A new ornithurine from the Early Cretaceous of China sheds light on the evolution of early ecological and cranial diversity in birds

Diagnosis

The placement of Changzuiornis ahgmi within the clade Ornithurae is supported by seven unambiguously optimized synapomorphies (listed in the Phylogenetic Analysis section below). It diagnosed by a combination of morphologies not seen in other described ornithurines: The rostrum is elongate, comprising greater than 60% of the total skull length. Most of this elongate rostrum is made up of maxilla, a characteristic not present in the avian crown clade in which most of the rostrum and nearly the entire facial margin is made up by premaxilla. The only other avialans known to exhibit an elongate rostrum with the facial margin comprised primarily of maxilla are Xinghaiornis, Juehuaornis and Dingavis and derived Late Cretaceous ornithurines previously placed phylogenetically as among the closest outgroups to the avian crown clade (i.e., Ichthyornis, hesperornithine taxa). It differentiated from Xinghaiornis by its much smaller size, many tiny teeth on the lower jaw, U-shaped furcula, metacarpal III sub-equal to metacarpal II in distal extent, a carpometacarpus with both proximal and distal fusion, and a tarsometatarsus that is completely fused. The skull is about 15% longer in Changzuiornis and Juehuaornis compared to Dingavis (Table S1). In addition, the scapula is proportionally longer in Changzuiornis, Juehuaornis and Dingavis have a manual phalanx II:2 shorter than manual phalanx II:1 while in Changzuiornis this phalanx is longer (Table S1). The new species is differentiated from Hesperornithes and Ichthyornis by significantly smaller teeth with less recurved crowns, the presence of a distinct dorsal process or “forking” of the posterior dentary, and the presence of a pubic symphysis. It is additionally differentiated from Ichthyornis by robust and more abbreviate furcular rami, a narrower sternal margin of the coracoid, and a significantly more elongate scapular acromion. While if new data shows that Xinghaiornis, Juehuaornis and Dingavis form a clade that constitutes the same genus, the genus name Juehuaornis would have priority for Dingavis longimaxilla and Changzuiornis ahgmi.

Skull

The skull (Fig. 3) is preserved in right lateral view. The rostrum is elongate (48 mm), comprising ∼68% of the total skull length. It is longer than those reported in the only other long-rostrum Jehol taxa, the longipterygid enantironithines, Longipteryx (64%), Rapaxavis pani (65%), Longirostravis (60–64%) and Shanweiniao (62%), and comparable in proportions to those of extant woodcocks. The dorsal processes of the premaxillae are not fused to each other posteriorly (Fig. 4). While the right dorsal process is missing, the left one visibly extends posteriorly to contact the frontal. The posterior tip of the facial margin of the left premaxilla is missing while the articulating tip of the maxilla is visible and sharply tapered rostrally (Fig. 4). From the length of the exposed maxilla, the premaxilla comprised less than one half of the facial margin (rostrum), a condition also seen in enantiornithines with long rostra (e.g., Longirostravis and Shanweiniao; Hou et al., 2004; O’Connor et al., 2009).

The right nasal is laterally exposed with a descending process that lies adjacent to a small dorsal process of the maxilla (Fig. 4). The nasals appear not to have contacted along the dorsal midline but were likely separated by the frontal processes of the premaxillae as seen in Confuciusornis (Chiappe et al., 1999; Fig. 3) or underlay parts of these processes. However, the right nasal does not appear preserved in life position. The external nares are elongate and relatively narrow. A thin vertical sheet-like element visible in the narial region extends the length of the external nares and is interpreted as an internarial septum (Figs. 3 and 4). The preservation of an internarial septum in Changzuiornis is the first known occurrence in a Mesozoic bird. A similar sheet like element is visible below several the preserved scleral ossicles, consistent with at least a partial interorbital septum formed by the mesethmoid; Fig. 4). A mesethmoid is present in an array of basal avialans including Confucisuornis and ornithurines such as Yixianornis as well as Enantiornithes (e.g., Longipteryx haoyangensis and Schizooura lii; Zhou & Zhang, 2001; Clarke, Zhou & Zhang, 2006; Zhou, Zhou & O’Connor, 2012).

Parts of the rostral jugal and posterior-most quadratojugal are preserved. Just dorsal to the small rostral portion of the jugal (preserved at its contact with the maxilla), the poorly preserved remnants of the right lacrimal are visible (Fig. 3). The quadrate is poorly preserved but shows an arcuate posterior margin and a relatively elongate orbital process (Fig. 3). The frontoparietal suture appears to have been open (Fig. 3). The posterior portion of the skull is severely crushed. A small roughly t-shaped element preserved nearly the posterior margin of the orbit could represent a remnant of a postorbital; however, this cannot be ascertained with confidence and a postorbital is currently unknown in Ornithurae (Martin, 2011).

