Old and Cosmopolite: Molecular Phylogeny of Tropical–Subtropical Kites (Aves: Elaninae) with Taxonomic Implications

: Kites of the Elaninae group are small and medium-sized, mostly tropical raptors traditionally considered as an early diverged subfamily of the Accipitridae. We used nucleotide sequences of three genetic markers (mitochondrial Cyt b and COI , nuclear RAG-1 ) to reconstruct the phylogenetic relationships of the Elaninae, other kites, and representatives of di ﬀ erent families of diurnal raptors. Our results conﬁrm the basal position of Elaninae, separated the latest in Early Miocene, including Chelictinia riocourii, which was not sequenced before and belongs to this group. Not only DNA data but also cytological, morphological, and ecological data show the singularity of Elaninae. We suggest elevating this group to family level as Elanidae within the order Accipitriformes. It includes Gampsonyx swainsonii as a monotypic subfamily because of distinctive traits and DNA sequence data. Taxonomic implications for other macrogroups of Accipitriformes are discussed.

Sanger sequencing was carried out on ABI 3730 automated capillary sequencer (Applied Biosystems, Carlsbad, CA, USA) with the ABI Prism Big Dye Terminator Cycle Sequencing Ready Reaction Kit 3.1 (executed by STARSEQ GmbH, Mainz, Germany).

Phylogenetic Analysis
Mitochondrial DNA sequences were controlled in order to exclude the possibility of stop codons, indicating that these sequences did not come from nuclear copies. Alignments were created in BioEdit 7.2.6 [45], nucleotide sequences were aligned using the MUSCLE algorithm [46].
We analyzed 1039 aligned base pairs (bp) of Cyt b, 604 bp of COI, and 878 bp of RAG-1 genes of Elaninae. Some sequences remained unsuccessful. A total of 17 sequences from GenBank and Barcode of Life Data System (BOLD) databases were added in the analysis, thus, we examined all taxa of Elaninae except E. scriptus (Table 1, Figure 1). Maximum Likelihood and Neighbor-Joining trees for each locus and average pairwise p-distances were calculated with MEGA 6.06 [47]. Relative branch support was evaluated with 1000 bootstrap replicates for the Maximum Likelihood and 10,000 for Neighbor-Joining trees.
Bayesian concatenated mitochondrial and nuclear gene trees were calculated using BEAUti 1.8.0 [61]. For finding the best evolution model, jModelTest 2.1.10 was applied [62,63] with the Akaike information criterion [64], which showed a GTR+I+Γ model for Cyt b and COI complete datasets, TrN+Γ for the RAG-1 dataset, TPM1uf+Γ for Cyt b, and TIM2+I+Γ for separate COI kite datasets. Because the three last (and many other) models are not available in the BEAST/BEAUti package, they were changed to the most similar by ∆ of corrected AIC, and finally GTR+I [65,66], GTR+Γ, and GTR+I+Γ [65,67,68] were chosen for RAG-1, separate kites Cyt b, and all other datasets, respectively. To set the rates of the clock model, we used the half-value of the 2.1% sequence divergence per million years for Cyt b, which is accepted for birds [69], and 1.8% for COI [70], the estimated rate with lognormal prior distribution of the substitution rate for nuclear alignment. We chose the strict clock model, Yule speciation process [71,72], for the Markov chain Monte Carlo method to measure the length of the chain set with 10 million runs, sampling every 1000th run. Phylogenetic trees were performed in BEAST 1.8.0 [61], log files were checked for convergence of runs in Tracer 1.7.1 [73], and the trees with the best likelihood were summarized using TreeAnnotator 1.8.0 [61], and the first 10% of trees were discarded as burn-in. Completed phylogenetic trees were arranged and visualized in FigTree 1.4.3.

