Relating form to function in the hummingbird feeding apparatus

High-resolution X-ray computed tomography (microCT)

I dissected three salvaged specimens, a Ruby-throated Hummingbird, an Anna’s Hummingbird, and a Short-tailed Woodstar to scan their heads. Such dissections consisted of separating the head from the rest of the body, which allowed a more expedited and low-cost staining procedure (see below) and a better positioning of the specimens for the scanning process (closer to the X-ray source to achieve higher resolution). To obtain detailed morphological data at the micrometric scale and visualize the tongue soft tissues, I employed a staining protocol with osmium tetroxide (OsO₄, cf. Metscher, 2009) with the difference that I did not embed my samples in resin, but instead placed them in small vials that could be positioned as close to the X-ray emitter as required for the desired resolution. I opted for osmium instead of iodine (e.g., Lautenschlager, Bright & Rayfield, 2014) because, although they both seem to bind to lipids (Bozzola & Russell, 1999; Gignac & Kley, 2014), osmium stabilizes tissue proteins, which then do not coagulate during dehydration with alcohol (Hayat, 2000). The heads were kept in 10% neutral buffered formalin and fixed with a solution containing 2.5% (wt/vol) glutaraldehyde and 2% (wt/vol) formaldehyde in 0.1 M sodium cacodylate trihydrate buffer (pH 7.4 adjusted with NaOH) for 8 h at 4 °C. After two washes in distilled water, the heads were fixed/stained with 2% (wt/vol) OsO₄ in 0.1 M cacodylate buffer water for 4 h at 4 °C. Samples were washed three times in distilled water (20 min apart at 4 °C) and then dehydrated in a graded series of ethanol solutions. The specimens were stored in 100% ethanol at 4 °C and scanned at The University of Texas High-Resolution X-ray Computed Tomography Facility. Scans were performed at 70 kV and 10 W, with Xradia 0.5 and 4X objectives, and 1 mm SiO₂, or no filter. Specimens were scanned in three parts, scans were stitched using Xradia plugins, and voxel size was between 15.5 and 5.2 µm. I obtained 16bit TIFF images that were reconstructed by Xradia Reconstructor, and the total number of slices per specimen was between 2,223 and 2,854, with scan times between 4 and 7 h. Using the data from the microCT scans I digitally decoupled the feeding apparatus components (segmenting in Avizo©) and constructed three-dimensional models to study the bill and tongue match.

Histological preparations

I dissected two Ruby-throated Hummingbirds to extract their tongues, which were cut into ∼3-mm long sections and fixed with 1.5% (wt/vol) glutaraldehyde—1.5% (wt/vol) paraformaldehyde in standard buffer (0.1 M HEPES, 80 mM NaCl, 3 mM MgCl₂, pH 7.4 adjusted with NaOH) for a total of 9 h at 4 °C with one change into fresh fixative after one hour. The sections were then fixed in a solution of 1% OsO₄—0.8% potassium ferricyanide—0.1 M sodium cacodylate—0.375 M NaCl for 2 h at 4 °C and then washed in distilled water. The sections were dehydrated in a graded series of ethanol solutions, and embedded in epoxy resin (a mixture of Embed812, Araldite 502 and DDSA, blocks polymerized at 60 °C for 48 h). I obtained semi-thin cross sections (1 µm) that were stained with methylene blue/azure II (1:1) followed by counterstaining with fuchsine for light microscopy. Photomicrographs were captured using a JVC High Resolution CCTV digital camera on an Olympus BX51 compound microscope at different magnifications (up to 1,000x). I used Auto-Montage software (Syncroscopy Inc.) to compile images of multiple optical planes, thereby obtaining pseudo-planar fields of view with improved visualization of the tissue structures.

Transmission electron microscopy (TEM)

I used one Ruby-throated Hummingbird for TEM. Using some of the fixed and embedded sections (epoxy resin processed in a Microwave Tissue Processor, Pelco Biowave Pro) of the tongue from the histological preparations, I obtained thin (80-nm) cross sections using a diamond knife on a Leica Ultracut UCT Ultramicrotome. The sections were put on Formvar support films for TEM and stained with either 2% uranyl acetate (UA) and lead citrate (LC, Reynolds, 1963), UA LC and RuO₄ vapors, or RuO₄ vapors only (Xue, Trent & Osseo-Asare, 1989). These sections were then imaged at the Bioscience Electron Microscopy Laboratory at the University of Connecticut, with a FEI Tecnai G2 Spirit BioTWIN transmission electron microscope at an accelerating voltage of 80 kV and at direct magnifications up to 120,000×.

