|Home | About | Journals | Submit | Contact Us | Français|
The midline pattern of sternal ossification characteristic of the Cretaceous enantiornithine birds is unique among the Ornithodira, the group containing birds, nonavian dinosaurs and pterosaurs. This has been suggested to indicate that Enantiornithes is not the sister group of Ornithuromorpha, the clade that includes living birds and their close relatives, which would imply rampant convergence in many nonsternal features between enantiornithines and ornithuromorphs. However, detailed comparisons reveal greater similarity between neornithine (i.e. crown group bird) and enantiornithine modes of sternal ossification than previously recognized. Furthermore, a new subadult enantiornithine specimen demonstrates that sternal ossification followed a more typically ornithodiran pattern in basal members of the clade. This new specimen, referable to the Pengornithidae, indicates that the unique ossification pattern observed in other juvenile enantiornithines is derived within Enantiornithes. A similar but clearly distinct pattern appears to have evolved in parallel in the ornithuromorph lineage. The atypical mode of sternal ossification in some derived enantiornithines should be regarded as an autapomorphic condition rather than an indication that enantiornithines are not close relatives of ornithuromorphs. Based on what is known about molecular mechanisms for morphogenesis and the possible selective advantages, the parallel shifts to midline ossification that took place in derived enantiornithines and living neognathous birds appear to have been related to the development of a large ventral keel, which is only present in ornithuromorphs and enantiornithines. Midline ossification can serve to medially reinforce the sternum at a relatively early ontogenetic stage, which would have been especially beneficial during the protracted development of the superprecocial Cretaceous enantiornithines.
Over the past several decades, the known fossil record of Mesozoic birds has grown explosively, largely due to specimens from the Lower Cretaceous Jehol Group of north-eastern China (Zhou et al., 2003). Typically, vertebrates are represented in the fossil record only by remnants of the boney skeleton; however, the Jehol Group, a stratigraphic unit which includes the Yixian Formation and the overlying Jiufotang Formation, continues to produce exceptional specimens that commonly preserve soft tissue and other rare traces, offering a unique glimpse into the biology of an Early Cretaceous fauna (Zhou et al., 2003). Recognition of fossilized melanosomes has resulted in the first accurate partial reconstructions of an extinct dinosaur’s plumage coloration (Zhang et al., 2010; Li et al., 2012), preserved stomach contents have provided direct evidence of diet (Zhou & Zhang, 2002; O’Connor et al., 2011c; Zheng et al., 2011, 2014a), and fossilized impressions of ovarian follicles have documented the absence of a right ovary in basal birds (Zheng et al., 2013). The Jehol is also unique in commonly preserving juvenile avian specimens (Chiappe et al., 2007), although these almost invariably belong to the extinct clade Enantiornithes. This lineage is characterized by protracted juvenile development, providing a unique opportunity to study the evo-devo of complex skeletal elements during early avian evolution (Zheng et al., 2012). By utilizing information from studies of ontogeny in birds and other vertebrates, it is possible to draw inferences about the underlying changes in developmental programmes that might have produced the major phenotypic changes observed during the early evolution of birds as they became increasingly modified for flight. This approach has made it possible to infer homology of the fibular crest between birds and nonavian theropods (Müller & Streicher, 1989) and address the controversy regarding the homologies in carpal skeletal anatomy between birds and other dinosaurs (Botelho et al., 2014).
Enantiornithes is the dominant clade of Mesozoic birds in terms of both species diversity and abundance of known specimens and represents the first major avian radiation (O’Connor et al., 2011b). Following confirmation of the monophyly of Enantiornithes through the discovery of nearly complete articulated specimens from Spain and China (Sanz et al., 1988; Zhou et al., 1992), the phylogenetic position of the clade as sister taxon to Ornithuromorpha, the clade that includes Neornithes (crown group birds) and some of their close fossil relatives, has been widely accepted and has received empirical support from numerous phylogenetic analyses (Chiappe, 2002; Clarke et al., 2006; Zhou et al., 2008a; O’Connor et al., 2011b; O’Connor & Zhou, 2013). (It should be noted that we consider strict node-based definitions for taxa premature at this stage in our understanding of Mesozoic bird phylogeny. For convenience and clarity, we use Ornithuromorpha to refer to all birds crownward of Enantiornithes, even though the two formal published definitions of Ornithuromorpha (Chiappe, 2001, 2002) are node based and technically apply to somewhat smaller clades.)
However, it has also been argued that enantiornithine and neornithine birds are much less closely related, belonging to lineages that arose separately from among the early archosauromorph reptiles of the Triassic, and evolved feathers and other flight-related features in parallel (Martin, 1983; Kurochkin, 2006). A recent study found evidence for a novel pattern of sternal ossification, previously unknown in Ornithodira, in several juvenile enantiornithine specimens from the Jehol Biota (Zheng et al., 2012). This apparent discrepancy in sternal development between enantiornithines and other ornithodirans, including living ornithuromorphs, provides intriguing if tentative support for the hypothesis that enantiornithines and ornithuromorphs are only distantly related.
Early ontogenetic stages are poorly documented for most clades among maniraptoran dinosaurs and nonenantiornithine basal birds. However, the available evidence suggests that the body of the sternum (i.e. the main plate-like portion of the bone, excluding the keel and trabeculae when these structures are present) ossifies in these taxa, when at all, as a bilateral, symmetrical pair of plates that may fuse along the midline in later ontogeny (Norell & Makovicky, 1997; Clark et al., 1999; Zheng et al., 2012). This pattern is strongly conserved and occurs even in living ratites. Considerable modifications are present in at least some derived birds (Parker, 1867; Livezey & Zusi, 2007b), although direct data on avian sternal ossification are somewhat sparse because postnatal patterns of ossification have only been studied in a handful of taxa (Apteryx, Gallus, Chrysolophus) (Parker, 1891; Hogg, 1980; Wen & Yang, 1991) and the sternum typically remains cartilaginous at the time of hatching. Regardless of the prevalence of novel modes of ossification among extant birds, however, bilateral ossification of the sternum clearly represents a maniraptoran norm.
In contrast, the juvenile Jehol enantiornithines studied by Zheng et al. (2012) reveal a unique pattern not seen in any other fossil or modern bird: the sternal body ossifies primarily as two median elements, one cranial and one caudal. Because the juvenile enantiornithine specimens available to Zheng et al. (2012) were limited in number and taxonomically indeterminate within Enantiornithes, this previous study was unable to determine whether their unique pattern of sternal ossification was present only in some derived enantiornithines or throughout the clade. The former alternative would indicate that a major reorganization of sternal development took place during enantiornithine evolution. The latter could point to a developmental shift coinciding with the origin of Enantiornithes, but would also be compatible with a scenario in which enantiornithines inherited their distinctive sternal ossification pattern from an ancestral group not closely related to ornithuromorphs.
A wide phylogenetic separation between enantiornithines and ornithuromorphs would imply that an enormous amount of parallel evolution must have taken place as both lineages became independently adapted to flight, a hypothesis defended by some (Martin, 1983; Kurochkin, 2006) but seemingly highly unparsimonious. On the other hand, the morphological similarity of the adult sternum in the two ornithothoracine clades implies that, if a major and relatively sudden shift in sternal development indeed took place either within Enantiornithes or at the very base of the clade, it had astonishingly little impact on the adult phenotype. Because both alternative explanations for the unusual ossification pattern seen in juvenile enantiornithines seemed intuitively implausible, Zheng et al. (2012) had no clear basis for preferring either based on the data available at the time of their study.
