Macroevolutionary dynamics of dentition in Mesozoic birds reveal no long-term selection towards tooth loss.

Several potential drivers of avian tooth loss have been proposed, although consensus remains elusive as fully toothless jaws arose independently numerous times among Mesozoic avialans and dinosaurs more broadly. The origin of crown bird edentulism has been discussed in terms of a broad-scale selective pressure or trend toward toothlessness, although this has never been quantitatively tested. Here, we find no evidence for models whereby iterative acquisitions of toothlessness among Mesozoic Avialae were driven by an overarching selective trend. Instead, our results support modularity among jaw regions underlying heterogeneous tooth loss patterns and indicate a substantially later transition to complete crown bird edentulism than previously hypothesized (∼90 mya). We show that patterns of avialan tooth loss adhere to Dollo's law and suggest that the exclusive survival of toothless birds to the present represents lineage-specific selective pressures, irreversibility of tooth loss, and the filter of the Cretaceous-Paleogene (K-Pg) mass extinction.

Introduction

The toothless beak is among the most characteristic features of the crown bird bauplan, and toothlessness has been hypothesized to increase the evolvability of crown bird jaws, partially underlying their sensational morphological disparity (Louchart and Viriot, 2011). Numerous potential drivers of avialan tooth loss have been proposed, including selection for flight-related weight reduction (Dilger, 1957; Feduccia, 1998; Proctor and Lynch, 1998) (though see [Zhou et al., 2019; Lautenschlager et al., 2013]), selection for rapid developmental rates precluding tooth formation (Yang and Sander, 2018), dietary shifts favoring alternative modes of food processing (Louchart and Viriot, 2011; Lautenschlager et al. 2013; Wang et al., 2017; Ksepka, et al., 2019), and selection for preening efficacy (Naish 2014; Mayr 2016). However, consensus remains elusive, as fully toothless jaws arose independently numerous times among Mesozoic avialans and dinosaurs more broadly (Louchart and Viriot, 2011; Dilger, 1957).

The origin of crown bird edentulism has been discussed in terms of a trend or selection toward toothlessness (Dilger, 1957; Zhou et al., 2019; Lautenschlager et al., 2013; Yang and Sander, 2018), implying an overarching selective pressure favoring edentulism throughout avialan evolutionary history as opposed to localized selection acting on subsidiary lineages as interpreted in other studies (O’Connor, 2019, 2020). For example, Zhou et al. (2019) justified their assumption of overarching selection favoring tooth loss in Avialae on the basis that it evolved numerous times independently: “Tooth loss occurred independently in several early avian lineages (Zhou and Zhang, 2006; Zhou et al., 2009; Davit-Béal et al., 2009), suggesting that reduction and eventual disappearance of the dentition might have conferred some broadly applicable selective advantage” (Zhou et al., 2019, pg 39). However, no studies had previously tested whether avialan edentulism was associated with increased rates of diversification (i.e., higher speciation rates and/or lower extinction rates), as would be predicted by overarching directional selection for a toothless bill through the Mesozoic, nor had they evaluated whether patterns of tooth loss across Avialae actually follow trend-like patterns. Such an interpretation does not align with other clades in which edentulism has arisen numerous times independently, such as crown mammals (Davit-Béal et al. 2009), which have never been suggested to exhibit a generalized trend toward edentulism.

The hypothesis of long-term selection or a trend toward edentulism in Avialae can be tested in two ways: identifying a shift in diversification rates associated with tooth loss could indicate selection for toothlessness on a broad phylogenetic scale [Jablonski, 2008; Rabosky and McCune, 2010), whereas demonstrating that regional losses of teeth within the jaws are more likely to lead to further losses of teeth would support the progression of edentulism throughout the jaws in a trend-like manner. We apply state-dependent diversification methods and evolutionary modeling to show that tooth loss had a negligible impact on avialan diversification, providing no evidence for a link between toothlessness and accelerated cladogenesis. We also found no evidence for models whereby iterative acquisitions of toothlessness among Mesozoic Avialae were driven by an overarching selective trend.

