Ventilatory mechanics from maniraptoran theropods to extant birds


Jonathan Codd, Faculty of Life Sciences, University of Manchester, Michael Smith Building, Oxford Rd, Manchester, M13 9PT, UK.
Tel: +44 161 275 5474; Fax: +44 161 275 3938;


Shared behavioural, morphological and physiological characteristics are indicative of the evolution of extant birds from nonavian maniraptoran dinosaurs. One such shared character is the presence of uncinate processes and respiratory structures in extant birds. Recent research has suggested a respiratory role for these processes found in oviraptorid and dromaeosaurid dinosaurs. By measuring the geometry of fossil rib cage morphology, we demonstrate that the mechanical advantage, conferred by uncinate processes, for movements of the ribs in the oviraptorid theropod dinosaur, Citipati osmolskae, basal avialan species Zhongjianornis yangi, Confuciusornis sanctus and the more derived ornithurine Yixianornis grabaui, is of the same magnitude as found in extant birds. These skeletal characteristics provide further evidence of a flow-through respiratory system in nonavian theropod dinosaurs and basal avialans, and indicate that uncinate processes are a key adaptation facilitating the ventilation of a lung air sac system that diverged earlier than extant birds.


The origin of Aves is one of the great transitions in vertebrate evolution. The discovery of Archaeopteryx lithographica provided Thomas Huxley with evidence to support his theory that birds were the direct ancestors of dinosaurs (Huxley, 1868). Current evidence such as the fossilized remains of feathered dinosaurs (Norell & Xu, 2005), evidence of brooding behaviour (Norell et al., 1995), osteological traits (Padian & Chiappe, 1998; Codd et al., 2008) and histological analysis (Chinsamy et al., 1995; Erickson et al., 2007, 2009) indicate that the origin of Aves is to be found in nonavian maniraptoran theropods. Therefore, by studying the anatomy and physiology of modern birds, we can begin to understand evolution from theropod ancestors and the evolutionary pressures that have shaped Aves.

Modern birds possess a highly derived respiratory system that consists of air sacs connected to a rigid, dorsally fixed lung (Powell, 2000). Caudoventral rotation of the ribs together with ventral excursion of the sternum facilitates inspiration. Upon expiration, the ribs move cranio-dorsally, causing a dorsal displacement of the sternum (Zimmer, 1935; Claessens, 2009). This reduction in abdominal air sac volume facilitates the unidirectional flow of air across the parenchymal tissue, where gas exchange takes place. Movements of the sternum that are normally referred to as scissor-like (Kardong, 2006) have been identified as elliptical in three basal bird species (Claessens, 2009). Regardless of the sternum’s pathway during respiration, it is clear that its displacement is accomplished by an increase in the angle between sternal and vertebral ribs. This increased angle causes a ventral displacement of the distal portion of the sternal ribs. As the caudal sternal ribs are elongate compared to the cranial sternal ribs, the caudal sternal margin is moved further ventrally than the cranial margin (Claessens, 2009). Vertebral ribs in all extant birds except the screamers (Anhimidae) have ossified uncinate processes (UP) that extend caudodorsally. The Mm. appendicocostales project from the proximal margin of the UP and insert onto the following vertebral ribs (Schufeldt, 1890; see figure 1 Codd et al., 2005). Codd et al. (2005) demonstrated that UP facilitate inspiration and expiration. Furthermore, a geometric model of the thoracic skeleton in extant birds has also shown that UP function as levers by improving the mechanical advantage of rib and therefore sternal movement during respiration (Tickle et al., 2007). Variations in process length do occur; diving birds have the longest, flying and swimming birds have intermediate, and walking/running species have the shortest UP (Tickle et al., 2007). Exceptions to this general pattern of morphology do occur in some species, perhaps to assist functions such as egg laying and vocalizations. The kiwi (Apteryx haastii), a flightless species, has UP that are unusually broad at the base and are longer than would be expected in other walking birds. Coupled with the generally flattened profile of kiwi vertebral ribs, the UP may assist egg laying (Codd, 2010) because the eggs of this species are the largest of any bird relative to body size (Colbourne, 2002). Uncinate process morphology appears to be fine-tuned and related to differences in the sternum and rib cage associated with adaptations to different forms of locomotion in birds (Codd, 2010). Process length also scales with metabolic rate in birds, being longer in species with a relatively high metabolic rate (Tickle et al., 2009).

