which we describe here in detail as Danionella cerebrum new species. We discuss anatomical differences between D. cerebrum new species and the other four species of Danionella and comment on the anatomical changes in the Weberian apparatus and associated structures during the evolution of this genus. We conclude by stressing that the study of Danionella presents a promising system to understand major morphological changes as the five species show anatomical differences far greater even than between species of only remotely related genera within the subfamily Danioninae.
No visible pigmentation in preserved specimens, except a line above anal-fin base in some specimens. In life, body colourless and largely translucent (Fig. 1a,b), except for a number of melanophores and yellowish colouration covering dorsal surface of skull. Melanophore pattern including an irregular scattering on top and sides of head, a row following the posterior margin of shoulder girdle, oblique melanophore rows along neural and haemal arches and spines in deeper layers of body, a horizontal row along insertion of anal-fin muscles starting above vent and extending posteriorly along caudal peduncle to anterior ventral procurrent rays, a row of melanophores at base of anal fin itself and along first anal-fin rays, and melanophores marking end of the hypural plate. Females with eggs, with numerous, large melanophores in lining of abdominal wall.
The cleared and double stained specimens ( Fig. 2a,b) revealed that, as in other species of the genus Danionella, the skull, hyopalatine arch, gill arches, endoskeletal shoulder girdle and pterygiophores are mostly cartilaginous with only thin perichondral ossifications giving the skeleton an overall larval appearance. The following bones are absent in D. cerebrum: kinethmoid, preethmoid, vomer, nasal, parietal, intercalar, extrascapular, infraorbitals 2-5, angular, ectopterygoid, metapterygoid, urohyal, hypobranchials 1-3, posttemporal, postcleithrum, mesocoracoid, pectoral radials 3-4, pelvic radials 1-3, all supraneurals behind supraneural 3, epineurals, epipleurals, uroneural 2, and scales. Exceptions to this theme of bone loss and reduction in the skull are the heavily ossified and toothed ceratobranchial 5, which is essential in food processing in conjunction with the well-ossified basioccipital, which carries the masticatory plate that ceratobranchial 5 works against. The basioccipital along with the equally wellossified exoccipital houses the intracranial part of the Weberian apparatus (sinus impar, capsules for lagena and asteriscus). Especially well-developed are also the Weberian ossicles and the os suspensorium, whose inner arms are fused in the midline with a bifurcated tip. There is a strong sexual dimorphism in the os suspensorium with males having the outer and inner arms more massively developed, the inner arms covering the roof of the swimbladder via posterior processes, the inner and outer arms connected via a broad bony flange and having the outer arms connected to the transverse processes of the second vertebra by a bony process. In addition males have a drumming cartilage associated with the fifth rib and swimbladder. Females lack all these modifications. Also the fifth rib is sexually dimorphic, stout and well ossified in the male and with a ventromedially directed process near its base, and feeble and poorly ossified in the female and lacking such a process.
Etymology. The species name cerebrum, Latin for brain, a noun in apposition, makes reference to the fact that this fish with one of the smallest adult brains among vertebrates has become a promising new model species for neurophysiological studies.
Distribution. Danionella cerebrum is known from a number of streams on the southern and eastern slopes of the Bago Yoma mountain range (Fig. 3) of Myanmar: Thandabin Chaung and Bala Chaung in Yangon Division, and from Daikme Chaung (type locality of Danionella translucida) and an unnamed stream northwest of Daikme Chaung in Bago Division.
Habitat. All water bodies in which Danionella cerebrum was found, are turbid low altitude streams (Fig. 3) with visible flow, surface temperatures of around 30 °C, pH 7.4-7.5 and soft water of 20-100 micro Siemens. Danionella cerebrum was not found at the surface, but at a depth below ca. 30 cm where the water is significantly  15 , the significance of which is unknown. Interestingly, we were also able to identify two adenin insertions in the coxI gene in the whole genome sequence of D. cerebrum provided by Kadobianskyi et al. 16 (GenBank accession SRMA00000000 as D. translucida) as well as the expected mt DNA tRNAs flanking the coxI, hundreds of base pairs up and downstream of the coxI region sequenced in our study, suggesting a genuine mitochondrial origin of our coxI sequences rather than a nuclear pseudogene (NUMT). For the phylogenetic analyses, however, the two insertions were excluded. In order to make sure that our phylogenetic analyses were not affected by the unlikely inclusion of NUMTs, we also conducted phylogenetic analyses based on an additional mitochondrial marker (16S rRNA) and three nuclear DNA markers (erg2b, rag1, rho). The different loci resulted in the same Our initial result based on morphological information that the model organism used in Schulze et al. 11 is not Danionella translucida, but a separate species, D. cerebrum, is supported by our molecular analysis. Samples from three different localities around the southern end of the Bago Yoma mountain range, including the type locality of D. cerebrum, clustered with samples obtained from the stock kept at Bolton Aquarium, from which the individuals used in Schulze et al. 11 , in Penalva et al. 12 and Kadobiansky et al. 16 originated and with the sample used in Britz et al. 6 and labeled Danionella sp. "South Myanmar" LR1707.