We describe a new subadult basal enantiornithine specimen with a sternum comprising unfused left and right plates. The new specimen indicates that the unusual ossification pattern documented by Zheng et al. (2012) evolved within Enantiornithes and that a major, relatively sudden developmental shift must have taken place within this lineage. We summarize evidence from the embryology of extant birds that indicates a similar transformation also occurred in parallel at some point in ornithuromorph evolution. In an effort to partially reconstruct the evo-devo of the avian sternum, we explore the nature of these developmental shifts in two respects: first in terms of their possible selective advantages, based on morphological information from the sternum across all ontogenetic stages, and second with regard to the possible molecular mechanisms involved.
Institutional Abbreviations: IVPP, Institute of Vertebrate Paleontology and Paleoanthropology, Beijing, China; LH, collection from Las Hoyas, housed in the Universidad de Paleontologia, Universidad Autonoma de Madrid, Spain; STM, Shandong Tianyu Museum of Nature, Pingyi, China.
Aves Linnaeus 1758
Ornithothoraces Chiappe 1995
Enantiornithes Walker 1981
Pengornithidae Wang et al., 2014
‘Protopteryx-horizon’ of the Huajiying Formation, Hebei Province, north-eastern China; Jiufotang Formation, Lower Cretaceous Jehol Group, Liaoning Province, north-eastern China.
Medium to large enantiornithines with the following unique combination of shared features: numerous small teeth (> 10) in the both the upper and lower jaws; short robust scapula with large, hooked acromion process; flat cranial surface of proximal humerus; and fibula nearly reaching distal articular surface of tibiotarsus. Further characterized by simple sternum with single pair of trabeculae and V-shaped caudal margin, pubes articulating along distal fifth of their length, femur long (approximately 95% as long as the tibia), metatarsal V present, and hallux long (modified from Wang et al., 2014b).
STM 29–15, a slab (dorsal view; Fig. 1) and counterslab (ventral view; Fig. S1) with an articulated partial skeleton of a subadult individual lacking the skull and cervical vertebrae, and preserving carbonized stains of the remiges. The immaturity of the specimen is inferred from the lack of fusion in compound bones such as the sternum, carpometacarpus, and tibiotarsus.
The specimen is referable to Enantiornithes based on the presence of a synsacrum formed of seven vertebrae, a straight scapula, a humerus with a weakly saddle-shaped proximal profile, and a minor metacarpal that projects distally further than the major metacarpal. The presence of large, recurved pedal unguals and the combination of large body size and a low degree of skeletal fusion further support this identification. Because of imperfect preservation and subadult ontogenetic status, the specimen cannot be positively identified at the species level. STM29-15 is clearly referable to the Pengornithidae based on sternal morphology alone (Hu et al., 2014). Other preserved pengornithid features include the following: short scapula with robust, hooked acromion process, pubes with extensive distal articulation, ischium more than 1/2 length of distally expanded pubis, femur nearly as long as tibia, elongate fibula and long hallux (Zhou et al., 2008b; Hu et al., 2014; Wang et al., 2014b). Differences in pedal morphology and inferred adult body size indicate that pengornithid sp. IVPP V18632 (smaller adult size) and STM29-15 are probably not conspecific. As a large subadult, STM29-15 is perhaps more likely to be a specimen of P. houi, the largest known Early Cretaceous enantiornithine. However, comparison between STM29-15 and the holotype of P. houi is hindered by differences in preservation and ontogeny. Furthermore, the bones of STM29-15 are preserved in a way that renders them unsuitable for histological analysis, preventing a more precise determination of the ontogenetic age of the specimen. Because the specimen is incomplete and subadult, we consider STM29-15 to be a taxonomically indeterminate member of the clade Pengornithidae.
The specimen preserves a nearly complete sternum composed of two medially articulating plates exposed in dorsal view in the main slab (Fig. 2a). The plates are entirely unfused medially, and in fact are separated by a narrow gap running the entire length of the sternum. In much younger and smaller previously described juvenile enantiornithines (e.g. IVPP V15564, STM34-1,-2,-7,-9; Fig. 2c), the only visible sternal ossifications are located on the midline, and no midline articulation or suture is present at any known stage in the ossification of the sternum. In STM29-15, the cranio- and caudomedial margins of each plate are convex, forming cranial and caudal clefts on the midline of the sternum; the cranial midline cleft is much more distinct than the caudal one. Similar clefts between the cranial and caudal ends of two sternal plates that are fused medially along most of their length are observed in subadult confuciusornithiforms and jeholornithiforms and are inferred to close later in ontogeny. Most commonly only a cranial cleft is present, indicating that the caudal cleft closes before the cranial cleft. A cranial midline cleft is present in one Spanish enantiornithine (Eoalulavis hoyasi Sanz et al., 1996 LH13500) with an ‘autapomorphic’ sternum lacking ossified coracoidal facets (see Discussion) (Sanz et al., 1996). In STM29-15, the straight cranial margins of the two plates converge at an angle of approximately 120°, comparable to that estimated for Eopengornis; the cranial margin of the sternum is not preserved in pengornithid IVPP V18632 (Fig. 2b). Craniolateral processes are absent, as in other known pengornithids. The lateral trabeculae are straight, strap-like and distally unexpanded, as in pengornithid IVPP V18632. The left and right trabeculae are truncated at the same level, just cranial to the caudal margin of the sternal body, but both show clear signs that the distal ends have been broken away; the trabeculae extend beyond the caudal margin in pengornithid IVPP V18632. In the xiphial region, the median trabeculae form a wide V (inner angle 70°; 72° in IVPP V18632); intermediate trabeculae are absent as in other pengornithids and Protopteryx. The morphology preserved in STM29-15 is nearly identical to that preserved in pengornithid IVPP V18632 (Hu et al., 2014) except that the sternum is fully fused in the latter specimen, even though its femora are the same length (35 mm) as those of STM29-15. The sternum is also fully fused in the holotype of Eopengornis (STM24-1; femoral length 27 mm), which histological analysis has confirmed to be a subadult (Wang et al., 2014b). The holotype of P. houi is much larger than any other specimen of this genus, but unfortunately the sternum is not preserved. The only other known enantiornithine with comparable sternal morphology is the basal taxon Protopteryx fengningensis Zhang et al., 2000, but the sternum of this genus differs from that of Pengornis in that the lateral trabeculae are shorter, a small lateral process is present (as in the basal ornithuromorphs Archaeorhynchus spathula Zhou & Zhang, 2006 and FRDC-05-CM-02; Fig. 2e), and the distal half of the xiphial region is constricted (Zhang & Zhou, 2000). The sternum of Protopteryx is only preserved in a single specimen (IVPP V11844), in which it is fully fused.
The dorsal vertebrae are disarticulated, revealing large neural openings that exceed the articular surface in area. The lateral surfaces of the vertebrae are deeply excavated by grooves, as in other enantiornithines (Chiappe & Walker, 2002). The synsacrum appears to be formed by seven partially fused vertebrae and bears a spinous crest along the dorsal midline. The transverse processes are all laterally directed except in the distalmost two sacral vertebrae, in which they are caudolaterally oriented. A pygostyle is preserved overlying the pubes in the main slab; it appears to be proportionately short, measuring approximately half the length of the tarsometatarsus (whereas the pygostyle is at least subequal in length to the tarsometatarsus in longipterygid enantiornithines) (O’Connor et al., 2011a) and resembles that of P. houi in having a blunt distal margin (Zhou et al., 2008a). Two sternal ribs are preserved in articulation with the right costal margin of the sternum. Several complete dorsal ribs are also preserved; they are long and gently curved throughout their length. At least seven pairs of gastralia must have been present.