Understanding the evolution of respiratory systems is complicated, as soft tissues such as the lungs are rarely preserved in the fossil record. Assuming that characteristics are shared in basal and derived species in a clade, evolution of the respiratory apparatus can be predicted according to the known ancestry of taxa using extant phylogenetic bracketing (Witmer, 1995; Perry & Sander, 2004). Theropod dinosaurs are phylogenetically bracketed by birds and crocodiles (Sereno, 1997), meaning characters common to both can be assumed to have occurred in the most recent common ancestor. The respiratory system of both birds and crocodiles is characterized by having a heterogeneous lung with an asymmetric branching pattern and it follows that theropod dinosaurs had a similar respiratory anatomy (Perry & Sander, 2004). Furthermore, similar to the pattern of respiratory airflow found in birds (Hazelhoff, 1951), unidirectional flow was recently identified in the crocodilian lung (Farmer & Sanders, 2010), indicating that this condition was present in the most recent common ancestor of birds and crocodilians. The structure of soft tissues can also be predicted according to the osteology of extinct species (Perry & Sander, 2004). For example, pneumatized vertebrae and ribs, and in some cases axial bones, are indicative of a heterogeneous lung structure and a highly compliant respiratory system (Perry, 2001), and have been used to indicate the presence of flow-through ventilation in pterosaurs (Claessens et al., 2009) and theropods (O’Connor & Claessens, 2005).

Reconstructing the movements of the thoracic skeleton using an analysis of the osteological features in fossils and application of biomechanical principles further adds to our understanding of the mechanics of ventilation in extinct species. Well-developed ossified UP (Codd et al., 2008) and gastralia (Claessens, 2004) have been described in several nonavian maniraptoran dinosaurs. For example, in Citipati osmolskae and Velociraptor mongoliensis these processes are long and thin, meaning that they most closely resemble those of flying and diving birds (Codd et al., 2008). Citipati osmolskae and V. mongoliensis are also characterized by the presence of ossified gastralia. Gastralia are a series of narrow cranially pointing chevrons that are formed by the association of slender bony elements angling laterally and caudally from the midline (Carrier & Farmer, 2000a,b). Retraction and protraction of the gastralia have been suggested to act as an accessory aspiration pump in dinosaurs and primitive archosaurs (Carrier & Farmer, 2000a,b; Claessens, 2004; O’Connor & Claessens, 2005; Schachner et al., 2009). This model of ‘cuirassal breathing’ was updated (Codd et al., 2008) to introduce a mechanism whereby UP facilitated respiratory movements in nonavian maniraptoran dinosaurs. Movements of uncinate process–associated musculature coupled with the narrowing and widening of the gastralia by the ischiotruncus and rectus abdominus abdominal muscles may therefore have contributed to the airflow within a maniraptoran air sac respiratory system (Codd et al., 2008). Support for this idea comes from the tuatara (Sphenodon spp.), a primitive lepidosaur and the most basal extant amniote clade to possess ribs actively involved in ventilation (Perry et al., 2010). Projecting from the ossified parts of the vertebral ribs of the tuatara are elongate cartilaginous UP connected to ossified gastralia by the external oblique muscle. Crocodiles also have cartilaginous UP and ossified gastralia (Hofstetter & Gasc, 1969). The role of UP and gastralia in crocodilians is intriguing given unidirectional airflow has been confirmed in both avian and crocodilian lungs (Farmer & Sanders, 2010). Therefore, in addition to pelvic aspiration, facilitated by a rotation of the pubes and displacement of the viscera by shortening of the diaphragmaticus muscle (Gans & Clark, 1976; Carrier & Farmer, 2000a,b; Farmer & Carrier, 2000; Claessens, 2004), it is possible that UP and the gastralia assist costal breathing in crocodilians.

Understanding the evolution of the highly derived avian respiratory system requires an idea of how basal morphologies may have functioned. The geometric model of UP function (Tickle et al., 2007) provides a method that can be used to estimate the efficacy of the UP for movement of the ribs, sternum and/or gastral basket. Here we apply this geometric model to a selection of taxa that are phylogenetically basal to extant birds and exhibit a mosaic of primitive and modern avian characteristics. A better understanding of the ventilatory mechanics in extinct birds will provide insight into the evolution of the highly specialized avian respiratory system. We hypothesize that irrespective of morphology, the UP improve the mechanical advantage for rib and sternal movements. Furthermore, we hypothesize that difference in process morphology is tuned to morphological differences in thoracic anatomy.

