Reg. Charity No. 1168330
Among dinosaurs, the pubis has convergently retroverted four times in Maniraptora (Theropoda) and once in Ornithischia. Although a clear correlation has not been demonstrated, it has been previously proposed that two traits were related to pubic retroversion: the reduced importance of cuirassal ventilation, and a herbivorous diet. Here, we analyse the possible influence of these traits on pubis orientation. Cuirassal ventilation was plesiomorphically present as an accessory ventilation mechanism in Dinosauria and was powered by the M. ischiotruncus, which was probably connected to a propubic pelvis. Cuirassal ventilation was reduced in both Ornithischia and Maniraptora, some of which also evolved herbivory. According to our results, cuirassal ventilation is more strongly correlated with pubic orientation than herbivory. The retroversion of the pubis during the evolution of birds resulted in major changes in the musculature of the tail. These changes increased the efficiency of the pubocaudalis muscles, which enhanced the birds’ capability for take‐off from the ground. The release of the evolutionary constraint on pubic orientation through changes in the ventilatory system can therefore be considered to be an important step in bird evolution.
The classic subdivision of Dinosauria into Ornithischia and Saurischia (even if this might not represent true phylogenetic relationships, see Baron et al. 2017a) recalls differences in the structure of the pelvis: ornithischians have retroverted pubes similar to the condition in birds, whereas most saurischians have anteriorly projecting pubes (Seeley 1887). However, more recent finds have shown that several groups of non‐ornithischian dinosaurs also evolved a retroverted, bird‐like pubis, in particular within the theropod subclade Maniraptora. Following a recent phylogenetic tree of theropods combining information from numerous detailed studies (Hendrickx et al. 2015), at least the basal members of every subclade of Maniraptora have anteroventrally (propubic condition) or ventrally projecting pubes (mesopubic condition), whereas retroverted pubes (opisthopubic condition) evolved only subsequently within four of those subclades. These four subclades include the Alvarezsauroidea (with, e.g. the propubic Haplocheirus, mesopubic Patagonykus, and opisthopubic Shuvuuia; Hutchinson & Chiappe 1998; Choiniere et al. 2010), Dromaeosauridae (e.g. the propubic Austroraptor and opisthopubic Velociraptor; Norell & Makovicky 1999) and birds (e.g. the mesopubic Archaeopteryx and opisthopubic Confuciusornis; Mayr et al. 2005; Dalsätt et al. 2006). Among Troodontidae, Sinovenator is the only member known to have evolved an opisthopubic pelvis (Xu et al. 2002). In total, the retroversion of the pubis happened four times in Maniraptora and once in Ornithischia, and it seems most plausible to interpret all five changes within dinosaurs as convergences (Figs 1, 2). Herein, we analyse two of the potential causes previously proposed in the literature, for these convergent acquisitions.
Figure 1 Open in figure viewerPowerPoint Pelvic structure of opisthopubic dinosaurs. A, Lesothosaurus (modified from Sereno 1991). B, Shuvuuia (modified from Hutchinson & Chiappe 1998). C, Segnosaurus (modified and mirrored from Zanno 2010a). D, Microraptor (modified from Gong et al. 2012). E, Sinovenator (modified from Xu et al. 2002). F, Apteryx (modified from Evans 1900). Note the different orientations of the pubes, the varying developments of the pubic boots and the ischium length relative to the pubis. The measurement protocol for the pubic angle is indicated in D. Pelves scaled to the same dorsoventral height to highlight differences in configuration. Scale bars represent 10 mm. Colour online. CaptionPelvic structure of opisthopubic dinosaurs. A, Lesothosaurus (modified from Sereno ). B, Shuvuuia (modified from Hutchinson & Chiappe ). C, Segnosaurus (modified and mirrored from Zanno ). D, Microraptor (modified from Gong et al. ). E, Sinovenator (modified from Xu et al. ). F, Apteryx (modified from Evans ). Note the different orientations of the pubes, the varying developments of the pubic boots and the ischium length relative to the pubis. The measurement protocol for the pubic angle is indicated in D. Pelves scaled to the same dorsoventral height to highlight differences in configuration. Scale bars represent 10 mm. Colour online. Figure 2 Open in figure viewerPowerPoint Phylogenetic tree of selected taxa indicating synapomorphies concerning pubic orientation, feeding strategy and ventilation modes. Nodes: (1) Sauropodomorpha; (2) Maniraptoriformes; (3) Maniraptora; (4) Pennaraptora. Abbreviations: C, carnivory; Cu, cuirassal ventilation; H, herbivory; Me, opisthopubic with a pubis angle > 90°; Nc, non‐cuirassal ventilation; Op, opisthopubic with a pubis angle > 115°; P, propubic. Phylogenetic tree based on Carballido et al. (2012), Boyd (2015), Hendrickx et al. (2015), and McPhee et al. (2015). Colour online. CaptionPhylogenetic tree of selected taxa indicating synapomorphies concerning pubic orientation, feeding strategy and ventilation modes. Nodes: (1) Sauropodomorpha; (2) Maniraptoriformes; (3) Maniraptora; (4) Pennaraptora. Abbreviations: C, carnivory; Cu, cuirassal ventilation; H, herbivory; Me, opisthopubic with a pubis angle > 90°; Nc, non‐cuirassal ventilation; Op, opisthopubic with a pubis angle > 115°; P, propubic. Phylogenetic tree based on Carballido et al. (), Boyd (), Hendrickx et al. (), and McPhee et al. (). Colour online.The first proposed evolutionary driver causing pubic retroversion was a switch to a herbivorous diet, requiring more space in the abdominal cavity than carnivorous, piscivorous or omnivorous ones (Weishampel & Norman 1989; see also Zanno & Makovicky 2011) given that an entirely plant‐based diet generally needs a longer intestinal tract compared to others (Chivers & Hladik 1980). Alternatively, Rasskin‐Gutman & Buscalioni (2001) noted that all ornithischians combined the occurrence of a retroverted pubis with the lack of gastralia, small bones that are not strongly connected to the remaining skeleton, but embedded in the ventral trunk musculature, and which only occur in crocodiles and the tuatara today (Claessens 2004a). In extant crocodiles, these bones form a gastral basket, which is connected to the sternum anteriorly and to the pubis and the ischium posteriorly (Claessens 2004b; Uriona & Farmer 2008). The main muscle that includes the gastralia is the M. rectus abdominis, which attaches to the sternum and to the pubis via a ligamentous sheet; other abdominal muscles lie lateral and superficial to the M. rectus abdominis (i.e. M. obliquus externus and M. internus abdominis; Claessens 2004a). Two pelvic muscles are attached to the ischium: (1) from the ischium to the anterior margin of the pubis (M. ischiopubis); and (2) to the posterior‐most gastralia (M. ischiotruncus; Claessens 2004a). Given that gastralia also occur in dinosaurs, pterosaurs, and a wide range of Triassic archosauriforms (Claessens 2004a; Tschopp & Mateus 