First Complete Wing of a Stem Group Sphenisciform from the Paleocene of New Zealand Sheds Light on the Evolution of the Penguin Flipper

: We describe a partial skeleton of a stem group penguin from the Waipara Greensand in New Zealand, which is tentatively assigned to Muriwaimanu tuatahi . The fossil includes the first complete wing of a Paleocene penguin and informs on previously unknown features of the mandible and tibiotarsus of small-sized Sphenisciformes from the Waipara Greensand. The wing is distinguished by important features from that of all geologically younger Sphenisciformes and documents an early stage in the evolution of wing-propelled diving in penguins. In particular, the wing of the new fossil exhibits a well-developed alular phalanx and the distal phalanges are not flattened. Because the wing phalanges resemble those of volant birds, we consider it likely that the wing feathers remained differentiated into functional categories and were not short and scale-like as they are in extant penguins. Even though the flippers of geologically younger penguins may favor survival in extremely cold climates, they are likely to have been shaped by hydrodynamic demands. Possible selective drivers include a diminished importance of the hindlimbs in subaquatic propulsion, new foraging strategies (the caudal end of the mandible of the new fossil distinctly differs from that of extant penguins), or increased predation by marine mammals.


Introduction
Penguins (Sphenisciformes) are flightless, wing-propelled diving birds, in which the forelimbs attained a flipper-like shape. In the extant species, the wing forms an essentially flat and stiff unit, which exhibits a strongly restricted motility of the carpal joint [1]. The distal wing muscles of crown group Sphenisciformes are greatly reduced [2] and most bony attachment structures are poorly developed. The wing feathers are, furthermore, greatly shortened and not differentiated into functional categories, such as primaries, secondaries, and coverts. However, even though penguins have a comparatively extensive fossil record that dates back to the earliest Cenozoic Era [3,4], little is known about the evolutionary transitions that occurred in the evolution of the penguin wing.

Material and Methods
The fossil is deposited in the Canterbury Museum, Christchurch, New Zealand (CM). Use of open nomenclature follows the recommendations of Bengtson [15].

Taxonomic Remarks
CM 2018.124.4 is tentatively assigned to Muriwaimanu tuatahi, but we note some differences to other M. tuatahi-like bones in the collection of the Canterbury Museum. The carpometacarpus of the new fossil, for example, is craniocaudally somewhat narrower and more gracile than that of CM zfa 34 ( Figure 3) and the femur to ulna ratio is 1.20 in the new fossil, whereas it is 1.34 in CM 2009.99.1. Whether these differences indicate intraspecific or sexual variation or whether they suggest distinctness on a species level will have to be determined once the small-sized penguins from the Waipara Greensand are subjected to a revision. Abbreviations: cmc, carpometacarpus; ext, processus extensorius (carpometacarpus); hum, right humerus; min, phalanx digiti minoris; ocr, os carpi radiale; ocr, os carpi radiale; ocu, os carpi ulnare; pda, phalanx digiti alulae; pdm, phalanx distalis digiti majoris; ppm, phalanx proximalis digiti majoris; rad, radius; ses, sesamoid bones associated with the tendon of musculus scapulotriceps; uln, ulna. Scale bars equal 50 mm.

Description and Comparisons
The proportions of the mandible indicate a long and narrow beak, which is also characteristic of other early Cenozoic stem group Sphenisciformes [6,11]. The mandibular symphysis exposes the ventral surface and the mandibular rami are closely appressed in their rostral sections. The caudal portions of both mandibular rami are broken and displaced. The caudal section of the left ramus mandibulae is situated next to the rostral section of the right ramus mandibulae (at first sight, Figure  2A may give the impression of a continuous connection, but closer examinations clearly shows that the ventral surface of the rostral portion of the mandible is exposed, whereas the caudal portion is seen in dorsal view; the rostral and caudal mandibular sections are not connected by bone). The caudal section of the right ramus could not be unambiguously identified, but may be a bone fragment situated on top of the right radius ( Figure 1A). For the first time, the caudal end of the mandible is preserved in a Paleocene penguin, with the dorsal surface of the caudal end of the left ramus mandibulae being well exposed. Unlike in the crown group ( Figure 4D) and Eocene stem group Sphenisciformes [19,20], the mandible of CM 2018.124.4 lacks a retroarticular process and the processus medialis is proportionally larger and more pointed and is medially, rather than caudomedially, directed (amongst extant penguin, the caudomedial orientation of the processus medialis is less pronounced in Eudyptula and Spheniscus than in Aptenodytes). With regard to the above features and overall morphology, the caudal end of the mandible of CM 2018.124.4 actually more closely resembles the caudal end of the mandible of extant Fregatidae (frigatebirds; Figure 2C) than that of geologically younger Sphenisciformes. Muriwaimanu tuatahi). Abbreviations: cmc, carpometacarpus; hum, humerus; min, phalanx digiti minoris; ocr, os carpi radiale; ocu, os carpi ulnare; pda, phalanx digiti alulae; pdm, phalanx distalis digiti majoris; ppm, phalanx proximalis digiti majoris; rad, radius; uln, ulna. The scale bar equals 50 mm.
