Aspects of the cephalic vascular system have been studied in a handful of bird species, including domestic chickens (Gallus gallus) (Kaku, 1959; Richards, 1967, 1968; Baumel, 1975), herring gulls (Larus argentatus) (Midtgård, 1984b), helmeted guinea fowl (Numida meleagris) (Crowe and Crowe, 1979), turkey vultures (Cathartes aura) (Arad et al., 1989), and mallard ducks (Anas platyrhynchos) (Arad et al., 1987; Sedlmayr, 2002). From what is known from the available sample, avian cephalic vasculature appears generally conservative (Baumel, 1993; Sedlmayr, 2002). However, little is known about the head vessels of flamingos, a conspicuously apomorphic bird with a highly divergent feeding apparatus. The greater (or Caribbean) flamingo, Phoenicopterus ruber, is one of five species of extant Phoenicopteridae, an avian group most recently allied to Podicipedidae (van Tuinen et al., 2001; Mayr and Clarke, 2003; Mayr, 2004), although relationships with Anseriformes (Feduccia 1976, 1978; Olsen and Feduccia, 1980; Hagey et al., 1990) and Ciconiiformes (Sibley and Ahlquist, 1990) also have been proposed.
Perhaps the most divergent aspect of flamingo natural history is their feeding behavior, during which they invert their heads, holding their extremely ventroflexed bills near horizontal to strain shallow waters for food items such as algae, small invertebrates, and fish. Flamingos separate food items from water and unwanted particles using a large, fatty, highly sensitive tongue with numerous fleshy spines complemented by a keeled, lamellate bill. In general, the tongue is used as a rostrocaudally oriented pump that quickly (5–20 beats/min) sucks in particulate-laden water and expels unwanted solutes via a complex set of movements (Jenkin, 1957; Zweers et al., 1995). Food is then ingested using the typical inertial throw-and-catch behavior of most birds (Zweers et al., 1994). Whereas P. ruber has a rather small maxillary keel, other flamingos have more heavily keeled maxillae (Mascitti and Kravetz, 2002). This keel aids in the mediolateral movement of water and solutes toward the lamellate margins of the tongue and bill.
We report here on the general vascular anatomy as well as a novel vascular structure of the flamingo head that potentially relates to this unusual feeding apparatus based on vascular injection, high-resolution CT imaging, three-dimensional digital reconstruction, and gross dissection.
MATERIALS AND METHODS
An adult male specimen of Phoenicopterus ruber [Ohio University Vertebrate Collections (OUVC) 9756] was donated to Ohio University from the Brevard Zoo, Melbourne, Florida. The specimen, which died of natural causes unrelated to the system under study here, was frozen very shortly after death and subsequently thawed for study. The specimen was in excellent condition, showing no signs of putrefaction, postmortem coagulation, or freezing artifacts. To promote visualization in dissection and CT scanning, the cephalic blood vascular system was injected with a barium/latex medium using a technique modified from Sedlmayr and Witmer (2002). Although it would have been optimal to have had completely fresh, unfrozen material, such was not possible in that the specimen was not sacrificed for the purpose of this study. Our team, however, has done literally dozens of vascular injections on very similar fresh-frozen specimens with excellent results (Witmer, 2001; Sedlmayr, 2002; Sedlmayr and Witmer, 2002; Clifford and Witmer, 2004; Holliday, 2006). The common carotid artery (aCC; Fig. 1A) and jugular vein (vJG; Figs. 1F and 2) were isolated in dissection and separately cannulated, and the vessels were flushed with tap water for 10 min. Separate 60 cc volumes of both blue and red (Fig. 1E and F) latex injection solution (Ward's, Rochester, NY) were each mixed with Liquid Polibar Plus barium sulfate suspension (BaSO4; E-Z-Em, Westbury, NY) to an approximately 30% barium solution for arteries and 40% barium solution for veins. Different barium concentrations were used to provide a distinction in the densities between arteries and veins so that they can be distinguished in radiographic analyses. Given that arterial lumina tend to be smaller than venous lumina, the concern arose that arteries might be underdetected using these barium concentrations. As documented below, however, this concern was unwarranted in that arterial visualization was excellent; in fact, small supply vessels had to be digitally “pruned” for the illustrations here to simplify and clarify the vascular tree. About 30 cc each of the barium-latex solutions were successfully injected into their respective vessels (red for arteries, blue for veins). The particle size of the barium essentially precludes passage of the injection medium through capillary beds, and thus radiographic signals for arteries and veins remain distinct. In the few cases where there was mixing (e.g., within the sinusoids of the paralingual sinuses), this was readily apparent both in the CT scans (intermediate density) and on dissection (purple color). The specimen was refrigerated for several days and refrozen. It was then CT-scanned, dissected, and compared to several osteological specimens of flamingos. For comparison, we also examined injected specimens available in the OUVC of ostrich (Struthio camelus), domestic duck (Anas platyrhynchos), American alligator (Alligator mississippiensis), and green iguana (Iguana iguana). These specimens were subjected to CT scanning as well as dissection. Moreover, literally dozens of specimens of birds and reptiles have been dissected by the authors in the past several years, and data from these specimens were consulted.