Scanning electron microscopy (SEM)

I dissected one Ruby-throated Hummingbird and one Rufous Hummingbird to extract their tongues. The tongues were flattened with microslides and fixed with a solution containing 2.5% (wt/vol) glutaraldehyde and 2% (wt/vol) paraformaldehyde in 0.1 M sodium cacodylate trihydrate buffer (pH 7.4 adjusted with NaOH) for 8 h at 4 °C. After six washes (30 min apart) with the 0.1 M cacodylate buffer, the tongues were fixed/stained with 2% (wt/vol) OsO₄ (2.5 ml) in 0.1 M cacodylate buffer (1.7 ml) + distilled water (0.8 ml) for 8 h at 4 °C. The tongues were cleaned by washing them three times in the cacodylate buffer and then dehydrated in a graded series of ethanol solutions. For all of these washes I used jets of fluid (using droppers immersed in the liquids) to ensure that the tongues were free of debris (and remaining nectar) in both dorsal and ventral surfaces; I did not scrape the tongue surfaces in order to keep them intact for posterior visualization. The first tongue was dried with a critical point dryer (Polaron E3000) for 2 h. Unfortunately, critical point drying (CPD) caused the edges of the tongue in the rostral region (where it forms the grooves) to spiral inward while drying, and only a small proportion of the dorsal surface of the tongue was visible after CPD. For the second tongue, I opted to use nylon mesh biopsy capsules and tissue cassettes to keep the tissue from spiraling inward. I inserted the tissue between layers of filter paper (chemically stable and allows adequate fluid exchange) to prevent mechanical damage from the mesh. By employing SEM, I could visualize and photograph the regions of interest, including equal access to both dorsal and ventral surfaces.

After CPD, I sputter coated (Polaron E5100) the tongues with gold and palladium, and attached them to aluminum SEM stubs using double-sided carbon tape, coated the caudal ends of the tongues with silver paint, and connected them to the aluminum stubs in order to reduce charging effects. I imaged the tongues at the Bioscience Electron Microscopy Laboratory at the University of Connecticut, with a Zeiss DSM982 field emission scanning electron microscope operated at an accelerating voltage of 2 kV and at direct magnifications up to 50,000×.

Ex-vivo experiments

I dissected one Ruby-throated Hummingbird to examine tongue-nectar interactions post-mortem. Under an Olympus SZX-12 dissecting microscope, I attached a Micro-Manipulator Model FX-117 (Electron Microscopy Sciences©) via surgical micro clamps to the epibranchial bones of the hyobranchial apparatus (Fig. S1). I held the skull in place with articulating arms coupled to a soft “helmet” made out of a polyvinyl chloride sheet and an Irwin© Quick-Grip Mini Handi-Clamp with swiveling clamping pads provided with longitudinal and transversal furrows that matched the hummingbird’s bill basal diameter without compressing it. At the tip of the bill I positioned a Mitutoyo© Digimatic Digital Caliper connected to a laptop to compare the compression of the tongue by the bill tip in this artificial setting and match it with previous estimates in living hummingbirds (Rico-Guevara, Fan & Rubega, 2015). The end result was the ability to precisely control tongue flattening and protrusion (Video S2). I attached a second Micro-Manipulator to a reservoir filled with artificial nectar (18.6% sucrose concentration) in order to control the bill tip to nectar surface distance without moving the fixed head. Lastly, I filmed the tongue-nectar interactions by coupling a high-speed camera (TroubleShooter HR), running up to 1,260 frames/s (1,280 × 512 pixels), to the dissecting microscope.

Activities were reviewed and authorized by the Institutional Animal Care and Use Committee at the University of Connecticut; Exemption Number E13-001.

High-resolution X-ray computed tomography (microCT)

I present the first complete cross-section series of a hummingbird feeding apparatus. I started with the most caudal section at the nasal operculum (Fig. 2, cross section [XS] 1) where the tongue is dorso-ventrally flattened, and the tongue body (corpus linguae) has started to divide medially due to an ingrowth (sulcus linguae) of the dorsal and ventral epithelia (Fig. 2, XS 1; cf. XS 11 in Weymouth, Lasiewski & Berger, 1964). The tongue body in hummingbirds encompasses the tongue from a distinct base, at the joint between the basihyale and the paraglossum, to the rostral grooves. I do not present a description of the structure of the lingual body in this paper given that this has been detailed previously (Weymouth, Lasiewski & Berger, 1964). At XS 2 there is a dark layer of cornified tissue almost completely surrounding the lingual body. Such layers become thicker at the ingrowth region and eventually connect, when moving rostrally through cross sections (Fig. 2, XS 2–5), effectively dividing the tongue body (cf. XS 13 in Weymouth, Lasiewski & Berger, 1964) and giving rise to a bifid tongue. At XS 3 the semi-cylindrical configuration characteristic of the tongue grooves is already conspicuous (cf. XS 14 in Weymouth, Lasiewski & Berger, 1964).