In size and morphology, the humerus, ulna, radius, and carpometacarpus resemble the corresponding bones of the Muriwaimanu tuatahi-type material [5]. As noted above, however, the carpometacarpus is more gracile than that of other specimens assigned to M. tuatahi ( Figure 3B). No sesamoid bones associated with the tendon of musculus scapulotriceps could be identified in the fossil. These sesamoids are located near the distal end of the humerus and are a derived characteristic of extant penguins ( Figure 4D). Their first occurrence in the fossil record is in the late Eocene Icadyptes salasi [12]; sesamoid bones associated with the tendon of musculus scapulotriceps are also present in the late Oligocene Kairuku grebneffi [13].
The phalanx digiti alulae, which was for the first time preserved in an early Cenozoic penguin, is detached from the carpometacarpus and is well developed. In crown group Sphenisciformes, this phalanx is fused with the carpometacarpus. However, based on the presence of an articular facet on the alular metacarpal of the carpometacarpus, it was hypothesized that the late Eocene Icadyptes salasi had an unfused phalanx digiti alulae [12]. In the late Oligocene Kairuku grebneffi, by contrast, the phalanx already seems to have been fused with the carpometacarpus [13].
The phalanx proximalis digiti majoris exposes its dorsal surface. Unlike in crown group Sphenisciformes, the bone is not dorsoventrally flattened, and it is furthermore craniocaudally wider and less elongated than in extant penguins. The phalanx proximalis digiti majoris of CM 2018.124.4 is also craniocaudally wider than that of the recently described Kupoupou stilwelli from the late early to middle Paleocene of Chatham Island. Whereas the phalanx proximalis digiti majoris has an essentially flat surface in crown group Sphenisciformes, it exhibits a shallow fossa dorsalis and a pila cranialis in the fossil. The facies articularis metacarpalis is a shallow cotyla. The phalanx distalis digiti majoris is likewise barely flattened and has a similar shape to the corresponding phalanx of volant birds.
The phalanx digiti minoris is not greatly elongated as it is in extant penguins, in which it surpasses the phalanx proximalis digiti majoris in length. In CM 2018.124.4 this phalanx measures only 50% of the length of the phalanx proximalis digiti majoris. Even though this phalanx is also proportionally shorter in Eocene and Oligocene stem group Sphenisciformes, it already measures 60% of the length of the phalanx proximalis digiti majoris in Kairuku grebneffi [13] and 65% in Icadyptes salasi [12]. Unfortunately, the bone surface of the phalanx is damaged in the fossil, so that further details were not clearly visible. Unlike in Kupoupou stilwelli, however, the caudal margin of the bone does not form a proximally directed process (see [9] (figure 6)).
Osteological details of the carpal bones were not well visible in the fossil. The os carpi ulnare (Figure 3) is proportionally smaller and less flattened than in crown group Sphenisciformes, in which it forms a large, plate-like structure of subtriangular shape. The os carpi radiale is situated between the distal end of the ulna and the os carpi ulnare (Figure 3) and no details of its morphology are discernible.
Elements of the pelvic girdle preserved in the specimen include the synsacrum and the caudal portion of the left side of the pelvis. A synsacrum is also part of the M. tuatahi-type material [5] and in CM 2018.124.4 it consists of 12 or 13 fused synsacral vertebrae. As far as comparisons are possible, the pelvis closely resembles that of Waimanu manneringi [5]. As in crown group Sphenisciformes, the ilium and synsacrum are not co-ossified.
Whereas the femur is preserved in the Muriwaimanu tuatahi fossils described in the original description of the species [5], the tibiotarsus has not yet been described. In its proportions and the shapes of the proximal and distal ends, the tibiotarsus of CM 2018.124.4 resembles that of Waimanu manneringi and Sequiwaimanu rosieae. However, a notable feature of the new fossil is the fact that the distal end of the fibula is co-ossified with the tibiotarsus, which does not occur in extant penguins and has not been reported from any other fossil sphenisciform taxon.
