We visualized and compared the ossified elements of the male (N = 12) and female (N = 5) zebra finch syrinx μCT datasets. Without staining of soft tissues, only ossified tissue is clearly visible in μCT scans. Figure 2 shows the 2D projections of a male and female 3D syringeal skeleton. Because no apparent asymmetries were observed in the number of bones comprising the syringeal skeleton (N = 17), we consider the structure symmetrical in our description. The zebra finch syringeal skeleton is clearly sexually dimorphic, with the male syrinx being larger and more robust (Figures 2 and 3, and Additional file 1). We observed considerable larger variation in the shape of the ossified elements in the female skeleton compared with the male skeleton (Figure 3). However, we found no differences in number and orientation of ossified elements between sexes (Figure 3). Therefore, we assume the following description to be accurate for both sexes.

The zebra finch syringeal skeleton consists of modified tracheal rings and bronchial half-rings. It is therefore considered a syrinx of the tracheobronchial type 16. Two bronchial half-rings and four to six tracheal rings are fused to form a rigid cylinder named the tympanum (Figure 2 and Additional file 1). At the tympanum's caudal end, two bronchial half-rings form a medial dorso-ventral bridge called the pessulus (Figure 2 and Caudal view in Additional file 1). The trachea consists of tracheal rings (T1, T2, and so on) rostral to the tympanum. Caudal to the tympanum are bronchial half-rings (B1, B2, and so on), of which the first three (B1 to B3) are highly modified (Figure 2). Despite the above-mentioned observation that the most caudal part of the tympanum consists of two bronchial half-rings, we decided not to rename them B1 to avoid excessive alterations to existing nomenclature. Figure 4 lists our nomenclature together with the alternative nomenclature of previous authors.

The syringeal muscles that attach directly onto bone surface leave clearly visible imprints. This is especially evident on the ventral side of the tympanum and also on the dorsal side of rings B1 to B3, where deep concavities indicate muscle attachment sites (Figure 2 and Additional file 1). These impressions are apparent, because bone is a dynamic tissue that is remodeled continuously as a result of local compressive and tensile stress 70. Consequently, local forces exerted by muscles shape the surface microstructure of skeletal elements.

Bronchial half-ring B1 is the first of the three heavily modified bronchial half-rings, whose geometry can be best understood by rotating the 3D models in Additional file 1. Bronchial half-ring B1 is hollow and arc-shaped seen from a lateral view and flattened in transverse section (Figure 5A,B). It fits tightly to the tympanum and its internal walls form a smooth surface (Figure 5A). In addition, several trabeculae structurally connect the inner and outer walls of B1 (Figure 5). The dorsal end of B1 contains no trabeculae (Figure 5C,D), is rostro-caudally flattened, and has a concave muscle attachment site (Figure 2C and Additional file 1). The dorsal apex of B1 has a lateral protuberance of which the caudal side forms a smooth arc with the pessulus and B4 when seen from a dorsal viewpoint (Additional file 1, Cut dorsolateral view with labels). The second bronchial half-ring (B2) is also arc-shaped, hollow and slightly flattened. Both the ventral and dorsal side have cup-shaped muscle attachment sites (Figure 2). In contrast with B1, B2 has no internal trabeculae (Figure 5). The third bronchial half-ring (B3) is the least arced of the three half-rings but also hollow, laterally flattened and heavily reinforced with trabeculae. The dorsal end widens to about four times its medial width and is characterized by a domed surface on the rostral end, which forms a joint with the flattened dorsal cup of B2 (Additional file 1, Rostral view).

The geometry and internal structure of the ossified skeleton determine how external forces are conducted by, through and in-between bones forming linkages and articulation points. As such, the skeleton provides a basis for the understanding of syringeal biomechanics. Several muscle attachment areas are clearly delineated by depressions and ridges on the syringeal skeleton.

To image soft tissues in 3D at high resolution, we compared results derived from three methods: high-field (17.6 T) MRI with added contrast agent; μCT with two different types of staining; and conventional histology. Figure 6A-C shows a histological section next to two virtual sections through datasets based on MRI and μCT. With classical histology it is rather problematic to construct accurate 3D datasets, because of section alignment inaccuracies as well as unavoidable tissue distortions resulting from fixation, embedding, cutting and staining. The μCT technique provided the highest resolution (5 μm) datasets and was therefore used to generate 3D datasets for further study and annotation. Additional iodine-based staining (see Methods) provided sufficient contrast for unambiguous distinction of all syringeal soft tissues (Figure 6C). We took special care to scan the excised syrinx in situ by leaving all surrounding tissues, such as the lungs, esophagus, crop and large blood vessels, attached and intact (Figure 6D-G).