We found that Danionella cerebrum differs significantly from its close congeners and the uncorrected p-distances in the coxI gene between this species and D. translucida (Electronic Supplementary Table 2), the species with which it has been previously confused, ranged from 22.2-22.9%. Even between D. cerebrum and its closest relative D. mirifica, p-distances still range from 10.2-10.9%.
We recovered the following topology among the five species of Danionella independent of the genes analysed ( Fig. 4): (D. dracula,(D. priapus,(D. translucida,(D. mirifica,D. cerebrum)))). The age for the split of D. priapus from the remaining three taxa was estimated at ~ 17.

Discussion
Alive or preserved, Danionella cerebrum is difficult to distinguish from D. translucida, because the two look very similar and the distinguishing count of anal-fin rays is challenging to obtain, as the most posterior fin rays are very small and often pressed close to the body rendering their number extremely difficult to count. Reliable anal-fin ray counts, however, can be easily obtained in c&s specimens (see Fig. 2a,b) and distinguish the majority of specimens of both species, with 15-18 rays in D. cerebrum, but 12-14, rarely 15, rays in D. translucida. The difficulty to distinguish the two species with external characters is contrasted by marked differences in their internal anatomy. The most obvious one that led us to reinvestigate our samples of Danionella from the Bago Yoma is the flange of cartilage that extends ventromedially from the taenia marginalis anterior towards the trabecula communis (Fig. 2c,d). Its posterior and posteroventral margin is covered by the perichondrally ossified orbitosphenoid, a bone that is restricted to the taenia marginalis anterior in D. translucida (Fig. 2d). There is also a difference in the medially fused ossa suspensoria, the tip of which is bifurcated and ends at the anterior curvature of the anterior swim bladder chamber in D. cerebrum, but is pointed in D. translucida and ends in a single tip at the middle of the ventral curvature of the swimbladder chamber (Fig. 2e,f). A third clear difference is in the shape of the lateral process of the second centrum (Fig. 2g,h). This is blade-like in D. cerebrum (Fig. 2g) but has a bulging axe-shape in D. translucida (Fig. 2h). A less conspicuous though consistent difference is the presence in D. cerebrum of a small medioventrally directed process at the base of the fifth rib in the male (Fig. 1g), which potentially plays a role in the production of sound (see below). Such a process is absent in D. translucida.
Like the other species of Danionella, D. cerebrum shows a conspicuous sexual dimorphism putatively related to the production of sound. In mature males, the os suspensorium is greatly expanded by a lateral bone flange that bridges the gap between the outer and inner arms and an anterior flange that originates from the anterior margin of the outer arm. This anterior flange is confluent with an anterior process, which in turn is fused to the lateral process of the second centrum. This forms a large rigid basket-like structure in front of the anterior swimbladder chamber. A large drumming muscle originates from these flanges and inserts on the enlarged fifth rib and covers the large conical drumming cartilage like a cup. This apparatus has been hypothesized to be related in the production of sounds that aid in intraspecific communication of 60 and 120 Hz and amplitudes of around 140db with a duration of tens of milliseconds to minutes 11 . Females do not possess these additional flanges, the drumming muscle or the drumming cartilage and their fifth rib is smaller and only poorly ossified.
The most striking sexual dimorphism in Danionella is developed in D. dracula, in which mature males have huge odontoid fangs that resemble teeth in their arrangement and appearance 6 , while in other species of Danionella including D. cerebrum, the head skeleton and jaws are not sexually dimorphic.
A sexual dimorphism that is developed in all species of Danionella concerns the skeleton of the Weberian apparatus. Roberts 3 illustrated the skeleton of the Weberian apparatus in a male of D. translucida, but he seems to have been unaware of its dimorphic structure in males and females. This was later clarified by Britz 4 in Danionella mirifica, and the sexually dimorphic anatomy of the Weberian apparatus was subsequently also described for D. dracula 6,9 and D. priapus 5 .
A comparison of the os suspensorium and the associated skeletal structures responsible for sound production shows that there is an increase in complexity of the apparatus and its sexually dimorphic structure during the evolution of the genus Danionella (Fig. 4). All species of Danionella share the presence of a drumming muscle and drumming cartilage associated with a sexually dimorphic stouter fifth rib and os suspensorium, all putative synapomorphies of the genus. However, in Danionella dracula, the sister taxon to all other Danionella 6 ( Fig. 4), the apparatus of males is similar to that of females, as the anterior flange on the outer arm, its connection to the lateral process of the second vertebra, and the flange connecting outer and inner arms are all missing (Fig. 4). Its main dimorphism is thus the robustness of the different bones, which are wider, better ossified and stouter in males 9 . This is true of the fifth rib as well as the os suspensorium, from which the drumming muscle originates 9 .