A pilot CT scan of the flamingo specimen (OUVC 9756) was conducted using a diagnostic scanner (General Electric HiSpeed FX/i helical CT scanner at 1 mm slice thickness, 120 kV, 80 mA) at O'Bleness Memorial Hospital in Athens, Ohio. These test data revealed that that the small size of the specimen necessitated the use of a high-resolution CT scanner to obtain optimal data. CT scanning was performed at the University of Texas High-Resolution X-Ray Computed Tomography Facility at 180 kV and 0.2 mA with 1,000 views per rotation (with two samples averaged per view), an aluminum filter, and an air wedge. The field of view was 73.0 mm across 1,024 × 1,024 pixel images. Slice thickness and spacing were 0.264 mm (yielding a voxel size of 0.0713 × 0.0713 × 0.264 mm, which is highly anisotropic). Reconstruction involved filtered back-projection using the Laks convolution filter. The data were imported into a Dell Precision 360 workstation with 2 GB of RAM and analyzed with Amira 3.1.1 (Mercury-TGS, San Francisco, CA). To ease the handling of the data set, the 16-bit data were converted to 8-bit (256 gray values) and cropped to eliminate extraneous data (e.g., air, uninjected areas of the specimen). The disparities in gray values between injection media, bone, and soft tissue (Fig. 3) were significant enough that the 8-bit data were entirely suitable for our purposes. The data were then resampled using a Lanczos filter to yield a voxel size of 0.12 × 0.12 × 0.18 mm. This resampling provided the dual advantage of making the large data set easier to manage and bringing the voxels much closer to isotropy. In order to provide topographic references, the endocranial (brain) cavity and inner ear labyrinths were segmented and rendered in 3D (Fig. 1C and D). A line probe was used in Amira to measure and plot the gray-scale differences between arteries and veins (Fig. 3). The arteries and veins were then segmented based on specified gray-scale values and connectivity. Because of the partial volume effect, a CT artifact that in cross-section creates a penumbra of lower density around the edge of a blood vessel caused by the averaging of the gray-scale values within a single voxel (Fig. 3) (Ketcham and Carlson, 2001), it was necessary to segment some vessels manually as they passed through the more complicated anastomoses to verify accuracy. As a workaround, the area attributed to veins was then swelled, and this larger area was then removed from the artery labels. The segmented arteries, veins, cerebral endocast, inner ear labyrinths, and head were subsequently rendered in 3D (Fig. 1A–D).
The specimen subsequently was dissected under a stereomicroscope to confirm its vascular anatomy (e.g., artery vs. vein, branching pattern), emphasizing anatomy in the orbital, temporal, encephalic, and pharyngeal regions. Thus, all of the CT-based visualizations illustrated here have been validated by gross observation. Although it would have been advantageous to have had an injection/dissection sample larger than this single individual, the anatomical, radiographic, and analytical approaches are so labor-intensive that a larger sample would have been impractical. This problem was ameliorated by identifying key osteological correlates for known vascular structures, thus expanding the sample by including dried skull specimens of P. ruber [e.g., Stony Brook University (SBU) 107161], as well as other flamingo species such as P. minor [Field Museum of Natural History (FMNH) 108183, OUVC 9757, 9758]. These specimens were scored for the presence of specific bony foramina, canals, grooves, and fossae associated with the vessels observed in the injected fleshy specimen (OUVC 9756); many of these same bony features are widely distributed in birds (Baumel and Witmer, 1993).
