Unappreciated diversification of stem archosaurs during the Middle Triassic predated the dominance of dinosaurs

Background Archosauromorpha originated in the middle–late Permian, radiated during the Triassic, and gave rise to the crown group Archosauria, a highly successful clade of reptiles in terrestrial ecosystems over the last 250 million years. However, scientific attention has mainly focused on the diversification of archosaurs, while their stem lineage (i.e. non-archosaurian archosauromorphs) has often been overlooked in discussions of the evolutionary success of Archosauria. Here, we analyse the cranial disparity of late Permian to Early Jurassic archosauromorphs and make comparisons between non-archosaurian archosauromorphs and archosaurs (including Pseudosuchia and Ornithodira) on the basis of two-dimensional geometric morphometrics. Results Our analysis recovers previously unappreciated high morphological disparity for non-archosaurian archosauromorphs, especially during the Middle Triassic, which abruptly declined during the early Late Triassic (Carnian). By contrast, cranial disparity of archosaurs increased from the Middle Triassic into the Late Triassic, declined during the end-Triassic extinction, but re-expanded towards the end of the Early Jurassic. Conclusions Our study indicates that non-archosaurian archosauromorphs were highly diverse components of terrestrial ecosystems prior to the major radiation of archosaurs, including dinosaurs, while disparity patterns of the Ladinian and Carnian indicate a gradual faunal replacement of stem archosaurs by the crown group, including a short interval of partial overlap in morphospace during the Ladinian. Electronic supplementary material The online version of this article (doi:10.1186/s12862-016-0761-6) contains supplementary material, which is available to authorized users.


Taxon sampling
. List of Permian, Triassic and Early Jurassic taxa. Table S2. List of additional taxa from the Middle and Late Jurassic and Cretaceous used to calculate ancestral shapes. Table S3. Additional taxa for time calibration with data of occurrences. Table S4. Description of landmarks and semi-landmark description.  Table S5. Sources for phylogenetic position of taxa included in the informal supertree.   Table S7. Sum of variances of Archosauromorpha through time. Table S8. Statistical differences of sum of variances and morphospace position between subsequent time bins for all Archosauromorpha. Table S9. Statistical differences of sum of variances and morphospace position between subsequent time bins for non-archosaurian Archosauromorpha.

Disparity and NPMANOVA results
Table S10. Sum of variances of non-archosaurifom Archosauromorpha and non-archosaurian Archosauriformes through time. Table S11. Statistical differences of sum of variances and morphospace position between subsequent time bins for non-archosaurifom Archosauromorpha and non-archosaurian Archosauriformes when phytosaurs are members of the crown-group of Archosauria. Table S12. Statistical differences of sum of variances and morphospace position between subsequent time bins for non-archosaurifom Archosauromorpha and non-archosaurian Archosauriformes when phytosaurs are members of the stem-line of Archosauria. Table S13. Statistical differences of sum of variances and morphospace position between subsequent time bins for Archosauria. Table S14. Statistical differences of sum of variances and morphospace position between subsequent time bins for Pseudosuchia. Table S15. Statistical differences of sum of variances and morphospace position between subsequent time bins for Ornithodira/Dinosauria. Table S16. Statistical differences of sum of variances and morphospace position between different groups within time bin when phytosaurs are crownarchosaurs.           Table S24. Results of ordinary least squares (OLS) and generalised least squares (GLS) regression of skull shape disparity sample size per bin. Table S3. Additional taxa for time calibration with data of occurrences (in millions of years). The most dorsal point of the postorbital along the orbital margin.

11
The contact between jugal and postorbital along the margin of the lateral temporal fenestra/opening. For Simosuchus we plotted the most dorsal contact of the jugal along the anterodorsal margin of the lateral temporal fenestra.

12
The ventral contact between postorbital and squamosal (This contact is usually located along the dorsal margin of the lateral temporal fenestra/opening, but is different in some rhynchosaurs, aetosaurs, Dakosaurus, Postosuchus, Protosuchus and Yinlong).