Materials and methods

The preserved thoracic skeletons of C. sanctus Senckenberg Forschunginstitut und Naturmuseum (SMF) Av 418, 419, 421, 422, 525 and C. osmolskae Institute of Geology, Mongolia (IGM) 1004 were photographed using a digital camera and analysed using Leica® application suite software (Leica Microsystems, Milton Keynes, UK) to calculate the mechanical advantage provided by the UP for ventilatory movements of the ribs [using the model of Tickle et al. (2007)]. Images of The Institute for Vertebrate Paleontology and Paleoanthropology (IVPP) 15900 Zhongjianornis yangi, an avian species that is phylogenetically basal to C. sanctus (see Fig. 4a, Zhou & Li, 2009), and IVPP 13631 Yixianornis grabaui, an early ornithurine specimen (see Fig. 4 in Clarke et al., 2006) were taken from the literature for comparison. The fossils used in this study have excellent preservation of the ribs and UP making possible an accurate calculation of mechanical advantage. The model of Tickle et al. (2007) uses detailed measurements of thoracic geometry to estimate whether the UP increase leverage for inspiratory muscles to facilitate cranial rotation of the ribs during inspiration. Key measured parameters in this model are the distance between vertebral ribs at the backbone, angle between vertebral rib and backbone, and the location on the rib of Mm. appendicocostales insertion (Fig. 1). Using the equations derived in Tickle et al. (2007), the effect of UP is described by the relative change in length of Mm. appendicocostales per unit change in angle between rib and backbone. Because the muscle architecture is not preserved in the specimens examined in this study, a selection of extant species (diving: razorbill, Alca torda; nonspecialist: barnacle goose, Branta leucopsis, and kestrel, Falco tinnunculus; walking: red-legged partridge, Alectoris rufa) were also used for comparison. These birds were dissected and the mean angle at which Mm. appendicocostales fibres projected from the UP recorded. This angle was then applied during computer analysis to the tip of the fossil UP as it affects the predicted point of insertion of the muscle onto the following rib. The predicted location of muscle insertion and values measured directly from photographs were then used to calculate the mechanical advantage (described as the change in mechanical advantage with and without an UP) using the equations of Tickle et al. (2007).

Figure 1.

 Schematic diagram of the ribcage and uncinate process system. Measurements were taken of the rib angle to the backbone (θ), distance between vertebral ribs (D), length of the uncinate process and location of Mm. appendicocostales (dashed arrow) insertion on the next most caudal vertebral rib. To derive the mechanical advantage, or leverage, of the UP, these parameters were entered into the geometric model of Tickle et al. (2007). Cranial is to the right of the diagram.

The geometric model requires an estimation of muscle length (which depends upon the distance between ribs and the projection angle of the process); therefore, data were only used from fossil forms in which the UP remained attached to the vertebral rib. To minimize inaccuracy in estimated muscle length arising from relative rib position, data were also only collected from UP where the rib from which they extended and the following caudal rib had similar projection angles from the backbone (Table 1). Only the preserved ribs and processes were numbered (i.e. missing bones were not taken into account when assigning a number), and this was performed according to their position beginning at the cranial end. Mean values for mechanical advantage were calculated based on the measurements from three C. sanctus specimens (SMF Av 419, 421 and 422), whereas single fossils of C. osmolskae, Y. grabaui and Z. yangi were analysed for comparison.

Table 1.   Measurements of uncinate process and rib morphology in three basal bird species, Confuciusornis sanctus, Yixianornis grabaui and Zhongjianornis yangi and a maniraptoran theropod, Citipati osmolskae. Estimated improvement in the mechanical advantage imparted by UP for movements of the ribs is shown.
SpeciesSpecimenUP locationUPUP length (mm)UP base width (mm)Rib length (mm)Angle of parent ribAngle of insertion rib% improvement of M.A.
C. sanctusSMF Av 418Left-hand side14.301.5812.24Rib is displacedRib is displaced
24.661.6227.77Rib is displacedRib is displaced
36.651.4327.25Rib is displacedRib is displaced
Right-hand side15.111.27Process is displacedRib is displacedRib is displaced
24.281.11Process is displacedRib is displacedRib is displaced
SMF Av 419Right-hand side1Obscured1.0910.39Obscured55.1°
SMF Av 421Left-hand side1Obscured1.8926.2850.5°49.6°
Right-hand side16.761.1535.9151.9°48.9°235
SMF Av 422Right-hand side112.661.1531.2673.3°74.3°127
Y. grabauiIVPP 13631Left-hand side17.582.13Obscured56.0°51.5°244
Z. yangiIVPP 15900Left-hand side27.092.66Obscured61.3°56.2°234
C. osmolskaeIGM 1004Right-hand side183.8921.77184.7083.1°79.8°274