CT scanning, image analysis, and dissection all reveal that the overall vascular pattern of flamingos (Figs. 1A, B, and F, 2, 4, and 5) closely resembles that of other birds, confirming that avian cephalic vasculature is evolutionarily conservative. Several vascular plexuses are present in the orbitotemporal (e.g., rOP; Fig. 1E), suborbital (e.g., vMX; Fig. 1E), and nasal (e.g., plCH; Fig. 1A and B) regions. In addition, vessels from the pharyngeal, suborbital, and submandibular regions reach the rictus (corner of the mouth), contributing blood flow to structures such as the Harderian gland, palatal plexus, pharyngeal mucosa (Fig. 1F), and rhamphotheca. Moreover, we discovered a pair of novel paralingual vascular sinuses or cavernous tissue masses. Vascular nomenclature generally follows Baumel (1993) and Sedlmayr (2002). The following description generally proceeds in a caudal-to-rostral direction beginning near the occiput.
Overview of Cephalic Vasculature
The primary trunks of the common carotid artery (aCC) are easily identifiable, although smaller neighboring tertiary branches are more difficult to discern because of the complexity of the vascular trees. The first major dorsal branches of the common carotid artery include the internal (cerebral) carotid (aCI) and stapedial (external ophthalmic; aST) arteries, both of which pass through the tympanic cavity (Fig. 4B). The internal carotid artery anastomoses with the palate via the sphenopalatine (aSP) and sphenomaxillary arteries (Fig. 1D), the basilar artery via the caudal encephalic artery (aEU; Fig. 1C), and the orbit via the ethmoidal (aET) and cerebral ophthalmic arteries (Fig. 1B). As in many other birds (Baumel and Gerchman, 1968), there is an extensive internal carotid anastomosis lying within the carotid canal, prior to the entrance into the hypophyseal fossa. The stapedial artery supplies most of the temporal and orbital regions of the head, including the ophthalmic rete (rOP) via the lateral temporal artery, jaw muscles via the superficial temporal artery (aTS; Fig. 1E), the orbit via the profundus artery and its bulbar branches, and the supraorbital salt glands via the supraorbital artery (aSO; Fig. 1A). The occipital artery (Fig. 1A) branches from the stapedial artery near the caudal part of the middle ear cavity. This relationship is similar to that found in herring gulls (Larus argentatus) (Midtgård, 1984b) and cormorants (Phalacrocorax auritus; personal observation), where it also branches from the stapedial artery, although closer to the rostral part of the middle ear. However, this pattern differs from the one found in Anas platyrhynchos (Sedlmayr, 2002), where the occipital artery branches from the internal carotid artery caudal to the middle ear cavity, and Gallus gallus, where the artery arises from the external carotid artery (Baumel, 1975). The occipital artery in flamingos and other birds then passes through a bony canal within the diploe of the skull to exit into the dorsal portion of the epaxial cervical musculature.
Ventrally, the caudal hyoidal (aHC) and hyolingual arteries branch from the common carotid artery and supply throat and hyolingual structures, sending both caudal and rostral branches into the oropharyngeal viscera (Fig. 1F). The lingual (aLI) and submandibular (aSM) arteries branch off the hyolingual artery and course rostrally into the tongue and the lateral wall of the pharynx, respectively (Figs. 1B and F and 2A and C). The submandibular artery supplies the highly vascular pharyngeal wall as a complex arteriovenous rete. It eventually anastomoses with the maxillary vein and with the pterygoid, sphenopalatine, and jugal arteries at the rictus of the mouth and contributes to the palatal plexus (Figs. 1A and 4B). The external carotid artery continues rostrally as the temporomandibular artery, which ramifies into the pterygoid, jugal (aJU), and mandibular (aMN) arteries (Figs. 1A and 4B). These vessels anastomose with numerous other vessels from the submandibular, sphenopalatine, and sphenomaxillary arteries along the lateral aspect of the orbit and palate.