Males of the Indian Danionella priapus share with D. translucida, D. mirifica, and D. cerebrum the flange connecting both arms of the os suspensorium, the anterior bone flange on the outer arm and the connecting process to the lateral process of the second vertebra (Fig. 4), putative synpomorphies of this Indo-Burmese clade. Interestingly some mature males of D. priapus lack the process connecting it to the lateral process of the second vertebra and, if present, it is much thinner than in the three Myanmar species of this clade. In contrast all mature males of D. translucida, D. mirifica, and D. cerebrum show a well-developed connecting process between the os suspensorium and the lateral process, and their bony flanges on the outer and inner arm are more extensive than in D. priapus, potentially providing a larger surface for the origin of the drumming muscle.
The striking structural differences in the arrangement of the Weberian apparatus in males of the different species of Danionella may equate to differences in the sounds produced by males. Unfortunately, sound in Danionella has been recorded and characterized to date only for a single species, D. cerebrum (as D. translucida 11 ). We expect additional comparative studies on sound production in the genus Danionella to demonstrate that species-specific differences in the frequency, amplitude and duration of sounds exist as a result of the differences in anatomical details of the putative sound producing apparatus that we observed.
The five species of Danionella are textbook examples for the link between miniaturization via developmental truncation and evolutionary morphological novelty, as proposed by Hanken and Wake 2 and discussed in detail for this genus by Rüber et al. 17 and Britz and Conway 9 . Specifically for Danionella, Conway et al. 10 have shown that it is the dramatic disparity in heterochronic shifts between different parts of the skeleton that has resulted in a vertebrate with an open, larval skull roof, but a fully developed Weberian apparatus. Both features render species of Danionella ideal candidates as neurophysiological model organisms. Unlike the equally tiny cyprinid species of the genus Boraras with more or less the same skeleton as their larger relatives of the Rasborini 18,19 , Danionella have a developmentally truncated neurocranium, but a precociously developing Weberian apparatus, which in the latter not only aids in the perception of sound, but also evolved into a unique apparatus for sound www.nature.com/scientificreports/ production. The remarkable combination of features of Danionella sets them apart from other cyprinids of the same or larger size and provide character combinations that are of interest to various vertebrate researchers. Whether miniaturization of the body of Danionella and its truncated anatomical condition go hand in hand with a miniaturization of the genome and loss of hox and other developmental genes, as recently demonstrated for the equally miniaturized and developmentally truncated danionine Paedocypris 20 will need to be demonstrated. We expect that in the future Danionella cerebrum will become another important vertebrate model organism beyond its current significance for neurophysiological studies.

Methods
Morphology. Methods, counts and measurements follow Britz et al. 9 , including the clearing and double staining (c&s). All sizes are provided as standard length (SL). Studied specimens are deposited in the collections of the Senckenberg Naturhistorische Sammlungen Dresden (MTD), the Natural History Museum, London (NHM), the Swedish Museum of Natural History (NRM), Stockholm, and the National Museum of Natural History (USNM) Smithsonian Institution, Washington DC. Because two species of Danionella co-occur at the type locality of D. translucida and the ranges in anal-fin counts reported by Roberts 3 indicate that the type series of D. translucida may have included two species, it was necessary to verify the taxonomic identity of the holotype of D. translucida (NRM 32232). Unfortunately Roberts 3 only provided a drawing of the holotype with little useful information and we had to request and study a photo of the holotype reproduced herein (Electronic Supplementary Fig. 1). Its close examination confirmed that it has an anal-fin ray count of 13 27 under the − f a setting and 1000 bootstrap replicates. The optimal partition schemes for the four data sets consisting of protein coding genes were generated using PartitionFinder 1.0.1 28 using initial partitions according to codon position. The setting model_selection = BIC and search = greedy was used for the different PartitionFinder runs (models = raxml).
Divergence times for data set 1 were estimated in BEAST v2.6.4 29 . After initial test runs and to ensure convergence and obtain ESS values > 200 we used a strict molecular clock and substitution. Clock and tree models were linked. The Yule process was used as speciation model. Based on the results of Britz et al. 6 , who obtained an age of the most recent common ancestor (MRCA) of Danionella of 29.5 (95% confidence interval 27.4-31.9) we used a calibration priors for the MRCA of Danionella assuming a normal distribution with a mean of 29.5 and a Sigma of 1.0. Two separate MCMC chain were run for 200 Million iterations sampling every 20'000 steps and were combined in LogCombiner 29 using a conservative burnin of 50%. Trees were annotated in TreeAnnotator 29