The bilateral sound-producing vibrating tissues in the songbird syrinx 24263032 are located in the primary bronchi (Figure 7). Additional file 2 contains the interactive 3D model in which all structures are switched on initially. Although both left and right sides can be interactively accessed in Additional file 2, we will here focus on the right side of the male syrinx because it is predominantly used during zebra finch vocalization 71. The lateral labium is connected to the medial side of half-ring B3 from dorsal to ventral (Figure 7A,B and Additional file 2, view Lateral labium) and thickens in two rostro-caudal running bands (Figure 7B), as also observed in all our fresh tissue micro-dissections (Figure 7C,D). On its medial side, each bronchus is sealed airtight from the pessulus to the lungs by the medial vibratory mass (Figure 7E and Additional file 2, view Medial vibratory mass). This tissue continuum 32 is roughly triangular-shaped from a medio-lateral view and consists of a thicker rostral part, the medial labium that connects dorso-ventrally to the pessulus, and a thinner caudal part, the medial tympaniform membrane (Figure 7F). We found no evidence of a membrana semilunaris on top of the pessulus in the zebra finch syrinx.

Dissections and μCT scans reveal the presence of four paired structures consisting of cartilaginous tissue (Figure 7 and Additional file 2). First, bronchial half-ring B2 has a large medial ventral cartilaginous extension in both males and females, which shows some ossification at the ends in some individuals. This structure connects to the ventral side of the ML. Setterwall 18 named this structure the cartilago tensor, but we here adopt a terminology modified after King 72 by naming this cartilaginous pad the medial ventral cartilage (MVC). Second, more lateral, a cartilaginous rostro-ventral extension of B3 can be observed, the lateral ventral cartilage. In addition, two cartilaginous elements are embedded in the medial labia on the dorsal side. Third, a disc-shaped dorsal element, the medial dorsal cartilage, is situated in the medial labia just caudal of the pessulus. Fourth, an elongated element, the lateral dorsal cartilage, is located on the border between the medial labia and the medial tympaniform membrane. This latter structure was named cartilago tensor in 32.

Summarizing, we provide, to our knowledge, a first 3D description of the sound-producing vibratory tissues including their embedded cartilaginous linkages in situ using the complementary approaches of conventional histology, micro-CT and MRI.

Syringeal muscles translate the motor commands from the central song system into syringeal behavior. Therefore it is critical to identify and locate all syringeal muscles. We identified the muscles controlling the zebra finch syrinx and established their orientation and attachments sites using a combination of micro-dissection and μCT scan annotation (see Methods). Vertebrate skeletal striated muscle tissue is hierarchically organized. A muscle fiber is a single multinucleated cell containing contractile elements, a fiber bundle is a collection of muscle fibers, and numerous fiber bundles make up a muscle. In contrast to most other skeletal muscles, we observed that syringeal muscle fiber bundles are not organized in cylindrical bundles, but instead in well-defined parallel sheets containing dozens to hundreds of muscle fibers. We will first focus on the muscles that have both attachment sites on the syrinx, the intrinsic syringeal muscles 16, and subsequently include the muscles with one attachment site outside of the syrinx, the extrinsic syringeal muscles. Figure 8 shows the syringeal muscles and their attachment sites for the male and female zebra finch syrinx. Additional file 2 contains the 3D model of the male syrinx morphome with all annotated structures and predefined views as seen in Figure 8. Although sexual dimorphism in muscle volume is obvious, we found no differences in muscle definition and attachment sites between male and female specimens. Therefore the description below applies to both sexes.

We identified two muscles on the ventral side of the syrinx: musculus syringealis ventralis or ventral syringeal muscle (VS) and musculus tracheobronchialis ventralis or ventral tracheobronchial muscle (VTB), of which the latter consists of two parts: musculus tracheobronchialis ventralis profundus or deep ventral tracheobronchial muscle (DVTB) and musculus tracheobronchialis ventralis superficialis or superficial ventral tracheobronchial muscle (SVTB) (Figure 8A,B). The VS is the largest syringeal muscle and its caudal attachment site is located in the cup of bronchial half-ring B2 and on the MVC, continuous with the ventral end of B2 (Figures 8C,D and 9A). Rostrally, the attachment site, which can be clearly seen as an impression in Figure 2 and Additional file 1 starts on the mid part of the tympanum and extends to the top of the tympanum (Figure 8C,K). The VS is organized as a series of muscle sheets that run parallel to the central axis of the syrinx (Figure 9B). The VTB has previously been considered as a single muscle 73, but using our μCT data supported by micro-dissection, it became apparent that the VTB is made up of two parts with distinct rostral attachment sites. The caudal attachment site for SVTB fibers is located on the rostro-ventral tip of B3 (Figures 8C,D and 9B). The rostral attachment site is not on bone but instead on the clavicular air sac membrane (CASM) at tracheal rings T4 and T5 (Figures 8C,D and 9C). The caudal attachment site for the DVTB is located on B3 more proximal to the attachment site of the SVTB (Figures 8CD and 9B). Both the SVTB and the DVTB are externally apparent at this point, but, more rostrally, the DVTB twists underneath the SVTB and is no longer externally visible. The rostral attachment site of the DVTB is a very thin lateral strip on the tympanum, running parallel to the central axis of the syrinx (Figure 8D,K).