The cerebral arterial circle is visible within the endocranial cavity (Fig. 1D). The basilar artery splits into the paired caudal encephalic arteries (aEU), each of which anastomoses with the ipsilateral common encephalic artery (aEC). The common encephalic artery, the terminal branch of the internal carotid artery, branches into the common rostral cerebral (aRC) and caudal cerebral arteries (aCA; Fig. 1D). The common rostral cerebral subsequently branches into the frontal cerebral (aCF), ethmoid (aET), and cerebral ophthalmic arteries (Fig. 1B). The caudal portions of the dural venous sinuses are present and likely overinflated in the occipital region of the cranial cavity. Just rostroventral to the dural sinuses, the basilar plexus (plBA) is formed by the basilar artery and the anastomoses of the jugular veins (Fig. 1B and D). Although not well visualized in the scan data, the caudal cerebral artery arises from the right common encephalic artery, a common asymmetry present among avian taxa (Baumel, 1967).
The palatal vessels (sphenopalatine medially, sphenomaxillary laterally) exit the cranial base along the basisphenoid and pterygoid bones. These vessels form dense vascular plexuses with the large maxillary vein and their contralateral counterparts in the suborbital region (e.g., Harderian gland) and palatal mucosa (e.g., palatal plexus). From these plexuses, the sphenopalatine artery anastomoses with the nasal branches of the ethmoid artery as part of the caudal and middle conchal plexuses (plCH) of the nasal cavity (Fig. 4A) and then combines with its contralateral counterpart to form the median palatine artery (aPM), which courses rostrally along the midline of the palate (Figs. 1A and 4A). The sphenomaxillary artery anastomoses laterally with the facial and dorsal alveolar branches of the temporomandibular artery to form the antorbital vascular plexus around the antorbital air sinus.
Accompanying the arteries are numerous veins, several of which are anatomically significant. For example, the palatine, rictal, dorsal alveolar, and nasal veins all drain the numerous vascular plexuses of the orbitonasal region of the head. These vessels and the ophthalmic vein from the eye drain into the large maxillary vein (vMX), which courses along the dorsal surface of the pterygoid (Fig. 2C). The maxillary vein merges with the temporomandibular vein (which drains the ophthalmic rete and mandible), forming the rostral cephalic vein (vCR), which then merges with the sphenopalatine and sphenomaxillary veins along the ventral aspect of the braincase (Fig. 2C). The lingual vein is the last cephalic contribution to the large bijugular anastomosis that lies alongside the common carotid artery at the craniocervical junction.
A pair of novel paralingual sinuses were discovered upon CT imaging (Figs. 1, 2, 4, 5, 6, and 7). The sinuses clearly are not an artifactual “blow out” because they are bilaterally symmetrical and have a characteristic vascular architecture, as confirmed by dissection. The sinuses are large enough in the injected specimen (OUVC 9756) to excavate a large ovate bony depression on the medial surface of the lower jaw, just ventral to the medial mandibular fossa and rostral to the medial mandibular process and the insertion of the pterygoideus musculature (Fig. 6A and B). This medial depression is apparently a consistent and highly reliable osteological correlate of the paralingual sinus in that similar fossae were identified in the skulls of other P. ruber specimens (e.g., SBU 107161), as well as skulls of other flamingo species (e.g., P. minor: FMNH 108183, OUVC 9757, 9758; Fig. 6C). Moreover, fossae on the inner surface of the mandible that could potentially be confused with the paralingual sinus fossa, such as those for jaw muscles and the mandibular neurovascular bundle, are well understood (Holliday, 2006) and can be shown to be separate from the paralingual sinus fossa (Fig. 6). Turning to potentially related taxa such as anseriforms and ciconiiforms (Fig. 6D and E) as well as the basal avian taxon Struthio (Fig. 6F), although all relevant muscular attachments and neurovascular fossae can be identified in these taxa, none bears a fossa similar to the paralingual sinus fossa of flamingos. Likewise, no such vascular sinus has been found in our dissections of birds other than flamingos.