Results
The numbers of parsimony-informative and variable uninformative sites for Elanin kites were 91 and 118 bp of a total 1039 bp for Cyt b, 114 and 31 bp of 604 bp for COI, and 5 and 16 bp of 878 bp for RAG-1 gene loci, respectively. Sequence divergences using pairwise distances are relatively high and consist up to 2.2% for COI between subspecies, 8.7-10.9% for Cyt b and 8.8-9.6% for COI between species of the genus Elanus, and 11.4-13.0% for Cyt b and 11.9-15.2% for COI between genera of the group ( Table 3).
As can be seen from Figure 2, the families Sagittaridae (Finsch and Hartlaub, 1870), and Pandionidae (Sclater and Salvin, 1873), represent the basal branches of the order Accipitriformes. The group of all Elanin kites forms a separate clade basal to other Accipitridae, and its monophyly is confirmed with high support values for Bayesian Inference for mitochondrial gene sequences ( Figure 2a) as well as for the nuclear RAG-1 gene ( Figure 2b). As expected, other kites demonstrate polyphyly of this artificial group, forming clades together with different subfamily representatives, except Elanoides, which clusters with other non-Elaninae Accipitridae, with the genus being basal to the non-Elanin clade ( Figure 2). Phylogenetic trees, which were constructed with ML and NJ, provide mostly identical topologies for separate mitochondrial and nuclear datasets (Figures S1-S3).
According to the molecular clock model used in Bayesian analysis, we can estimate the separation of Elaninae from other diurnal raptors at Early Miocene, 21.3 Mya (with 95% interval of highest posterior density 24.9-17.6 Mya; Figure 3). The divergence within the group between Elanus and two other genera passed 16.7-15.3 Mya (95% HPD = 20.1-12.7 Mya for both nodes), and the speciation inside Elanus occurred in Late Miocene 8.6-7.5 Mya (95% HPD = 10.2-6.1 Mya for both nodes).
Chelictinia riocourii, which was never sequenced before, unequivocally, belongs to the Elaninae kite clade. The position of Gampsonyx swainsonii is more distant to the two other Elaninae genera. The separate position of Elanus axillaris in the genus as revealed by mitochondrial sequences is not confirmed by nuclear RAG-1 sequences. There are several different haplotypes inside Elanus caeruleus belonging to nominative subspecies supported by mitochondrial and nuclear data ( Figure 2).

Morphological Features of Elanin Kites
Among evolutionary features, which distinguish elanins from other diurnal raptors, several morphological traits are found. Some of them are absent in other Accipitridae. Firstly, it has a primitive structure of the skull resembling birds of prey embryos [74]: the palate remains completely schizognathous [18]. This feature was later found in several Accipitridae genera [19,21], but are most strongly separated in Elaninae skull maxillopalatine processes [20,21]; the boundary of the foramen olfactorii is ossifying [74]; very low basicranial angle for Elanus and curved falcon-type form of the orbital process for Gampsonyx were also notified [22]. Features of limb bones: manus is longer than humerus, the inner spur of hypotarsus starts from the inner board of the shaft [19], distant positions of metatarsal scars (like Osprey) with a specific form of the middle trochlea, pillar-like medial process of calcaneal tuberosity, and "elanin subtype" of the tarsus in general as result of specialization for easy flexing of the toes. [22]. Other osteological traits are an absence of hypocleidium in furcula, specific form of coracoids-high position of coracoid foramen on procoracoid process and short sternocoracoid process, large depth of the keel and constitution of the sternum like Osprey, and the first cervical rib fused with the vertebra for Elanus [18,22,23].
Elanin kites equally have myological peculiarities. Among those related with limbs: the fully developed piriformis serves as origin for the parts of the flexor digitorum superficialis, outstanding flexor digitorum longus, wide tendons of internal flexor hallucis brevis, very weak peroneus longus, completely separated heads of the pars perforatus digiti III, as well as only medial and no lateral head of the pars perforatus digiti IV [23]. Other unique muscle types: ceratoglossus arise to the medial insertion of the massive transverses hyoideus like cathartids, Secretarybird, and in part Osprey; lack of pars mandibularis; muscles of innominate bones reach forward beneath the anterior plates of the ilia, and the caudilio-femoralis arising strongly forward by a tendon from the ventral point of the pygostyle. For Gampsonyx, the pars metapatagialis of latissimus dorsi is related with ribs not only three and four but equally five and six by aponeurosis like falcons [23].
Other unique morphological features of Elaninae are absent grooves on the claws [20]; almost complete absence of plumules on the crown for Gampsonyx [21], most distinct coloration parts among all Accipitriformes and absence of the median black throat stripe; weakly developed tympanic box in syrinx; and a simple gut: open duodenum, undeveloped ileum with absent supracecal loop [23]. The peculiarities which appear apparently independently during the evolution of elanins and other Accipitriformes should be added: lack of feathers in back patagium, also for the Hooded Vulture Necrosyrtes Gloger, 1841 [21]; perforated-type pelvis and reducing of zygomatic process like Macheirhamphus [22,23]; reducing of the insertion upon the third bronchial ring as for Buteo (Lacépède, 1799), hawks [23].
In recent study [75], Elanus, the well-investigated genus of this group is characterized unlike other diurnal birds of prey by owl-like features as a result of convergent evolution: zygodactylism (like Osprey), large frontally-placed eyes, vibrissae around the beck, velvet structure of upper parts of primary and secondary feathers as an adaptation for silent flight, and the producing of pellets with undigested bones as result of low stomach acidity [75]. The Elanus representatives are more or less nocturnal, the most night-active species is Elanus scriptus, which dispose an asymmetrical position of ear openings like several owls, which helps them to detect prey by sound [75]. Moreover, elanin eggs are intensively pigmented as falconid, and elements of their vocalization resemble the Barn Owl Tyto (Billberg, 1828)-screech, and falcons-chatter, maternal crock [32].