Thoracic morphology of Confuciusornis sanctus

Detailed examination of the Senckenberg C. sanctus specimens revealed striking similarities with modern avian ribcage morphology, in accordance with previous accounts of Confuciusornis (e.g. Chiappe et al., 1999). UP are clearly present in specimens SMF Av 418, 419, 421 and 422. The characteristic distended base morphology of UP in extant birds is also present in C. sanctus (Fig. 2). Furthermore, the fossil UP taper along the length of the shaft and fully span the intercostal spaces, overlapping the lateral face of the following rib (Fig. 2). The excellent ribcage preservation in SMF Av 421 demonstrates that UP occur on at least five successive vertebral ribs, which bears a strong similarity to modern bird anatomy. In addition, vertebral ribs that bear an uncinate process articulate with the remains of an ossified sternum in two specimens (SMF Av 419 and 421: Fig. 2). Interestingly, although preservation of the ribcage is limited and UP are absent from the fossil (consequently, application of the geometric model is impractical), a keeled sternum is present in specimen SMF Av 525 (previously described as SMF Av 423 by Peters & Qiang, 1999) (Fig. 3). Although the depth of the keel is reduced compared to those in the majority of modern birds, it may have been extended by a cartilaginous outgrowth as often seen in extant species (Padian, 1983). In addition to the ossified sternum in C. sanctus, caudally projecting gastralia are present in the abdominal region of specimens SMF Av 418 and 525 (Fig. 3). Preserved medial gastral elements form a zigzag pattern in accordance with the description provided by Chiappe et al. (1999), whereby the gastralia on the left-hand side adjoin and slightly overlap the corresponding right-side bones, which extend to the preceding (i.e. more cranial) left element.

Figure 2.

 Comparison of UP in Confuciusornis sanctus (SMF Av 421: a; SMF Av 422: c) and a modern bird, the barnacle goose (Branta leucopsis): (b). Scale bars, 1 cm.

Figure 3.

 Thoracic morphology in Confuciusornis sanctus SMF Av 525. An ossified sternum (highlighted by the dashed line) that has a nascent keel is well preserved. There is an area of probable gastral elements (dashed circle) in the abdominal region caudal to the sternum. Scale bar, 1 cm.


Phylogenetic distribution of uncinate processes

The characteristic morphology of the UP in C. osmolskae, Z. yangi (Zhou & Li, 2009), Y. grabaui (Clarke et al., 2006) and C. sanctus is similar to that found in extant flying and swimming birds (Codd et al., 2005). Chiappe et al. (1999) has previously documented long caudodorsally extending UP in several other C. sanctus specimens (GMV-2130, GMV-2146, GMV-2147 and GMV-2149). Therefore, the reported lack of UP in some C. sanctus specimens (Hou et al., 1996; Peters, 1996) seems most likely to be a preservational artefact. Indeed, the seemingly common disarticulation and loss of UP from fossil specimens indicate that these elements were connected to the ribs by a weak cartilaginous attachment and therefore easily lost during preservation or fossil extraction (Codd et al., 2008). Ossification and fusion of UP and ribs are dependant upon developmental stage (Maxwell, 2008; Maxwell & Larsson, 2009; Tickle & Codd, 2009) and phylogeny [UP in extant birds directly ossify to the rib except in Apteryx spp. and Sphenisciformes, where processes are connected via a chondric attachment (Codd, 2010)]. It is worth noting that a cartilaginously attached process may still be functional; in extant birds, the processes are attached to the rib from which they extend by a strong aponeurotic membrane (Tickle et al., 2007), which may stiffen the process enough to allow it to be effective as a lever even if it is not ossified to the rib.

