Atheriniform caudal-fin development
The development of the caudal-fin skeleton is largely consistent throughout the examined atheriniforms. In most of the examined species, five hypurals develop as separate entities. The lower hypural plate is then formed by fusion of hypural 1 and 2. The upper hypurals (hypural 3, 4 and 5) show different grades of fusion in different species, e.g. hypural 4 and 5 fuse in G. incisus and hypural 3 and 4 fuse in I. werneri and Atherina harringtonensis [39]. An exception is Pseudomugil furcatus in which two hypural plates (lower and upper hypural plate) are present but no separate hypurals develop as individual entities during any stage of ontogeny.
In all examined atheriniforms (i.e., Atherina presbyter, Bedotia geayi, Iriatherina werneri, Leuresthes tenuis, Marosatherina ladigesi, Melanotaenia lacustris, Pseudomugil furcatus), we observed two separate ural centra in late flexion to early postflexion stages. These initially separated centra fuse in later stages to form the compound centrum. This was also reported by Parenti LR [16], who described that in developmental stages of Phenacostethus and Dentatherina two ural centra are present. Two studies, on the development of the caudal skeleton in atheriniforms, i.e., Atherina harringtonensis [39] and Leuresthes tenuis [40], did not report this detail specifically. Neither the depicted specimens of Atherina harringtonensis nor the description [39] gave information on the presence of two separate ossifications. This may be the result of the relatively short frame during development in which the separate ossification centers are observable and the limited material available. However, the stipplings in the drawing of the latter study ([40]: Fig. 1) indicate the formation of two separate ural centra, thereby supporting our findings. We therefore conclude that the presence of two separate ural centra during ontogeny is a general atheriniform character. Parenti LR [16] assumed that preural centrum 1 and ural centrum 1 fuse into the anterior of these centra. In our specimens, there was no sign of preural centrum 1 and we conclude that preural centrum 1 is never developed.
The PH develops as an autogenous cartilage that initially has no connection to the notochord/ural centrum 1/compound centrum or hypural 1/lower hypural plate. During development a common cartilaginous base is formed that connects the parhypural and the lower hypural plate proximally and further articulates both structures with the notochord and subsequently with ural centrum 1 and then with the compound centrum. This cartilage is later reduced and the parhypural is separated from the lower hypural plate again and articulates with the compound centrum. In few species, e.g., G. incisus, the parhypural fuses to the lower hypural plate distally. After the reduction of the cartilage connecting the parhypural and the lower hypural plate, the latter fuses to the compound centrum (or ural centrum 1) in all herein examined species and in A. harringtonensis [39].
Beloniform caudal-fin development
The herein documented development of the caudal-fin skeleton of Oryzias woworae is consistent with that of Oryzias latipes as described by Fujita [41]. Despite the availability of several smaller specimens, we could not find separate hypural 1 and 2 and suspect that the lower hypural plate of Oryzias is a product of evolutionary fusion of hypural 1 and 2. An evolutionary fusion of hypural 1 and 2 therefore seems to characterize Adrianichthyidae. In the hemiramphid Hyporhamphus cf. limbatus hypural 1 and 2 develop as separate entities before they fuse to form the lower hypural plate, and we suspect a similar development occurs in Belone belone (Belonidae), and Hyporhamphus sajori (Hemiramphidae), where hypural plate 1 and 2 are already fused distally in the examined specimens [41, 42]. For Cypselurus doederleini (Exocoetidae) [43] and Cololabis saira [44] it was reported that a lower hypural plate formed by the fusion of hypural 1 and 2, but at hatching the lower hypural plate was already formed and it is unclear if hypural 1 and 2 develop separately. The character state in the grundplan of beloniforms is therefore debatable. In the evolutionary framework of Atherinomorpha either two evolutionary fusions of hypural 1 and 2 must have occurred (stem groups of Cyprinodontiformes and Adrianichthyidae) or one evolutionary fusion in the stem group of the Cyprinodontea and a subsequent separation in Belonoidei. We believe that the evolution of such a fusion is more likely than an evolutionary separation with a developmental fusion. We therefore consider the developmental pattern of separately developing hypural 1 and 2 and a subsequent fusion during development, as shown for Hyporhamphus cf. limbatus, as part of the grundplan of Beloniformes.