2013; Fechner & Gößling 2014), their presence is most parsimoniously interpreted as a plesiomorphic condition of Archosauria. Thus, the anatomical and muscular connections of the gastralia in dinosaurs were also probably the same as those in extant crocodiles (Claessens 2004a), with the possible exception of the ischiopubis muscle; this muscle mostly acts on the pubis, which is kinetic in crocodiles, but not in other archosaurs. This kinetic pubis is connected to the liver, and forms an important accessory ventilation system supporting the costal ventilation by working like a hepatic piston (Claessens 2004a, 2015; Uriona & Farmer 2008). Gastralia have only a marginal ventilatory function in extant crocodiles, presumably facilitating deformation of the abdominal wall and storing elastic energy during respiration (Claessens 2015). The hepatic piston pump of extant crocodiles appears to be a derived condition of Crocodyliformes (Uriona & Farmer 2008), given that no extinct non‐crocodyliform pseudosuchian shows any indication of a kinetic pubis (Claessens 2004b). Because birds also use a form of pelvic aspiration (their pelvic musculature helps to inflate the abdominal air sacs during inspiration and expiration, elevating and depressing the pelvis (and synsacrum) relative to the trunk vertebrae and the tail relative to the pelvis and synsacrum; Baumel et al. 1990; Carrier & Farmer 2000; Farmer & Carrier 2000) and considering that pelvic aspiration appears to be rare among vertebrates, Farmer & Carrier (2000) deduced that the common ancestor of birds and crocodilians must have had some form of pelvic aspiration. It has been suggested that this kind of pelvic ventilation was cuirassal, and therefore dependent on an ischiotruncus muscle with a similar pattern of activity to that of alligators, causing the expansion of the abdominal cavity by pulling the gastralia posteriorly (Claessens 2004b, 2015; Farmer 2006). Thus, although basal archosaurs were likely to have relied extensively on costal ventilation, the presence of accessory cuirassal ventilation is generally considered to be a plesiomorphic condition of Archosauria, which was probably secondarily lost in birds, and reduced in crocodiles. Therefore, the hypothesis of Rasskin‐Gutman & Buscalioni (2001) is that the propubic and mesopubic conditions were associated with cuirassal accessory ventilation, whereas opisthopubic pelves were correlated with the loss of cuirassal ventilation. Carrier & Farmer (2000) suggested that an important switch concerning the muscular control of the cuirassal ventilation occurred within Maniraptora, when the M. truncocaudalis (or M. caudotruncus, as they call it) replaced the M. ischiotruncus as main rotator of the gastralia during the evolution of the clade. A propubic or mesopubic pelvis appears to have been necessary for the M. ischiotruncus to work, because it used gastralia (Carrier & Farmer 2000; Fechner & Schwarz‐Wings 2013) and the ischium as sites of insertion, and the pubis as pulley to stabilize it (Carrier & Farmer 2000), whereas the functional shift towards an M. truncocaudalis did not constrain the pelvis in such a way. If the ventilation was strictly connected with the structural composition of the propubic pelvic girdle arrangement, this could be a good explanation for the fact that the pubis has never rotated in non‐maniraptoran theropods and non‐eusauropod sauropodomorphs. Cuirassal ventilation powered by the ischiotruncus muscle could have been an evolutionary constraint on retroversion of the pubis. Certain taxa may have been able to rely only on basic costal ventilation for their respiratory needs. Nevertheless, although crocodiles use solely costal ventilation while resting (Munns et al. 2012), the fact that they resort to accessory ventilation modes to support this during periods of high activity (Munns et al. 2012) implies that the presence of an accessory ventilation system is advantageous. The constraint imposed by a cuirassal ventilation system would have been much weaker if there were more efficient skeletal or muscular structures capable of supporting ventilation of the lungs, either by improving the costal ventilation or forming additional accessory ventilation modes. Therefore, we should expect that natural selection would not lead to the reduction of cuirassal ventilation without any other modification in the ventilatory structure of Maniraptora and Ornithischia. These modifications should include an improvement of costal ventilation through additional structures or novel muscular connections, which happened contemporaneously with the reduction of the ischiotruncus muscle. The fact that gastralia have never been found in ornithischians, despite the abundance of well‐preserved specimens, indicates that the ventilation was no longer cuirassal. It is possible that costal ventilation remained the only ventilation mode or that some unknown structures helped to inflate the lung. For example, Boyd et al. (2011) proposed that intercostal plates found in some ornithischian dinosaurs (Hypsilophodon, Othnielosaurus and in Parksosauridae: Talenkauen, Thescelosaurus, Parksosaurus, Macrogryphosaurus) could be connected with an unknown ventilation mode (see also Codd et al. 2008). Another theory was put forward by Carrier & Farmer (2000), who suggested that the pubis–ischium structure had an accessory ventilatory function, using the propubic process of the pubis in an analogous way to the mobile pubis of crocodiles. Among Maniraptora, as in birds, costal ventilation was improved by the ossification of the uncinate processes, which are even fused to the ribs in derived avians, and serve as the insertion points along the ribcage for the M. appendicocostalis, a ventilatory muscle (Codd 2004; Codd et al. 2008). Uncinate processes also occur in more basal archosaurs, such as crocodiles and Sphenodon, but they remain cartilaginous structures throughout their lives, are more plate‐like in shape, and are connected to a number of different muscles (Codd et al. 2008; Sullivan et al. 2016). Thus, cartilaginous uncinate processes were probably plesiomorphically present in archosaurs, but apparently served a variety of functions, including ventilation (Sullivan et al. 2016). Even though it is probable that they had a ventilatory role in some non‐pennaraptoran archosaurs (Sullivan et al. 2016), their ossification in Maniraptora suggests an increased use for ventilatory functions in this clade, which led to a higher amount of induced stress on the cartilaginous processes, and thus finally to their ossification. While uncinate processes were gradually ossifying along the stem‐birds line, some maniraptoran groups retroverted their pubis and reduced the length of the ischium. In both Ornithischia and Maniraptora, the role of the ischiotruncus muscle in cuirassal ventilation therefore seems to have been substituted (or at least reduced) by an improved costal ventilation mode using either intercostal plates or uncinate processes, and/or that this system was complemented by an accessory pelvic ventilation mode.