The dorso-lateral musculature consists of three intrinsic syringeal muscles that consist of five different parts. Although these muscles run in close proximity to each other, they attach to separate syringeal bones and therefore actuate the syringeal skeleton differently. First, the musculus syringealis dorsalis or dorsal syringeal muscle consists of three parts: musculus syringealis dorsalis medialis or medial dorsal syringeal muscle (MDS), musculus syringealis dorsalis lateralis or lateral dorsal syringeal muscle (LDS) and musculus syringealis dorsalis profundus or deep dorsal syringeal muscle (DDS). The most medially located muscle, the MDS, attaches caudally on the medial dorsal cartilage that is embedded in the dorsal part of the ML (Figures 8G and 9E). Rostrally, the MDS ends in connective tissue, continuous with the CASM, and not on the tympanum directly. The LDS attaches caudally onto the ridge around the cup on the ventral end of bronchial half-ring B1. Both MDS and LDS run parallel to the central axis of the syrinx and are organized in a series of muscle sheets (Figure 9E). Their rostral attachment sites are not on the tympanum, but on the CASM near tracheal ring T2. The DDS is the only syringeal muscle that is not visible externally. The caudal attachment site of the DDS is interior to the attachment site of the LDS in the cup of half-ring B1 and the rostral attachment site is at the robust bony ridge on the tympanum running parallel to the central axis of the syrinx (Figure 8G,H).

Second, the relatively large musculus tracheobronchialis dorsalis or dorsal tracheobronchial muscle (DTB) attaches caudally in the cup at the dorsal end of half-ring B2, and the rostral attachment site is not on bone but on the CASM near tracheal ring T3. Third, the musculus tracheobronchialis brevis or short tracheobronchial muscle (STB) attaches caudally on the lateral edge of the dorsal portion of half-ring B2 just outside the ventral cup. Some STB fibers also insert directly into the lateral labium (Figure 9D). Moving rostrally, the STB runs from dorsal to ventral on the lateral surface of the tympanum. Its rostral attachment site is on the tympanum just dorsal to the attachment site of the DVTB (Figure 8D,H).

Two syringeal muscles have one attachment site outside of the syrinx and are therefore traditionally referred to as extrinsic (versus intrinsic) syringeal muscles. First, the musculus sternotrachealis or sternotracheal muscle (ST) attaches on a lateral portion of tracheal ring T1 (Figure 8D). The ST runs through the clavicular air sac and may therefore be a unique example of a vertebrate muscle running through air inside the body. Distally, the ST attaches on a dorsally oriented protrusion of the sternum (see below). Second, the musculus tracheolateralis or tracheolateral muscle (TL) attaches onto the CASM near tracheal ring T4 (Figure 9C) and its muscle mass runs along the trachea towards the larynx. We did not investigate where the TL muscles fibers attach rostrally. The length of the trachea greatly exceeds the length of regular muscle fibers, suggesting that the TL consists of serially connected fibers or that its fibers span only several tracheal rings.

In summary, we found that syringeal muscles are organized as intriguing parallel sheets of fibers. We redefined the intrinsic syringeal muscles based on precise localization of their attachment sites. These results permit us to conclude that closely apposed syringeal muscles potentially cause differential actuations via distinct attachment sites.

The two bilateral pairs of labia are the principal sound generators in songbirds 24263032. Our data show that two intrinsic syringeal muscles attach directly to cartilages that are embedded in the ML (Figure 10). Ventrally, the VS attaches to the MVC (Figure 8A) and, dorsally, the MDS attaches to the medial dorsal cartilage (Figure 10B). Because these cartilages are firmly embedded in the ML, our data suggest that, based on the orientation and location of the VS and MDS, shortening by either muscle would stretch the ML predominantly in the dorso-ventral axis. This would increase ML tension without changing ML ab- or adduction into the bronchial lumen (Figure 10C-E), which is consistent with earlier endoscopic observations during muscle micro-stimulation 23.

All other intrinsic syringeal muscles attach directly to bronchial half-rings B1 to B3 (Figure 11). Each half-ring has a parallel-running pair of muscles of which at least one muscle attaches to the tympanum and one more rostrally to the CASM. Two dorsal muscles actuate the bronchial half-ring B1: the LDS attaches to the tympanum and the DDS to the CASM (Figure 11A,B). Three muscles actuate the bronchial half-ring B2: on the dorsal side the parallel-running muscles DTB and STB (Figure 11C,D), of which the STB attaches to the tympanum and the DTB to the CASM. On the ventral side of the syrinx, some fibers of the VS attach to the ventral side of half-ring B2 (Figure 10A). Therefore both DTB and STB could act as functional antagonists to the VS, rotating half-ring B2 around a medial pivot (Figure 11C). Finally, two parts of the VTB actuate bronchial half-ring B3 of which the DVTB attaches to the tympanum and the SVTB to the CASM.