The components of the upper hypural plate of Oryzias are not that easy to determine as it could either comprise hypural 3, 4 and 5 or only hypural 3 and 4, which would include the presumption that hypural 5 is reduced [41]. In the belonids B. belone and Cololabis saira and the hemiramphids H. sajori and H. limbatus hypural 3, 4 and 5 develop separately and hypural 3 and 4 fuse to form the upper hypural plate [42, 44]. In the exocoetid Cy. doederleini the upper hypural plate is present at hatching and its components remain unclear [43], while in another exocoetid, Parexocoetus mento, two upper hypurals, presumably hypural 3 and 4, are present and fuse to form the upper hypural plate [45]. Hypural 5 is not developed in either of these two taxa. It seems likely that the upper hypural plate in Oryzias is a product of fusion of hypural 3 and 4 and that hypural 5 is completely reduced.
The CC in all examined Oryzias species is a product of the fusion of ural centrum 1 and 2. While Fujita [41] assumed that preural centrum 1 is part of the anterior ural centrum, we inferred it to comprise only ural centrum 1, as there are no signs of the occurrence of a separate preural centrum 1 during ontogeny. In C. saira, Cy. doederleini and H. sajori only one ural centrum supposedly develops [42, 44]. However, studying the development of H. limbatus we found two ural centra, which fuse to form the compound centrum. This contradicts these previous results and at least supports the assumption that in hemiramphids two ural centra are present during development. Comparing the late developmental stages of B. belone to H. limbatus, it seems possible that the compound centrum is also the product of fusion of ural centrum 1 and 2. However, the developmental data for C. saira contradicts this assumption, leaving the presence of two ural centra at the evolutionary base of the belonids in question. The condition in the grundplan of the Beloniformes, however, still seems to be the presence of two ural centra, as the reduction of one centrum or the evolutionary fusion of both centra seems more likely than the resurgence of one centrum within two families of beloniforms.
The development of other caudal-fin skeleton structures is similar to that of Oryzias and the other studied beloniform species [41,42,43,44,45]. Exceptions are the development of a uroneural as well as the presence of a third epural. While the latter is lacking in adrianichthyids, it is present in all other beloniforms [41,42,43,44,45,46]. A uroneural develops in all beloniforms dorsal to the posterior portion of the compound centrum. In adrianichthyids it is reduced and in O. woworae it is absent (Fig. 4e) [41,42,43,44,45,46]. The development of an extra caudal ossicle is restricted to Adrianichthyidae and is an autapomorphy of this family [41, 46].
Cyprinodontiform caudal-fin development
A variation in the pattern of hypural formation was observed among the cyprinodontiform species studied herein. While in Aplocheilus lineatus the lower hypural plate and hypural 3 and 4 develop, only two separate elements, the lower and upper hypural plate, develop in Aphyosemion striatum, Epiplatys annulatus and Poropanchax normani. For Fundulus xenicus it is reported that only a single hypural plate develops [47]. In the examined species HYP5 is not present during any point of ontogeny. It can be assumed that in the grundplan of cyprinodontiforms hypural 5 was already reduced and that hypural 3 and 4 developed as separated entities, much like in Aplocheilus lineatus. A common feature of cyprinodontiform development is the development of only one ural centrum, which emerges centered anterior to the lower hypural plate and upper hypural plate/hypural 3 & 4.
Grundplan of the caudal-fin skeleton in Atherinomorpha
The independent development of the lower hypurals (hypural 1 and 2) is a shared character of atheriniform species [39, 40] and beloniform species [41,42,43,44]. In these taxa hypural 1 and 2 fuse to form the lower hypural plate during ontogeny (Fig. 9). In the examined adrianichthyids [41] and cyprinodontiforms the lower hypural plate seemingly develops without prior separated hypurals. As we concluded that in beloniforms and atheriniforms hypural 1 and 2 develop separately, the evolutionary fusion of hypural 1 and 2 apparently evolved in parallel in adrianichthyids and at the base of the cyprinodontiforms (Fig. 9). In the grundplan of the Atherinomorpha hypural 1 and 2 develop separately and fuse later in ontogeny. A difference in the way the lower hypural plate is developed is not evident between adult atheriniforms (Fig. 8a-d) [1, 11, 12, 14,15,16, 18, 20, 21, 48,49,50,51,52], most adult beloniforms (Fig. 8e,f,h) [1, 11, 14, 16, 53] and those adult cyprinodontiforms in which the lower and upper hypural plate are not fused (Fig. 8i-l) [11, 14, 54]. In adult specimens of the beloniform B. belone a foramen in the LHP indicates the fusion of two formerly separated bones (Fig. 8g).