Herein, we aim to compare the two hypotheses described above, trying to understand if herbivory or the disappearance of cuirassal ventilation driven by the ischiotruncus muscle was more strongly correlated with the evolution of the opisthopubic pelvis.
Material and methodA new method was developed to analyse the plausibility of a certain trait being an evolutionary driver for another trait (Fig. 3). This method aims to assess whether one or both of two traits (A and B in Fig. 3) are correlated with a third one (C in Fig. 3) and could thus have potentially been its evolutionary driver or constraint. Here, we compared correlation of pubic retroversion and the two traits previously proposed in the literature to lead to retroversion of the pubis: feeding strategy or ventilation. The method consists of five steps, described in detail below and summarized in Figure 3.
Figure 3 Open in figure viewerPowerPoint Diagram representing the steps of the comparison method conceived herein. See text for details. DiD, Difference in Distribution. Colour online. CaptionDiagram representing the steps of the comparison method conceived herein. See text for details. DiD, Difference in Distribution. Colour online.
The phylogenetic framework was based on recent publications on specific subclades of Dinosauria: Hendrickx et al. (2015) reviewed the most recent phylogenetic analyses of Theropoda, creating a composite phylogenetic tree of the entire clade; Boyd (2015) provided a new, extensive analysis of Ornithischia, including both basal and advanced taxa; Sauropodomorph interrelationships were taken from Carballido et al. (2012) and McPhee et al. (2015). We combined these studies without a new phylogenetic analysis because they each analysed a different clade of dinosaurs and were thus not in conflict with one another. Given the ongoing discussion about the relationship between these major clades (see Baron et al. 2017a, b; Langer et al. 2017), we compiled the information into the classic view of dinosaur relationships, instead of following the still controversial topology found by Baron et al. (2017a).
Pubic orientation (Macaluso & Tschopp 2018, appendix S1, S2) was taken from the literature describing the pelvic structure of the taxa included here (see Macaluso & Tschopp 2018, appendix S2 and references therein, for data sources), following a modified measurement protocol of Rasskin‐Gutman & Buscalioni (2001). Modifications concerned the landmark used for the orientation of the pubis, for which we chose the anterior‐most point of the pubic boot instead of the distal‐most point of the pubis. This change was introduced because, according to us, if a pubis has the anterior‐most part of its boot extending beyond the line that delimits the two categories of propubic and opisthopubic pelves (see below for its definition), it should not be considered retroverted in its entirety. We used two different thresholds to define the opisthopubic condition, given that the term ‘opisthopubic’ has been used differently in the literature. Rasskin‐Gutman & Buscalioni (2001) only considered a pelvis to be opisthopubic if the pubic angle is greater than 108°, whereas Xu et al. (2002) described the pelvis of Sinovenator (with an angle of 102°) to be opisthopubic. Therefore, we performed two test series, one considering the pelvis to be opisthopubic if the pubic angle was greater than 108° (termed ‘Op’ hereafter), and another one including all taxa with a pubic angle greater than 90° (termed ‘Me’ hereafter, because this includes many species classified as ‘mesopubic’ by Rasskin‐Gutman & Buscalioni, 2001). Following the phylogenetic framework mentioned above, the plesiomorphic condition in Maniraptora is probably propubic or mesopubic, because basal members of all groups show this kind of pelvis. On the other hand, all ornithischians are opisthopubic. The measurement protocol we followed had the peculiar effect of defining therizinosaurs as propubic (e.g. Segnosaurus pubic angle 80°), although they are generally considered to have a retroverted pubis (Zanno 2010a, b). In fact, their pubis is posteriorly curved in its proximal part, but their pubic boot projects strongly anteriorly (Zanno 2010a; Fig. 1C), resulting in a lower pubic angle. Implications of this peculiarity will be discussed below.
Our interpretation of the feeding strategy is based on previous studies which mainly analysed tooth morphology. Because of the similar gut structure between carnivorous and omnivorous animals (only strictly herbivorous animals need a particularly long digestive tract; Chivers & Hladik 1980), we do not need to distinguish between omnivorous and carnivorous dinosaurs herein. The earliest dinosaurs, as well as basal theropods and basal Maniraptora appear to have been omnivorous or carnivorous (Zanno & Makovicky 2011). Even though it has recently been proposed that the diet of basal ornithischians like heterodontosaurids and basal thyreophorans, such as Lesothosaurus, was omnivorous (Baron et al. 2017a; Sciscio et al. 2017), we herein prefer to consider herbivory as a synapomorphic feature of Ornithischia, in part following Sereno (2012), who convincingly argued for nearly exclusive herbivory in the basal ornithischian clade Heterodontosauridae. Even if at least some basal sauropodomorphs could have been omnivorous (Barrett 2000), all sauropods were clearly herbivorous (Upchurch et al. 2004). Among theropods, definitive herbivory probably occurred in derived Therizinosauria (Zanno & Makovicky 2011). As far as Alvarezsauroidea and Oviraptorosauria are concerned, we decided to follow Zanno & Makovicky (2011) in considering them to be herbivorous, even though they stated that this interpretation was based on weak evidence; in this way, we avoided a possible bias against a correlation between pubic retroversion and herbivory.