As in other Paleocene stem group Sphenisciformes, no patella is preserved in CM 2018.124.4. This is particularly notable, because the right femur and tibiotarsus are preserved in close association and appear to have been embedded in an articulated state.

Discussion
The new Paleocene stem group penguin CM 2018.124.4 (cf. Muriwaimanu tuatahi) has a more plesiomorphic wing morphology than all geologically younger Sphenisciformes, of which the corresponding wing bones are known. In particular, the alular phalanx is well developed and the phalanges of the major digit are not flattened so that the bones of the hand section of CM 2018.124.4 are strikingly similar to those of volant birds. Even though the greatly abbreviated ulna and radius and the thickened bone walls of the major limb elements clearly show that the fossil species already lost its flight capabilities, it must have used its wings in a different way than extant penguins.
Penguins are the only avian group in which a phalanx for the alular digit is absent. Ontogenetic evidence suggests that this phalanx is co-ossified with the carpometacarpus, and from the presence of a facies articularis alularis of the carpometacarpus it was concluded that some stem group Sphenisciformes had a free alular phalanx [12]. However, the actual development of this phalanx in early Cenozoic penguins remained unknown and in juvenile crown group Sphenisciformes it is a vestigial, splint-like rudiment (see [12] (figure 15)). CM 2018.124.4 not only provides the first direct evidence for the presence of a freely movable phalanx digiti alulae in stem group Sphenisciformes, but also reveals that this phalanx is of similar length and proportions to that of volant birds with a functional alula.
In flying birds, the alula functions as a vortex generator at high angles of attack, especially during start and landing [21,22]. In all extant birds except penguins, the alular phalanx bears welldeveloped feathers and usually there are four alular remiges [23]. The well-developed alular phalanx of CM 2018.124.4 suggests the presence of such remiges since the bone is unlikely to have maintained its full length if the alula feathers were greatly reduced. In turn, this indicates that the wing feathers of at least the hand section of the fossil species were differentiated into functional groups and were distinguished from the short, scale-like feathers of crown group Sphenisciformes. The close resemblance of the phalanges of the major wing digit (phalanx proximalis digiti majoris and phalanx distalis digiti majoris) to the corresponding wing elements of volant birds also suggests that the primaries of the fossil species were better developed than in extant penguins.
Based on the morphology of the articular surfaces of the distal end of the ulna, it was hypothesized that the wings of Paleocene stem group Sphenisciformes could be folded to a greater degree than those of extant penguins and did not form stiff flippers [6]. This assumption appears to be supported by the fact that the wrist of CM 2018.124.4 is preserved in a deflected position, with the hand section forming an angle of almost 90° with the ulna (Figures 1A,3). Such a high degree of flexion is not possible in crown group Sphenisciformes, in which the longitudinal axis of the hand section is much more in line with that of the humerus and ulna ( Figure 4D). Certainly, however, some caution is appropriate in the interpretation of the relative position of the wing bones of the fossil, which were embedded in an advanced state of decay.
In any case, we conclude that Muriwaimanu-like stem group Sphenisciformes did not yet have flippers and differed from extant Sphenisciformes in wing beat kinematics. As in auks, the hand section of Paleocene stem group Sphenisciformes was probably kept in a flexed position during the propulsive stroke, whereas it is extended in penguins [1,24]. In the proportions of the ulna, radius, and the bones of the hand section, the wing of CM 2018.124.4 most closely resembles the extinct Great Auk, Alca impennis (Alcidae, Figure 4C). In this flightless species, the ulna and radius are greatly abbreviated, but the distal wing elements do not differ from those of volant auks in their proportions. With regard to its humerus morphology, Muriwaimanu is distinct from A. impennis and other Alcidae and exhibits the characteristic derived features found in other penguins, including a globose humerus head and a very large pneumotricipital fossa.
Muriwaimanu tuatahi is one of the oldest and, phylogenetically, most early-diverging stem group representatives of the Sphenisciformes known so far [3,5,6]. The new fossil is about 23-24 million years older than previously known penguin wings with sufficiently complete distal sections, that is, fossils of Anthropornis and Icadyptes from the late Eocene (36 Ma) of Antarctica and Peru, respectively [10,11]. In these taxa, the hand section already shows the flipper-like morphology characterizing the crown group representatives, which suggests that the distinctive paddle-shaped wing of penguins evolved between the late Paleocene and the late Eocene.
It is not unexpected to find a plesiomorphic wing morphology in one of the earliest penguins. However, it is less trivial to scrutinize the reasons for the presence of highly advanced flippers in all geologically younger species. There were, and still are, other evolutionary successful groups of wingpropelled diving birds with forelimb morphologies similar to that of Muriwaimanu, such as auks and the extinct penguin-like Plotopteridae, which occurred in the North Pacific from the late Eocene to the early Miocene [4]. Accordingly, there must have been strong selective forces in the evolution of penguins, which shaped their forelimb morphology and led to the extinction of species with less specialized wings.