Results and discussion

No long-term selection toward toothlessness in Avialae

State-dependent diversification (SDD) methods (FiSSE [Rabosky and Goldberg, 2017)] and HiSSE [Beaulieu & O'Meara, 2016]) reject the hypothesis that edentulism provided a selective advantage throughout the evolutionary history of Avialae (operationally defined here as the clade uniting Archaeopteryx and crown birds). Speciation rates inferred by FiSSE were higher in toothed avialans than in toothless avialans (Figure 1A). However, these differences were never significant (p < 0.05), implying that diversification rates were not substantially affected by the presence or absence of teeth (Table S2). HiSSE analyses fit a null model better than a state-dependent model, suggesting that avialan patterns of speciation and extinction were unrelated to tooth loss (Figure 1B). These results are consistent when considering transitions to complete toothlessness, as well as transitions to regional edentulism in the premaxilla, maxilla, and dentary (Figures S1–S6). We confirmed that these results are not an artifact of incomplete sampling with a taxonomic jackknifing approach (Close et al., 2015). Although the range of rate values and Akaike weight scores are greater in jackknifed datasets, we still obtained overwhelming support for a null model over a trait-dependant model of diversification (Figure S7) and found no significant difference between speciation rates of toothed versus toothless taxa (Figure S8).

Figure 1

Results of state-dependent diversification analyses

(A) Violin plots showing speciation rates of toothed and toothless avialan lineages through the Mesozoic inferred from FiSSE analysis. Width of violins indicates the density of speciation rate values inferred from 100 time-calibrated trees.

(B) Violin plots showing support for Null (no connection between diversification rates and presence or absence of teeth) and Trait-dependent (diversification rates depend on presence/absence of teeth) models inferred from HiSSE analysis. Width of the violin indicates the density of Akaike weights scores inferred from the 100 trees. Ichthyornis (toothed stem avialan) skull model modified from (Field et al., 2018b), and Asteriornis (toothless crown bird) skull model modified from (Field et al., 2020).

We also found no evidence for a general trend toward edentulism throughout avialan evolutionary history. Loss of teeth in one region of the upper jaw (e.g., premaxilla) did not increase the likelihood of tooth loss in another region of the jaw (e.g., maxilla), and vice versa. Instead, model fits supported the premaxilla and maxilla as discrete modules with largely independent trajectories with respect to tooth loss and tooth retention (Figure 2A), which is understandable in light of the differing populations of primordia, tissues, and cells from which these elements arise (Helms and Schneider, 2003; Schneider and Helms 2003; Wang et al., 2020a). The independent patterns of tooth loss in the premaxilla and maxilla provide evidence against strong directional selection driving transitions to toothlessness across avialans. Although there are several instances of tooth-loss in the premaxilla coinciding with that in the maxilla (e.g., Confucuisornithidae, Gobipteryx, Schizooura), there are also numerous instances where the evolution of toothlessness in one region occurred independently of the other (e.g., Jeholornis, Longipterygidae), or where the evolution of toothlessness in one region preceded that in the other by more than 30 million years (Ornithurae). These idiosyncratic patterns, combined with the results of our model-fitting analyses, indicate that transitions toward regional or complete tooth loss in Avialae reflect lineage-specific patterns, rather than a broad-scale trend toward toothlessness across Avialae.

Figure 2

Violin plots showing support for models of dependent (integrated) versus independent (modular) evolution of discrete dental characters

Width of the violin indicates the density of Akaike weights scores inferred from the 100 trees. ER: Equal Rates; ARD: All rates different

(A) Results for degree of integration of presence/absence of teeth in the premaxilla (red bone) and maxilla (green bone).