Cytogenetics of Elanus caeruleus
The study of the Black-winged Kite reveals its remarkable cytogenetic structure not observed in other Accipitridae [76]. Diurnal raptors have karyotype organized by medium-sized chromosomes with few numbers of small chromosomes, instead of macro-and dot-chromosome karyotypes of other birds. Elanus caeruleus show an even more specific organization with only one pair of microchromosomes. At the same time, nucleolus organizer regions of the Black-winged Kite are located on two chromosome pairs instead of one which is atypical for Accipitridae [77]. Finally, a comparison of its chromosome banding pattern with other raptor taxa demonstrates that high rearrangement and sex chromosomes of E. caeruleus are almost equal in size in contrast even to most other birds. All these particularities are considered as a result of chromosome evolution [76].
The specimens from India which should belong to E. c. vociferus (Latham, 1790) have been ascribed a diploid number of 66 chromosomes [78], while samples from Morocco and South Africa, which should be part of nominative subspecies that contain 68 [76] and 64 [79] chromosomes, respectively. However, the differences in the number of chromosomes between the latter two studies may be also erroneous in distinguishing and reporting microchromosomes [76]. This explanation seems more probable than the existence of populations in Southern Europe and Maghreb, which are significantly distinct in genetics and cytology from populations in the main range. Further research on the topic is necessary.

