Abstract
Despite recent advances in aerodynamic1, 2, neuromuscular3, 4, 5 and kinematic6, 7 aspects of avian flight and dozens of relevant fossil discoveries8, the origin of aerial locomotion and the transition from limbs to wings continue to be debated9, 10. Interpreting this transition depends on understanding the mechanical interplay of forces in living birds, particularly at the shoulder where most wing motion takes place. Shoulder function depends on a balance of forces from muscles, ligaments and articular cartilages, as well as inertial, gravitational and aerodynamic loads on the wing11. Here we show that the force balance system of the shoulder evolved from a primarily muscular mechanism to one in which the acrocoracohumeral ligament has a critical role. Features of the shoulder of Mesozoic birds and closely related theropod dinosaurs indicate that the evolution of flight preceded the acquisition of the ligament-based force balance system and that some basal birds are intermediate in shoulder morphology.
Steady-speed gliding, in comparison to flapping, is more amenable to estimating the interplay of forces stabilizing the shoulder in flight. The minimum forces needed to hold the wing outstretched in a gliding pigeon are provided by a mathematical model12. Each wing produces a vertical aerodynamic force at its centre of pressure13 of approximately 0.5BW (0.5 
body weight; Fig. 1a, b). This force engenders a moment about the shoulder (tending to elevate the wing relative to the body) that is countered by a downward force from the pectoralis muscle. The pectoralis acts with such a short lever arm (Lp, Fig. 1b) that a force of 7BW is required to balance the aerodynamic moment.
Figure 1: Forelimb skeleton and pectoral girdle of a pigeon.
a, b, Anterolateral (a) and anterior (b) views. During a steady speed glide the bird's body weight (downward grey vector shown below the chest, FBW) is supported by aerodynamic forces of 0.5BW from each wing (upward blue vectors, Fa). Fa has a lever arm (La) about the shoulder approximately 11 that of the pectoralis (Lp) requiring a pectoralis force (red vector, Fp) of about 6BW to balance the elevation moment. c, Anterolateral stereo pairs showing the saddle shaped glenoid (light blue, G), which lacks a ventral shelf to support the humeral head. The acrocoracohumeral ligament (AHL), spanning from the elevated acrocoracoid process of the coracoid to the transverse sulcus on the proximal humerus, is ideally situated to prevent ventral dislocation by the pectoralis. H, humerus, Sc, scapula (green), C, coracoid (blue), F, furcula (red), St, sternum (yellow). Scale bar in b equals 1 cm.
However, this model12 inaccurately represents the shoulder as a sturdy ball-and-socket joint capable of fully supporting the wing. In reality, a large, ovoid humeral head articulates with a relatively shallow, saddle-shaped glenoid with open dorsal and ventral margins14 (Fig. 1c). The glenoid is structurally incapable of resisting the downward pull on the humerus as the pectoralis counters the elevation moment. Antagonistic muscles and/or ligaments must provide an upward force to preclude ventral dislocation.
The primary upstroke muscle, the supracoracoideus, might be expected to pull upward on the humerus, but its tendon actually pulls ventromedially (rotating the wing by an elevation/supination moment15 rather than by raising the humeral head). Only three of the twelve muscles crossing the shoulder are oriented to produce an upward force (scapulohumeralis, latissimus dorsi and the subcoracoscapulares complex; Table 1, Fymax) in gliding. The acrocoracohumeral ligament (AHL, Fig. 1c), spanning from the acrocoracoid process of the coracoid to the transverse sulcus of the humerus, is also situated to preclude ventral dislocation. The ligament was predicted to limit wing pronation (in pigeon16), or constrain the normal range of flapping movements (in starling14). In swimming penguins, the AHL was inferred to be taut in all wing positions, effectively coordinating shoulder movement and preventing disarticulation17. To explore the relative contributions of these four structures to joint stability, we undertook a force balance analysis of a wing in a gliding posture.
Three-dimensional force balance calculations that include the AHL yield a family of viable solutions in which all forces and moments equilibrate, requiring ligament forces of about 6–9BW and only small inputs from other muscles (Table 1; Fig. 2a, b). No combination of the three upward-pulling muscles can balance the system without the AHL. Higher forces (about 13BW) are generated by the pectoralis during flapping18, implying even greater demands on the AHL. Mechanical testing of the pigeon AHL reveals that the ligament can withstand tensile loading of 39BW (
5BW, n = 10) without failure, far exceeding potential flight-induced forces. Following transection of the AHL, stimulation of the pectoralis or manipulation of the humerus in fresh specimens produces ventral shoulder dislocation, as expected. We conclude that the AHL is critical to maintaining the integrity of the shoulder joint during aerial locomotion. Rather than being supported from below by articular cartilage, the humeral head is suspended from above by the AHL.