At XS 4 it is apparent that the tissue inside the lingual body chambers is thinner, leaving an empty space dorso-laterally (cf. XS 15–17 in Weymouth, Lasiewski & Berger, 1964). At this section, the dorsum linguae is made of cornified tissue and it forms a pair of dorsal cornified rods of the lingual tip (cf. Weymouth, Lasiewski & Berger, 1964). These dorsal rods become thicker and more robust when moving rostrally through cross sections (Fig. 2, XS 2–5), probably because they are the sole structural support of the rostral half of the tongue. By XS 5 there is no tissue inside the cornified semi-cylindrical grooves, and the two sides of the lingual body are completely separated (i.e., bifurcated tongue). There is almost no change between the tongue appearance and size between XS 5 and 6, which is about 3 mm corresponding to about half of the total groove length. From XS 6 to 8 there is no ostensible change in the tongue shape besides an overall reduction in size (∼25%). The rostral portion of the tongue is characterized by a reduction of the rods and a thinning in the cornified tissue comprising the grooves (Fig. 2, XS 9–10). It is worth noting that from XS 1 to 4 it is evident how the tongue fills the internal buccal spaces (when the bill is shut), leaving only a small space dorso-laterally. Such space matches the position of tongue base projections (Scharnke, 1931; XS 2 in Weymouth, Lasiewski & Berger, 1964). A reduction in the internal space outside the grooves and a tighter coupling between bill internal walls (oropharyngeal roof, or palatum, and oropharyngeal floor, or interramal region) and tongue shape is evident in the rostral portion of the feeding apparatus (Fig. 2, XS 5–10). A more in-depth description of the bill structures, such as the salivary ducts openings in the oropharyngeal floor (Fig. 2, XS 7), will be provided elsewhere.

Histology and electron microscopy

I focused on the rostral half of the tongue (e.g., Fig. 3A) to complement the work of Weymouth, Lasiewski & Berger (1964) that focused on the caudal half. At its basal region, the tongue is a cylindrical structure containing bones, muscles, vessels, blood cells, loose connective tissue, nerves, and sensory structures (e.g., taste buds), all surrounded by stratified squamous epithelium (Weymouth, Lasiewski & Berger, 1964). Moving rostrally, the tongue shape transitions into two distinct bean-shaped chambers running parallel to each other (Fig. 2, XS 1; Weymouth, Lasiewski & Berger, 1964), the paired paraglossum becomes cartilaginous and thins until it finally disappears along with the muscles, vessels, nerves, and other abovementioned structures, whereas the stratified squamous epithelium becomes thicker and a strongly cornified layer appears in between two layers of epithelium (analogous to the human nail matrix covered by the cuticle, Fig. 2, XS 2–3, 3C; Weymouth, Lasiewski & Berger, 1964). In the rostral half of the tongue all the connective tissue is absent, the bean-shaped chambers become hollow, and the remaining cornified epithelium (stratum corneum) is shaped like two extended ‘commas’ mirroring each other and forming the paired grooves or semi-cylinders at the tongue tip (Fig. 2, XS 4–10, 3B; Weymouth, Lasiewski & Berger, 1964; Ortiz-Crespo, 2003). The growing tissue seems to be abundant at the base of the grooves (cf. Fig. 2; Weymouth, Lasiewski & Berger, 1964), but to disappear in the rostral portions with few remaining cells at the interior of the cornified rod (Fig. 3D).

I found elliptical-to-circular dark corpuscles distributed evenly throughout the tongue tissue (black arrow head, Fig. 4A). The cell boundaries are continuous lines of corneo-desmosomes (e.g., black arrow, Fig. 4B). I found structures of ∼35 Å diameter that possibly are microfibrils (e.g., white arrow, Fig. 4C). Regarding the different staining methods, I found that staining with uranyl acetate and lead citrate provided the best imaging of the elliptical dark corpuscles and the most external layers of keratin, especially in the dorsal surface of the grooves (Fig. S2). However, vapor-staining with RuO₄ offered the best visualization of the corneo-desmosomes necessary to study the cell architecture (Fig. S2).

In the grooved (rostral) half of the tongue, two layers of the stratum corneum can be distinguished: a thicker one underlying the ventral (convex) surface of the grooves, which I refer to as ‘cornified layer’, and a thinner one underlying the dorsal (concave) surface of the grooves (Fig. 3B). The cornified layer is made of larger cells, it is less densely packed, and it contains less granules than the layer closer to the dorsal surface (Fig. 4A). This latter layer may contain some flattened granular-cornified cells but I do not refer to it as stratum granulosum because that name is mostly applied to mammal tissues (Baumel et al., 1993). It is plausible that some of the germinative layers of this keratinized stratified squamous epithelium could be found at the basal portions of the dorsal rods (Fig. 3B), but most of it is restricted to the caudal half of the tongue (Weymouth, Lasiewski & Berger, 1964).