(B) Results for degree of integration of presence/absence of teeth in the anterior and posterior portions of the dentary (blue bone). Ichthyornis skull models modified from (Field et al., 2018b)

In contrast to the largely modular behavior of the anterior and posterior regions of the upper jaw, the anterior and posterior regions of the lower jaw exhibit tighter integration, most likely because the only ancestrally tooth-bearing element (the dentary) comprises a single bone derived from paired mandibular primordia during embryonic development (Schneider and Helms 2003) (Figure 2B). However, several avialan fossils illustrate that teeth may be present or absent throughout the entire lower jaw, or in either the anterior or posterior region of the dentary (e.g., Jeholornis only exhibits teeth in the anterior portion of the dentary [Zhou and Zhang 2002]), demonstrating that, although more integrated than the upper jaw, the lower jaw may exhibit comparable patterns of regional edentulism (Figure 2B).

Idiosyncratic trajectories toward full edentulism are observed across Avialae. For example, within Enantiornithes, instances of maxillary tooth loss preceding premaxillary tooth loss are observed (Figure 3) and are further substantiated by the recent description of the probable latest Cretaceous enantiornithine Falcatakely (O’Connor et al., 2020). By contrast, the route to neornithine toothlessness began with the loss of premaxillary teeth from otherwise fully toothed jaws: Hesperornithes and Ichthyornis (the immediate outgroups to Neornithes) exhibit only maxillary teeth in the upper jaws (Figure 3). However, this directionality is not universal among ornithuromorphs; taxa within Schizoouridae appear to have lost their maxillary teeth first (Schizooura is entirely edentulous, whereas Mengciusornis exhibits only premaxillary teeth) (Wang et al., 2020b).

Figure 3

Evolution of complete tooth loss in ancestrally tooth-bearing jaw bones across Avialae

(A–C) Maximum clade credibility trees with branch colors corresponding to the proportion of stochastic maps indicating a toothless state along phylogenetic branches through time. (A) Results for premaxilla; (B) results for maxilla; (C) results for dentary.

(D and E) Histogram indicating estimates of the point in Earth history when the transition to toothlessness homologous with that of crown birds took place in various jaw bones: (D) the maxilla (the last tooth-bearing bone in the upper jaw along the neornithine stem lineage) and (E) the dentary (lower jaw). Dashed vertical lines represent median age estimates for these transitions. Ichthyornis skull models modified from (Field et al., 2018b)

The origin of crown bird edentulism has been discussed in terms of a broad-scale selective pressure or trend toward toothlessness (Dilger, 1957; Zhou et al., 2019; Lautenschlager et al., 2013; Yang and Sander, 2018), although the hypothesis of such dynamics underlying the origin of crown bird toothlessness has not been quantitatively tested. In light of our results, we contend that the numerous iterative transitions to toothlessness in Avialae reflect selection on a phylogenetically localized scale combined with an underlying developmental propensity for tooth loss, instead of the outcome of long-term directional selection.

We posit that explorations of the evolutionary dynamics of Mesozoic avialan edentulism have been unduly biased by the fact that all crown birds are toothless. Indeed, the ubiquity of edentulism in crown birds—characterizing all >10,000 living species, and the entirety of the neornithine fossil record going back to the latest Cretaceous (Field et al., 2020)—appears to be a mere artifact of survival. Avian survivorship to the present day was strongly influenced by the end-Cretaceous mass extinction 66.02 Ma (Clyde et al., 2016), a single event through which only Neornithes are known to have survived (Feduccia, 1995; Longrich et al., 2011; Prum et al., 2014; Field et al., 2018a). Although the toothless bill of avian survivors may have been well suited to exploiting resources such as insects and seeds in the extinction's aftermath (Larson et al., 2016; Field et al., 2018a), the fact that only toothless avialans have persisted to the present day appears to have more to do with patterns of extinction and survival across the end-Cretaceous mass extinction event—which may have been unrelated to dentition—than with any long-term directional trend toward avialan edentulism throughout the Mesozoic.