The scapulae are short and robust; the acromion process is robust and hooked, a feature unique to pengornithids among enantiornithines. The coracoids are poorly preserved. The neck of the coracoid is proportionately long, with lateral and medial margins that are proximally concave but distally become convex as in Pengornis. The dorsal surface of the coracoid is weakly concave. As in other pengornithids, the proximal margin of the humerus lacks the deep dorsal concavity seen in other enantiornithines. The weakly developed deltopectoral crest is less wide than the humeral shaft and tapers distally, resembling that of the subadult pengornithid IVPP V18632. The humeral shaft is robust with an expanded distal end, as in Pengornis. The radius and ulna are poorly preserved. Both hands are preserved, but morphological details are clearer on the left (Fig. 1). The semilunate carpal and alular metacarpal appear unfused, indicating subadult ontogenetic status, but the degree of proximal fusion between the major and minor metacarpals is unclear. The major metacaral is straight; the minor metacarpal is bowed, enclosing a small intermetacarpal space, and extends distally further than the major metacarpal, as in other enantiornithines (Chiappe & Walker, 2002). The alular meta-carpal is poorly preserved, but the alular digit is short, ending approximately at the level of the distal end of the major metacarpal. The alular ungual is larger than that of the major digit, as in other pengornithids. The first phalanx of the major digit lacks the dorsoventral compression and extreme caudal expansion seen in ornithuromorphs, is longer and much more robust than the penultimate phalanx and bears a small caudal projection proximal to the distal articular surface (also present in IVPP V18632). The minor digit preserves one phalanx, which tapers to a blunt distal end.
The entire pelvic girdle is preserved in dorsal view in the main slab (Fig. 1), and the pelvic bones are unfused. The ilia are poorly preserved, but their pre-acetabular wings are clearly longer than their post-acetabular wings. The pubes are distally expanded and articulated but remain unfused, as in other pengornithids. The ischium is slender and measures 56% the length of the pubis (60% in IVPP V18632). It has a small but distinct proximally located dorsal process, distal to which the dorsal margin is straight. The ventral margin is concave, and the distal end is tapered.
The femur is long, measuring approximately 95% the length of the tibia as in other pengornithids. The tibia is poorly preserved, but is not fused to the distal tarsals; the degree of fusion between the astragalus and calcaneum is unclear. The fibula is long, nearly reaching the level of the distal condyles of the tibiotarsus, as in Pengornithidae (Wang et al., 2014b). The distal tarsals and metatarsals are unfused; it is unclear how many free tarsals were present, but a single large distal tarsal may cap metatarsals III and IV on the left, a condition similar to that reported in a referred specimen of Parabohaiornis martini (Wang et al., 2014a). The metatarsals lie in a single dorsoventral plane throughout their lengths as in other enantiornithines and basal birds. Metatarsal IV is slightly reduced in width relative to metatarsal III, which in turn is thinner than metatarsal II. A metatarsal V, like that observed in Eopengornis, is not preserved. Metatarsal I is long, exceeding 1/3 the length of metatarsal II, as in other pengornithids (Wang et al., 2014b), and ends proximal to the metatarsal II trochlea. The distal surfaces of the trochleae of metatarsals III and IV are not clearly preserved. The hallux is long; the first phalanx of this digit is the longest in the foot, and the ungual appears to be larger than those of the other digits. The first phalanx of the second digit is approximately 30% shorter than the penultimate phalanx. The second digit is the most robust in the pes, and the ungual appears to be comparable in size to that of the first digit. The third and fourth pedal digits are not clearly preserved, but their unguals are more delicate and less recurved than those of digits I–II. The ungual of digit IV is the smallest in the foot.
The only visible feathers are primary remiges, and even they are preserved only as faint impressions (Fig. 1). The two wings overlap and the feathers are not preserved in situ, being displaced in such a way that the visible rachises are not parallel but oriented at various angles to each other. The attachment sites of the feathers and the positions of their distal ends are not clear, preventing estimates of length.
In contrast to the juvenile enantiornithines studied by Zheng et al. (2012), STM29-15 reveals that some enantiornithines possessed the typical, ancestral bilateral pattern of sternal ossification in which a midline articulation is present between the paired plates until late in ontogeny, comparable to the pattern observed in nonavian maniraptorans (e.g. Citipati osmolskae Clark et al., 2001, Bambiraptor feinbergi Burnham et al., 2000, Microraptor gui Xu et al., 2003), basal birds (Jeholornis prima Zhou & Zhang, 2002a, Confuciusornis sanctus Hou et al., 1995) and extant flightless palaeognaths (e.g. Rhea americana Linnaeus, 1758) (Figs 1 and Fig 2). The basal position of the Pengornithidae (O’Connor, 2009; Wang et al., 2014b) is supported morphologically by the presence of a metatarsal V and an unreduced fibula. Bilateral ossification of the sternum appears to represent the maniraptoran primitive condition, which was retained in pengornithids but lost in at least some derived enantiornithines (Fig. 3). The morphology of STM29-15 therefore refutes suggestions based on sternal development that enantiornithines did not share the dinosaurian ancestry of other birds, by showing that basal enantiornithines retained a typical dinosaurian pattern of sternal development that was only abandoned in the later evolutionary history of the clade. Accordingly, the atypical ossification pattern of the sternum observed in most juvenile enantiornithines from the Jehol (Zheng et al., 2012) is now most parsimoniously interpreted as an autapomorphy of a subclade of Enantiornithes. This re-evaluation has significant implications for the evolutionary history of sternal ossification in enantiornithines and other avian groups. Below we explore these implications in the light of available information on the development of the sternum in extant and fossil birds.
….The size [of the sternum], both in length and breadth, its outgrowths, its histological consistence, the number of its ossific centres, and, above all, the number of its unfinished clefts–all these things conspire to make the study of this part of the Bird’s skeleton difficult in the highest degree. (Parker, 1868)
Over a hundred years later, scientific understanding of sternal development in extant birds remains surprisingly limited. In the literature, the number of ossification centres contributing to the neornithine sternum is reported to vary widely, but the centres are generally described as occurring in bilateral pairs (Fell, 1939; Livezey & Zusi, 2007a). Nevertheless, there exist reports of unpaired ossification centres in some taxa (Hogg, 1980; Wen & Yang, 1991), and midline ossification of the body of the sternum may be the norm in extant neognaths. Attempts to understand the ossification sequence of the sternum are hindered by the predominance of embryological studies in the literature, as the sternum typically only begins to ossify after hatching (Starck, 1993). Furthermore, chondrification and ossification sequences appear to be decoupled in most studied taxa, in the sense that centres of chondrification do not necessarily correspond in number or position to the centres of ossification that appear subsequently (Maxwell, 2008b).
Embryological studies of both mice and chicks indicate that the sternum is part of the appendicular skeleton, in that the sternal precursor cells reside within the somatic region of the lateral plate mesoderm (LPM) in the limb bud and migrate to the ventral midline (Bickley & Logan, 2014); thus, medial migration of precursor cells is an intrinsic part of sternal formation in amniotes. In Aves, the sternum universally begins to develop as a bilateral pair of condensations that are mediolaterally well separated when they begin to chondrify (Fell, 1939; Klima, 1962; Nakane & Tsudzuki, 1999; Maxwell, 2008b; Bickley & Logan, 2014). The chondrified plates begin to fuse at their cranial ends (Fig. 4), and growth and fusion of the cartilaginous sternum proceed in a primarily cranial to caudal direction (Fell, 1939), as in mouse embryos (Bickley & Logan, 2014).