Probably related to the abovementioned differences in underlying tissue, I found qualitative differences between the dorsal and ventral surfaces of the tongue grooves (Fig. 5). These surfaces were cleaned in the same manner (see Methods: SEM), therefore differential accumulation of nectar or dirt residue does not appear to be a confounding factor. In addition, given that the accelerating voltage can alter the level of surface detail visualized, I kept constant 2 kV for all the comparisons. While capturing the EM images, I tried to compare corresponding points on the dorsal and ventral surface, but I did not perceive noticeable differences on the tongue surfaces depending on the relative position on the groove wall (e.g., relative distance to the cornified rod, or at the lancinated portions, Figs. 5A and 5B). At the 10-µm scale the ventral tongue groove surface (Fig. 5C) seems to have more granulated regions in comparison with the dorsal side that appears smoother (Fig. 5D). Furthermore, at the 500-nm scale the ventral surface (Fig. 5E) presented a rougher aspect than the dorsal surface (Fig. 5F).

Ex-vivo experiments

I recorded expansive filling (sensu Rico-Guevara, Fan & Rubega, 2015) in the post-mortem experiments (Fig. S1, Video S2). This observation indicates that physical (structural) rather than muscular forces are responsible for the expansion and filling of the tongue. I flattened the grooves by closing the bill tips and leaving only a small aperture to extrude the tongue through (see methods), reproducing previous observations in free-living birds (Rico-Guevara & Rubega, 2011; Rico-Guevara, Fan & Rubega, 2015), and registered that the flattened grooves expanded spontaneously upon contact with nectar in tongues of deceased specimens (Video S3). Additionally, I observed that the separation of the tips and the relaxation of the fringed regions occurred in post-mortem experiments (Video S4). Consequently, nectar trapping (sensu Rico-Guevara & Rubega, 2011) would be the first step of the fluid collecting system and is immediately followed by expansive filling. I hypothesize that the main force driving the expansive process and therefore the filling of the tongue with nectar is the elastic energy that can be stored in the cornified groove walls.

I explain the hypothesis as follows: (1) The process starts when the tongue is dorso-ventrally compressed upon protrusion; when the tongue is extruded, only a thin layer of nectar remains inside the grooves. Such a thin layer acts as an adhesive (Stephan adhesion) maintaining the dorsoventrally flattened (elliptical) configuration of the grooves even after they pass the extrusion point (bill tip). The attractive forces between the nectar and the tongue (adhesion, cohesion, and surface tension) are able to resist the elastic energy stored in the grooves’ walls (cornified layers), and thus keep the grooves flattened. This stable flattened configuration is conserved during the trip of the tongue across the air space between the bill tip to the nectar pool. In the dorsal portion of the tongue, where the groove’s inside upper edge meets the rod, the free (outer) edge of the groove is prevented from rolling outward by a narrow sheet of nectar joining it to the rod. The surface tension at this exposed nectar sheet keeps the grooves “zipped up” by preventing air from entering the groove itself. Surface tension at the tip of the tongue also keeps the grooves stuck to each other, forming a unitary structure. (2) Once the tongue passes the compression point at the bill tips, there is a slight expansion in the tongue grooves (because of the cessation of compressive forces). The expansion of the grooves is arrested at the point in which the attractive forces between the tongue walls and the nectar balance out the elastic forces of the grooves’ walls. This creates an initial transient equilibrium that maintains the flattened configuration (cf. Rico-Guevara, Fan & Rubega, 2015). (3) Once the tongue tip contacts the nectar surface, the free supply of fluid eliminates the surface tension that was holding the grooves together, allowing the area of the grooves that is inside the nectar to open (cf. Rico-Guevara & Rubega, 2011). This opening of the ends of the grooves allows the nectar molecules from the nectar pool to start interacting with the nectar molecules inside the grooves (i.e., elasticity-induced flow, Fig. 6). On the dorsal surface of the length of the grooves still outside the nectar pool (more proximal to the bird’s mouth), the surface tension of the fluid sheet between the rods and the groove walls holds the grooves in the rolled, flattened position. (4) Molecules of liquid entering the tongue grooves, at the boundary where the tongue enters the nectar pool, start moving proximally through the grooves, creating a jet of fluid that fills the grooves following their expansion (cf. Rico-Guevara, Fan & Rubega, 2015). This continued destabilization of the initial transient equilibrium causes the area of the grooves outside the nectar to expand which in turn causes them to fill, creating a positive feedback that forces the grooves open along their entire length. This creates a filling front wave because the expansive process happens from the point of contact with the nectar backwards (Fig. 6). (5) The expansion stops when most of the potential elastic energy is released (and the grooves are fully reshaped into their cylindrical configuration) and when the remaining elastic energy is counteracted by the surface tension at the zipped dorsal slit (cf. Rico-Guevara & Rubega, 2011). At this point the grooves have achieved their maximum capacity, and they are completely filled with nectar.