Timing the origin of crown bird toothlessness

Crown birds comprise a single clade (Neornithes) to the exclusion of all known toothed avialans, and their present-day edentulism presumably reflects a single evolutionary transition to toothlessness in the Late Cretaceous (Meredith et al., 2014a). Our median estimate for the timing of the transition to full edentulism in the upper and lower jaws homologous with the condition in crown birds is slightly less than 90 mya (Figure 3). The median age estimate for the transition to a toothless maxilla homologous with that of crown birds is 88.46 mya (range of estimates = 104.33–67.25, with a skew toward earlier ages), and the median age estimate for the transition to a toothless dentary homologous with that of crown birds is 87.55 mya (range = 104.28–67.19, with a similar skew). Both of these are substantially younger than previous estimates (116 mya [Meredith et al., 2014a]).

Dollo's law and edentulism

Dollo's law (Gould, 1970) posits that complex features lost through evolutionary change are unlikely to be re-acquired in the same form. Tooth loss has been cited as a conspicuous example of Dollo's law: of the numerous avialan lineages that have evolved complete edentulism, none are thought to have undergone a reversal to a toothed state (Louchart and Viriot, 2011; Collin and Miglietta, 2008) (although complete dentition in Enantiornithes optimized as an evolutionary reversal in at least one parsimony analysis [Turner et al., 2012], this inference has been interpreted as unlikely to be accurate [Mayr 2016]). In a notable non-avian example, teeth are thought to have re-evolved in the previously toothless dentary of the frog Gastrotheca guentheri (Wiens 2011). However, unlike the situation in crown birds, the ancestral condition for G. guentheri was not complete edentulism, because frogs maintain teeth in their upper jaw. Despite independently acquiring tooth-like projections formed by the rhamphotheca and sometimes the underlying jaw bones several times (e.g., Pelagornithidae, mergansers [Anatidae: Mergini], plantcutters [Phytotoma: Cotingidae] (Louchart et al., 2013, 2018; Kennedy, 1948; Kirwan and Green 2012), crown birds have never reacquired teeth. Talpid mutant chicks can develop tooth buds (Harris et al., 2006), illustrating that at least some birds retain the genetic underpinnings of tooth development, but this is a lethal mutation. Indeed, the genome of the red junglefowl shows that genes necessary for enamel formation have been lost, precluding the full development of teeth (International Chicken Polymorphism Map Consortium, 2004).

Our model-fitting analyses support the irreversibility of tooth loss across Avialae (Figure 2, Table 1): Patterns of toothlessness across avialan phylogeny best fit models of evolution where transitions from a toothless to a toothed state have instantaneous transition probabilities of zero (Table 1). Stochastic mapping across our maximum clade credibility tree under the best-fitting evolutionary model suggests at least five independent transitions to complete edentulism among the taxa sampled in our phylogeny (Figures 3A–3C): preceding the origin of Confuciusornithidae, within Enantiornithes, separately along the lineages leading to the ornithuromorphs Archaeorhynchus and Schizooura, and preceding the origin of crown birds (Neornithes). Indeed, Mengciusornis—a recently described Cretaceous ornithuromorph inferred to be a close relative of Schizooura—exhibits teeth in its premaxilla (Wang et al., 2020b), helping to corroborate our inference that the transition leading to toothlessness in Schizooura occurred independently from that of Archaeorhynchus. Additional potentially independent transitions to toothlessness among taxa not sampled in our phylogeny include the ornithuromorph Eogranivora (Zheng et al., 2018), and the ornithothoracine Xinghaiornis (Wang et al., 2013).

