Morphological evidence for sister group relationship between ﬂamingos (Aves: Phoenicopteridae) and grebes (Podicipedidae)

A recent molecular analysis strongly supported sister group relationship between ﬂamingos (Phoenicopteridae) and grebes (Podicipedidae), a hypothesis which has not been suggested before. Flamingos are long-legged ﬁlter-feeders whereas grebes are morphologically quite divergent foot-propelled diving birds, and sister group relationship between these two taxa would thus provide an interesting example of evolution of different feeding strategies in birds. To test monophyly of a clade including grebes and ﬂamingos, I performed a cladistic analysis of 70 morphological characters which were scored for 17 taxa. Parsimony analysis of these data supported monophyly of the taxon (Podicipedidae + Phoenicopteridae) and the clade received high bootstrap support. Previously overlooked morphological, oological and parasitological evidence is recorded which supports this hypothesis, and which makes the taxon (Podicipedidae + Phoenicopteridae) one of the best supported higher-level clades within modern birds. The phylogenetic signiﬁcance of some fossil ﬂamingo-like birds is discussed. The Middle Eocene taxon Juncitarsus is most likely the sister taxon of the clade (Podicipedidae + (Palaelodidae + Phoenicopteridae)) although resolution of its exact systematic position awaits revision of the fossil material. Contrary to previous assumptions, it is more parsimonious to assume that ﬂamingos evolved from a highly aquatic ancestor than from a shorebird-like ancestor.


INTRODUCTION
Avian higher-level systematics is only poorly understood and in particular, the phylogenetic affinities of flamingos (Phoenicopteridae) are among the 'most controversial and long-standing problems' (Sibley & Ahlquist, 1990). These birds show a strange mosaic of a duck-like beak with stork-like legs and were indeed considered to be most closely related to either anseriform or ciconiiform birds by earlier authors (see the review of the history of avian classification in Sibley & Ahlquist, 1990). Neither of these hypotheses, however, has been convincingly supported with derived characters and the evidence has been critically discussed by Olson & Feduccia (1980a).
There is strong morphological and molecular support for monophyly of Galloanseres to the exclusion of the Phoenicopteridae (e.g. Sibley & Ahlquist, 1990;Livezey, 1997;van Tuinen, Sibley & Hedges, 2000;Livezey & Zusi, 2001;this study), and monophyly of Phoenicopteridae and Anseriformes was not supported by the study of Mindell et al. (1997). The only cladistic analysis which did support sister group relationship between Phoenicopteridae and Anseriformes was by Ericson (1997) who did not, however, list the characters supporting this result. Re-analysis of the data by Ericson, Parsons & Johansson (2001) did not support monophyly of the taxon (Phoenicopteridae + Anseriformes). Cracraft (1981) considered a clade comprising flamingos, storks and ibises as only 'moderately welldefined', and most of the few synapomorphies supporting sister group relationship between storks and flamingos (Cracraft, 1988) are characters of the hindlimb and might well be functionally related to the elongated legs of these birds. Olson & Feduccia (1980a) suggested that the Phoenicopteridae are most closely related to the charadriiform stilts and avocets (Recurvirostridae). These authors did not evaluate the proposed synapomorphies in a cladistic context and the assumed relationships were not depicted in some kind of phylogenetic tree. As already discussed by Cracraft (1981), there is considerable character conflict with this hypothesis, which has not been supported by subsequent analyses (e.g. Sibley & Ahlquist, 1990;Livezey & Zusi, 2001), because flamingos lack many derived characters that support inclusion of the Recurvirostridae in a clade with other charadriiform birds such as gulls (Laridae).
Recently, a completely new hypothesis on flamingo relationships was set up by van Tuinen et al. (2001), who analysed mitochondrial and nuclear DNA sequences as well as DNA-DNA hybridization data and found strong support for sister group relationship between flamingos and grebes (Podicipedidae). Although both DNA-DNA hybridization studies and mitochondrial sequence data have been considered problematic for the study of higher level relationships among birds (e.g. Houde, 1987;Lanyon, 1992;Groth & Barrowclough, 1999), the study of van Tuinen et al. (2001) is notable for its high statistical support and the high degree of congruence between these different data sets.
A closer relationship between flamingos and grebes has not been proposed before, and grebes were traditionally (e.g. Fürbringer, 1888; Beddard, 1898) considered to be most closely related to loons (Gaviidae). With the exception of Cracraft (1982,1988), most recent authors have considered the shared derived similarities between Podicipedidae and Gaviidae to be due to convergence, and the systematic affinities of grebes to be uncertain (Sibley & Ahlquist, 1990).
Whereas flamingos are long-legged, wading filterfeeders, grebes are foot-propelled diving birds which mainly feed on fish and small arthropods (del Hoyo, 1992;Llimona & del Hoyo, 1992). As noted by van Tuinen et al. (2001), sister group relationship between grebes and flamingos would thus provide an interesting example of evolution of divergent feeding strategies among birds.
Van Tuinen et al. (2001) stated that flamingos and grebes 'show no resemblance', but no detailed morphological comparisons between these taxa have ever been made. The single cladistic analysis of morphological data that included both flamingos and grebes is an analysis of cranial and vertebral characters by Livezey & Zusi (2001), which resulted in sister group relationship between flamingos and a taxon including grebes, loons, penguins (Spheniscidae) and procellariiform birds. However, these authors considered the results of their study to be preliminary and the position of flamingos not to be convincing; derived characters supporting the resulting phylogeny were not listed.
Here, I test the hypothesis of a flamingo-grebe sister group relationship by a cladistic analysis of 70 morphological characters, and present previously overlooked morphological, oological and parasitological evidence for its support.