13
The dorsal contact between postorbital and squamosal. fenestra (For those taxa where the fenestra is ventrally open due to a reduction of the quadratojugal, the most posterior point of the posterior process of the jugal was marked. In the reconstruction of Dysalotosaurus, we plotted this landmark at the most posterior point of the jugal along the ventral margin of the lateral temporal fenestra). S16-S17 Two semi-landmarks between LM1 and LM2 along the ventral margin of the premaxilla. S18-S20 Three semi-landmarks between LM2 and LM3 along the ventral margin of the maxilla.
S21-S35 15 semi-landmarks between LM1 and LM14 along the dorsal margin of the skull S36-S38 Four semi-landmarks between LM6 and LM7 along the anterior and posterior margin of the antorbital fenestra. In taxa lacking am antorbital fenestra, the semi-landmarks are placed along the suture contact of maxilla and lacrimal (Fig. S1).
S39-S43 Four semi-landmarks between LM8 and LM10 along the anterodorsal margin of the orbit.
S44-S45 Two semi-landmarks between LM10 and LM9 along the posterior margin of the orbit.

S46
One semi-landmark between LM8 and LM9 along the ventral margin of the orbit.
S47-S48 Two semi-landmarks between LM11 and LM15 along the anteroventral margin of the lateral temporal opening.
S49-S53 Five semi-landmarks between LM12 and LM14 along the ventral margin of the squamosal.

Phylogeny
To reconstruct ancestral shapes, we created two informal, time-calibrated supertrees based on recent literature. The main topology was based on Nesbitt (2011), but for one tree the position of phytosaurs were placed within crown Archosauria (Sereno 1991;Brusatte et al. 2010;Ezcurra 2016), while in the other phytosaurs are treated as non-archosaurian archosauromorphs following Nesbitt (2011). The sources of the phylogenetic positions of taxa not included in the analyses of Nesbitt (2011) are listed in Table S5. The correlation between shape and phylogeny was tested using multivariate K statistics on the basis of Procrustes-fitted landmark coordinates using the program R (R Development Core Team) and the package geomorph (Adams and Otárola-Castillo 2013). This method estimates the strength of a phylogenetic signal in a data set in relation to a simulated Brownian motion model and was performed with 1,000 replications (Blomberg et al., 2003;Paradis, 2012;Adams, 2014). The strength of the phylogenetic signal is expressed as K and p-values. The test reveals that skull shape of archosauromorphs correlates significantly with phylogeny, while the position of phytosaurs has no impact on the signal (Table S6). However, the K-value < 1 implies that the skull shapes of the terminal taxa resemble each other less than expected under Brownian motion evolution (Blomberg et al., 2003). To confirm these results we performed an additional permutation test in MorphoJ (Klingenberg 2011), shuffling the Procrustes-fitted landmark coordinates of each taxon randomly across the tree (10,000 times), while the topology is held constant (Laurin 2004;Klingenberg and Gidaszewski 2010). As found using K statistics, cranial shape correlates significantly with phylogeny as the squared length of the original supertree (= tree length) occurs in over 95% of the randomly generated trees ( Fig. S3; Table S6).

General disparity trend within archosauromorph cranium
The current study shows that after the end-Permian mass extinction, cranial disparity of archosauromorphs increased slowly during the Early Triassic, which is in agreement with a general delayed recovery of ecosystems and taxon diversity after the extinction event (Irmis and Whiteside 2012; Lau et al. 2016). However, the shift in morphospace (Table S8) (Table S16, S17), supporting a process of radiation during the Middle Triassic. This radiation corresponded also with an increase in body size (Sookias et al. 2012a,b) and reveals an ecological diversification into different niches and probable dietary preferences, including piscivory (e.g. Phytosauria, Proterochampsia), omnivory (basal Sauropodomorpha, Aetosauria), carnivory (e.g. Rauisuchidae, Theropoda) and herbivory (e.g. Rhynchosauria, Ornithischia) (see Butler et al. 2011;Nesbitt et al. 2013). The mass extinction at the end of the Triassic resulted in a decrease in taxonomic diversity, including the extinction of all stem line archosaurs and all noncrocodylomorph pseudosuchians, which in turn led to a decrease in cranial disparity.
During their coexistence, the cranial morphospace of non-archosaurian archosauromorphs and archosaurs overlaps during the Ladinian (this result is found with both phylogenetic positions of phytosaurs, Table S16, S17). However, this overlap is not unambiguous evidence for ecological competition between the two groups during this period, but could also result from a statistical artefact as the crown group members in this time bin tend not to show extremes of skull shape (e.g. Gracilisuchidae and various basal Loricata), while the stem line representatives include taxa with short and high (Allokotosauria) and long and flattened skulls (Protochampsia), resulting in similar medians (centroids) in morphospace. However, whether this overlap is a real signal needs to be tested in the future with more detailed analyses including multiple lines of evidence.