The ventilatory structures are mainly composed of soft tissue, which are rarely found in fossils. Therefore, we had to identify osteological correlates to infer ventilation modes. The reduced importance of the ischiotruncus muscle for cuirassal ventilation was inferred by either the absence of gastralia, the absence of a distally expanded ischium or a shorter ischium compared to the pubis (ischium to pubis ratio < 0.66; Table 1). We did not consider the ossification of uncinate processes here, because their record is very fragmentary and all the species that preserve ossified uncinate processes had a reduced ischium as well (see below).
Table 1. Osteological correlates interpreted as evidence for the reduction of the ischiotruncus muscle Osteological correlate Clade References Absence of gastralia Ornithischia Carrier & Farmer (2000); Rasskin‐Gutman & Buscalioni (2001); Claessens (2004a) Eusauropoda Claessens (2004a); Tschopp & Mateus (2013); Fechner & Gößling (2014) Complete absence of a distal expansion of the ischium Ornithischia Carrier & Farmer (2000); Rasskin‐Gutman & Buscalioni (2001) Alvarezsauroidea (except for basal member Haplocheirus) Hutchinson & Chiappe (1998); Choiniere et al. (2010) Ischium to pubis ratio < 0.66 Falcarius Zanno (2010b) Pennaraptora Macaluso & Tschopp (2018, appendix S2) (Uncinated processes)a Pennaraptora? Codd (2004); Codd et al. (2008) Oviraptoria: Caudipteryx, Oviraptor Dromaeosauridae: Velociraptor, Microraptor Avialae: most of the membersThe absence of gastralia, although the most logical evidence of a lack of cuirassal ventilation, could be due to preservation bias. Thus, we only interpreted it as strong evidence of the absence of cuirassal ventilation when even the best‐preserved specimens of extensively known groups completely lacked gastralia, as is the case in ornithischians (Carrier & Farmer 2000; Rasskin‐Gutman & Buscalioni 2001; Claessens 2004a). Gastralia also seem to be absent in most derived eusauropods (Claessens 2004a; Tschopp & Mateus 2013; Fechner & Gößling 2014). Where they have been reported (mostly in non‐neosauropod sauropods and diplodocids; Tschopp & Mateus 2013), gastralia have a very different shape to those used for ventilation in theropods and basal sauropodomorphs (Claessens 2004a; Fechner & Gößling 2014) indicating that they might have lost their ventilatory function during the evolution of sauropods.
Among Maniraptora, an interpretation of the accessory ventilation system is more complicated, because gastralia were only lost in derived avialans (Claessens 2004a; O'Connor et al. 2015). The transition from the basal reptilian ventilation to the derived, avian one involved a substantial, progressive change in muscular structure. The main muscles active during ventilation among birds are the M. appendicocostalis (Claessens 2009) and the M. pubocaudalis (Baumel et al. 1990), thus we considered the increased importance of these muscles to be an indicator for the decoupling of the ventilation from the predominant role of the cuirassal one. The M. appendicocostalis of birds attaches to the ossified uncinate processes, which can therefore be used as an osteological correlate. The presence of ossified uncinate processes has been confirmed in some members of Pennaraptora (see Table 1), but it remains unclear whether these represent independent acquisitions, or if their presence can be interpreted as synapomorphic for Pennaraptora (Codd et al. 2008). Indeed, the most recent reviews discussing the gradual assembly of avian traits all ignore the appearance of ossified uncinate processes (Zanno & Makovicky 2011; Xu et al. 2014; Brusatte et al. 2015). It is highly probable that this difficulty arises from a preservational and collection bias concerning the recovery of uncinate processes, given that these bones are generally small, and only connected to the ribs with connective tissue, so that they are easily detached by taphonomic processes, or overlooked or misidentified during excavation. We would therefore expect that additional, exceptionally preserved specimens, or detailed searches for osteological correlates for the presence of uncinate processes on the dorsal ribs (like scars or rugose areas as found in crocodilians; Sullivan et al. 2016) would confirm these structures as synapomorphic for a more inclusive clade (Pennaraptora, or maybe even the more inclusive Maniraptora) instead of representing several independent acquisitions.
Because of these difficulties, we preferred to focus on the pelvic area to find osteological correlates for the reduction of the importance of the M. ischiotruncus for ventilation. For the M. pubocaudalis, we considered the same characters as Hutchinson (2001) to be osteological correlates: the complete absence of a distal expansion of the ischium, and a shortening of the ischium with respect to the pubis. It is likely that, as the pubocaudalis muscles increased their ventilatory role, the dependence on the plesiomorphic cuirassal ventilation driven by the ischiotruncus muscle was reduced. Even though a hypothetical truncocaudalis muscle could have coexisted with the M. pubocaudalis, it would probably have constrained the pelvis much less than the ischiotruncus muscle, being inserted on the tail rather than the ischium (Carrier & Farmer 2000).
Among alvarezsauroids, preservation bias plays an important role, given their generally incomplete and poor preservation. However, the ischium of the most complete specimens is without any, even slight, distal expansions, except for the basal‐most member Haplocheirus (Choiniere et al. 2010). No reduction of the ischium occurs in any of the members of this clade. Among Therizinosauria, the only member with an osteological correlate indicating a potentially non‐cuirassal ventilation is the basal‐most known member of the clade, Falcarius, which has a shortened ischium, whereas more derived therizinosaurs possess a long ischium (Zanno 2010a, b; Fig. 1). Thus, it is probable that the shortening of the ischium in Falcarius is an autapomorphy with no wider implication for the evolution of ventilation in its clade. Every well‐preserved member of Pennaraptora (see Macaluso & Tschopp 2018, appendix S2 for references) shows a reduction of the ischium.
We analysed the distribution of feeding strategy and ventilation modes among propubic and opisthopubic dinosaurs. As mentioned above, these analyses were made using the two possible thresholds for the pubic angle (Op and Me).
In order to perform this kind of comparison, an objective sample choice is essential and an approach to avoid possible bias is needed. Firstly, for studies concerning causal relationships of specific traits, any potential evolutionary innovation that happened after the first occurrence and fixation of the trait it should have caused should not be considered.