TAXA AND CHARACTERS
Apart from representatives of all other higher avian taxa, skeletons of the following taxa have been examined in the collection of Forschungsinstitut Senckenberg (SMF Although the phylogenetic affinities of some fossil taxa are briefly discussed, revision of the fossil record of flamingo-like birds was beyond the scope of this study and these were not included in the cladistic analysis. Coding of the osteological characters is based on personal study, and all non-osteological characters were taken from the literature. Anatomical terminology follows Baumel & Witmer (1993) and Vanden Berge & Zweers (1993).

PHYLOGENETIC ANALYSIS
Seventy morphological characters of 17 ingroup taxa were coded for the phylogenetic analysis with PAUP 3.1 (Swofford, 1993) (see character matrix in Appendix 2). The shortest tree was found with the heuristic search option. The consistency index (CI), retention index (RI) and rescaled consistency index (RC) were calculated. The robustness of the tree was tested with a bootstrap analysis of 1000 replicates.
Three characters were coded as ordered. Two of these concern vertebrae counts which were grouped to avoid an overly great influence on the analysis (which would have strengthened monophyly of Phoenicopteridae and Podicipedidae). Calculation with all characters unordered did not change the resulting tree topology.

OUTGROUP
Due to the poorly resolved higher-level phylogeny of birds, choice of an appropriate outgroup has proved difficult. Outgroup comparisons were initially made with the palaeognathous Tinamidae, which were established as the sister taxon of neognathous birds by most recent phylogenetic analyses (e.g. Groth & Barrowclough, 1999;Livezey & Zusi, 2001). This analysis confirmed monophyly of Galloanseres to the exclusion of the Phoenicopteridae. A second analysis was then performed using Tinamidae, Galliformes and Anseriformes as outgroup taxa.

PHYLOGENETIC ANALYSIS
Cladistic analysis of the character matrix in Appendix 2 with PAUP 3.1 resulted in three most parsimonious trees, the consensus tree of which is shown in Figure 1. In all the resulting trees, Phoenicopteridae was optimized as sister taxon of Podicipedidae, and this clade received a bootstrap support of 80%. Monophyly of (Galliformes + Anseriformes) and (Recurvirostridae + Laridae), to the exclusion of the Phoenicopteridae, received high bootstrap support of 92% and 99%, respectively. Sister group relationship between Threskiornithidae and Ciconiidae was supported with a bootstrap value of 64%. Three other taxa received weak bootstrap support (see Fig. 1). An analysis with Phasianidae/Cracidae and the two anseriform taxa as additional outgroup taxa did not change the resulting tree topology.

SYNAPOMORPHIES OF THE TAXON (PHOENICOPTERIDAE + PODICIPEDIDAE)
In all the resulting trees, the following characters were optimized as synapomorphies of the taxon (Phoenicopteridae + Podicipedidae); the numbers in parentheses refer to character numbers in Appendix 1: (30) At least fourth to seventh cervical vertebrae strongly elongate, processus spinosus forming a marked ridge (CI = 1.0). The shape of the cervical vertebrae of flamingos is very characteristic and strongly resembles that of grebes.
(32) At least four thoracic vertebrae fused to a notarium (CI = 0.25). In flamingos, four vertebrae are fused, in grebes the number varies between four and five (Storer, 1982). A notarium is absent in other taxa that were previously considered to be related to the Phoenicopteridae (Anseriformes, Ciconiidae and Recurvirostridae).
(41) Humerus with a marked oval depression at insertion site of musculus scapulohumeralis cranialis (CI = 1.0). This character was regarded by Olson & Feduccia (1980a) to be a unique feature of flamingos (including the fossil taxa Juncitarsus and Palaelodidae), but it is also present in grebes (Fig. 2).