In the vast majority of living birds, including all members of Neognathae (the clade containing all extant birds other than the flightless ‘ratites’ and tinamous, which belong to the clade Palaeognathae), a conspicuous feature of the sternum is a large ventral carina or keel. In Melopsittacus undulatus Shaw, 1805 and other neognaths whose sternal development has been examined, the mesenchyme tissue that will eventually chondrify to form the main body of the carina becomes apparent just around the time the plates begin to fuse. The fused chondrified sternal plates and the precartilaginous cells of the carina rapidly consolidate cranially into a single integrated structure in most birds (Fell, 1939). The sternal plates form the dorsal portion of the carina (Fell, 1939; Klima, 1962). The body of the keel begins chondrification from a single distinct centre ventral to the sternal body (Parker, 1867; Klima, 1962). Ventral outgrowth of the sternum (formation of the carina) follows medial fusion between the cartilaginous plates – both fusion of the plates and development of the carina proceed from cranial to caudal, but the former runs slightly ahead of the latter (Fell, 1939). These observations suggest that the formation of the keel must always lag behind fusion of the sternal midline.
In nearly all birds in which sternal development has been investigated, most or all of the sternum has formed by the time of hatching but remains cartilaginous (Parker, 1867; Starck, 1993; Maxwell, 2008b). In the flightless palaeognaths (the ‘ratites’), sternal ossification follows the same bilateral pattern as sternal chondrification, so that a medial gap or suture between the two incompletely ossified sternal plates persists through much of post-natal ontogeny (Parker, 1891; Maxwell & Larsson, 2009). This resembles the inferred mode of sternal ossification in basal Cretaceous birds (including STM29-15) and nonavian maniraptorans, taxa whose sterna are without an ossified ventral keel. Notably, ‘ratites’ are the only extant Neornithes that completely lack a keel (Kaiser, 2010), their flat ‘raftlike’ sternum earning the group its name (from Latin ratis meaning raft).
Studies of ossification centre formation during postnatal ontogeny in extant neognaths suggest that the sternal body and carina ossify from a single centre located on the midline, in striking contrast to the ‘ratite’ condition. Recent studies have focused primarily on phasianids (Galliformes). Unfortunately, this clade is unique in that ribs (each represented by an ossification centre) are incorporated into the body of the sternum, although they are not true sternal anlagen (Hogg, 1980; Wen & Yang, 1991; Starck, 1993; Nakane & Tsudzuki, 1999; Maxwell, 2008a). The six ossifications (three pairs) that later become incorporated into the sternum ossify much earlier than the body of this element, consistent with their identification as part of the rib complex (Starck, 1993). In phasianids (e.g. Gallus gallus Linnaeus, 1758; Coturnix coturnix Linnaeus, 1758; Chrysolophus amherstiae Leadbeater, 1829), the sternal body begins to form as two mediolaterally separated chondrifications, but ossifies together with the carina from a single centre located near the cranial end of the sternal midline where the sternal body joins the carina. This ossification appears around the time of hatching (stage 40 of Hamburger and Hamilton, 1951), after the cartilaginous sternum is fully formed, and no additional ossifications appear during juvenile (post-natal) ontogeny (Hogg, 1980). Ossification sequences are reportedly conserved within modern birds (Starck, 1993), and this general rule might be expected to apply to the sternum.
Although data are limited, it appears that a median carinal centre is also the first ossification centre to appear in the Melopsittacus (Psittaciformes) (Fell, 1939) and is present at the time of hatching in some individuals of the common eider, Somateria mollissima Linnaeus, 1758 (Anseriformes) (Maxwell, 2008b). Although all features of the sternum may be bilaterally paired in a developmental sense, being derived from cell populations that originate on the left and right sides of the embryo and never cross the midline (Fell, 1939), the carina and corpus ossify from a single visible median centre (Melopsittacus undulatus Shaw, 1805, Gallus, Coturnix, Chrysolophus). Parker (1867) described the sternal carina and corpus as ossifying from distinct centres (Corvus monedula Linnaeus, 1758; Turnix suscitator Gmelin, 1789; Gallus, Phalacrocorax sp.), but this is clearly incorrect in the case of Gallus, the only neognath discussed by Parker for which modern data are available. Parker mistakenly identified the rib-derived centres that form craniolateral processes as paired ossification centres within the corpus. Furthermore, in all juvenile neognaths that Parker illustrated in his compilation, he depicted the median centre as making by far the greatest contribution to the sternum by area, forming the keel and most of the corpus.
Despite the limited information in the literature, several conclusions can be reached regarding sternal development in extant birds: first, despite retaining a plesiomorphic chondrification pattern, neognaths apparently begin ossification of the sternum from a cranially located median centre that also forms the ventral keel, so that chondrification and ossification can be described as decoupled; second, ossification of the sternum in Neornithes universally proceeds from cranial to caudal; and third, ‘ratites’ are the only modern birds in which a midline suture has been documented at any point during post-natal development of the sternum. The first two points describe a pattern nearly opposite, in terms of craniocaudal orientation, to that observed in the juvenile derived enantiornithines studied by Zheng et al. (2012). These specimens indicated that the sternal body initially began to ossify from a centre positioned on the caudal midline, from which projected the ventral keel or carina. A second ossification centre, situated cranial to the first, subsequently appeared on the midline. Notably, the keel in Early Cretaceous enantiornithines is caudally restricted, when it can be observed at all, and we suggest that this is because the keel grew only from the caudal ossification centre. Based on the apparent association with the keel, the caudal centre is presumably analogous to the single median one seen in neognaths, whereas the cranial one is presumably a novel feature of some enantiornithines. Notably, in STM29-15, the separation between the sternal plates is more pronounced cranially than caudally (Fig. 2c), and specimens of Confuciusornis (Fig. 2b) in which the sternum is fused except for a rostral cleft are common, suggesting that caudal-to-cranial ossification of the sternum also characterizes pengornithids and basal birds.
The fact that extant flightless palaeognaths are the only extant birds in which bilateral ossification of the sternum has been documented suggests a connection between bilateral ossification and the absence of a sternal keel. In all neognaths in which sternal development has been studied, the keel and sternal body ossify entirely or primarily from a single cranially located midline centre. At no stage in the process of sternal ossification is the ossifying carina observed as an element separate from the corpus (except possibly in Corvus, in which the carina is illustrated as a separate ossification and not contributing to the sternal body; Parker, 1867: plate XV.18), and no carinal ossification ever coexists with a midline suture between paired boney plates. Even with regard to the cartilaginous stage of sternal development that precedes ossification, there is no documented case in living birds in which a cartilaginous carina and midline suture coexist in the same craniocaudal position on the sternum (as noted above, closure of the suture and growth of the keel both take place from cranial to caudal, with the former process running ahead of the latter).
In extinct taxa, as discussed in more detail below, the mode of sternal ossification often cannot be determined directly from fossil evidence. However, no known fossil avian or nonavian theropod specimen combines a sternal keel with a midline suture, or combines an unkeeled sternum with preserved evidence of midline ossification. Unfortunately, sternal ossification has not been studied in tinamous, which are volant living palaeognaths that possess a fully developed sternal keel. If they prove to resemble ratites in having a bilateral pattern of sternal ossification, the hypothesis that the presence of a keel is invariably linked to midline ossification will be falsified. Presently available evidence, however, supports this linkage and suggests that any fossil avian sternum with a well-developed keel can be presumed to have ossified primarily from one or more midline centres.