The effect of the phylogenetic position of Phytosauria
The disparity trajectories for non-archosaurian archosaurormorphs are greatly affected by the phylogenetic position of Phytosauria. When phytosaurs are considered the sister-taxon of Archosauria (Nesbitt 2011), the cranial disparity of non-archosaurian archosauromorphs increases from the late Permian to the Carnian, in which the changes between the Ladinian and the Carnian are significant. After a Carnian peak their cranial disparity decreases significantly in the early Norian, but re-expands slightly in the late Norian, before the group goes extinct at the end of the Late Triassic (Table S9). In contrast, the cranial disparity of Archosauria increases from the Early Triassic until the late Norian. From the late Norian into the Hettangian cranial disparity decreases, and then it expands again until the Toarcian (Fig. S3, Table S13).
The changes between the Hettangian and Sinemurian are significant. In this scenario, cranial disparity of non-archosaurian archosauromorphs is higher than that of Archosauria throughout the interval over which they coexist, but only significantly higher from the Early Triassic to the Carnian (Table S17). Non-archosaurian archosauromorphs exhibit a significant shift in morphospace from the late Permian to the Early Triassic and from the Early Triassic to the Anisian (Table S9), while in Archosauria these shifts occur from the late Norian to the Hettangian and from the Sinemurian to the Toarcian (Table S13). When compared to each other, both groups are significantly separated from each other in morphospace over the entire Late Triassic (Fig. S4), but not in the Early and Middle Triassic (Table S17). In sum, if phytosaurs were stem-archosaurs, non-archosaurian archosauromorph disparity far exceeds that of Archosauria during the whole Triassic, before the sudden extinction of all non-crown taxa at the end of the Triassic. In addition, the cranial disparity of Pseudosuchia and Ornithodira would be relatively equal during the Late Triassic (Fig.   S3), although both groups occupy different areas of morphospace (as found by Brusatte et al. 2008). In this scenario, the "true" radiation of crown archosaurs began

Cranial disparity trends within Ornithodira
Due to their poor record in earlier time bins, the oldest time interval for which we can estimate disparity for ornithodirans is the Carnian. From that interval, cranial disparity increases continuously until the Sinemurian, with both Ornithodira and Dinosauria showing significant shifts in morphospace from the Carnian to the early Norian (Fig.   4, S5). Disparity changes from the late Norian to Hettangian could not be estimated due to poor sampling in the later bin, but a significant increase is observed from the late Norian to the Sinemurian. In the Toarcian, cranial disparity decreases again (Table S15). When compared to each other, Pseudosuchia (including phytosaurs) and however, the morphospace of pseudosuchians and ornithodirans overlaps in the early Norian (Table S17). The general differences in cranial morphospace over time between ornithodirans and pseudosuchians suggests that the two groups probably did not compete extensively with each other, e.g. for similar food resources. However, due to the poor fossil record of ornithodirans prior to the Late Triassic (Langer et al. 2013), nothing can be said about potential ecological overlap in previous time periods.

The impact of ancestral shapes for disparity analyses
To study disparity changes through time we included the shapes of hypothetical ancestors to increase temporal resolution. To estimate the impact of the ancestors, we calculated the temporal disparity curves for all groups (when phytosaurs are members of crown Archosauria) on the basis of terminal taxa only and compared them with the results presented in the main text (Table S18-S23). Using only terminal taxa results in slightly higher disparity. However, as the magnitude of sum of variance, the disparity metric used in this study, depends on the sample size, the higher disparity values are a mathematical artefact as the sample size is reduced due to the exclusion of ancestral shapes. The shapes of the disparity curves through time are generally similar for both approaches (Fig. S6), so that we assume that the basic trends described in this study are still valid. The inclusion of ancestral shapes furthermore allowed us to estimate disparity for earlier time bins and for those with low sample sizes. This is especially true for the Early Jurassic, where the fossil record of pseudosuchians and ornithodirans is generally low, complicating the estimation of statistically meaningful disparity values (Fig. S6).

The age of the Chañares Formation
During the writing of this manuscript, Marsicano et al. (2015)