The paralingual sinuses are located anatomically be tween the ipsilateral serpihyoideus muscle (mSH) (Vanden Berge and Zweers, 1993) and ceratohyal ventrolaterally (Fig. 1F), the pterygoideus ventralis muscle (mPTV; Fig. 6) and pharyngeal mucosa dorsomedially, the trachea, laryngeal artery, laryngopharynx, and tongue medially, and the lower jaw rostrolaterally (Figs. 1, 2, 4, and 5). Each sinus is suspended between the lingual artery (a large branch of the hyolingual artery) dorsomedially and lingual vein ventromedially (Figs. 1 and 2A). The lingual vein invests the ventromedial portion of the sinus and drains caudally into the jugular vein on the right side and into the rostral cephalic vein on the left side. Each sinus can be divided into two fairly discrete lobes, a caudolateral lobe (sPC) and a rostromedial (sPR) lobe connected by a thin isthmus of glandular tissue (Figs. 4B, C, and G, 5A and B, and 7). The sublingual artery (aSL) branches off from the lingual artery just dorsal to the caudal portion of the caudolateral paralingual lobe and runs rostrolaterally between the two paralingual lobes, where it continues rostrally as a helicine artery along the lateral sides of the basihyale and paraglossus muscles of the tongue (Homberger, 1988). The left sinus is primarily arterial, whereas the right sinus is partitioned into a venous portion in the caudolateral lobe and an arterial and mixed (i.e., purple latex) region in the rostromedial lobe (Fig. 1F). A large laryngeal artery that supplies the laryngopharynx branches off the lingual artery dorsomedially within the rostromedial lobe of the paralingual sinus. The lingual artery then continues into the distal end of the tongue, where it ends in a dense lingual plexus.
Flamingos share a number of similar vascular patterns with other birds, including an asymmetrical caudal cerebral artery and a complex assortment of vascular devices (e.g., the ophthalmic rete). However, particular differences in the branching patterns of some vessels, such as the occipital artery, may eventually shed light on resolving the relatively ambiguous taxonomic relationship flamingos share with other avian groups. In addition, this study revealed the presence of an unanticipated and previously undescribed vascular device, the paralingual sinus. The new vascular injection protocols used here hold promise as a means of distinguishing and then visualizing the three-dimensional anatomy of arteries and veins in CT data sets. This approach successfully demonstrates and, for the first time, allows documentation of the arterial and venous patterns of flamingo head vessels. It is an important advantage to be able to study and analyze all of the vascular plexuses and devices in the virtual realm prior to engaging in destructive and yet ultimately valuable dissection.
Techniques aside, the paralingual sinuses themselves are intriguing structures. They clearly are vascular devices, yet their function has not been experimentally tested. In general, arterial blood supplies the sinuses via the hyolingual system (lingual and sublingual arteries), and the sinuses are drained via the lingual vein. The caudolateral lobes are tightly interwoven with the serpihyoideus muscle, a strap muscle of the hyoid skeleton. Ostensibly glandular material (nonmuscular, nonvascular tissue) also permeates the caudolateral lobe, whereas the rostromedial lobe is almost completely vascular and surrounds the lingual artery. Based on this anatomical arrangement, possible functional hypotheses include a secondary salinity regulatory structure (the primary excretory organ of the head being the supraorbital salt glands, the vasculature of which also is visible on this scan); a countercurrent heat exchanger [e.g., as part of the circumpharyngeal vascular ring (Baumel et al., 1983; Arad et al., 1989; Sedlmayr, 2002)]; and, perhaps most likely, a cavernous mass of erectile tissue providing a mechanical adjunct to the peculiar lingual-pump feeding mechanism by allowing controlled adjustment (via variable engorgement) of mechanical stiffness of the system. Whatever the specific role of the vascular device [see Midtgård (1984a) and Sedlmayr (2002) for elaboration on avian vascular devices], its association with the quickly pumping hyolingual system and its proximity and vascular communication with the jaw muscles, pharynx, and lower jaw (all of which likely physically constrain the sinus) certainly indicate the importance of this previously unknown device and merits further investigation.
The authors thank Tobin Hieronymus and Lisa Cooper for assistance with the injection; Heather Mayle and O'Bleness Memorial Hospital, Athens, Ohio, as well as Richard Ketcham, Matthew Colbert, and Timothy Rowe at the University of Texas at Austin's High-Resolution CT Facility for providing access to and technical assistance with CT scanning; Julian Humphries for assisting with the accompanying Web site on digimorph.org; Chris DeLorey (Brevard Zoo) for donation of the specimen; and Hauke Bartsch (Template Graphics Software) for providing helpful suggestions for approaching the CT data. Funding was provided by NSF grants to L. M. Witmer (Ohio University) and to T. Rowe (University of Texas), and by Ohio University and the College of Osteopathic Medicine.