Notes on gross tongue morphology relevant to feeding in hummingbirds

Hummingbird tongues may look as a fishing line due to their extreme slenderness, but are truly complex structures well adapted to particular tasks. Hummingbirds can extend their tongues beyond their bill tips up to about two times the bill length (e.g., Fig. 7A), given that most hummingbird tongues are only slightly longer than their bills (Fig. 2, Rico-Guevara, 2014), the tongue base can be extended pass the bill tip (transition visible in Fig. 7A). This remarkable lingual protraction can be achieved by the rostral displacement of the elongated hyoid apparatus (e.g., Video S5 ), and hummingbirds can protrude their tongues with their bills closed because of the presence of an elastic envelope between the larynx and the tongue base (e.g., Fig. 7B), which allows lingual protraction without dragging the trachea inside the bill. Lingual protrusion serves to increase the range of the tongue tips, and also to reach the bill tips with the tongue base, which is important for the intraoral transport of the food. At the tongue base, hummingbirds present two caudal-facing flaps without conical papillae (e.g., Fig. 7C), which may aid during intraoral transport. I did not find papillae neither through macroscopical observations of the entire tongue nor through microscopical observations at the rostral regions.

Gross morphology of hummingbird tongues

Hummingbird tongues entirely lack papillae, a rare condition in vertebrate tongues (Schwenk, 2000; Iwasaki, 2002) and even among birds (review in Erdoğan & Iwasaki, 2014). Avian lingual papillae are involved in manipulation of solid food (e.g., prey apprehension, holding, cutting, filtering, shelling, Iwasaki, Asami & Chiba, 1997; Kobayashi et al., 1998; Jackowiak et al., 2010; Jackowiak et al., 2011; Guimarães et al., 2014; Skieresz-Szewczyk & Jackowiak, 2014) and caudal intraoral transport of solid items (review in Parchami, Dehkordi & Bahadoran, 2010). Hummingbirds have remarkable feeding modes; first, about half of their diet (cf. Stiles, 1995) is composed of floral nectar that is collected inside the tongue grooves; this process does not involve adhesion of the liquid to intra-papillar spaces, as in the case of bats (Birt, Hall & Smith, 1997; Harper, Swartz & Brainerd, 2013) or lorikeets (Homberger, 1980, p. 41). Second, the other half of their diet (cf. Stiles, 1995) consists of arthropods, which most hummingbirds capture by flycatching (Stiles, 1995; Rico-Guevara, 2008). Yanega & Rubega (2004) showed that the flycatching mechanism in hummingbirds involves an expansion of the gape (see also Smith, Yanega & Ruina, 2011) and most of the aerial prey are captured at the base rather than at the tip of the bill; therefore, little or no intraoral lingual transport is necessary. Other hummingbirds, especially from the subfamily Phaethornithinae (‘hermits’), consume mostly substrate-captured prey (e.g., spiders, Stiles, 1995). This is also the case of reproductive females of many species across the family, which have higher protein requirements (Rico-Guevara, 2008; Hardesty, 2009). In the process of consuming substrate prey or prey that are generally captured near the bill tip, hummingbirds, as other birds, can use inertial transport (cf. Mobbs, 1979; catch and throw, Zweers, Berge & Berkhoudt, 1997; or cranioinertial feeding, Tomlinson, 2000; Gussekloo & Bout, 2005; also called ballistic transport, Baussart et al., 2009; Baussart & Bels, 2011; Harte et al., 2012) while flying, or lingual transport (Yanega, 2007). Hummingbirds have evolved the ability to protract their tongues past the bill tips to feed on nectar, but the purpose of the extreme protrusion that they can achieve (e.g., Fig. 7A) is still a mystery. Thus, hummingbirds can reach the rostral portions of their bills with the tongue base (to perform lingual transport for instance), without dragging their tracheae rostrally, because of the development of an accordion-like tube (tuba elastica, Zusi, 2013) between the epiglottis and the tongue base which can contain a large part of the hyobranchial apparatus during tongue protrusion (cf. Weymouth, Lasiewski & Berger, 1964; Fig. 7B). This tuba elastica appears to be a modification of the fibrous attachment between the rostral process of the cricoid cartilage and the rostral process of the basihyale (Soley, Tivane & Crole, 2015). Hummingbirds’ lack of lingual papillae and protrusion abilities may be explained by their arthropod hunting and consumption strategies, as well as their liquid food collecting method: grooves with smooth surfaces are easier to extrude nectar from, and protrusible tongues not only to reach but also to transport food intraorally.