Our data support the hypothesis that shortening of the paired ST muscles can mechanically stabilize the syrinx during song 192274. Caudally to the syrinx, the primary bronchi are firmly anchored to the lungs, which are held in place against the vertebral column by a spiny intrusion of the second thoracic vertebra (Figure 12A). Micro-dissection shows that the syrinx is attached to this spine on the dorsal side by collagenous tissue. Furthermore, a thin, inflexible membrane, the interbronchial ligament (IBL), or bronchidesmus (Figures 8A and 9A), connects bronchial half-rings B4 and B5, thereby constraining rostral and lateral movement of the primary bronchi. These structures provide the caudal side of the syrinx with a rigid frame of reference relative to the spine (Figure 12A). In addition, on its dorsal side, the syrinx attaches to the esophagus and the vertebrae by collagenous sheets. We observed that ventral to the syrinx, the sternum protrudes into a Y-shape structure, the external spine (Figure 12). The ventral shape of the syrinx corresponds to the external spine (Figure 12B). Based on the orientation and attachment sites of the ST in situ (Figure 12C,D), our data suggest that co-activating the paired ST, leading to ST shortening, will pull the syrinx against the dorsal surface of the external spine, structurally stabilizing the syrinx relative to the skeleton.

Figures 2, 3 and 5 and Additional file 1 show that the structure of the syringeal skeleton is more complex than previously thought and inform us about the loads the skeleton experiences in vivo. The presence of hollow thin-walled bones in the syrinx (Figure 5) indicates an optimization of the skeleton through weight reduction and strength preservation 75. Hollow cylindrical bones are advantageous because the bone's weight decreases with decreasing wall thickness while being able to resist equal bending moments without breaking 75. In addition, lateral flattening of thin-walled bones further increases resistance to bending to perpendicular forces, because of increased second moment of area 7677. Thus bronchial half rings B1 to B3 provide high resistance to bending by forces (for example, F⊥ in Figure 5D) perpendicular to the horizontal axis of their cross-section (indicated by the plane in Figure 5D). In addition, we found strut-like elements, trabeculae, that internally connect inner and outer walls. Bones this thin do not break but rather fail due to abrupt wall deformation, also known as buckling 75, and trabeculae prevent buckling. In constantly remodeling bone tissue 70, trabeculae indicate the need to prevent mechanical failure due to buckling, and therefore predict the direction of in vivo loads on the bone. In conclusion, the laterally flattened thin-walled hollow structure of the ossified bronchial half-rings of the zebra finch syrinx indicates that these bones are optimized for low weight while maintaining strength and are subjected to perpendicular loads in vivo 757677.

The numerous trabeculae, especially in B1 and B3 (Figure 5B inset), prevent bending deformation, and as such strongly argue against bending of B1 and B2 as a mechanism for sound frequency modulation 43. Their presence also indicates that high perpendicular forces are present. These can be attributed to extreme acceleration by superfast syringeal muscles, which can move syringeal elements cyclically up to 200 Hz and 250 Hz in zebra finch and starling males, respectively 36. The absence of trabeculae in B2 suggests that the bending forces are lower here, which is most likely due to much slower force modulation by syringeal muscles.

We propose that the weight reduction of the syringeal skeleton and particularly of the bronchial half-rings B1, B2 and B3 is predominantly a result of physiological constraints on song production. In several songbirds species, females prefer song notes consisting of fast trills 787980818283, suggestive of a selective pressure for the evolution of faster song elements that require faster movement of syringeal structures. The superfast muscles found in the songbird syrinx are pushed to the performance limits of vertebrate skeletal muscle so that their work output is 10 to 20 times lower than regular skeletal muscle 364984 and a distinct trade-off exists between the work they can generate and the maximum frequency at which they can cyclically actuate objects 498586. Compared with the weight of a songbird (ranging from about 4.5 g (the goldcrest Regulus regulus) to 100 grams), the weight reduction of filled versus hollow syringeal bones (in the order of milligrams) must have very little effect on flight energetics. Lighter syringeal bones, however, can be moved at higher maximum speeds, enabling faster modulation and more precise temporal control of acoustic song features. Sexual selection pressure on syringeal performance therefore imposes additional costs on the syringeal skeleton, forcing it to adapt to avoid bone fracture.

We suggest that the medial ventral and medial dorsal cartilages embedded in the vibrating medial labia (Figures 7 and 10) are crucial to prevent tissue damage. They sum and transfer the forces produced by muscles to the labia in a graded manner. Instead, if the syringeal muscles VS and MDS (Figure 10) inserted directly on the ML, differential motor unit recruitment would result in large local strain discontinuities at the apex between two adjacent muscle fibers if only one fiber contracts. Without these cartilages distributing the applied forces evenly to the labia, the chance of severe tissue damage to the ML would be increased. The cartilages diffuse such large local strain discontinuities, a prerequisite for smooth modulation of tension.