DISCUSSION
As shown by van Tuinen et al. (2001) and in this study, monophyly of the taxon (Phoenicopteridae + Podicipedidae) is supported by molecular, morphological and oological evidence, which makes it one of the best supported higher-level clades within modern birds. Virtually none of the morphological characters presented above has been recognized as potential synapomorphies of a flamingo-grebe clade before, although the derived similarities shared by flamingos and grebes are difficult to explain by convergence, given the very different living habits of both taxa.
Grebes and flamingos are parasitized by a taxon of cestodes, the Amabiliidae, which is unique to these two taxa (Storer, 2000). Cestodes have a high degree of host specificity (Olson & Feduccia, 1980a;Storer, 2000), and the presence of a taxon which is found exclusively in grebes and flamingos is further evidence for their close relationship.
The earliest, well preserved fossils of grebes are from Miocene deposits ( vec, 1982, 1984Olson, 1985Olson, , 1995, but they are very similar to modern Podicipedidae and thus do not contribute to an understanding of the phylogenetic relationships of the group. Flamingo-like birds, however, are known from earlier deposits and have a fairly extensive fossil record (see Olson & Feduccia, 1980a).
Van Tuinen et al. (2001) hypothesized that evolution of grebes and flamingos started from an ancestor with a 'typical shorebird habitus and lifestyle', and based this assumption on the morphology of the Middle Eocene Juncitarsus which was thought to provide an evolutionary link between flamingos and shorebirds (Olson & Feduccia, 1980a;Peters, 1987). Except for the strongly elongated tarsometatarsus, however, all characters that were listed as evidence for phoenicopterid affinities of Juncitarsus by Olson & Feduccia (1980a) are also present in the Podicipedidae. Moreover, Juncitarsus lacks several of the derived characters which define monophyly of the clade (Podicipedidae + Phoenicopteridae). Most notably, a notarium is either completely absent (Olson & Feduccia, 1980a) or consists of only two vertebrae (Peters, 1987 the proximal phalanx of the second digit is not greatly elongated and narrow (Fig. 2), and the hindtoe is not strongly reduced as in extant Podicipedidae and Phoenicopteridae (Fig. 4). Assessment of the exact systematic position of Juncitarsus awaits revision of the fossil material, which is beyond the scope of this study. However, if this taxon can be shown to be closely related to grebes and flamingos, it is most likely the sister taxon of a clade including Phoenicopteridae and Podicipedidae.
The Palaelodidae, another taxon of early Tertiary flamingo-like birds, are known from numerous skeletal remains from the Tertiary of Europe, both Americas, and Australia (Olson & Feduccia, 1980a;Cheneval, 1983;Alvarenga, 1990;Boles, 1991). Shared derived characters which support monophyly of Palaelodidae and Phoenicopteridae to the exclusion of the Podicipedidae include a well developed spina externa sterni with a slightly bifurcate tip and very deep mandibular rami, which Cheneval & Escuillié (1992) mentioned as evidence for the presence of a primitive filter-feeding apparatus in palaelodids. Palaelodids also resemble flamingos in many aspects of their overall osteology, but detailed comparisons with grebes have not yet been made. Most parts of the skeleton, especially the wing bones, are equally similar to the corresponding elements of grebes (Fig. 2).
Most interesting with regard to flamingo evolution is the fact that the 'foot bones of Palaelodus show many similarities with those of a foot-propelled diving bird such as Podiceps' (Cheneval & Escuillié, 1992), e.g. the tarsometatarsus is mediolaterally compressed, the hypotarsus is complex and some flexor tendons are enclosed in bony canals (the arrangement of which is, however, slightly different, see Fig. 5). Olson & Feduccia (1980a) assumed that palaelodids 'may have occupied a more duck-like swimming niche than do typical flamingos', whereas Cheneval & Escuillié (1992) noted that palaelodids 'have been better diving birds than swimming birds'. However, some of the characters that the latter authors listed as diving adaptations of the Palaelodidae are also present in the Phoenicopteridae, i.e. a 'short femur with a double articulation in the region of the acetabulum' and a 'long tibiotarsus with a very well-developed cnemial crest' (Fig. 4). Contrary to virtually all modern diving birds, the humerus of palaelodids was pneumatic, as evidenced by the presence of pneumatic foramina, which also makes it unlikely that these birds were specialized divers. Irrespective of whether palaelodids were adapted to swimming or diving, they were unquestionably highly aquatic birds. Recognition of sister group relationship between flamingos and grebes shows that an aquatic way of living is present in two successive sister taxa of modern flamingos, Podicipedidae and Palaelodidae. It is thus more parsimonious to assume that an aquatic way of living is primitive for the clade (Podicipedidae + (Palaelodidae + Phoenicopteridae)) than to assume that it evolved independently in stem group representatives of Podicipedidae and Palaelodidae, particularly as extant flamingos still occasionally feed while swimming (del Hoyo, 1992). The shared similarities in the foot structure of grebes and palaelodids are best interpreted as plesiomorphic resemblances which were already present in the last common ancestor of Podicipedidae, Palaelodidae and Phoenicopteridae.