Moreover, a potential bias can be created by differential diversification rates in the clades under study, as repeatedly articulated in the literature on comparative methods (Ridley 1983; Felsenstein, 1985; Höglund 1989). Unequal species diversification among these clades might not necessarily be correlated with the appearance of the morphological innovation in question. Assuming that a convergently acquired or lost feature (such as pubic retroversion or cuirassal ventilation) is equally advantageous for every clade in which it appeared, subsequent diversification in only a subset of these clades should have been triggered by different innovations, or new, unique combinations of traits (which in our case might or might not include ventilation modes and pubic orientation). Therefore, the inclusion of a number of taxa proportional to the diversity within a clade could be a misleading approach in a quantitative analysis of trait correlation. In our case, for instance, the inclusion of a high number of derived ornithischian taxa would inevitably lead to a strong correlation between an opisthopubic condition and herbivory, just because they all are herbivorous and opisthopubic, and because their diversity is very high, being a very long‐living clade. The lower number of known species for Dromaeosauridae, compared to Ornithischia, does not imply that the convergent evolution of an improved costal ventilation was less advantageous in this clade than for ornithischians.
Another important bias is the fragmentary nature of the fossil record, which makes it difficult to sample an equal number of taxa per clades, especially considering that groups important for our study (Alvarezsauroidea) are poorly represented. The new methodological approach developed and used herein attempts to address both of these biases simultaneously, by combining an extensive literature review with a very selective taxon sampling as explained below.
In order to identify an adequate sample of taxa, we compiled a matrix with a large number of species within the studied groups (Macaluso & Tschopp 2018, appendix S2), and collected morphological information on the osteological correlates identified above, excluding those for which the available data were too fragmentary. Subsequently, we plotted them into the phylogenetic framework, to detect all the evolutionary changes. Then, to avoid the biases described above, we chose a representative group per trait change (grey cells in Fig. 4; Macaluso & Tschopp 2018, appendix S1), and one group with the plesiomorphic condition for the traits (Dinosauromorpha was used in both cases, herbivory and ventilation). The aim here was to pick one group whenever an innovation occurred in the studied traits. Thereby, the plesiomorphic condition represents a single initial state that can be altered when evolutionary events occur. A different taxon sampling was used for every comparison: one set of taxa was used to analyse potential correlation of herbivory and pelvic structure (Fig. 4A), and another set for the possible correlation between ventilation and pelvic structure (Fig. 4B). This approach is somewhat similar to that used in comparative biology to avoid bias due to arbitrary sample choice and relatedness of the chosen taxa (Ridley 1983; Höglund 1989), and we think it is also useful to simultaneously circumvent incompleteness of the fossil record. It is important that every independent evolutionary innovation is represented by a group in the analysis. In the case of Ornithischia, for instance, we could include the entire clade to represent the opisthopubic and herbivorous condition typical for this group, because it did not experience any reversal before its extinction. The same reasoning is valid for every group or taxon chosen (the derived character for which the single groups were chosen is highlighted in grey in Fig. 4).
Figure 4 Open in figure viewerPowerPoint Taxon sampling for each test. A, herbivory and pubic orientation using the categories Op (opisthopubic if pubic angle (pa) > 108°) and Me (opisthopubic if pa > 90°). B, ventilation and pubic orientation using the categories Op and Me. Grey cells indicate the synapomorphy for which that taxon or group was chosen as a representative. CaptionTaxon sampling for each test. A, herbivory and pubic orientation using the categories Op (opisthopubic if pubic angle (pa) > 108°) and Me (opisthopubic if pa > 90°). B, ventilation and pubic orientation using the categories Op and Me. Grey cells indicate the synapomorphy for which that taxon or group was chosen as a representative.The steps of the comparison methodology and the calculations are summarized in a diagram in Figure 3 and in Figure 5. For each comparison, we included the groups identified as representatives for the specific test in a 2 × 2 contingency table, with pelvic condition in rows and feeding strategy (Fig. 5A) or ventilation mode (Fig. 5B) in columns. These tables indicate how many groups show a particular trait combination (Fig. 5). In order to normalize these values, we summed the two values in every row and column, calculating how many herbivorous and carnivorous, cuirassal and non‐cuirassal, propubic and opisthopubic taxa there are in every comparison table (‘Row sum’ columns and ‘Column sum’ rows in Fig 5). Then, we calculated percentage values relative to the sum of the row (or column) of the derived state of the trait, quantifying the relative number of groups within those with the same derived state of one trait (herbivory, reduced importance of ischiotruncus muscle, or opisthopubis), which had a plesiomorphic or derived state of the other trait (Fig 5A, B: % Opisthopubic, % Herbivory, % Non‐cuirassal). An evolutionary force which acted on the opisthopubic condition should have the following two attributes:
Figure 5 Open in figure viewerPowerPoint Contingency tables showing the distribution of trait combinations, percentages and differences in distribution among dinosaurs with different pelves. A, comparing carnivorous and herbivorous dinosaurs. B, comparing those with cuirassal ventilation (i.e. dependent on the ischiotruncus muscle) and non‐cuirassal ventilation (i.e. independent of the M. ischiotruncus). Percentages are derived from the normalization of the respective row or column with its sum (e.g. in the Op test comparing feeding mode with opisthopubic groups, % Opisthopubic refers to the Opisthopubic row normalized with respect to the total of 4 opisthopubic groups; % Herbivory refers to the 2 opisthopubic groups normalized to the total of 7 herbivorous groups). Difference in Distribution is by derived subtracting the two percentage values in the respective row. CaptionContingency tables showing the distribution of trait combinations, percentages and differences in distribution among dinosaurs with different pelves. A, comparing carnivorous and herbivorous dinosaurs. B, comparing those with cuirassal ventilation (i.e. dependent on the ischiotruncus muscle) and non‐cuirassal ventilation (i.e. independent of the M. ischiotruncus). Percentages are derived from the normalization of the respective row or column with its sum (e.g. in the Op test comparing feeding mode with opisthopubic groups, % Opisthopubic refers to the Opisthopubic row normalized with respect to the total of 4 opisthopubic groups; % Herbivory refers to the 2 opisthopubic groups normalized to the total of 7 herbivorous groups). Difference in Distribution is by derived subtracting the two percentage values in the respective row.