Our data support the contention that the MVC of the syrinx also plays a crucial role in the mechanism to control sound frequency (Figure 10). Songbirds produce sound by flow-induced oscillations of the medial and lateral syringeal labia 24263032, which generate tracheal pressure fluctuations radiated as sound. Sound frequency seems to be controlled predominantly by muscles effecting tension changes in the medial, but not the lateral labium: VS activity correlates positively with fundamental frequency during song 21, and pulling on the MDS increases sound frequency in vitro 32. By contrast, the amplitude (and gating) of sound appears to be predominantly controlled by muscles affecting the position of the lateral labium 222336. Our data therefore support the hypothesis that the primary mechanism controlling sound frequency by modulating ML tension is the bending of the flexible cartilaginous extension of bronchial half-ring B2, the MVC (Figure 10). Because bones easily break or buckle when bent, deformation of the half-rings B1 and B2 themselves 43 is not a suitable mechanism to modulate ML tension.

This MVC bending mechanism leads to two further hypotheses. First, energy stored in temporary elastic deformation of the MVC by VS shortening will act as a VS antagonist. Vertebrate striated muscle fibers need to be stretched back to resting length after contraction, which is often accomplished by contraction of an antagonistic muscle. In the zebra finch syrinx, release of the elastic energy stored within the MVC by VS contraction would restore the resting length of VS fibers and also the resting tension in the ML. Secondly, the MVC bending mechanism will enable the songbird syrinx to uncouple control of ML tension and position. In the mammalian larynx, tension in the vocal folds determines their oscillation frequency and, hence, the sound frequency produced, while the sound amplitude is determined by a combination of bronchial pressure and the amount of adduction of the vocal folds into the laryngeal lumen 8788. In two songbird species, the brown thrasher (Toxostoma rufum (Linnaeus, 1758)) and the northern cardinal (Cardinalis cardinalis (Linnaeus, 1758)), micro-stimulation of the VS bends the MVC and increases tension in the ML dorso-ventrally, but does not abduct the ML into the bronchial lumen 23. We suggest that bronchial pressure (a result of respiratory muscle action) and the position of the lateral labium predominantly control sound amplitude in songbirds, whereas the tension in the ML locks the oscillation frequency for both labia and hence sets sound frequency. As such, the MVC is in a prime position to allow for the uncoupling of sound frequency and amplitude.

This uncoupling function of the MVC may have been key to the evolutionary success of songbirds or oscines (Passeriformes: Passeri). They first appeared about 55 million years ago 89 and now form the largest radiation within birds, including almost half of all 10,000 extant bird species 90. Song has played a crucial role in the evolution of songbirds, driving sexual selection both intersexually (females choose mates based on song quality and male-male contests) and intrasexually (males use song to establish and defend territorial boundaries). Our comparative knowledge on song production is very limited. Most of our understanding of the physiology and different mechanisms of song production comes from a handful of species. The songbirds alone produce a wide variety of song, ranging from almost tonal to broadband sounds. This range of sounds might reflect differences in production mechanisms, syringeal tissue properties and morphology, as well as differential recruitment of the syringeal musculature.

Compared with the huge variation in syringeal morphologies found in non-songbirds, the morphology of the songbird syrinx is quite conserved 161920. Gaunt previously proposed that more elaborate intrinsic muscles in songbirds allowed for independent control of song parameters 20. Field recordings 91, excised preparations 92, mechanical models 9394 and numerical models of sound production in non-songbirds, such as ringdoves (Streptopelia risoria) 95, and sub-oscines (Passeriformes: Tyranni), such as the great kiskadee (Pitangus sulphuratus) 96, show that sound amplitude and frequency are often coupled to some degree, thereby severely limiting the potential repertoire of syllables that can be produced physically. To our knowledge, none of these species appears to have evolved a structure comparable to the MVC found in the songbird syrinx that could allow for tension control without affecting position. We speculate that the MVC contributed to the extensive diversification of songbirds because they evolved a syringeal morphology that allowed for largely independent control of sound parameters, opening up a wealth of song syllables not previously available.

The morphome data (Figure 8 and Additional file 2) reveal that closely apposed muscles attach to different syringeal elements, which predicts distinct mechanical effects when recruited. In particular the dorsal musculature superficially looks like a continuous mass of muscle tissue (Figures 8 and 9, and Additional file 2), but our detailed analysis clearly demonstrates that spatially closely located fibers attach to (and therefore actuate) the ML, bronchial half-rings B1 and B2, and attach to six different locations on the tympanum and higher up the trachea in the CASM (Figures 8 and 9). The resulting subtle, but distinct, subdivision of syringeal muscles has significant implications for experiments addressing the function of specific muscles, because recording the activity or the individual stimulation of such physically closely located muscles is very sensitive to electrode design. The twisted fine-wire electrodes commonly used in syringeal electromyography (EMG) studies 213797 and micro-stimulation studies 233661 function as antennas that detect or cause cell membrane depolarization of many fibers and could easily cross boundaries of muscles with distinct biomechanical effects. The EMG signal strength depends upon wire diameter, impedance, length of stripped end and fiber size 98. Because this effect is scale-dependent, using a species with a large syrinx partially circumvents this problem 212223. Nevertheless, one must be cautious when measuring and interpreting EMG recordings and micro-stimulation from a syrinx of a bird the size of the zebra finch.