The upper hypurals (hypural 3, 4 and 5) develop separately at the base of atheriniforms and at the base of beloniforms. At the base of the cyprinodontiforms presumably only two upper hypurals (hypural 3 and 4) develop. We conclude that in the grundplan of the Atherinomorpha three separate upper hypurals develop and that the reduction of hypural 5 occurred at the base of the Cyprinodontiformes (Fig. 9). In adult specimens separate upper hypurals persist in many atheriniform taxa (Fig. 8a-c) [20, 49]. In a few adult beloniforms, i.e., B. belone and Tylosurus crocodilus, separate upper hypurals remain [1]. In zenarchopterids and exocoetids hypural 3 and 4 are fused to form the upper hypural plate (in many species only partially) and in some species hypural 5 is part of the upper hypural plate [53]. In the scomberesocid Cololabis saira hypural 5 remains separated from the upper hypural plate [1, 44]. In adrianichthyids hypural 5 seems to be reduced [1, 11]. In most cyprinodontiforms an upper hypural plate is present and composed of hypural 3 and 4. Exceptions are Aplocheilus lineatus (some specimen) and Epiplatys steindachneri in which hypural 3 and 4 remain separate [11]. No hypural 5 is distinguishable in any cyprinodontiform species.
A common ontogenetic character of atheriniforms and beloniforms is the development of two ural centra that fuse to form the compound centrum during ontogeny. In cyprinodontiforms only one ural centrum develops. In the grundplan of the Atherinomorpha two ural centra develop and fuse to form the compound centrum (Fig. 9). We cannot be sure if the one ural centrum that is developed in cyprinodontiform species is the result of evolutionary fusion of both or due to the reduction of either ural centrum 1 or ural centrum 2. The position of the developing ural centrum, centered anterior to the lower and upper hypural plate, would support the first case, as in atheriniforms and beloniforms ural centrum 1 and ural centrum 2 develop anterior to the lower hypurals and upper hypurals respectively. The fusion of the two ural centra could be expected to develop in an intermediate state. If the second case applies, it would be impossible to unequivocally homologize the developing ural centrum with either ural centrum 1 or ural centrum 2 in atheriniforms and beloniforms. In adult specimens of all three taxa, the compound centrum of atheriniforms and beloniforms and the ural centrum of cyprinodontiforms are not distinguishable by their shape, which can be described as an anterior half centrum and a posterior upward-pointing cone (Fig. 8). This would also support the hypothesis that the ural centrum of cyprinodontiforms is the result of evolutionary fusion. A preural centrum 1 is neither developed separately in any of our examined species nor in any of the previously studied species [39,40,41,42,43,44, 47]. Although it was hypothesised by some authors that preural centrum 1 is part of the compound centrum in some species, we found no evidence that would support this hypothesis.
Further similarities of atheriniforms, beloniforms and cyprinodontiforms which are also part of the grundplan of the Atherinomorpha include the autogenous development of the parhypural and the epurals as well as the autogenous development of at least the haemal and neural spines of preural centra 2–5 (Fig. 9).
To recap, the grundplan of the caudal-fin development of the Atherinomorpha includes: 1) development of five individual hypurals of which hypural 1 and 2 subsequently fuse to form the lower hypural plate; 2) development of two separate ural centra which fuse to form the compound centrum; 3) absence of preural centrum 1 during ontogeny, 4) development of an autogenous parhypural and autogenous haemal spines and neural spines of at least preural centra 2 to 5; 5) development of two autogenous epurals and 6) development of inter-haemal spine cartilage 3 (Fig. 9).