Phylogenetics, Distribution, and Ecology of Elanins
The molecular clock estimates of elanins based on our mitochondrial DNA data do not correspond to calculations on the appearance of different Accipitridae based on the calibration of three previously described raptor fossils, Osprey and two Buteo hawks [80], Snake Eagle and two vultures [81], both assuming a split for Elaninae of circa 34 Mya. The consistence between molecular and morphological data issued from fossils is usually not very high, and the retention index between phylogenetic trees [82] of two types of information for birds is 0.6-0.7 [83]. At the same time, our data are not far from previous estimations of the node Elanus/Gampsonyx with an age of~20.5 Mya (95% HPD = 29.4-14.4 Mya) by Nagy and Tökölyi [80] and 16.7 Mya (95% HPD = 20.1-13.3 Mya) by this study (Figure 3). We did not find any data on Elaninae fossils, and the fossilization is often not usual for tropical biomes, but its existence eventually could highlight the origin of this group.
The position of Chelictinia riocourii has until recently been considered as provisional in absence of molecular data [7], and, traditionally, the Scissor-tailed Kite is placed in Elaninae [2]. Our study can thus confirm this assumption. The pairwise p-distances between genus Elanus and the two other genera, especially Gampsonyx, are comparable with the distances between representatives of different subfamilies in the order Accipitridae for Cyt b as well as for COI (Table 3) genes.
The phylogenetic position of subspecies of Elanus spp., including currently accepted and three formerly valid subspecies of E. caeruleus [84], cannot be determined with our dataset because of the lack of sufficient data. Several Cyt b and RAG-1 haplotypes of the African subspecies of the Black-winged Kite E. c. caeruleus (Desfontaines, 1789) were found across the entire continent (Figure 2a, Figures S1  and S2). This may be explained by the nomadic active movements as subadults, and adults across their wide range may exceed 1000 km [85], and the dispersal of nestlings (probably together with adult movements) even 3000 km [86]. The suggestion of Negro et al. [75] appears relevant, stating that the relatively nocturnal Elanus feeding on small rodents may occupy the niche which is mostly used by nomadic owls in the Northern Hemisphere. The difference between haplotypes of E. c. caeruleus, even for the more variable Cyt b gene, is low and the same haplotype appears in Egypt and South Africa. At the same time, two subspecies of E. leucurus show identical haplotypes for the accessible sequences of COI (Table 3). These subspecies are distinguished by small but significant size differences of wing, tail, and culmen [87,88], and their range is disrupted by the Amazonian basin [5], thus, the validity of these subspecies is not in doubt. However, we observed no distinction by the DNA-barcoding marker, but it apparently may be found with high-variable parts of the mitogenome as control region.
The Pearl Kite, the smallest Neotropical raptor, is morphologically and ecologically distinct from other elanins, being the smallest and most sedentary kite of this group (Table 4), and phylogenetically, this is the most distant taxon (Figures 2 and 3, Figures S2 and S3). These facts could suggest that the origin of Elaninae might be the Neotropic realm, and their migration trait probably appeared after a divergence of Gampsonyx from the common ancestor. The neotropical ancestor distribution for Elaninae is also estimated by Nagy and Tökölyi [77]. When the Pearl Kite was described, Vigors [89] notified its remarkable appearance to be close to falcons and hawks, but placed it together with hawks cause of the form of its bill and size of the wings. Later, this genus was placed to Falconidae by different authors [2,[90][91][92] and back to Accipitridae [20,[93][94][95] families. Three subspecies were described: nominative G. s. swainsonii (Vigors, 1825), G. s. leonae (Chubb, 1918), and G. s. magnus (Chubb, 1918). According to the geographic origin of the samples, our dataset should have molecular data for the two first subspecies. However, he difference in COI haplotypes between them is low or absent, and only birds from Guyana, which belong to the leonae race differ by one substitution ( Table 2, Figures 2a and  3). The range of these subspecies adjoin each other, and swainsonii and leonae differ slightly only by their plumage colors: gray upper surface darker, straw-color instead of the orange-beige of forehead and sides; and the magnus race, which has no rufous on its flanks and is also separated geographically, differing by wing, and tail measurements [5,96,97]. We cannot reject that a clinal variation may exist for this species with a wide range, thus, inviting detailed morphological and genetic studies on Pearl Kite subspecies in the future. Among the ecological features of elanins, parasites specific for these birds are worth mentioning separately. Clay [103] named the species group elani of Degeeriella (Neumann, 1906), a genus of Mallophaga (Nitzch, 1818), common for all genera of Elaninae, Elanoides, and Aviceda, that each of chewing lice species correspond to its own kite genus. It seems interesting to follow the cophylogeny of these lice and their hosts, like it was shown for other ectoparasites of different bird groups [104][105][106][107].