Extreme values such as 0 and 100% are probably impossible to reach, because ecological and biological systems are influenced by a large number of factors. For instance, oviraptorosaurs are herbivorous, and have both a shortened ischium and uncinate processes, but no oviraptorosaur species has yet been found with an opisthopubic pelvis. Other factors (e.g. nesting or reproductive behaviour) could have influenced pubic orientation, and it is also possible that they simply would have had no real advantage by developing an opisthopubic pelvis. However, herein we compare two traits that have been reported as potential evolutionary drivers or constraints for pubic retroversion in the literature, and we do not intend to test all possible correlations. It is reasonable, though, that if the evolution of a certain derived state of a trait (e.g. herbivory or the reduced importance of the ischiotruncus muscle) had influenced the evolution of an opisthopubic structure, a large percentage of herbivorous (or ‘non‐cuirassal’) groups would have had an opisthopubic pelvis and a large percentage of opisthopubic groups would have had a herbivorous diet or a reduced ischiotruncus muscle, even if the end values of 0 or 100% are not reached. In other words, if a derived state of trait A (or B) drove the evolution of the derived state of another trait C (Fig. 3), we expect the derived state of A (or B) to be much more common in taxa with the derived state of C than in those with a plesiomorphic state of C, and the derived state of C to be much more common in taxa with the derived state of A (or B) than in those with the plesiomorphic state of A (or B). Thus, by subtracting the percentages of the groups with the plesiomorphic state of a trait (propubic pelvis, carnivory, or ischiotruncus muscle) from the percentages of the ones with the apomorphic state (opisthopubic pelvis, herbivory, or reduced importance of ischiotruncus muscle), we obtain a single value (here called ‘Difference in Distribution’) that represents how unequally distributed the derived state of a trait is among members with the plesiomorphic and derived states of the other trait. Difference in Distribution should be calculated for taxa with the derived state of all three traits (Fig. 5): in this case, among herbivores, among groups with a reduced ischiotruncus muscle, and among opisthopubic dinosaurs. The higher this single value, the less uniform is the distribution of the studied traits among dinosaurs with a similar, derived feeding strategy, ventilation mode, or pelvic structure. A clear correlation between two derived states in these comparisons should show positive differences in distribution (up to a maximum of 100%) in both of them.
The reasoning behind the development of this methodology in our specific study case is exemplified by two scenarios testing for a potential correlation between herbivory and an opisthopubic pelvis. If herbivory was not an evolutionary driver for an opisthopubic pelvis, all the herbivorous dinosaurs should have had the same probability to evolve an opisthopubic pelvis and thus, with the specific sample choice we made, there should be a subequal number of opisthopubic herbivores and propubic ones. This distribution of groups among propubic and opisthopubic taxa would result in Difference in Distribution Herbivory values of 0%. If other factors meant that there were more propubic herbivores than opisthopubic ones in the sample, this value could be less than zero. At the same time, if feeding strategy was irrelevant to pelvic structure, herbivorous taxa could have evolved an opisthopubic pelvis with equal probability to carnivorous ones. Such a distribution would result in Difference in Distribution Opisthopubic values of 0%. If for other reasons there were more groups of propubic herbivores than of opisthopubic ones, the value concerning the opisthopubic structure would be less than zero. On the contrary, if herbivory was indeed an evolutionary driver for the retroversion of the pubis, herbivorous dinosaurs would have had a higher probability of evolving an opisthopubic pelvis than carnivorous ones. This would result in more opisthopubic herbivores than propubic ones, and most of the groups with an opisthopubic pelvis should be herbivorous, resulting in positive values of Difference in Distribution Herbivory and Difference in Distribution Opisthopubic, instead of zero or less as in the first example.
As proposed, our methodology provides a simple but effective way to compare the influence of a number of well‐defined traits on a specific evolutionary innovation. It does not allow us to test for all possible correlations within a dataset of morphological traits, but can serve to rule out single traits as potential evolutionary drivers or constraints for another. If both hypotheses returned a potentially evolutionary correlation after this simplified test, a statistical analysis using modern, phylogenetic comparative methods should be carried out.
ResultsOur results differ considerably between the two potential evolutionary forces acting on pubic orientation that have previously been proposed.
As far as feeding strategy is concerned, the Difference in Distribution values (Fig 5A) are equal or less than zero (Difference in Distribution Herbivory = −42.86/−33.33; Difference in Distribution Opisthopubics = 0/−20), arguing against a correlation between herbivory and pubic retroversion (Fig. 6). Moreover, groups that evolved an opisthopubic condition are herbivorous in only 40–50% of cases (depending on whether the Me or Op criterion is used), and strict herbivory appears after the evolution of an opisthopubic pelvis in all clades but Ornithischia and tentatively Alvarezsauroidea. Groups that evolved herbivory are opisthopubic in only about 30% of the cases, because several herbivorous dinosaur clades retained an anteriorly projecting pubis (Sauropoda, Ornithomimidae, Oviraptorosauria and Therizinosauria). This distribution argues against a potential influence of herbivory on pubic orientation within dinosaurs in general. However, see below for the particular case of therizinosaurs.
Figure 6 Open in figure viewerPowerPoint Graphical visualization of Difference in Distribution, reported from Figure 5. Colour online. CaptionGraphical visualization of Difference in Distribution, reported from Figure . Colour online.In contrast to the results obtained with herbivory, the Difference in Distribution values concerning ventilatory structure (Fig. 5B) are higher than zero (Difference in Distribution Non‐cuirassal = 14.29/42.86; Difference in Distribution Opisthopubics = 100 with both pubic angle thresholds; Fig. 6). Furthermore, the appearance of osteological correlates for a change in the ventilation system precedes pubic retroversion in all five clades in which both traits co‐occur; groups that have lost cuirassal ventilation are opisthopubic in about 60–70% of cases, and only two clades which retained a propubic pelvis exist with evidence for a change in the ventilatory structure (Oviraptorosauria and Eusauropoda). Pubic retroversion therefore seems to be more strongly correlated with the ventilatory system than with feeding strategy.