Comparison to ovalentarian taxa
The Atherinomorpha have been considered a monophyletic group throughout the last 60 years [10, 12, 13, 16, 17, 19, 20, 22,23,24,25, 27,28,29, 55] but their phylogenetic position within Percomorphacea and their closest relatives remain uncertain, due to morphological [e.g., 13, 16, 56] and molecular [e.g., 25, 26, 28, 30, 31] analyses repeatedly retrieving varying results. Recently, Wainwright PC, Smith WL, Price SA, et al. [26], Betancur-R R, Broughton RE, Wiley EO, et al. [27] and Betancur-R R, Wiley EO, Arratia G, et al. [28] provided convincing molecular evidence for the Atherinomorpha as part of the Ovalentaria. The Ovalentaria-hypothesis suggests that many taxa, which previously were widely separated within the Percomorphacea, are closely related and form a monophylum and, therefore, provides new impulses for comparative analyses. Although molecular support values for the Ovalentaria are persuasive, the support values for ovalentarian intrarelationships for most cases are quite low. Possible sister-taxa relationships previously suggested for atherinomorphs by morphological and molecular data include the Mugilidae [13, 19, 20, 22, 24, 56], the Blennioidei and Gobiesocidae [23], the Cichlidae, Embiotocidae and Pomacentridae [25]. Recent studies suggest that the Cichlidae [29] or the group comprising Cichlidae, Polycentridae and Pholidichtyidae [27, 28] are more closely related to the Atherinomorpha.
For many of the contemplable taxa studies on the development of the caudal fin are scarce or missing. For blenniids [57], cichlids [58, 59] and clinids [60, 61] there is some ontogenetic data, and for mugilids [62] and pomacentrids [63] detailed descriptions are available. Similarities between the caudal-fin development of these taxa and the Atherinomorpha include autogenous development of the parhypural and the epurals, the autogenous development of some haemal and neural spines of the preural centra (i.e., preural centra 2 and 3 in mugilids and at least preural centra 2 and 3 in blenniids, cichlids and pomacentrids) [59, 60, 62, 63]. In the cichlids examined for this study (Amatitlania nigrofasciata, Geophagus sp., Hemichromis bimaculatus), the haemal spines and neural spines of preural centrum 2 and preural centrum 3 develop autogenously. A cartilaginous bridge connects the proximal tip of the parhypural to the proximal tip of hypural 1 during ontogeny in atheriniforms. Such a connection is also present in cichlids, clinids, mugilids and pomacentrids [57, 59, 62, 63] suggesting that at the base of the Atherinomorpha such a connection was present and was reduced within beloniforms and cyprinodontiforms.
At the base of the Atherinomorpha five hypurals are present during development and hypural 1 and 2 fuse to form the lower hypural plate. Five hypurals also can be seen during ontogeny in cichlids, some clinids, e.g., Clinus cottoides, mugilids and pomacentrids [59, 60, 62, 63]. While in cichlids no hypural fusion occurs, and the hypurals remain separate in adults, hypural 1 and 2 fuse to form the lower hypural plate in clinids, mugilids and pomacentrids. In clinids this fusion occurs early in development and additionally the parhypural fuses to the lower hypural plate. The fusion of the lower hypurals to form the lower hypural plate could be a character that positions the clinids, mugilids and pomacentrids closer to the Atherinomorpha. Fusion of the upper hypurals happens in clinids and mugilids, where hypural 3 and 4 fuse to form the upper hypural plate. Although such a fusion occurs in beloniforms and cyprinodontiforms too, it seems likely that this trait evolved independently within the atherinomorphs and clinids/mugilids based on the well supported monophyly of the Atherinomorpha. In blenniids, the lower and upper hypural plate develop without separate hypural-precursors [57]. Apparently, this is also a separately acquired character in blenniids and cyprinodontiforms.
The compound centrum in atherinomorphs develops by fusion of ural centrum 1 and ural centrum 2. Within the Ovalentaria a similar development is only known in mugilids, wherein ural centrum 1 emerges anterior to the lower hypurals and ural centrum 2 anterior to the upper hypurals and both fuse to form a compound centrum with an identical shape to the compound centrum of atherinomorphs [62]. In the other previously studied ovalentarian taxa, only one elongated ural centrum develops that covers the notochord from the beginning of the parhypural almost to the caudal tip of the notochord [57,58,59,60,61, 63]. During ontogeny this centrum also shortens and in adults has a similar shape as in atherinomorphs and mugilids [1]. The similar development in atherinomorphs and mugilids could indicate a closer relationship of these taxa or a shared plesiomorphic character absent in the remaining ovalentarians. The development of the ural centrum in the other taxa in contrast raises the question if this is the result of evolutionary fusion of two previously separated centra or if one ural centrum got reduced and the remaining centrum elongated and took the former’s place. The connection of these two developmental modes remains unanswered for now and needs more detailed developmental studies of a variety of ovalentarian taxa to be answered with more certainty. Subsequently, this would help to evaluate the validity of the Ovalentaria based on morphological data.