Our in situ data (Figure 12) show that the syrinx can be lodged into the Y-shaped external spine of the sternum and therefore be stabilized with respect to the skeleton. This stabilization could reduce the amount of mechanical noise affecting the sound-producing labia, thereby augmenting the production of stereotyped songs. Several sources of mechanical noise are present during song. The beating of the heart and its major arteries provide rhythmic mechanical perturbation close to the labia and primary bronchi (Figures 1 and 12). During song, postural changes of the body due to courtship dances 99100 affect the position of the trachea. To facilitate spectral changes of sound radiation, tracheal length is modulated by TL contraction 101 and changes in position of larynx and hyoid 3840. Additionally, the pressure of expired air from the lungs during vocalization can push the syrinx rostrally 23, and fast length modulations of the syringeal muscles can send vibrations up the trachea 74. Co-contraction of the STs would press the syrinx onto the rigid framework that is composed of sternum, ribcage and spine.

EMG data support the idea that the ST is active during song 34.The ST has also been hypothesized to function as an adductor of the syringeal labia 99 and as an antagonist to the TL in many non-songbirds 95102103104105106107108. The ST muscle is present in all songbirds, and is missing only in very few bird species that are either non-vocal or that produce song with little frequency modulation (for example, the new world vultures 109110 or Darwin's nothura (Nothura darwinii 111). Smith 109 reported that surgical lesioning of the ST did not have much effect on acoustic song parameters in several songbird species. However, due to the very low sensitivity of the analysis possible at that time, this conclusion needs to be reassessed using modern quantification of sound parameters 5456. An important role for the ST in fine-tuning sound production is also implied, because of the high energetic costs 49 associated with superfast contractile properties as even found in ringdoves 97. Furthermore, an activated ST can absorb vibration energy along the trachea by producing negative work when longitudinal vibrations are sent along the trachea due to fast shortening of intrinsic syringeal muscles 74.

The present refinement of muscle identity (Figure 8 and Additional file 2) also allows for more precise predictions about muscle co-activation known to increase accuracy and stability of position control in many complex motor systems such as arm reaching 112113114. We currently do not know the recruitment patterns of all syringeal muscles simultaneously during song, but EMG recordings can demonstrate strong correlation between muscles 21. With antagonistic control of a specific parameter, such as lateral labium position, different levels of recruitment of at least four relevant muscles (DTB, STB, DVTB, SVTB; Figure 11) could produce the same position of the lateral labium. Therefore different motor programs may result in similar acoustic features and, conversely, combinations of muscle activation, or synergies 656667, might be required to achieve control values. If muscles are activated synergistically, it is not individual muscle activation but the underlying synergy that functionally correlates to sound parameters. This would imply that we should not aim to correlate sound parameters to individual muscle activity, but to underlying muscle synergies instead.

The morphome presented here provides a crucial foundation for mapping motor projections of the song system onto the syringeal muscles and to study how syringeal muscle recruitment shapes sound production. To ultimately assess muscle function during song production, muscle stimulation and perturbation experiments in controlled environments and during song need to be performed. Such studies will ultimately link syringeal biomechanics (that is, the determination of pivoting points and mechanical properties of the bronchial half-rings, their mechanical coupling and kinetics) to sound and song parameters. In fact, it is impossible to meaningfully connect the song system at the level of hindbrain motoneurons to the syrinx without a thorough understanding of the number, orientation and function of syringeal muscles. Additionally, such a motor map will allow the translation of commonly used sound parameters 535456 to neural firing patterns. The syringeal geometry presented here provides a quantitative basis for implementing realistic 3D morphologies into computational models of sound production. The parallels between birdsong and human speech observed at the behavioral, neural 210115 and genetic level 116 are incomplete until the peripheral contribution to human speech and birdsong are understood at a similar level of detail. To grasp how central motor patterns are translated into vocal communication, it is important to understand peripheral mechanisms of sound production across the vocal vertebrates.

Male and female zebra finches (T. guttata (Vieillot, 1817); Aves: Passeriformes: Estrildidae) were kept in group aviaries at the Freie Universität, Berlin, Germany and the University of Southern Denmark, Odense, Denmark on a 12 h light:dark photoperiod and given water and food ad libitum. All procedures were carried out in accordance with the Landesamt für Gesundheit und Soziales (Berlin, Germany) and the Danish Animal Experiments Inspectorate (Copenhagen, Denmark).