It may be that absence of UP in some fossil birds more primitive than C. sanctus is also due to poor preservation (Chiappe et al., 1999; Codd et al., 2008). Despite acknowledging these difficulties, variation in the occurrence of UP amongst the ancestors of modern birds remains enigmatic. UP are missing from Archaeopteryx; although reported in the 10th Archaeopteryx (Codd et al., 2008), we consider this a misinterpretation of the fossil (originally described in Mayr et al., 2005) and conclude that none are present. Archaeopteryx also lacks a sternum, suggesting it may have possessed substantially different mechanics of ventilation. Saperornis (Zhou & Zhang, 2002a), Jeholornis (Zhou & Zhang, 2002b) and Zhongornis (Gao et al., 2008) are also reported to lack uncinates and are considered basal to the avian specimens in this study. Furthermore, the occurrence of UP varies significantly amongst the comparatively more derived enantiornithines and ornithurines (O’Connor et al., 2009). However, the description of well-developed UP in, for example, the Early Cretaceous Z. yangi (Zhou & Li, 2009) and ornithurine Y. grabaui together with evidence from C. sanctus and nonavian maniraptoran taxa [UP occur in a minimum of four oviraptorosaur and dromaeosaurid specimens (Codd et al., 2008)] provides evidence for conservation of UP in the lineage from nonavian theropods to extant birds (Chiappe et al., 1999).

Evolutionary implications of uncinate process function in C. osmolskae and basal birds

Our estimation of the mechanical advantage of the UP for rib movement in three basal bird taxa and C. osmolskae suggests that the leverage would be improved, just as is found in extant birds (Table 1). Furthermore, the magnitude and the pattern of relatively decreased effectiveness moving caudally are also similar (Tickle et al., 2007; Codd, 2010). This suggests that UP were able to function as integral respiratory structures before the origin of modern birds (Codd et al., 2008). Because the improved mechanical advantage conferred by UP is of a similar magnitude across the species studied, it would seem that process morphology is optimized to enable an improved leverage for movements of the ribs and sternum. Therefore, for a given body shape, process length and shape are fine-tuned to function (Codd, 2010). This optimized model may explain why the superficially similar cursorial ostrich and C. osmolskae specimens have contrasting UP: whereas the ostrich has UP of reduced length, C. osmolskae has long UP reminiscent of modern flying birds (Codd et al., 2008). Perhaps the functional requirements of moving large sternal plates (Clark et al., 1999) and gastralia in the body wall required a longer lever arm than does the ostrich, which has a relatively small sternum and no gastral elements to rotate during breathing. Furthermore, UP in nonavian theropods may have articulated with the ribs via a cartilaginous joint symphysis. The inherent flexibility in this joint compared to a bony symphysis of rib and process may have caused reduced lever force (Tickle & Codd, 2009) therefore leading to a requirement for longer UP.

Species carrying a large sternal mass are expected to incur relatively expensive ventilatory movements due to the ventral displacement of the large flight muscles during inspiration (Tickle et al., 2010). Recent research demonstrates that the metabolic cost of a sternally applied mass is double that of an identical back load (Tickle et al., 2010). These relatively high energetic costs are driven by an increase in the work required to move the sternum and attached muscle mass up and down. Therefore, development of enlarged flight muscles most likely required accessory respiratory structures (UP) to be in place beforehand. These skeletal structures are indeed found in early diverging birds. It seems that the development of UP as a means of facilitating cuirassal ventilation first appeared in nonavian theropods (Codd et al., 2008) and represent an exaptation for the bellows-like sternal pump breathing mechanics of their avian descendants. The morphology of extant avian UP is optimized to sternal morphology and represents a progression on the form necessary to assist movements of the functionally analogous gastral basket in theropods and the early birds (Codd et al., 2008). Interestingly, the occurrence and effectiveness of UP in C. sanctus indicates that maximal improvement in mechanical advantage was provided in the most cranial portion of the ribcage (Table 1). This area is adjoined to the lightly keeled sternum, presumably where the pectoral musculature was attached. The L-shaped morphology of the UP in C. sanctus, identical to that of extant flying birds (Tickle et al., 2007), would have provided a large insertion area for inspiratory muscles that could offset the work necessary to move the sternum. Furthermore, evidence to support the idea that UP initiated movements of the gastral elements in C. sanctus is found in specimen SMF Av 419 where UP occur on ribs caudal to those that directly interact with the sternum. Similar preservation is also visible in specimen GMV-2147 (Fig. 31, Chiappe et al., 1999). Relatively high mechanical advantage in the cranial ribcage compared to the reduced effectiveness of the lever arm moving caudally is consistent with the hypothesis that UP effected gastral movement, because the mass of these elements would most likely have been lower than the ossified sternum. In the only extant animal to have ossified UP and gastralia, Sphenodon, the external oblique muscles extend from the gastral basket to the UP (Maurer, 1896; Codd et al., 2008), suggesting a possible ventilatory role for gastralia (Lambe, 1917; Perry, 1983; Carrier & Farmer, 2000a,b; Claessens, 2004).