Besides lacking papillae, hummingbird tongues are also unique because of their alae linguae (cf. Weymouth, Lasiewski & Berger, 1964; Homberger, 2017), which are flattened projections at the base of the tongue (Fig. 7C). These two flaps are located and oriented at the same place and in the same general direction as the papillary crest in other birds. Nevertheless, these structures do not present caudally directed conical papillae, as is usual in avian tongues (e.g., Erdoğan & Alan, 2012; Erdoǧan, Pèrez & Alan, 2012). In comparison to the width of the tongue, these flaps are greatly elongated laterally in Sicklebill hummingbirds (Eutoxeres, Fig. 7C), which have strongly decurved bills. These flaps are thin and flexible at touch, as well as positioned dorso-laterally forming a V-shaped structure. These flaps in hummingbirds have no parallel among nectar-feeding birds (Lucas, 1894; Scharnke, 1932; Scharnke, 1933; Rand, 1961; Rand, 1967; Bock, 1972; Morioka, 1992; Pratt, 1992; Downs, 2004; Chang et al., 2013), or birds in general (e.g., Erdoğan & Alan, 2012; Erdoğan, Sağsöz & Akbalik, 2012; Erdoǧan, Pèrez & Alan, 2012; Erdoğan & Iwasaki, 2014; Erdoğan & Pérez, 2015). I hypothesize that the alae linguae could aid to move the nectar backwards during its intraoral transport (Rico-Guevara, 2014) and to drag proximally arthropod prey that are caught at different places along the bill length (cf. Yanega, 2007). In terms of general shape, hummingbird tongues are not triangular and dorsoventrally flattened as in most birds (review in Erdoğan & Pérez, 2015), instead, as it is the case in other nectarivorous birds, these tongues are cylindrically shaped (e.g., Bock, 1972; Downs, 2004; Chang et al., 2013). Lastly, I found that hummingbird tongues near the tip also lacked taste buds and salivary glands (found in other birds, review in Erdoğan, Sağsöz & Akbalik, 2012), in agreement with previous work by Weymouth, Lasiewski & Berger (1964).

Ultrastructural characteristics of hummingbird tongues

The rostral portions of the hummingbird tongue, the ones that collect the food, are mostly transparent and their tissues are extremely thin (Figs. 2, 8A and 8C), a rare condition in vertebrates. The species studied with TEM had transparent tongues and also presented few and small dark corpuscles (Fig. 4A), which possibly are melanin granules (e.g., Dummett & Barens, 1974). I expect that species with darker tongues (tongue color varies across the family, (Rico-Guevara, 2014)) will have more and/or larger dark corpuscles of the kind reported here. The ∼35 Å diameter structures that I found in the tissue (Fig. 4C) are likely to be microfibrils; the ventral layers of cornified tissue are more similar to those found in feathers (β-keratin) than to that of tissues with α-keratin (cf. Filshie & Rogers, 1962). Specifically, the diameter of the putative microfibrils is within the range of other β-keratin tissue microarchitectures (Parakkal & Alexander, 1972, p. 33), and almost a third of the diameter of α-keratin microfibrils (Filshie & Rogers, 1962; Johnson & Sikorski, 1965). In most avian tongues the stratum corneum at the ventral surface comprises less than 10% of the lingual tissue in a cross section (Erdoğan, Sağsöz & Akbalik, 2012; Erdoğan & Iwasaki, 2014). Different from most birds, the cornified ventral layer in hummingbirds accounts for between 50%, near the cornified rod and near the groove base, and 100%, at the edge of the groove wall and at the tongue tip, of the tissue in cross sections (Figs. 2, 3A, 8B and 8D, S2). I suggest that most of the germinative layers of this keratinized stratified squamous epithelium (including the layers of dead cells, the stratum corneum) disappear before reaching the most rostral portions of the hummingbird tongue; similar to what would be expected in cross sections of human nail overhangs. Therefore, the caudal half of the hummingbird tongues is made of dead cornified tissue that is shaped by the interaction with the bill, and it is constantly replaced from the rostral half. A thick (cornified) layer of β-keratin can increase mechanical resistance on a surface that is compressed and scraped by the serrated edges of the bill tip ∼14 times a second (Ewald & Williams, 1982) and literally tens of thousands of times a day (Rico-Guevara, 2014). Future experiments to test the hypothetical high percentage (50–100%) of β-keratin in the hummingbird tongue grooves could use in situ hybridization, immunolabeling for β-keratins (e.g., in Alibardi et al., 2009) or selective biodegradation of β-keratin (e.g., Lingham-Soliar, Bonser & Wesley-Smith, 2010; Lingham-Soliar & Murugan, 2013).