The inferred link between midline ossification and the presence of a carina may have either a developmental or a mechanical basis, or (most likely) may arise from a combination of these factors. Developmentally, the intimate coupling between the formation of the carina and fusion of the median part of the sternal corpus seen in extant carinate birds suggests that midline ossification of the corpus may be necessary in order for the carina, a prominent midline structure, to form correctly and/or achieve proper integration with the rest of the sternum (Fell, 1939). Mechanically, a carina anchored to a strip of persistent cartilage between two progressively ossifying plates, rather than to bone, would be less adequately braced against forces exerted by contracting flight muscles.
The mechanical and developmental considerations both suggest that the appearance of the carina in enantiornithines, and separately in ornithuromorphs, is unlikely to have preceded the shift from bilateral to midline ossification. They do not weigh against the alternative possibility that the appearance of midline ossification might have preceded that of the carina, which would imply the existence of a stage at which the sternum was unkeeled as in basal taxa but nevertheless ossified from the midline. However, no fossil bird preserves evidence of this condition, and it is also possible that midline ossification and the sternal keel appeared simultaneously as part of a single developmental innovation.
Presumably because of selection for various aspects of flight performance, a remarkable range of complex sternal shapes rapidly evolved within Aves (Fig. 3); the sternum is arguably the most variable and diagnostic element among Early Cretaceous ornithothoracines (You et al., 2010; O’Connor, 2012), although it is notably absent in the basal birds Archaeopteryx lithographica Meyer, 1861 and Sapeornis chaoyangensis Zhou & Zhang, 2002b (Zheng et al., 2014b). In nonavian maniraptorans and basal birds, the sternum appears to have been a developmentally conservative structure, forming bilaterally from a single pair of left and right plates that articulate along the midline in oviraptorosaurs, dromaeosaurids, and the basal bird Confuciusornis (Zheng et al., 2012). In these taxa, the fully ossified sternum is plate-like and approximately polygonal, although relatively short lateral processes may be present, and lacks a well-developed keel. A midline suture or even gap between the two halves of the sternum, sometimes widening into clefts at the rostral and caudal ends, tends to persist into late ontogeny. Although a carina has been reported in Confuciusornis (Confuciusornithiformes + Ornithothoraces = Carinatae) (Chiappe et al., 1999), this feature is limited to a low ridge that is only present in the most mature known individuals, in which the midline suture is fully closed (e.g. STM13-231, STM13-30). We do not consider such ridges to be fully homeologous to the far more prominent carina that is present in most ornithothoracine birds. Rather we infer this feature evolved in the confuciusornithiform lineage in parallel to true ‘carinate’ birds; this relatively modest structure may have functioned as a rudimentary keel, acted as a site of muscle attachment in its own right, anchored a larger cartilaginous carina as previously suggested (Chiappe et al., 1999), or simply helped to reinforce the sternum, the apparent function of the similar midline thickening observed in adults of some living ratites (Parker, 1891; Elżanowski, 1988).
In the enantiornithine and ornithuromorph lineages, the primitive type of sternum became modified in two important ways: through the addition of new peripheral centres of ossification, often associated with trabeculae, and through the replacement of the paired centres forming the two halves of the sternal body with a midline centre associated with a keel. In derived enantiornithines, as noted above, a second, more cranially placed midline centre also appeared. Avian evolution appears to have been characterized by a general trend towards an increased number of sternal ossification centres, a phenomenon that occurred in parallel in Enantiornithes (Zheng et al., 2012) and Ornithuromorpha (Parker, 1867). At least some Neornithes, enantiornithines and even jeholornithiforms have trabeculae and craniolateral processes that develop from distinct peripheral ossification centres, indicating that increasing the number of centres has been a widespread mechanism for morphological elaboration of the sternum (Zhou & Zhang, 2002; Zheng et al., 2012). In both e-nantiornithines and ornithuromorphs, this was associated with a reduction in the amount of bone formed by each centre, and also made possible an increase in the number, size and morphological complexity of the sternal trabeculae. In addition to the keel, the appearance of the trabeculae expanded the area available for the attachment of muscles, and this adaptive shift in sternal architecture was underpinned by changes in the pattern of ossification.
Midline ossification of the sternal body is the second major innovation seen in large clades within Ornithothoraces and appears to be linked to the presence of a sternal keel as noted previously. Among enantiornithines, the existence of a keel cannot be confirmed in any pengornithid, because all known sterna are preserved in dorsal view (Eopengornis STM24-1, pengornithid IVPP V18632, STM29-15). A keeled sternum is present in Protopteryx, which like pengornithids is positioned close to the base of the enantiornithine radiation (until the recognition of the Pengornithidae, Protopteryx was fairly consistently resolved as the basal-most enantiornithine and is still resolved as such by some recent analyses)(O’Connor & Zhou, 2013). In this taxon, the keel is small but distinct and caudally located, as in other Early Cretaceous enantiornithines; although the keel appears to be fairly low, its ventral depth cannot be accurately determined or compared between most taxa due to preservational limitations. Although Protopteryx does not preserve direct evidence of the mode of sternal ossification, the presence of a caudally restricted keel suggests that this taxon possessed the derived midline ossification pattern. This in turn suggests that Protopteryx is more derived than the Pengornithidae, as resolved by some analyses (Wang et al., 2014b) and inferred from the presence of a suite of seemingly advanced characters. The clear presence of bilateral ossification in STM29-15 [and deep rostral clefts present in pengornithids STM29-11 (J. K. O’Connor, unpublished data) and IVPP V18687 (Hu et al., 2015)] and the lack of evidence for a well-developed keel in any pengornithid suggest that this clade of basal enantiornithines was characterized by a relatively flat sternum with minimal, if any, development of the keel. In taxa with midline ossification, the keel begins to form very early in the ossification process (Zheng et al., 2012); were a keel or similar structure present in pengornithids, we suggest it could only have formed later in ontogeny after at least the caudal half of the medial sternal suture had closed (e.g. at the ontogenetic stage at which STM29-11 and IVPP V18687 died; could not have been present in STM29-15, in which the sternal plates are entirely unfused).
Midline ossification centres can be directly observed in the juvenile enantiornithines described by Zheng et al. (2012), but unfortunately, these juvenile specimens are taxonomically indeterminate, and enantiornithine interrelationships are poorly resolved in any case (O’Connor & Zhou, 2013). These uncertainties make it difficult to determine how widespread midline ossification was in enantiornithines, and whether midline ossification appeared simultaneously with the sternal keel or whether one of these derived features (presumably midline ossification, for reasons given above) preceded the other. However, as most enantiornithines are known to possess a carina on at least the caudal part of the sternum (extends to cranial margin in the Late Cretaceous Neuquenornis volans Chiappe & Calvo, 1994), we suggest that midline ossification must characterize a rather inclusive subclade of enantiornithines.