The small, scale-like feathers of extant penguins have an insulating function [25]. Hence, it is conceivable that a reduction of the distal forelimb muscles as well as the formation of a uniformly feathered integument constitute adaptations for foraging in cold waters, because it is energetically favorable to concentrate muscle mass near the trunk in order to minimize heat loss through the vasculature supplying these muscles. However, with more specialized, and more "modern-type" wings being present in the ca. 35 million-years-old Icadyptes and Anthropornis, the evolution of a penguin flipper seems to have preceded the formation of the Antarctic Circumpolar Current, which induced cooling of the sea surface temperatures and is dated at about 33-34 million years ago [26]. Even though the flipper-like penguin wing may well constitute an exaptation for survival in cold climates, we concur with previous observations [27] that climatic cooling of the Antarctic Region was probably not a major driver in its formation.
The differences in the shape and orientation of the caudal end of the mandible as well as the absence of a retroarticular process indicate that Paleocene stem group Sphenisciformes had a different feeding ecology than more crownward penguins. This is also suggested by the fact that CM 2018.124.4 and other early Cenozoic stem group Sphenisciformes featured extremely long and pointed beaks [6,11]. Possibly, there existed a functional correlation between the length of the beak and the shape of the caudal end of the mandible (which is unknown in other long-beaked species), and it was assumed that the diet of early Cenozoic Sphenisciformes included larger prey items than those taken by crown group penguins [3]. A transition in the feeding ecology of penguins may have been related to the formation of the Antarctic Circumpolar Current, which was critical for the evolution of notothenioid fishes and krill [28,29], both of which today constitute a major part of the diet of Antarctic and Subantarctic penguins. However, long, spear-like beaks occur in penguins with flipper-like wings, such as Icadyptes [11], so that differences in foraging strategies probably did not account for the transformation of the forelimbs into stiff paddles.
In extant penguins, the feet only have a steering function in subaquatic locomotion, but the very long fibula of CM 2018.124.4, whose distal end is co-ossified with the tibiotarsus, suggests that the legs of Paleocene stem group Sphenisciformes played a greater role in aquatic locomotion. Fusion of the distal end of the fibula and the tibiotarsus is a characteristic feature of birds with a strongly developed leg musculature and, in addition to raptorial birds, it occurs in foot-propelled divers, such as the Phalacrocoracidae. That Paleocene stem group Sphenisciformes may have used their feet for underwater locomotion was also deduced by earlier authors from the morphology of the tarsometatarsus of these birds [3,6]. The formation of a flipper-like forelimb may have been correlated with a diminished importance of the hindlimbs in subaquatic propulsion and it is well possible that the stem species of penguins used both its wings and hindlimbs for underwater locomotion, as do extant Sulidae [30].
Finally, we note that the first occurrences of more specialized (more "modern-type") penguin flippers coincided with those of the earliest stem group cetaceans in the Southern Hemisphere, which were reported from the middle Eocene of Antarctica and Peru [31,32]. These toothed species may well have preyed on penguins and it is conceivable that dispersal of stem group cetaceans into the Subantarctic and Antarctic regions resulted in an increased selective pressure on the escape strategies of penguins. However, if predation pressure by cetaceans played a role in the evolution of the penguin flipper, it is difficult to understand why plotopterids and auks, with less specialized, Muriwaimanu-like wings, coexisted with cetaceans in the North Pacific [33]. Earlier analyses did not identify a correlation between the diversity of stem group cetaceans and stem group penguins [33], but the fossil record of both groups is continuously improving for the Subantarctic and Antarctic regions and future studies may provide new insights into competitive interactions. Even though it is not possible to identify a single evolutionary driver for the evolution of paddleshaped wings in penguins, the wing of CM 2018.124.4 documents an early stage in the evolution of wing-propelled diving in penguins. As such, the specimen will be of significance in future studies addressing the evolution of one of the most aberrant forelimb morphologies found in birds. Other Paleocene stem group Sphenisciformes from the Waipara Greensand probably had a similar wing morphology, even though future findings may eventually reveal species that are even closer to that of the flying ancestor of penguins. Funding: This study was supported by a grant to V.L.de.P., R.P.S., and G.M. from the Marsden Fund Council from Government funding, managed by Royal Society Te Apārangi (CTM1601). Access to neutron beamtime was supported by ANSTO ACNS Grant 6008.