The mandibles are relatively straight and taper anteriorly. A predentary bone lies in front of left mandibular ramus (Fig. 3). Such a structure has also been reported in Hongshanornis. Ichthyornis, Hesperornis, and Parahesperornis and an array of other ornithurine birds (e.g., Zhou & Zhang, 2005). The dentary is forked posteriorly whereas the posterior dentary lacks a dorsal process in Ichthyornis and Hesperornithes (Fig. 4). At least three large mental foramina are visible in a shallow groove near the dorsal margin of the rostral dentary. The mandible is obscured by partly crushing posterior to the dentary.

Both premaxillae are completely edentulous, unlike those in the enantiornithines Longipteryx, Rapaxavis, Longirostravis and Shanweiniao (O’Connor & Chiappe, 2011). Approximately seven tiny, narrow, pointed and unserrated loose teeth are exposed closely between the maxilla and dentary (Fig. 3). One appears to be in situ, indicating an association with the dentary. No teeth seem preserved on the tip of dentary. Small neurovascular foramina are visible on the dorsal surface of the premaxilla consistent with the presence of a rhamphotheca (e.g., Chiappe et al., 1999).

Parts of the hyoid elements are preserved. Two recurved ceratobranchials are preserved with breakage separating their rostral-most tips from the rest of their preserved length. These rostral tips lie close to the right mandible (Fig. 3). A small, isolated element preserved below their tips may comprise a very small, ossified basihyal (Fig. 3: bh).

Vertebral column

The entire cervical series is preserved. The anterior cervicals are visibly heteroceolic while posterior-most cervicals do not appear heterocoelic, but are poorly exposed. The mid-series (fifth–eighth) cervical vertebrae are elongate. Elongate costal processes are present on the second through eighth vertebrae of the eleven or twelve vertebrae in the series (Fig. 2). The thoracic series is obscured by the right scapula. However, approximately ten thoracic vertebrae are discernable with large lateral fenestrae (Fig. 2) and amphiplatyan centra articulations. The neural spines are craniocaudally elongate. The sacrum is badly crushed, and no morphologies can be discerned. Only the approximate outlines of several free caudal vertebrae are visible while the pygostyle is clearly exposed. It is strongly mediolaterally compressed and short, approximately two anterior caudal vertebrate in length (Fig. 2).

Pectoral girdle and limb

The right coracoid is exposed in ventral view (Fig. 2). The left coracoid is obscured by the furcula and the right humerus. A large flange-like procoracoid process and an abbreviated lateral process are present (Fig. 5). The furcula is U-shaped and relatively thin; a furcular apophysis appears to be absent. The omal tips of the rami are severely crushed. The left scapula, lying under the furcula and several ribs, preserves an elongate and hook-shaped acromion process (Fig. 5). The posterior ends of both scapulae are missing. However, their preserved length approaches that of the humeri. The sternum is not visible. Approximately five large gastroliths (∼30 mm in diameter; Fig. 2) are visible in the abdominal region, between the coracoid and the posterior thoracic vertebrae. A small quantity of fine grit is intermixed with these larger stones. Unfortunately, a textured depression just posterior to the five preserved stone may indicate a larger quantity of such stones were present but lost during collection or early preparation.

Both humeri are exposed in anterior view (Figs. 2 and 5), and are just slightly shorter than the ulnae. The dorsally-directed deltopectoral crest is recurved and projected approximately equal to shaft width (Fig. 5). It extends distally for just greater than 1/3 of the total humeral length and then grades gently into the shaft. The head is weakly globose, and a small m. acrocoracohumeralis ligament scar is visible. On the distal humerus, the dorsal condyle is more elongate than the ventral. The radii are narrower than the slightly-bowed ulnae. Few morphologies of the proximal or distal radii and ulnae are visible. The radiale is larger than the ulnare, and the ulnare appears differentiated into distinct dorsal and ventral rami.

The right carpometacarpus is exposed in dorsal view and the left, in ventral view. The metacarpals are fused proximally and distally. Metacarpal III is significantly narrower than metacarpal II. It is straight and closely appressed to metacarpal II. On the left carpometacarpus, a small piciform process is visible (Fig. 6). Metacarpal I bears a very weakly-projected extensor process (Fig. 6); its anterior margin is nearly straight. Digit I:1 extends over the ½ length of metacarpus and bears a small claw. Manual digit II:1 is approximately the same length as digit II:2. The posterior margin of digit II:1 is strongly compressed dorsoventrally, and digit II:2 shows a distinct fossa that in crown clade Aves is the attachment site for the leading edge primary feather (Fig. 6; Hieronymus, 2015). An impression of a small digit II ungual is preserved on the left side. It is only weakly recurved. Phalanx III:1 is not visible.