DiscussionAmong the two hypotheses tested for the possibility of being an evolutionary driver (in the case of herbivory) or constraint (in the case of the ischiotruncus muscle) of pubic retroversion in dinosaurs, the reduced importance of the ischiotruncus muscle during ventilation has been corroborated as an evolutionary constraint. However, this does not imply that a change in the ventilatory system was the only evolutionary force acting on the structure of the archosaurian pelvis. For instance, egg morphology, locomotion, nesting behaviour and reproductive organs could all have been equally influential. Because these characteristics are more difficult to recognize in the skeleton, we concentrated our study on the two previously proposed traits, for which relatively well‐supported osteological correlates exist.
The Difference in Distribution concerning ventilation strongly indicates that the ventilation mode was a more important evolutionary force acting on pubic orientation than the need for a larger gut during the evolution of herbivory. More specifically, our results indicate that pubic orientation is constrained towards a propubic condition as long as plesiomorphic cuirassal ventilation using the ischiotruncus muscle was present and any other way of helping to inflate the lungs (which could be truncocaudalis muscle taking the function of the M. ischiotruncus or a more efficient costal ventilation; e. g. using uncinate processes) is absent. This constraint is likely to have weakened progressively as other ventilation modes supporting the ischiotruncus muscle evolved or the costal ventilatory muscles improved their action on the ribs, allowing the pubis to rotate freely. The fact that our second test series including a larger angle in the ‘opisthopubic’ category (Me) produced even larger differences in trait distributions than the less inclusive angle (Op; Fig. 5) further corroborates the progressive weakening of the constraint, and the newly acquired freedom concerning pubic orientation in taxa with alternative ventilation systems.
On the other hand, the evolution of herbivory seems to be very weakly correlated with a posterior rotation of the pubis, when studying the innovations of the traits among dinosaurs in general. Given that all the values concerning feeding strategy are less than zero, they are not consistent with the hypothesis of herbivory being an evolutionary driver for the retroversion of the pubis. Moreover, the only clade with opisthopubic members in which herbivory is widely accepted is Ornithischia. However, a significant number of recent studies have questioned the herbivory of the basal‐most ornithischians, and suggested that they were more likely to have been facultative omnivores (Barrett 2000; Norman et al. 2004; Butler et al. 2010; Baron et al. 2017a; Sciscio et al. 2017). If this is true, the hypothesis that herbivory was not the evolutionary driver for the retroversion of the pubis is even more strongly confirmed, because it evolved only after the appearance of the opisthopubic pelvis. Conversely, the evolution of herbivory could possibly have been helped by the acquisition of a retroverted pubis, even if this hypothesis must be eventually confirmed by other studies.
Although feeding strategy might not be as strongly correlated with pubic orientation as the ventilation mode when analysing the entire clade Dinosauria, this does not necessarily imply that in some cases, herbivory could not have driven the retroversion of the pubis after the constraint was removed. Moreover, even if the ischiotruncus muscle was generally an evolutionary constraint on the pelvis, herbivory could have been a crucial evolutionary force in less inclusive clades. Indeed, therizinosaurs represent a particular case within Maniraptora. As mentioned above, the peculiar shape of their pelvis, with the posteriorly rotated proximal shaft and a strongly anteriorly projecting pubic boot (see above; Zanno 2010a; Fig. 1C), is generally considered to be opisthopubic, even if our measuring protocol defines it as propubic. Because of this issue, we repeated the analyses assigning this group as opisthopubic to see how this would have affected the results. The Difference in Distribution Opisthopubic value slightly increased as far as feeding strategy is concerned (by 20%) and slightly decreased in the case of ventilation (by 30–40%). The Difference in Distribution Herbivory increased but remained negative. The Difference in Distribution Ventilation remained the same. Thus, although the feeding strategy remains an improbable driver for the opisthopubic condition within dinosaurs in general, herbivory might have been a strong evolutionary force in Therizinosauria. Whereas the ventilatory muscles crucial for the cuirassal ventilation remained attached to the pubic boot, the larger gut helping to process a herbivorous diet were able to ‘push’ the proximal shaft backwards, simultaneously forming a well‐developed anterior projection of the pubic boot due to the constraint imposed by the cuirassal ventilation. The importance of cuirassal ventilation in Therizinosauria is confirmed by the absence of any osteological correlate for an alternative ventilation mode. However, it remains to be understood why herbivory became a strong enough evolutionary force to retrovert at least part of the pubis in therizinosaurs, but not in other dinosaurs such as sauropodomorphs, ornithomimids and oviraptorosaurs.
The identification of the presence of cuirassal ventilation as an important evolutionary constraint on pubic orientation, and the exclusion of herbivory as a general driver for pubic retroversion in dinosaurs leaves the question regarding evolutionary drivers of pubic retroversion open. However, we note that an opisthopubic condition evolved only in groups with bipedal locomotion, such as the earliest ornithischians and maniraptorans. Quadrupedality in ornithischians evolved later (Maidment et al. 2014) and had no influence on orientation of the pubis. Also, no member of Sauropoda (the only quadrupedal dinosaur clade in which cuirassal ventilation disappeared) has a retroverted pubis, even though they had probably lost the cuirassal ventilation system (see above), and even though they were strictly herbivorous (Upchurch et al. 2004). An opisthopubic pelvis might therefore have been advantageous for bipedal taxa, whereas in quadrupedal species there does not seem to have been any particular evolutionary preference for the orientation of the pubis. Alternatively, the highly variable pubic angles in maniraptorans which already acquired improved costal ventilatory muscles, might indicate that once the M. ischiotruncus became reduced, there was simply no strong selective force on pubic orientation anymore.
The subsequent fixation of a strongly retroverted pubis in derived avialans was probably influenced by other factors. As mentioned above, in addition to improved costal ventilation using uncinate processes, extant birds also actively use the pubocaudalis muscles to control ventilation of the posterior air sacs. The M. pubocaudalis internus and externus connect the tail with the pelvic girdle (Baumel et al. 1990) and probably derive from the M. ischiocaudalis, which connects the ischium and the tail in living reptiles (Hutchinson 2001). Within theropods, the M. ischiocaudalis was reconstructed as being connected to both the pubic and ischial boot in Avetheropoda (Hutchinson 2001). The subsequent loss of the ischial boot and gradual shortening of the ischium within Maniraptoriformes would imply that the attachment site of the M. ischiocaudalis was shifted completely onto the pubis in more derived maniraptorans, forming the pubocaudalis muscles (Hutchinson 2001).