Taxonomic Implications for Elanins and Other Accipitrids
The South Asian subspecies of the Black-winged Kite Elanus caeruleus vociferus from India was described as full species [108]. The p-distance between African and the Thai E. caeruleus COI sequences ( Table 2) is even higher than between several species of one genus in Accipitridae [109][110][111][112][113]. Their split occurred at 1.5 ± 0.8 Mya (Figure 3). We suppose that E. c. vociferus may be restored as species. This also may be confirmed by different karyotypes of this subspecies that we mentioned above. Nonetheless, this issue cannot be resolved without additional material gathered from different subspecies of E. caeruleus. Additionally, vociferus differs morphologically from other Black-winged Kites by well-defined black tips of underwing coverts [114]. It is curious that this race is not even recognized as a subspecies by Brown and Amadon [4] followed by Ferguson-Lees and Christie [5], merging it with caeruleus because of their close morphology, and Mees [115] leaves the validity of vociferus to be discussed. Currently, E. c. vociferus is present in all main bird checklists [6,7,14].
At the same time, the position of Elanus axillaris sequences is contradictory. It is separate from large species E. caeruleus on the mitochondrial tree (Figure 2a), but both sequences of this species form clade with the Black-winged Kite (Figure 2b), though its support is very low (the posterior probabilities do not exceed 0.1). Apparently, the resolution of nuclear sequences RAG-1 is not sufficient to separate these species. Nonetheless, the tentative separation in mitochondrial but not in nuclear sequences may offer support to the suggestions of Debus [32] that the Australian E. axillaris is a recent species, split from E. caeruleus of the Old World. E. axillaris as well as the American E. leucurus were merged by some authors with the "main" species of the group as subspecies [84,114,116,117], which was not supported by other researchers [4,20,24,[118][119][120]. Clark and Banks [88] found conclusive evidence for the validity of these three species based on morphometric differences between them. Currently, the Black-shouldered and White-tailed Kites are mentioned as full species in the checklists [6,7,14]. All three species are considered as allospecies, forming superspecies Elanus [caeruleus] [24,88,118]. The presence of E. axillaris in E. caeruleus clade may mean that the Black-shouldered Kite is not homogeneous, despite the fact that there are no morphological differences inside of the species. We recognize the current taxonomic status of the Elanus species with the remark that the final composition of this genus may be changed when data from more mitochondrial and nuclear markers or even whole genomes for each species and subspecies will be available.
The change of status of elanins invites to look at neighboring taxa: A large group of hawk-like diurnal birds of prey Accipitridae together with the families Elanidae and Pandionidae could be considered as suborder Accipitres (Linnaeus, 1758), proposed by Coues and Prentiss [121], whose composition was changed by several taxonomists [4,[122][123][124][125] in new frames.
If we consider Cathartiformes as separate order [14], phylogenetically, the most distant taxon of the order Accipitriformes is a morphologically distinct primarily terrestrial large predator and sub-Saharan African endemic the Secretarybird Sagittarius serpentarius (Miller, 1779). The time of his separation from the main branch is relatively close to other Accipitriformes by our calculations (Figure 3) but more distant by the measures using fossil calibrations [80,81], and the Secretarybird branch is well-supported by different analyses (Figure 2), [3,33]. Thus, it forms a monotypic suborder Sagittari (Seebohm, 1890). Other taxonomic ranks as infraorder Sagittarii (Seebohm, 1890) proposed by Cracraft [126] or Wetmore's superfamily Sagittariioidea (Finsch and Hartlaub, 1870) [127] seem less appropriate. Firstly, this taxon is phylogenetically (Figure 2) [3,31,33] and morphologically [21][22][23] more distant from other taxa of Accipitriformes and should not have the same taxonomic rank as those. Furthermore, Cracraft grouped the Secretarybird in suborder Accipitres together with infraorder Falconii, which consisted of all other diurnal raptors except New World vultures, and Wetmore considered all other diurnal birds of prey as second superfamily Falconoidea. Today, after the separation of former order Falconiformes in two [7] or even three [14] orders, these classifications are apparently outdated.

Gampsonychinae Starikov & Wink, subfam. nov.
Type genus: Gampsonyx Vigors, 1825 Type species: Gampsonyx swainsonii Vigors, 1825 LSID: urn:lsid:zoobank.org:act:D25DE61A-7AE6-46BC-919A-44071B7AB9EA Diagnosis: The new subfamily is distinguished by combination of molecular phylogenetic, morphological and ecological data. Small falcon-like predators. Males and females have black upperparts, crown on the head, patch on breast side, bill, white underparts, chin and throat, yellow to orange forehead and cheeks. Plumules on the crown are practically absent. Orbital processes in the skull are rounded. Pars metapatagialis of latissimus dorsi muscle is related with ribs three to six. Mild sexual dimorphism, females are 10-20% larger than males. Feeds on small lizards and large insects. Inhabits dry and arid regions, savannas, shrubs, snags, pastures, open woodlands, also edges of tropical evergreen forest. Non-migratory, nomads and juvenile dispersal. Distributed in Neotropics from El Salvador to North Argentina. Monotypic, one genus and species, three subspecies.

Conclusions
The examination of the molecular phylogeny of Elanin kites based on mitochondrial and nuclear markers confirms their position as sister taxa to other Accipitridae, and together with morphological, cytogenetic, and ecological data, lead to separating them in the family Elanidae. At the same time, the peculiarities of the Pearl Kite allow us to distinguish it as a monotypic subfamily Gampsonychinae, which forms Elanidae with a second subfamily Elaninae consisting of genera Elanus and Chelictinia. An accumulation of genetic data will permit us to clarify the phylogeny on a species-subspecies level of elanins in perspective. Phylogenetic relationships between Elanidae, Accipitridae, and Pandionidae from one side and the distant Secretarybird from the other allow us to classify them in two suborders of Accipitriformes.