Zebra finches (six males, six females) were deeply anesthetized with isoflurane and transcardially perfused with heparinized 0.1 M PBS followed by 4% phosphate-buffered paraformaldehyde (PFA). We performed MRI scans of whole animals (one male: one female) at the Institut für Klinische Radiologie in Münster, Germany using a Bruker Biospec 94/20, 9.4 T, high-field, horizontal-bore, small animal scanner (Bruker BioSpin GmbH, Ettlingen, Germany). MRI scans of dissected syrinxes were performed at the Institut für Experimentelle Physik 5 in Würzburg, Germany using a Bruker Avance 750WB, 17.6 T, high-field, vertical-bore, small animal NMR scanner equipped for imaging (Bruker BioSpin GmbH). Intact animals were scanned in PBS with 2 mM gadopentetate dimeglumine (Magnevist; BayerSchering GmbH, Berlin, Germany) as contrast agent. We used the following scanning parameters: Turbo RARE 3D protocol, 680 × 296 × 256 pixel matrix size, 6.5 × 2.85 × 2.5 cm field of view, 96 × 96 × 98 μm voxel resolution, T_R = 1500 ms, T_E = 14.7 ms, eight averages, RARE factor 16, and 15 h 21 min 36 s acquisition time. We additionally scanned the head of one male zebra finch using the following parameters: Turbo RARE 3D protocol, 288 × 288 × 256 pixel matrix size, 25.88 × 25.88 × 23.00 mm field of view, 90 μm isotropic voxel resolution, T_R = 1500 ms, T_E = 9.97 ms, eight averages, RARE factor 16, and 15 h 21 min 36 s acquisition time.

In addition, we scanned excised syrinxes (five males, five females). After fixation as described above, syrinxes were removed, post-fixed in 4% PFA overnight, and stored at -20°C in 30% sucrose with 0.01% sodium azide. For mechanical stabilization, we embedded the preparations into low-melting agarose (Carl Roth, Karlsruhe, Germany) with 0.01% sodium azide and 2 mM Magnevist as contrast agent. We then placed them inside 15 ml Falcon tubes (15 mm diameter) for scanning using the following scanning parameters: FLASH 3D protocol, 256 × 290 × 290 to 512 × 512 × 512 pixel matrix size, 1.2 × 1.2 × 1.2 to 1.4 × 1.4 × 1.4 cm field of view, 23 to 46 μm isotropic voxel resolution, T_R = 20 to 22 ms, T_E = 3.3 to 5.4 ms, 9 to 32 averages, and 3 h 38 min 27 s to 18 h 0 min 32 s acquisition time.

All MRI datasets were analyzed using the Volume Viewer plugin in ImageJ (NIH, Bethesda, MD, USA).

We performed whole animal μCT scans (one male, three females) at the Museum für Naturkunde (Berlin, Germany) using a Phoenix Nanotom X-ray tube tomography system with a 2,300 × 2,300 pixel detector (GE Sensing & Inspection Technologies, Wunstorf, Germany). Animals were overdosed with isoflurane and placed inside air-filled 50 ml Falcon tubes after respiration had stopped for 5 min. The following scanning parameters were used: 150 to 180 kV source voltage, 145 to 270 μA source current, no filter, 750 ms exposure time, 13 to 22 μm isotropic voxel resolution, 1,440 angular steps over 360° with three averaged images per rotation position, and about 54 min scan time.

We also made scans of isolated syrinxes (12 males, 5 females) to increase scan resolution with respect to whole animal scans. These scans were performed at the Center for Nanoscale Systems (Cambridge, MA, USA) using a MCT 225 X-ray tube tomography system with a 2,000 × 2,000 pixel detector (Nikon Metrology, Leuven, Belgium). Both μCT scanners used in this study were equipped with a tungsten target. Specimens were overdosed with isoflurane and the syrinxes were dissected with a large portion of esophagus, trachea and lungs attached to maintain structural integrity. We took special care to mount the preparations in their natural position onto 3 mm thick silicone sheets using 0.1 mm diameter stainless steel Austerlitz insect pins (Fine Science Tools, Heidelberg, Germany). The samples were fixed for seven days in 4% PFA in PBS stored at 4°C. After removal of the needles, the samples were transferred to 1 ml pipette tips filled with PBS and stored at 4°C until scanning.