Although it is unclear whether C. sanctus was capable of flapping flight (Chiappe et al., 1999; Nudds & Dyke, 2010; Paul, 2010), the fossil record indicates that these birds possessed a keeled sternum (Peters & Qiang, 1999). When considering this together with derived UP morphology and their role in facilitating respiration, it seems likely that early avian species were already ventilating their lungs using a mechanism that would go on to prove essential in moving the keeled sternum associated with later flapping flight. Therefore, while the origin of flapping flight in birds is unknown, the large energetic demands that are incurred as a result of later powered flight were not likely to have been constrained by limits on oxygen delivery. Indeed, the effective flow-through, UP-driven ventilatory mechanism was already in place in the earliest avian taxa.

Flapping flight has only evolved in three vertebrate lineages. In the Triassic, pterosaurs were the first vertebrates to develop flapping flight and exhibit skeletal features that show convergent similarities to the derived avian structures associated with powered flight. In addition to the increased surface area for flight muscle attachment provided by a small sternal keel (Padian, 1983), pterosaurs also have ossified dorsal and ventral rib processes, termed sternocostapophyses (Claessens et al., 2009). These projections are hypothesized to have worked as levers for respiratory movements (Claessens et al., 2009) in a mechanism similar to that described for the UP in birds and some nonavian theropods (Tickle et al., 2007), indicating that ventilatory levers attached to the ribs have developed across evolutionary lineages. Regions of post-cranial pneumaticity together with the proposed skeletal ventilatory pump signal the presence of flow-through lung ventilation in pterosaurs (Claessens et al., 2009), analogous to the system proposed for advanced nonavian theropods (O’Connor & Claessens, 2005) and observed in extant birds and alligators (Farmer & Sanders, 2010). Well-developed rib levers may therefore be integral components in the evolution of a skeletal aspiration pump that facilitates air movement around an air sac flow-through system. Alteration of process form in response to body shape changes, in part driven by locomotor behaviour (Tickle et al., 2007), may be a relatively simple mechanism to maintain sufficient skeletal excursions to aspirate the lung in a highly compliant air sac system.

A high rate of energy metabolism is characteristic of extant birds and helps to supply their relatively high energy requirements (Butler, 1991). As we cannot directly determine the anatomy and physiology of fossil species, we must use indirect methods to infer their respiratory biology and rates of energy metabolism. Determination of growth rates from fossil bone histology (Chinsamy et al., 1995; Erickson et al., 2001, 2009) and the discovery of feathered theropods (Norell & Xu, 2005) suggest an elevated metabolism, intermediate between the reptilian and extant avian condition, was likely to have been sustained in early birds and their close nonavialan ancestors. Given the positive correlation between process length and metabolic rate in birds (Tickle et al., 2009), the degree of rib lever development may be considered a correlate of the rate of energy use; the presence of UP (and an efficient flow-through lung) in nonavian theropod dinosaurs may have helped to sustain a metabolic rate that was higher than found in earlier dinosaurs (O’Connor & Claessens, 2005). Therefore, the development of flow-through breathing mechanics that encompass an UP-driven sternal aspiration pump together with rib and gastral movements may have been instrumental in the diversification from terrestrial to volant niches.


We would like to thank Rainer Brocke (Senckenberg Forschungsinstitut und Naturmuseum, Frankfurt, Germany) for access to Confuciusornis sanctus specimens and the American Museum of Natural History (USA) for facilitating access to the Citipati specimen. This work was funded by the Biotechnology and Biological Sciences Research Council (BBSRC, G01138/1) and the Deutsche Forschungsgemeinschaft (DFG, 721/1). PGT was supported by a BBSRC stipend during his PhD.

Data deposited at Dryad: doi: 10.5061/dryad.cp4tc417