Figure 2: Force balance for the glenohumeral joint of a gliding pigeon and comparison of alligator and pigeon shoulder morphology.
a, b, Anterior (a) and right lateral (b) views of primary forces balancing the shoulder (S1, Table 1). The ventromedial force of the pectoralis (Fp) is resisted by the dorsomedial pull of the acrocoracohumeral ligament (Fahl) and a lateral push from the glenoid (Fj). Aerodynamic and smaller muscle forces are not shown. Grey vector proportionately indicates one body weight. c, Comparison of the alligator (left) and pigeon (right) right scapulocoracoids in lateral view, showing the saddle-shaped glenoid cartilages (blue) and homologous coracohumeral and acrocoracohumeral ligaments (white, cut at humeral attachment). Scale bars equal 1 cm.
High resolution image and legend (162K)The avian coracoid is typically described as a compressive strut between the wing and the sternum12, 13 that resists loads generated by the pectoralis. Although the pectoralis vector generally aligns with the long axis of the coracoid, the laterally facing glenoid cannot by itself transmit a force from the humeral head down the coracoid's shaft. At the glenoid, our force balance predicts that the coracoid bears a medially directed load equal and opposite to the laterally directed joint force on the humeral head (Fj; Fig. 2a, b). If unopposed, this force would push the coracoids together, compressing the thorax. The medial push on the glenoid is primarily balanced by the lateral pull of the AHL. The acrocoracoid process experiences a ventrolateral force equal and opposite to the dorsomedially directed AHL force experienced by the proximal humerus (Fahl; Fig. 2a, b). Compression at the glenoid and tension on the acrocoracoid process sum to a net ventromedial force that approximately aligns with the shaft of the coracoid. This net force can be primarily opposed by a dorsolateral force from the sternum at the coracosternal joint. The AHL thus provides a mechanism for transmitting the pectoralis force through the coracoid as a compressive strut.
Has AHL function evolved as a specialization for flight in birds or is its role primitive for archosaurs? Crocodilians, the closest extant relatives of birds, possess a homologous coracohumeral ligament (CHL) and, like birds, retain the ancestral archosaurian saddle-shaped glenoid14. However, in alligators, as in most other diapsids, the long axis of the glenoid is oriented approximately 90° to that of birds, such that the open margins of the joint are anterior and posterior (Fig. 2c). The alligator's humerus is supported ventrally by the coracoid portion of the glenoid saddle, but during locomotion retractors such as the caudal part of the pectoralis pull the humeral head towards the open, posterior margin.
Unlike birds, crocodilians possess several muscles that arise from a broad scapulocoracoid plate anterior to the glenoid (Fig. 3a) and are thus capable of countering the pectoralis. Three-dimensional kinematic records of the shoulder girdle and humerus of alligators walking on the treadmill were compared with ligament-taut positions derived from manipulating dissections. Minimal overlap between ligament-taut and walking humeral positions demonstrates that the CHL is slack throughout the majority of the stride cycle and therefore not essential for joint stabilization during locomotion. Electromyographic data from varanid lizards19 show that several muscles, despite their protractive lines of action, are active while the limb is retracting, suggesting a stabilizing role. Homologous protractor muscles in Alligator mississippiensis are also likely to be primarily responsible for balancing the pectoralis. We propose that shoulder joint stabilization evolved within archosaurs from a primarily active, muscle-based balance system to a passive, ligament-based system in extant, volant birds.
Figure 3: Comparison of archosaur scapulocoracoids.
Lateral view of right scapulocoracoids of Alligator mississippiensis (a), Sinornithoides youngi (b), Sinornithosaurus millenii (c), Archaeopteryx lithographica (d), Confuciusornis sanctus (e) and Columba livia (f) on a simplified phylogeny23. The changing orientation of the CHL/AHL (dashed lines) and the reduction of the scapulocoracoid plate anterior to the glenoid for muscle attachment (orange surfaces) record the transition to a novel, ligament-based force balance mechanism in modern birds. The dashed lines do not represent ligament length.
High resolution image and legend (88K)On the basis of comparative coracoid morphology in Deinonychus, Archaeopteryx and extant birds, the coracoid tuberosity of theropod dinosaurs is considered to be homologous with the avian acrocoracoid process20. With reference to the tuberosity/process and other anatomical landmarks, we reconstructed the orientation of the scapulocoracoid and CHL/AHL for the non-avialan theropods Sinornithoides youngi and Sinornithosaurus millenii and for the volant avialans Archaeopteryx lithographica and Confuciusornis sanctus (Fig. 3b–e). The ancestral ligament orientation of the CHL is retained in all but C. sanctus (Fig. 3e), in which elevation of the acrocoracoid process above the glenoid and deepening of the humeral head yield an AHL orientation intermediate between that in crocodilians (Fig. 3a) and extant birds (Fig. 3f).
An intermediate condition is also found in the recently discovered Jeholornis21 and Sapeornis22, which diverged from the line to extant birds between Archaeopteryx and Confuciusornis22. Hence, we conclude that the common ancestor of Jeholornis and extant birds had likely developed an incipient stabilizing role of the AHL (Fig. 3). On the basis of coracoid morphology, full development of the ligament-based force balance traces back at least to the common ancestor of Enantiornithes and Aves23.