The Early Cretaceous Spanish enantiornithine Eoalulavis provides some additional, though difficult to interpret, data regarding the evolution of sternal ossification in this group. The holotype and only known specimen of E. hoyasi has an autapomorphic sternum that somewhat resembles the juvenile phenotype in derived enantiornithines, being a narrow midline element without ossified articular surfaces for the coracoids (Sanz et al., 1996). However, a cranial cleft is present, suggesting that at least the cranial part of the sternum ossified bilaterally as in basal enantiornithines and most other ornithodirans. The sternum of Eoalulavis presumably ossified in a caudal-to-cranial direction as in other enantiornithines, so the persistent cleft may simply indicate that fusion between the left and right sternal plates was never completed. This could imply that the only known specimen of Eoalulavis had not reached somatic maturity at the time of death or that the sternum in this taxon was paedomorphic. Unfortunately, it cannot be determined if compound bones such as the carpometacarpus or tarsometatarsus were already fused, but the articular surfaces of all elements are fully developed, suggesting the specimen was at least fairly mature. A well-developed keel is not present in this taxon. The fact that the only ossified portion of the sternum in the holotype of Eoalulavis is essentially a midline strip of bone despite the seemingly bilateral mode of ossification supports the inference that reinforcement of the sternal midline was functionally advantageous to enantiornithines (see below).
Phylogenetic analyses have not recovered a particularly close relationship between pengornithids and Eoalulavis (Hu et al., 2014), which may suggest that incorporating developmental data might radically alter currently accepted views of enantiornithine interrelationships. Given the current lack of clarity with regard to enantiornithine phylogeny, it seems possible that Eoalulavis is a relatively basal enantiornithine lying outside the clade characterized by midline ossification and a well-developed keel, however large this clade might turn out to be. However, it also seems possible that Eoalulavis is a member of that clade that underwent reversal to a flat, unkeeled, bilaterally ossifying sternum, although the impetus for the reversal would not be obvious given that Eoalulavis was almost certainly volant based on wing proportions. What emerges clearly from the enantiornithine record is that midline ossification and a sternal carina appeared in a derived subset of enantiornithines. The absence of these features in basal enantiornithines and the differences of detail between their expression in enantiornithines and ornithuromorphs both imply that they evolved independently in the Enantiornithes and Ornithuromorpha.
Unfortunately, it is difficult to determine when midline sternal ossification appeared in the ornithuromorph lineage, and even the question of whether midline or bilateral ossification represents the neornithine primitive condition is presently open. Neornithes, as discussed above, have morphologically complex sterna whose varying patterns of ossification appear to fall into two major types. In neognaths, the presence of a carinal centre provides sternal ossification with a predominant midline component. By contrast, ‘ratite’ sterna develop in a bilateral, ‘reptilian’ manner, lacking a carinal centre. However, sternal development has not been studied in the only extant flying palaeognaths, the tinamous (Tinamidae), which resemble both neognaths and most fossil ornithuromorphs in having an ossified keel. Similarly, a deep sternal keel is present in extinct basal volant palaeognaths such as the Palaeocene lithornithids (Houde, 1988), but no direct fossil evidence reveals the mode of ossification in these taxa.
The lack of information from volant palaeognaths, and particularly from extant tinamous, is an important source of uncertainty in attempts to reconstruct the primitive neornithine mode of sternal ossification. If the sternum turns out to ossify bilaterally in tinamous, despite the presence of a keel, bilateral ossification will be readily interpretable as the primitive condition for palaeognaths and possibly for Neornithes as a whole. Furthermore, bilateral ossification in tinamous would be consistent with the possibility that this mode of sternal development was inherited by palaeognaths from non-neornithine ornithuromorphs and ultimately from nonavian maniraptorans, with midline ossification potentially representing a neognath synapomorphy. However, if midline ossification proves to be present in the keeled tinamou sternum, the more plausible inference will be that this pattern appeared at some point in ornithuromorph evolution prior to the origin of Neornithes, only to be secondarily lost in large flightless lineages. The circumstantial evidence suggesting that bilateral ossification is associated with the absence of a keel strongly supports this second interpretation, but documentation of midline ossification in tinamous would represent important corroborating evidence.
Data from the fossil record of non-neornithine ornithuromorphs may shed some light on the neornithine primitive condition, even though direct evidence bearing on patterns of sternal ossification is virtually absent from the ornithuromorph fossil record. There are no Early Cretaceous ornithuromorphs in which the early ontogenetic stages of the sternum are known, making the presence or absence of midline ossification centres at this grade of avian evolution impossible to observe. However, no known Mesozoic ornithuromorph specimen at any ontogenetic stage possesses a medial sternal suture, and almost all have keels. In fact, the sterna of most Early Cretaceous ornithuromorphs (e.g. Yixianornis grabaui Zhou & Zhang, 2001; Fig. 2i) are superficially comparable to those of some living birds (type 5 sternum of Heimerdinger & Ames, 1967) in having not only a large ventrally extensive carina but also well-developed craniolateral processes and trabeculae that often caudally enclose a pair of sternal fenestrae.
The fully keeled basal ornithuromorph Archaeorhynchus is known exclusively from subadult specimens (e.g. IVPP V17075, V17091; Fig. 2g), which have sterna resembling those of some immature neognaths (e.g. Corvus). Although these individuals are less ontogenetically advanced in terms of their skeletal morphology (for example, in retaining an unfused pygostyle) than the pengornithid STM29-15, their sterna lack a midline suture and bear a large keel (Zhou et al., 2013). Unlike in more derived ornithuromorphs known from adult specimens, the caudal half of the sternal body is proportionately smaller in the Archaeorhynchus subadults, limiting the overall surface area of the boney sternum (Fig. 2g). If sternal ossification proceeded from cranial to caudal as in living birds, it is possible the sternum might have undergone considerable morphological change through further ossification of the caudal margin and development of the carina if growth had continued. The alternative is that this is the adult morphology of Archaeorhynchus, indicating a trend towards expansion of the caudal half of the ossified sternum in more derived ornithuromorphs.
The presence of a prominent carina, combined with the absence of a midline suture in all known specimens, implies that even the most basal known Mesozoic ornithuromorphs may have been like extant neognaths in having a single cranial midline centre from which the sternal body and keel both ossified. Midline ossification and the keel would then have appeared at an early stage in ornithuromorph evolution as part of a single developmental innovation that was lost multiple times within the Palaeognathae but retained in almost all other ornithuromorphs. Although enormously reduced in many flightless birds (e.g. the kakapo Strigops habroptilus Gray, 1845; the Titicaca flightless grebe Rollandia microptera Gould, 1868; the weka Gallirallus australis Sparrman, 1786), the keel is retained in all extant neognaths (Kaiser, 2010). This structure is notably absent in the Hesperornithiformes, a flightless clade of Late Cretaceous aquatic birds nested deeply within Ornithuromorpha (O’Connor & Zhou, 2013; Bell & Chiappe, 2015) and large graviportal galloanseriforms from the Palaeogene and Neogene (e.g. Dromornis, Gastornis) (Murray & Vickers-Rich, 2004; Mayr, 2009). The mode of sternal ossification in these groups is unknown, but it seems likely that they resembled flightless palaeognaths in having secondarily reverted to a keelless, bilaterally ossifying sternum. However, this scenario depends strongly on the inference that the presence and absence of a keel are linked to midline and bilateral sternal ossification, respectively. It must be emphasized that this inference requires further testing, particularly on the basis of data on sternal ossification in tinamous. Until such testing has taken place, it will not be possible to reject the alternative possibility that bilateral ossification was simply retained in all non-neognath ornithuromorphs, including palaeognaths. Nevertheless, we believe that current evidence favours the idea that midline ossification and a sternal keel evolved early in ornithuromorph history, although independently of the appearance of equivalent features in derived enantiornithines, and were simply lost in both ratites and hesperornithiforms.