I found differences between the layers of tissue underlying the dorsal and ventral surfaces of the tongue grooves (Fig. 3B). These differences may be explained by the organization of the tissues (Fig. 4A), but they may also be influenced by differential composition and organization between proteins (fibrous vs. matrix components) and/or the presence of β-keratin (reviewed by Alibardi et al., 2009), which has been found in the rostral ventral epithelium of other avian tongues (review in Carver, Knapp & Sawyer, 1990). On the ventral surface of the tongue grooves I found thick stratum corneum (cf. Fig. 4 in Kadhim et al., 2013; Figs. 5 and 6 in Jackowiak et al., 2015), but without the underlying lamina propria characteristic of heavily cornified areas in bird tongues (Farner, 1960; Kadhim et al., 2013). This stratum corneum in the tongue surface is common in birds (Farner, 1960; Erdoğan, Sağsöz & Akbalik, 2012; Erdoğan & Iwasaki, 2014), however, as opposed to hummingbirds, in several bird species the stratum corneum is better developed on the dorsal lingual surface (Iwasaki, 2002; Erdoğan, Sağsöz & Akbalik, 2012). I found more sloughing cell layers in the histology and TEM preparations in the dorsal compared to the ventral surface, which indicates that the ventral surface is underlain by harder keratin (cf. Lucas & Stettenheim, 1972). Interestingly, my results are consistent with the idea that dorsal and ventral surfaces of hummingbird tongues have different rugosities (Figs. 5 and 8D). To conclude that there are significant differences between dorsal and ventral surfaces of the hummingbird tongue, it would be necessary to quantify differences in roughness; the best way to do this is by using Atomic Force Microscopy (e.g., Ghosh et al., 2013). Alternative techniques (e.g., Nanda, Sarangi & Sahu, 1998; Fujii, 2011; Kremer et al., 2015) include the use of optical interferometry (e.g., white light scanner), and 3-D reconstructions of tilted SEM micrographs (stereomicroscopy). Differential rugosity between tongue surfaces would have direct implications for their hydrophobicity, i.e., increased roughness may significantly increase contact angle of a water droplet and decrease contact angle hysteresis, which would augment its hydrophobicity (e.g., Michael & Bhushan, 2007). Therefore, the dorsal tongue groove surface, which is less rugose, may be more hydrophilic than the ventral grove surface, and potentially facilitating the fluid trapping process described by Rico-Guevara & Rubega (2011).

Microanatomy of the hummingbird feeding apparatus

Hummingbird tongues, as well as most avian tongues, correspond to the shape of the interramal region (oropharyngeal cavity floor), although commonly not to its size (e.g., Abou-Zaid & Al-Jalaud, 2010; Tivane et al., 2011; review in Abumandour, 2014). Nevertheless, it is worth noting that avian tongues are not larger than the oropharyngeal cavity (as it is the case in some nectarivorous bats, Muchhala, 2006), instead, to reach farther away from the tip of their bills, the mobile bones of the hyoid apparatus in some avian taxa appear greatly elongated, allowing for tongue protrusion (e.g., Video S5). In hummingbirds, the tongue grooves fit perfectly in the rostral portion of the oropharyngeal cavity and match both lower and upper bill internal walls (Fig. 2), which is of vital importance for the efficient offloading of nectar (cf. Rico-Guevara & Rubega, 2011) and intraoral transport (Rico-Guevara, 2014). My study presents the first high-resolution (5-µm voxels) CT scan of a vertebrate tongue satisfactorily stained to highlight soft tissue. A study on flamingos presented detailed CT scans of the head (including the tongue) stained with a novel injection technique (Holliday et al., 2006), but it focused on vascular anatomy at lower resolution than in the present study. Within the last five years other studies have used a variety of techniques to enhance visualization of soft tissue in vertebrates (reviews in Gignac & Kley, 2014; Lautenschlager, Bright & Rayfield, 2014; Gignac et al., 2016), but they have not been focused on tongues or worked at the micro scale of the present study. This three-dimensional modeling of hummingbird tongues allows for the clarification of some misconceptions; for instance, it has been suggested that the mathematical model derived for capillary filling provides a rationale for the shape of hummingbird tongues (Kim et al., 2012). Specifically, that the semi-cylindrical shape of the grooves (cylinders with a dorsal slit) can be explained by an optimal opening angle of a cross section, which matches a peak of energy intake rates (Fig. 4 in Kim et al., 2012). I prefer a more parsimonious explanation: starting with a dorso-ventrally flattened tongue as an ancestral condition (cf. Emura, Okumura & Chen, 2010; Shah & Aziz, 2014), evolution would maximize the nectar-holding capacity by selecting for a cylindrical structure. In the same way in which a sphere is the shape with the lowest surface area to volume ratio, for an elongated structure (like a tongue), a cylindrical configuration achieves the greatest capacity for a given amount of tissue (in this case, the groove walls). It is worth noting that the tongue tip whilst outside the nectar ends in a conical shape (Fig. 1 in Rico-Guevara, Fan & Rubega, 2015), due to a shortening of the cross-sectional length of the groove wall (Figs. 2 and 3), which helps to trap and retain the nectar at high licking rates (Rico-Guevara & Rubega, 2011). Rostrally, the groove wall membranes exhibit diagonal to perpendicular cuts in the tissue starting from their lateral edges (Fig. 8C), forming lancinated walls in the distal portions of the grooves (Lucas, 1891; also called lamellae, Rico-Guevara & Rubega, 2011). Such cuts may originate by wear during the extruding action of the serrated bill tips on the rostral tongue portions (Lucas, 1891; Rico-Guevara, 2014), and may facilitate the bending of the tongue tip and trapping of fluid drops while mopping the inside of nectar chambers. Wearing at the tongue tip seems to counteract the continuous elongation of the tongue by the growing tissue at the base of the grooves (cf. Fig. 2), and unpublished descriptions of hummingbirds with ‘dislocated’ tongues (feeding from artificial feeders with the tongue always hanging to one side from the bill base) report that their tongues are unusually long and/or they become longer with time.