The selective impetus for transforming the coracoid tuberosity into an acrocoracoid process has been associated with the functional modification of three muscles. Displacement of the acrocoracoid directly affected the lines of action of the coracobrachialis and biceps brachii (coracoid head) because, like the AHL, both originate from the process20. In addition, enlargement of the tuberosity formed the triosseal canal, creating a pulley that deflected the tendon of the supracoracoideus such that this muscle became the primary humeral elevator/supinator powering upstroke15. We now also recognize elevation and expansion of the acrocoracoid process as directly related to the reorientation and increased loading of the AHL, which underlies the novel osseo-ligamentous mechanism for passive shoulder stabilization in modern birds.
Methods
Pigeon model
Bones of a wild-caught pigeon (Columba livia) were scanned at the High-Resolution X-ray Computed Tomography Facility at the University of Texas at Austin (UTCT). Three-dimensional polygonal models were created using VTK (Kitware), articulated into a virtual skeleton in Maya (Autodesk), and configured in a mid-downstroke posture (humeral shaft protracted 60° from the vertebral axis, elevated 0° and pronated 15° relative to a neutral position with distal condyles vertical). On the basis of dissections, muscular and ligamentous attachments were mapped onto the virtual skeleton to determine their positions and lines of action (Fig. 1).
Three-dimensional force balance
The wing was treated as a free rigid body articulating only at the shoulder (elbow and wrist joints fixed). Aerodynamic, gravitational, muscular, joint and ligament forces were mapped onto the virtual skeleton as 16 vectors in Maya. These 3D vectors were used to calculate three translational components (Fx, Fy, Fz) and three rotational moments (Mx, My, Mz) relative to the coordinate system's origin at the point of glenohumeral contact (Fig. 2). Two vectors are fully described. A net aerodynamic force (Fa) of 0.5BW acts vertically upward at the centre of pressure, which was estimated to be 13.2 cm lateral and 1.1 cm posterior to the origin by averaging the quarter-chord points of ten chord-wise strips of equal area along the wing's 29.3 cm shoulder–wingtip span13. The wing's weight of 0.074BW acts vertically downward at its centre of mass (3.2 cm lateral and 0.8 cm posterior to the origin, based on a frozen, outstretched wing). All other vectors are only partially characterized. Muscle and ligament lines of action are fixed by anatomy and wing posture in all but the pectoralis, for which a range of vector directions were tested owing to its broad origin. Likewise, joint contact force orientation is constrained to horizontal by glenoid morphology. Maximum isometric muscle force (Fmax) was estimated from the physiological cross sectional area of one pigeon and from published data24, 25, 26 (Table 1). However, the actual magnitudes of muscle, joint and ligament forces are unknown, creating an indeterminate system for which multiple answers exist.
A family of potential solutions was explored by interactively solving the 3D force balance using a MEL (Maya embedded language)-based expression to identify equilibrium. We discovered viable combinations of muscle activation using a simple workflow. All muscle forces began completely inactive (0BW), but could be increased up to Fmax. First, a pectoralis force (Fp) was added to balance the aerodynamic elevation moment (wing weight was included but was not significant). Second, horizontal forces created by the pectoralis were balanced by the joint force (Fj). Third, the pectoralis's vertical force was countered—primarily by the ligament force (Fahl) because muscles alone proved insufficient. Other muscles were then interactively recruited to balance remaining forces and moments, automatically adjusting previously entered Fp, Fj and Fahl values as required. We sought simple alternatives that did not invoke unnecessary and widespread co-contraction (Fig. 2a, b; Table 1). The necessity for high AHL loading was insensitive to variation in pectoralis orientation or wing posture.
Mechanical testing
The coracoid–AHL–humerus complex was dissected from frozen pigeon specimens, each bone potted in low melting-temperature metal alloy and then mounted on an Instron Mechanical Testing System (Model #4442) with a 500 N load cell. Tensile force was measured for 15 loading/unloading cycles from 0–70 N (5 ligaments from 3 birds), followed by failure tests (10 ligaments from 6 birds). Crosshead speeds (rate of loading) ranged between 25–1000 mm min-1, with data collected at either 6.67 or 20 points s-1. The ligament substance was not completely torn in any of the failure tests. Failure occurred by humeral (6) or acrocoracoid (4) fracture.
Alligator shoulder data
Three alligators (1.08–3.08 kg) were recorded at 30 frames per second with simultaneous light (Sony DCR-TRV30) and dorsoventral fluoroscopic video (Sony DCR-VX1000 digital handycam attached to a 27.94 cm image intensifier with a Siemens cineradiographic apparatus) as they walked on a motorized treadmill. Optical and radiographic distortions were removed by 'warping' images of square calibration grids in Maya. Shoulder and forelimb bone models were created using a 3D laser scanner (ShapeGrabber), articulated, and animated in Maya by aligning each bone to its fluoroscopic image27. Shoulder kinematic data during walking were compared with ligament-taut positions derived by matching models to digital photographs of manipulated specimens.
Fossil reconstructions
Three-dimensional models of the scapula, coracoid and humerus were created in Maya based on measurements, photographs and drawings of the original material as well as published descriptions14, 20, 28, 29, 30. The vertebral column, sternum, furcula and ribs (when present) were used to constrain the position and orientation of the scapulocoracoids. The thoracic vertebral column was oriented horizontally, with the humerus in a position comparable to that of the gliding pigeon model.