Pelvic girdle and limb

Most of the partially-disarticulated pelvic girdle is obscured by crushing. The rod-like pubes (Figs. 2 and 7) have separated from the rest of the pelvic elements and preserve a relatively elongate symphysis. Their distal ends are not visibly expanded. The femur is notably shorter than the tarsometatarsus (Table 1) and its shaft is straight. The trochanteric crest is weakly projected proximally. The attachment of the capital ligament is indicated by a distinct notch on the lateral surface of the proximal femur (Fig. 7). The left tibia preserves a slightly proximally-projected anterior cnemial crest (Fig. 7). The morphology of the distal condyles is poorly preserved, but they appear fused to the tibia and are visibly separated by an intercondylar incisure (Fig. 7). The distal tarsals are fused to the metatarsals.

The tarsometatarsus is mediolaterally compressed. A prominent midline m. tibialis cranialis tubercle is visible on the proximodorsal left tarsometatarsus (Fig. 7). The j-shaped metatarsal I is well-exposed in articulation with a short hallux; pedal digit I:1 is longer than the small ungual, I:2. Pedal phalanges are preserved in association with both tarsometatarsi. They are narrow with deep flexor pits and only weakly recurved unguals. In all digits (Fig. 2), the unguals are shorter than their penultimate phalanges. The flexor tubercles on the unguals are proximally located, unlike the more distally located tubercles used to infer the presence of webbing in Gansus yumenensis (You et al., 2006). In digit II, the second phalanx slightly exceeds the first in length. Both non-terminal phalanges of this digit are the longest of the pedal digits. In digit three, the penultimate phalanx is slightly shorter than the more proximal phalanges. By contrast, in the fourth digit, the penultimate phalanx is slightly longer than the more proximal phalanges.

Feathers

Feather remains are poorly preserved (Fig. 2). Remnants of body contour feathers are associated with the posterior cranium and the cervical area as well as near wing and leg elements. In addition, several primary feathers are associated with both forelimbs. Traces of rachis and barbs are distinguishable only in certain distal regions of these feathers, which are otherwise preserved as grey-white impressions. One primary feather associated with left manual digit II:1 is approximately 106 mm in length. The rachis of this asymmetrically veined feather is discernable. The lengths of the rest of the preserved remiges are impossible to determine with confidence. Scanning electron microscopy (SEM) results show that wing and leg/tail feather samples contain melanosome molds that are highly aligned, closely spaced and elongate in shape (Fig. 2). Their aspect ratio (length:width ratio:2.29–4.99) is typical of eumelanosomes seen in black feathers (Vinther et al., 2009; Li et al., 2010; Clarke et al., 2010).

Maturity

The following anatomical features present the holotype specimen have been proposed to indicate an adult at death (Forster et al., 1998; Xu & Norell, 2004; Turner et al., 2007; Gao et al., 2012; Godefroit et al., 2013): (i) the texture of the bones is regular and well ossified bearing articular facets and muscular scars (e.g., proximal and distal ends of the coracoid, the humerus, the ulna, the pubis and the femur); (ii) the frontals are fused to each other; (iii) cervical ribs and cervical vertebrae co-ossified to enclose transverse foramina; (iv) distal carpals coosified with metacarpals II and III; (v) metatarsals II, III and IV fused throughout their length.

Conclusions

With its combination of a long and slender bill, proximally-projected cnemial crests, relative long tibiotarsus, and small pedal unguals, Changzuiornis represents a distinct departure from other known Jehol birds and contributes to our understanding of morphological and ecological diversity. That highly specialized species usually occur in low numbers (Julliard, Jiguet & Couvet, 2004; Şekercioğlu, Daily & Ehrlich, 2004) in a given fauna, maybe one explanation for comparative paucity of previously known diversity in rostral morphology in the Jehol Biota. At the same time, Mesozoic avialans maybe be subject to developmental constraints on rostral shape (Bhullar et al., 2012) and most diversity in feeding ecology is reflected in diversity in dentition. If so, inferences of feeding ecology primarily from rostral shape in Mesozoic birds should be approached with caution (Li et al., 2014). Changzuiornis constrains further when this shift in rostral development (Bhullar et al., 2012; Bhullar et al., 2015) may have occurred. So far it appears to be a Late Cretaceous phenomenon that arose close to the timing of crown clade origin and as such should be further investigated as a possible key innovation in their remarkable radiation. Increased kinesis, possibly as a form of prokinesis, is estimated to predate this developmental shift.