Additionally, microCT data could inform the mathematical models necessary to make predictions about feeding efficiency across the varying morphology of hummingbird species. For instance, by calculating the total and partial groove capacities depending on immersion lengths (conditioned by the nectar pools on the flowers they visit) the expected amount of liquid extracted can be obtained, and then compared to performance measurements in the wild. Further calculations of the intraoral flow on nectar (based on the bill-tongue internal coupling) taking into account a range of liquid properties that vary in nature (e.g., composition, viscosity, temperature, etc.) will provide information on the limiting step of the fluid collection and transport system. Such an approach would generate falsifiable quantitative predictions about the action of the feeding apparatus, and the volumes of nectar that can be collected and the speed at which they can be transported, for nectars of different concentrations and at different temperatures (hummingbirds feed from flowers at elevations as high as 5,000 m, Carpenter, 1976). Results from this proposed approach will shed new light on the long-standing debate about the reason of the mismatch between hummingbird nectar concentration preferences (Hainsworth & Wolf, 1976; Roberts, 1996; Morgan et al., 2016) and the concentration of the nectar of the flowers they pollinate (review in Nicolson, Nepi & Pacini, 2007). The predictions from these mathematical models available only with the MicroCT reconstruction data, could be tested with additional experiments under controlled conditions using post mortem tongues (building on the ex-vivo experiments presented here), and by measuring nectar extraction rates (fluid volume uptake (µl/s)) in free-living nectarivores living under extreme environmental conditions.

Biophysics of nectar collection

The post-mortem observations (e.g., Videos S3, S4) are consistent with the idea that expansive filling and nectar trapping are processes that do not incur any extra energy than that necessary to squeeze the nectar out of the tongue and inside the bill, making this elastic micropump a highly efficient device (Rico-Guevara, 2014). This is because when the tongue tips enter the surface of the nectar pool, the attractive forces (adhesion and cohesion) holding the groove walls flattened get weaker because more molecules of fluid are available to fill the internal groove space. This creates an imbalance, with elastic forces dominating, which results in reshaping of the groove walls away from the flattened configuration at the tongue tips. Molecules of nectar are pulled inside the grooves through the release of the elastic energy initially stored on the flattening of the groove walls (Fig. 6). Because the grooves are sealed on top (by surface tension in the zipped dorsal slit), the release of the elastic energy (reshaping of the grooves) pulls more and more nectar molecules inside the grooves until they reach a stable cylindrical configuration, from the tips to the base of the grooves. The net result of this process is that the portions of the tongue that remain outside the liquid expand and are filled quickly with nectar, thereby improving fluid collection efficiency. Thus, the tongue filling is achieved through the transition from a high potential energy state (flattened grooves) to a low potential energy state (filled grooves). In summary, the elastic properties of the cornified layer make the elasticity-induced flow hypothesis plausible. This is ecologically relevant because when the bill tip is almost in contact with the nectar surface (most likely scenario in the wild given hummingbird flowers’ internal morphology), the process described above is sufficient to fully load the fringed distal portion of the tongue. Nevertheless, when the bill tip is not in contact with the surface of the nectar (e.g., hummingbirds visiting flowers with corollas longer than their bills), but instead there is a space between the bill tip and the nectar pool, the portion of the tongue that remains outside the liquid would be filled with fluid by the interaction of the aforementioned physical forces in a process I hypothesize as follows: As the tongue is protruded, the grooves are dorso-ventrally flattened by the bill tips, and once the tongue tip contacts the nectar surface the fluid starts to penetrate the flattened grooves (because of cohesion of water molecules in the nectar pool and water molecules in the nectar remaining trapped inside the tongue). When the grooves expand, their walls start releasing the potential energy stored by the bending (flattening by the bill tips). At this point, the excess Laplace pressure due to the nectar flowing inside the grooves plus the releasing of the potential energy whilst the grooves’ walls are recovering their semi-cylindrical shape, create a positive feedback between the groove’s internal space expansion and the nectar flow. The net result of this process is that the portion of the tongue that remains outside the nectar is also loaded with nectar (Fig. 6). Additionally, if there are empty portions of the tongue still inside the beak before passing through the squeezing zone at the bill tip, the nectar filling the grooves will reach them via cohesive and adhesive forces to all the internal surfaces; thereby allowing complete loading of the grooves. Alternatively, the complete filling of the tongue may be achieved by the bill-tongue interaction, involving mechanisms like suction, surface tension transport, hydrostatic pressure motion, etc. However, this would be dependent on, and pertains to, the intra-oral transport of the nectar, which remains understudied.