We made two scans of each preparation. First, we optimized scanning parameters to assess the anatomy of the ossified syringeal skeleton. We scanned the samples using the following parameters: 85 kV source voltage, 110 μA source current, 0.1 mm copper filter, 2 s exposure time, 5 μm isotropic voxel resolution, 1,440 angular steps over 360° with two averaged images per rotation position, and about 1 h 36 min scan time. Second, we added contrast agent and optimized scanning parameters to image soft tissues. We stained the samples inside glass vials with 0.1% Lugol's solution (aqueous I₂/KI, Sigma Aldrich, St. Louis, MO, USA) and placed them on a tube roller for 48 h following previous protocols 117118119. Prior to scanning, the samples were rinsed in distilled water for 10 min twice and then placed inside 1 ml pipette tips with distilled water. We scanned the samples using the following parameters: 85 kV source voltage, 110 μA source current, 0.1 mm copper filter, 2 s exposure time, 5 μm isotropic voxel resolution, 1,440 angular steps over 360° with two averaged images per rotation position, and about 1 h 36 min scan time. Dataset reconstruction was performed using the software provided with each scanner, that is, Phoenix DatosX Reconstruction 1.5 for the GE system and Metris XT 2.2 for the Nikon system. Reconstruction was performed with noise reduction activated, but without binning.

We quantified shrinkage by measuring the distance between the IBL and the top of tracheal ring T2 from photographs of the preparations taken immediately after dissection and after seven days of PFA fixation. Shrinkage averaged 1.3% and was less than 4.8% in all preparations.

We performed micro-dissection (six males, five females) after euthanasia by isoflurane overdose. The syrinxes were exposed by cutting through the pectoral muscles and sternum. They were carefully isolated and pinned down on Sylgard-covered Petri dishes in oxygenated Ringers solution (NaCl, 154 mM; glucose, 12 mM; KCl, 6 mM; MgCl₂, 1 mM; NaH₂PO₄, 1 mM; MgSO₄, 1 mM; HEPES, 10 mM; CaCl₂, 4 mM; adjusted to pH 7.4 with Trizma Base) at room temperature (18°C). Previous muscle mechanics experiments on syringeal muscles have shown that the muscles retain contractile abilities for up to 8 hours under these conditions 367497. The syrinx was separated medially into left and right hemisyrinxes that were separately stored in oxygenated Ringers solution either on icepacks or at room temperature (18°C). We studied syringeal tissues by careful micro-dissection under a M165-FC stereomicroscope (Leica Microsystems, Wetzlar, Germany) using a CLS-150XE light source (Leica Microsystems). Images were taken with a DFC425-F camera (Leica Microsystems). Muscle fibers were visualized using variable lighting conditions to achieve large contrast. Attachment sites of each muscle were photographed and noted on various 2D projections of previously generated μCT-based 3D volume renderings of the syringeal skeleton.

For histological examinations, we perfusion-fixed the syrinxes of six animals (four males, two females) in 4% PFA and decalcified them in a mixture of 4% PFA/ethylenediaminetetraacetic acid for one week. We then dehydrated them in an ascending alcohol series of 30%, 50% and 70% ethanol for 10 min per step on an automatic shaker to facilitate tissue penetration. The ethanol was gradually replaced with paraffin using a TP 1020 tissue processor (Leica Microsystems). Finally, syrinxes were embedded in melted paraffin in an aluminum tube covered with a perforated plastic cap. The embedded syrinxes were cooled down overnight to about -20°C. A rotary microtome (Microm HM 340e, Thermo Scientific, Billerica, MA, USA) was used to section the syrinxes into 10 μm thick sections. Sections were stained using Mayer's hematoxylin and eosin staining protocol and examined under a MDG30 microscope (Leica Microsystems). A DFC420C camera (Leica Microsystems) was used for imaging.

Out of all the μCT-based syringeal datasets, we chose one representative male and female specimen, based on optimal tissue contrast, presence of all structures and correspondence to in vivo orientation. All recognized muscles, muscle attachment sites, cartilaginous pads and extensions, and medial and lateral labia were annotated onto this dataset for the male and female syrinx. Over 10 different systems of muscle nomenclature exist to describe syringeal muscles in songbirds 19. Ames 19 studied the syrinx of 533 songbird species and concluded that the different nomenclatural systems were divided by two main issues: '1) whether muscles should be named based on the basis of function or on location, and 2) if a single mass of muscle fibers attached to several cartilaginous elements constitutes one muscle or more'. Because of insufficient data on syringeal muscle function, we decided to use a combined approach to nomenclature, based first on location (sensu Owen 120). Secondly, working towards a functional nomenclature, we considered distinct attachment sites to define muscles, because the recruitment of muscles with different attachment sites is the basis of differential actuation of syringeal elements, which in turn will lead to differential biomechanical effects on sound production.

The syrinx morphome was annotated in consensus, based on comparative analysis of histology, μCT scan projections, 3D models and micro-dissection and additional micro-dissections of two males and two females at the University of Southern Denmark, Odense.

3D volume and surface rendering and visualization were performed using myVGL 2.1 (Volume Graphics GmbH, Heidelberg, Germany), Amira 5.2 (Visage Imaging GmbH, Berlin, Germany), and ImageJ and its Volume Viewer plugin (NIH). All morphome annotation and data analyses were performed using Amira. The interactive 3D PDF models were created using Adobe Acrobat 9.4 and Adobe 3D Reviewer 9.4 (Adobe Systems, San Jose, CA, USA) following procedures described elsewhere 6869.