The reorganization of the pelvic musculature during the retroversion of the pubis in Maniraptora did not only have implications for ventilation. The newly evolved pubocaudalis muscles of birds depress the tail when contracted bilaterally, and rotate it when contracted unilaterally (Moreno & Møller 1996). They are highly bilaterally active during flight in pigeons, particularly during lift‐off and downstrokes (Gatesy & Dial 1993, 1996a), whereas the M. pubocaudalis externus is continuously active during flapping flight (Gatesy & Dial 1993). Thus, they have an important role in tail fanning and manoeuvring, which are particularly crucial movements during lift‐off from the ground (Maynard Smith 1952; Gatesy & Dial 1996a, b). The combined occurrence of an opisthopubic pelvis and a shortened tail with pygostyle as insertion points of the pubocaudalis muscles and as support for the rectrices, bulbs and M. bulbi rectricium, resulted in a shortening of the pubocaudalis muscles, thereby increasing contraction speed (Moreno & Møller 1996), allowing for a fast and controlled take‐off from the ground. This unique combination of features at the base of the avian clade Pygostylia (Fig. 7) correlates well with a significant increase in evolutionary rates along the bird lineage (Brusatte et al. 2014).
Figure 7 Open in figure viewerPowerPoint Gradual assembly of morphological features necessary for powered flight plotted on a composite phylogeny (following Xu et al. 2014, Brusatte et al. 2015 and Hendrickx et al. 2015). The figure shows the retroversion of the pubis, the appearance of ossified uncinate processes, tail shortening with the accompanying development of a pygostyle, and weight reduction (from left to right). The pubis, uncinate processes, pygostyle, and symbols for low body mass are drawn in black. Pelves are shown in left lateral view with dorsal towards the top (modified from Heers & Dial 2012), the dorsal rib and uncinate process is drawn in right anterolateral view (traced from Clark et al. 1999), and tails are shown in left lateral view with the distal end towards the top (modified from Rashid et al. 2014). The arrows indicate the appearance of the features along the phylogenetic lineage leading to birds. Dotted portions indicate a gradual backwards rotation in the pubis and the unknown presence of uncinate processes in Alvarezsauroidea and Therizinosauria. Asterisks indicate convergent occurrences. Weight indications are approximative, and were taken from Benson et al. 2014. Colour online. CaptionGradual assembly of morphological features necessary for powered flight plotted on a composite phylogeny (following Xu et al. , Brusatte et al. and Hendrickx et al. ). The figure shows the retroversion of the pubis, the appearance of ossified uncinate processes, tail shortening with the accompanying development of a pygostyle, and weight reduction (from left to right). The pubis, uncinate processes, pygostyle, and symbols for low body mass are drawn in black. Pelves are shown in left lateral view with dorsal towards the top (modified from Heers & Dial ), the dorsal rib and uncinate process is drawn in right anterolateral view (traced from Clark et al. ), and tails are shown in left lateral view with the distal end towards the top (modified from Rashid et al. ). The arrows indicate the appearance of the features along the phylogenetic lineage leading to birds. Dotted portions indicate a gradual backwards rotation in the pubis and the unknown presence of uncinate processes in Alvarezsauroidea and Therizinosauria. Asterisks indicate convergent occurrences. Weight indications are approximative, and were taken from Benson et al. . Colour online.Given that a highly effective take‐off from the ground would have allowed nearly immediate lift‐off without having to run or climb first, escape from possible predators could have been vertical (into the air) instead of horizontal, and would thus have been much faster, providing a significant increase in individual fitness. Such an increase in fitness might at least in part explain their subsequent radiation.
ConclusionsThe orientation of the pubis and its repeated, convergent retroversion among dinosaurs was influenced by a number of different evolutionary constraints and drivers. Herein, we have analysed the influence of the two previously proposed factors that might have acted on it: the ventilation system and feeding strategy. Of the two, the more influential evolutionary force acting on pubic orientation appears to have been the ventilation mode. Our study shows that the presence of the ischiotruncus muscle for the plesiomorphic cuirassal ventilation of dinosaurs functioned as an evolutionary constraint inhibiting the retroversion of the pubis. All clades with a retroverted pubis lost cuirassal ventilation and/or evolved alternative ventilatory strategies that reduced the importance of cuirassal ventilation as a support for the main, costal one. In all of these clades, changes in the ventilatory structure evolved before or at the same time as the opisthopubic pelvis. The only other evolutionary driver that has previously been proposed, a switch to a herbivorous diet, almost always appeared at a more derived node within opisthopubic clades, with the possible exceptions of Ornithischia and the peculiar theropod clade Alvarezsauroidea, where the evolution of herbivory might have played a role in pubic orientation.
After a gradual release from the evolutionary constraint on the propubic condition, the combined occurrence of a retroverted pubis and a shortened tail in the avian clade Pygostylia, with the resulting short, and thus highly effective pubocaudalis muscles, improved manoeuvrability and stability during lift‐off, possibly leading to an increase in individual fitness thanks to the newly acquired capacity of vertical escape from predators.
AcknowledgementsLM thanks Massimo Delfino (University of Torino, Italy) for the interest shown in her ideas, for the active support and for the helpful advice and Renzo Levi (University of Torino, Italy) for his enlightening lessons, which led to the research question studied herein. The authors also thank Steve Brusatte (University of Edinburgh, UK), Rhiannon Meaden (The Royal Society, UK), Kevin Padian (University of California, USA), Vivian Allen (Royal Veterinary College, London, UK), Leon Claessens (College of the Holy Cross, Worcester, USA) and two anonymous referees for critical comments on earlier drafts of the manuscript, Simone Colombero (University of Torino, Italy) for comments on the abstract, Lorenzo Fatibene (University of Torino, Italy) and Sergio Castellano (University of Torino, Italy) for critical discussions concerning the statistical methodology, Raffaello Casu (University of Torino, Italy) for active support and mathematical advice, Hollie Bean (University of Birmingham, UK) and Shane Webb (University of Birmingham, UK) for proofreading the text. Finally, we thank the editors of Palaeontology (in particular Philip Mannion and Sally Thomas) for their support.