In functional terms, the keel in extant ‘carinate’ birds makes up a large portion of the sternum (the keel is greater in area than the corpus in some taxa, including the tinamous), greatly increasing the surface area available for muscle attachment, and may be more important than the sternal body itself (Bickley & Logan, 2014). Flying ability increases with the size of the keel (Gill, 2007). Mechanically, the carina acts as a hypertrophied muscle attachment ridge or intermuscular line, increasing the surface area available for the origins of the pectoralis and supracoracoideus and also augmenting the moment arm of the former by moving its line of action further from the centre of rotation of the shoulder joint. The keel was of comparable extent in many extinct ornithuromorphs, and even in derived enantiornithines was large enough to suggest it played a critical role in anchoring the flight muscles.
The functional implications of midline ossification of the sternum in Mesozoic birds depend partly on whether a carina was present and would have been limited to immature ontogenetic stages before growth was complete. Regardless of a sternal keel, midline ossification would have stiffened and reinforced the midline of the plate-like sternum at an early ontogenetic stage, enabling it to better withstand forces exerted by the developing flight muscles. Conversely, late fusion of the paired boney sternal plates, as evident in STM29-15 and other basal birds (Jeholornis, Confuciusornis), would have left the pectoral girdle with a median line of weakness and potentially detrimental flexibility until growth was almost complete.
In birds with a sternal carina, midline ossification would have also provided this structure with a boney rather than cartilaginous base during early ontogenetic growth. By facilitating an increase in the masses and moment arms of the flight muscles, the presence of a carina would have compounded the structural disadvantages of a gap or suture situated on the sternal midline in a developing bird, if indeed it was even developmentally possible for a boney carina to appear in a sternum that ossified bilaterally. Midline ossification may have appeared prior to a well-developed carina in the enantiornithine and ornithuromorph lineages, and the most likely initial advantage of the new mode of ossification would then have been reinforcement of the sternal midline in juveniles. Fusion of the sternal plates not only increases the function of the sternum as a brace for the pectoral girdle, but also facilitates the absorption of forces directed through the coracoid through decreased flexion of the midline (Turner et al., 2012). The transition to a keeled sternum, as an adaptation for enlarging the flight muscles and enhancing their moment arms, would have been facilitated by the establishment early in ontogeny of a median strip of bone that could act as a rigid base for the carina. However, it is also possible that midline ossification and the carina appeared simultaneously, or even that the carina preceded midline ossification in enantiornithine and/or ornithuromorph evolution. Under the latter scenario, the carina would initially have been anchored on a weak sternal midline in juvenile individuals, presumably resulting in a strong selection pressure favouring a shift to midline ossification.
Regardless of a keel, the advantage provided to juveniles by midline ossification would have been particularly important to enantiornithines because of their unique developmental strategy. This additional ontogenetic factor may have helped to drive the evolution of the derived ossification pattern in this clade, whether or not this shift in sternal development was immediately accompanied by the appearance of a keel. Uniquely among Mesozoic birds, enantiornithines are inferred to have been superprecocial and to have become capable of flight soon after hatching (Chinsamy & Elżanowski, 2001; Zhou & Zhang, 2004; Chiappe et al., 2007). Locomotor activity tends to reduce growth rates, so it is unsurprising that the juvenile period of ontogeny is inferred to have been protracted in enantiornithines relative to other clades. The difference in maturation rate is reflected in the fact that numerous juvenile enantiornithine specimens have been collected (revealing the intermediate stages of sternogenesis), whereas no juveniles are known for any other Cretaceous avian clade. Slow enantiornithine growth is confirmed more directly by histology, which indicates that it took multiple years for an enantiornithine chick to reach sexual maturity (Chinsamy et al., 1995; Chinsamy & Elżanowski, 2001; O’Connor et al., 2014). However, no juvenile pengornithids are recognized (STM29-15 is the youngest known specimen) and their developmental strategy is unknown, so the general enantiornithine rule of precociality and slow growth may not apply to this basal clade.
The rapid onset of volant activity implies that the enantiornithine sternum would have been subjected to the large muscle-induced stresses associated with flight early in post-natal ontogeny, at least in derived taxa (Fig. 4). Thus, strong selection pressure would presumably have been directed at the juvenile phenotype, because an individual whose flight apparatus functioned well during this extended period of ontogenetic growth would also have been more likely to survive to adulthood.
The stresses of powered flight during prolonged early ontogeny may have created a selection pressure that led to the evolution of the novel ossification pattern in derived enantiornithines. Midline ossification would have stiffened the sternal midline, increasing the ability of the sternum to withstand the forces generated by the flight muscles in the juvenile phenotype, and sternal reinforcement could either have been accompanied by the appearance of a keel or facilitated the subsequent addition of this structure. If pengornithids did not share the precocial and prolonged ontogeny of derived enantiornithines, the selection pressure favouring midline ossification would have increased as precociality became established. Alternatively, it is possible that precociality and the resulting selection pressure existed in pengornithids, but that the mutation(s) underpinning midline ossification did not appear at the basal grade of enantiornithine evolution occupied by this group.
In this connection, it is worth noting that essentially nothing is known regarding the pattern of development in basal ornithuromorphs. It is possible they were also highly precocial, so that midline ossification would have been advantageous to developing juveniles to the same degree as in enantiornithines. If basal ornithuromorphs were indeed precocial, this ontogenetic mode must have quickly been abandoned, as derived rapid growth strategies are known to have evolved early in ornithuromorph history (O’Connor et al., in press).
Biological patterning is invariably initiated not just by a single molecular signal, but rather by an entire molecular network, which determines the topological configuration of localized morphogenetic zones (Li et al, 2015). That is, differences in cell behaviour across spatial domains are generally known to depend on molecular gradients that arise in the organizing centres via changes in regulatory elements of signalling molecules, forming new localized centres for morphogenetic processes such as cell proliferation, rearrangement, apoptosis and possibly chondrification (Wu et al. 2004; Li et al, 2013; Chang et al, 2004; Garzon-Alvarado et al, 2009). The duration of proliferation, arrangement of cells, cellular hypertrophy and pattern of tissue differentiation collectively determine the final shape of a bone. As is inferred for the sternum, the number of proliferation zones involved in forming the beak varies within Neornithes (Wu et al, 2004); differences in the positions, angles of extension and growth durations of these individual zones are responsible for the great diversity of beak morphologies seen in the clade (Wu et al, 2004). Localized proliferative zone activity involving BMP4 has been shown to be the driving force in the molecular network controlling development of the beak, as BMP4 regulates the range, rate and duration of growth at each locus (Wu et al, 2006). Although the beak is composed of membrane bone (also called ‘dermal’ bone and derived from intramembranous ossification), ossification of at least some endochondral bones is also driven by BMP signalling (Wang et al, 2013), and this presumably applies to the avian sternum; as in the beak, the diversity of sternal shapes observed among Neornithes could be produced through modulations in the location and duration of BMP4 positive zones, which can be determined by epigenetic mechanisms such as mutations of the cis-regulatory elements of BMP, BMP upstream molecules, BMP inhibitors or physical forces (discussed below).