Table 1

Instantaneous transition probabilities of tooth loss in different regions of the jaw

	Model	Transition probabilities	To
		From	0 | 0	0 | 1	1 | 0	1 | 1
Premaxilla | maxilla	Independent ARD	0 | 0	–	0.016	0.016	0
		0 | 1	0	–	0	0
		1 | 0	0	0	–	0.016
		1 | 1	0	0	0	–
Anterior dentary | posterior dentary	Dependent ARD	0 | 0	–	0.029	0.238	0
		0 | 1	0	–	0	0.052
		1 | 0	0	0	–	0
		1 | 1	0	0	0	–

Inferred for the best models of evolution of toothlessness in each region of the jaw, fit to the maximum clade credibility tree. Character score of 0 = presence of teeth; 1 = absence of teeth. ARD: All rates different.

Notably, our stochastic mapping approach did not identify any probable reversals from a toothless state to a toothed state in the premaxilla, maxilla, or dentary throughout avialan evolutionary history. Although repeated independent acquisitions of pointed jaw projections such as pseudoteeth provide prima facie evidence for the selective benefit of tooth-like projections in some neornithine subclades subsequent to tooth loss (Louchart et al. 2013, 2018), these instances have never led to a reversal to a toothed state. As such, avian edentulism adheres to the expectations of Dollo's law.

Limitations of the study

Limitations of the present study are primarily related to the nature of the Mesozoic avialan fossil record. All fossil records are incomplete, and that of Mesozoic birds has been shown to be biased by incomplete preservation and heterogeneous human sampling (Brocklehurst et al., 2012). The incompleteness of many Mesozoic avialan specimens (Brocklehurst et al., 2012) imposes limits on the methods we were able to use to evaluate evidence for sustained selective pressures or trends throughout avialan evolutionary history. With precise information on tooth numbers for a greater proportion of Mesozoic avialans it would have been possible to treat tooth number as a continuous character, enabling exploration of selective pressure toward tooth loss in the context of an Ornstein-Uhlenbeck model (in which a trait is drawn to an adaptive peak), or explicit testing of continuous trend-like models. However, as most of the taxa investigated here are represented by variably incomplete fossil remains, requiring such precise counts of tooth numbers would have necessitated the exclusion of many taxa from our analyses. Instead, discrete characters representing presence or absence of teeth in particular regions were generated, with SDD methods used to assess selection and analyses of character interdependence applied to assess trends. We consider this approach to represent a satisfactory compromise in light of the available data.

SDD methods have been shown to be impacted by incomplete sampling (Harvey and Rabosky, 2018; FitzJohn et al., 2009). However, by using a taxonomic jackknifing approach (where analyses were repeated on a dataset from which taxa were removed at random [Close et al., 2015]), we illustrated that variations in our taxon sample did not induce substantial changes in our results. We therefore consider our sample adequate to support robust conclusions on state-dependent diversification.

Conclusions

It may be tempting to assume that the origin of neornithine endentulism reflects the culmination of an overarching selective pressure or general evolutionary trend toward toothlessness across Avialae. However, transitions to toothlessness in all or part of the jaw evolved via alternative pathways among Mesozoic avialans (Figure 3), and these do not appear to have been associated with increased speciation rates or reduced extinction rates (Figure 1). It is clear that the apparently irreversible transition to neornithine edentulism reflects only one of several transitions to toothlessness throughout the Mesozoic evolutionary history of Avialae, Theropoda, and Reptilia more broadly (Louchart, and Viriot, 2011; Dumont et al., 2016; Pérez-Moreno et al., 1994; Ma et al., 2017; Li et al., 2008), and that at least some independently edentulous non-crown avialans were present in the Late Cretaceous (Elzanowski 1977; Chiappe et al., 2001). We suggest that the ubiquity of the toothless beak among Neornithes simply reflects the sole survival of this single edentulous avialan subclade across the end-Cretaceous mass extinction event, and that neornithine survival may have had little to do with any specific advantages conferred by toothlessness. This interpretation emphasizes the potential for historical contingencies to govern the origin of some of the most ubiquitous morphological patterns observable in the modern world.

Methods

All methods can be found in the accompanying transparent methods supplemental file.