For example, if the lateral trabeculae maintained cell proliferation longer before ossification (through localized enhanced BMP4 activity), the result would be lengthening of the trabeculae or formation of larger distal expansions, depending on the shape of the proliferation zone in each trabecula. Thus, by regulating BMP signal activities within the sternum in different ways, diverse sternal morphologies could easily be achieved without altering the fundamentals of the ancestral developmental model. Such modulations of pre-existing ossification patterns of sternal development can be inferred to have formed lateral processes and small craniolateral processes (jeholornithiforms, enantiornithines, ornithuromorphs), increased the size and complexity of the distal expansions of the lateral trabeculae (common in both enantiornithines and ornithuromorphs) and united the distal ends of the intermediate trabeculae with those of the median trabeculae to enclose large fenestrae (ornithuromorphs). Thus, many differences of detail in sternal morphology between basal ornithuromorphs and more primitive birds can be plausibly explained by the spatial/temporal dynamics of BMP signalling.
Besides BMP, the shape and size of a developing cartilaginous structure can be modulated by other molecular signalling pathways, such as FGF, Wnt and PTHrP/Ihh, due to their impact on chondrocyte proliferation. FGF and Wnt signalling generally inhibit chondrocyte proliferation, whereas BMP signalling probably acts earlier to induce the dermal condensation that defines the initial shape of the cartilage (Karsenty et al, 2009) (Fig. 4). Molecular studies on the mouse and chicken indicate that endochondral ossification, the process that produces the boney sternum, is partially regulated by the interaction and diffusion of two secreted molecules: parathyroid hormone-related peptide (PTHrP) and Indian hedgehog (Ihh) (Kronenberg, 2003). PTHrP secreted from perichondral cells at the ends of cartilage moulds, and from early proliferative chondrocytes, inhibits the production of Ihh and maintains the proliferation of undifferentiated chondrocytes. Chondrocytes distant from the PTHrP-producing cells start to differentiate and synthesize Ihh. Ihh from these prehypertrophic/hypertrophic chondrocytes feeds back to the PTHrP secreting cells to stimulate PTHrP production. Meanwhile, Ihh also stimulates neighbouring perichondral cells to become osteoblasts and start the process of ossification (Kronenberg, 2003; Vortkamp et al, 1996). A mathematical model based on this reaction-diffusion relationship between PTHrP and Ihh can predict the locations of secondary ossification centres during long bone development (Garzón-Alvarado et al, 2009). In this model, the shape and size of the cartilage mould will affect the spatial distributions of PTHrP and Ihh, and hence the placement and number of the ossification centres that form within the developing bone. The addition of a large carina to the cartilaginous sternum, a significant change in embryonic morphology that evidently took place independently in the enantiornithine and ornithuromorph lineages, may thus have altered the pattern of sternal ossification in each case simply by changing the distributions of PTHrP and Ihh.
Changes in the spatial/temporal expression patterns of genes such as FGF, Wnt, BMP, PTHrP and Ihh through mutations in the cis-regulatory elements in the promoter and enhancer regions (Chan et al., 2010; Chuong et al, 2013) provide one possible explanation for the unique sternal ossification pattern described by Zheng et al. (2012). However, simultaneous mutations in the cis-regulatory regions of multiple genes have a low probability of occurring, whereas mutations of a single gene rarely, if ever, directly produce evolutionary novelties (Müller 1990) such as the change in ossification pattern observed in Enantiornithes. Rather, the sudden appearance of a novel morphological feature is more typically the product of epigenetic mechanisms in which mechanical forces (such as those generated by body movement, cell contraction or extracellular matrix production) cause relatively minor quantitative shifts in such traits as body size and limb proportions, which in turn may have disproportionate effects if they cross some morphological threshold that has profound epigenetic importance. For example, the size of a limb bud in early ontogeny may influence the number of chondrogenic centres that can form within the developing limb (Müller, 1990), ultimately having a major impact on the adult morphology. Mechanical forces have been shown to have major effects on endochondral ossification; for example, mathematical modelling demonstrates that the combined effects of shear and hydrostatic stresses on the ribs can explain differential ossification patterns observed in the human sternum (Wong and Carter, 1988). A more recent study found that mechanical stimulation of the developing chick tibiotarsus by the associated musculature was concentrated near an eventual centre of ossification (Nowlan et al., 2008). Furthermore, mechanical forces have been shown to promote osteoblast differentiation, the initial step in ossification, in in vitro cell cultures (Kelly & Jacobs, 2010). Epigenetic mechanisms of this kind are often critical in producing rapid character transformations and morphological novelties.
In the developing chick, both the pectoralis muscle and the sternum become profoundly malformed if an embryo is experimentally paralysed (Hall & Herring, 1990). This implies that the pectoralis develops normally only when allowed to contract and that the mechanical stimulation provided by the pectoralis and perhaps other muscles is a critical factor in morphogenesis of the sternum. During her embryonic studies of Melopsittacus, Fell (1939) also noted that the formation of the keel was linked to the pectoralis muscle. As superprecocial birds, enantiornithines were probably capable of flight as young juveniles or even hatchlings, well before sternal ossification was complete. The flight musculature was probably at least as active in ovo as is the case in extant chickens, and the pectoralis would presumably have continued to undergo frequent bursts of activity and exert strong forces on the sternum after an individual had hatched. If the level of pectoralis activity at early ontogenetic stages was greater in derived enantiornithines than in pengornithids, it is compelling to hypothesize that the change in the mechanical environment of the sternum may have stimulated the growth of the carina, changing the shape of the cartilaginous sternum and thus producing shifts in PTHrP/Ihh secretion which induced midline rather than bilateral ossification (Fig. 4). If early ornithuromorphs shared the precocial developmental pattern of enantiornithines, a possibility noted above, increased muscular activity at early ontogenetic stages may also have driven the initial appearance of the keel in the ornithuromorph lineage. Alternatively, and particularly if ornithuromorphs were more altricial than enantiornithines even in the period following the initial divergence of the two groups, shifts in PTHrP/Ihh secretion may have produced the ornithuromorph keel.
The pengornithid enantiornithine STM29-15 indicates that basal enantiornithines possessed the typical ornithodiran pattern of bilateral sternal ossification, implying that the unusual midline ossification pattern observed in previously described juvenile Jehol enantiornithines is an autapomorphy of a more exclusive clade. This refutes suggestions based on purported differences in sternal ossification between Ornithuromorpha and Enantiornithes that these clades are not closely related. A form of midline ossification has also been documented in living neognaths and may have a much deeper history within Ornithuromorpha. In both enantiornithines and ornithuromorphs, a shift to midline ossification appears to have been linked to the appearance of a sternal carina that provided a large attachment surface for flight musculature. Midline ossification of the sternal body provided the carina with a strong, stiff base in early ontogeny and may have either preceded the evolution of the carina itself or appeared together with the carina as part of a novel developmental module. We further suggest that the superprecocial developmental mode of the volant enantiornithines could have resulted in both in ovo and in vivo mechanical forces that would have strongly affected the development of the sternum, driving the shift to midline sternal ossification that occurred within Enantiornithes and producing a sternum that was well equipped to meet the demands of flight at a relatively early ontogenetic stage. If midline ossification indeed first appeared as an epigenetic response to muscle forces acting on the sternum, this episode in avian evolution represents a striking example of how epigenetic mechanisms can reinforce and amplify an incipient functional change. The palaeontology findings here point out potential new evo-devo mechanisms for future investigation.
This research was supported by the National Basic Research Program of China (973 Program, 2012 CB821906), the National Natural Science Foundation of China (41172020, 41172016, 41372014) and the Chinese Academy of Sciences. CMC is supported by U.S. National Institute of Health (RO1 AR 47364). We thank D. Marjanović for his review of two earlier versions of this manuscript.
Additional Supporting Information may be found in the online version of this article:
Table S1 Comparative measurements of published pengornithids.
Figure S1 Counterslab of pengornithid indet. STM29-15.