Timing of organogenesis support basal position of turtles in the amniote tree of life

Background The phylogenetic position of turtles is the most disputed aspect in the reconstruction of the land vertebrate tree of life. This controversy has arisen after many different kinds and revisions of investigations of molecular and morphological data. Three main hypotheses of living sister-groups of turtles have resulted from them: all reptiles, crocodiles + birds or squamates + tuatara. Although embryology has played a major role in morphological studies of vertebrate phylogeny, data on developmental timing have never been examined to explore and test the alternative phylogenetic hypotheses. We conducted a comprehensive study of published and new embryological data comprising 15 turtle and eight tetrapod species belonging to other taxa, integrating for the first time data on the side-necked turtle clade. Results The timing of events in organogenesis of diverse character complexes in all body regions is not uniform across amniotes and can be analysed using a parsimony-based method. Changes in the relative timing of particular events diagnose many clades of amniotes and include a phylogenetic signal. A basal position of turtles to the living saurian clades is clearly supported by timing of organogenesis data. Conclusion The clear signal of a basal position of turtles provided by heterochronic data implies significant convergence in either molecular, adult morphological or developmental timing characters, as only one of the alternative solutions to the phylogenetic conundrum can be right. The development of a standard reference series of embryological events in amniotes as presented here should enable future improvements and expansion of sampling and thus the examination of other hypotheses about phylogeny and patterns of the evolution of land vertebrate development.

• Table S1 Species used in this study. •

Table S2
List of events observed. •

Table S3
Relative timing of events and stages for species studied. •

Table S4
Matrix listing the temporal ranks of the events for each species.
• Table S5 Step-by-step protocol for a Parsimov analysis. •

Table S6
Comparison of Parsimov results and tree lengths (PAUP mapping) for each tested topology (unordered characters). •

Table S7
Comparison of Parsimov results and tree lengths (PAUP mapping) for each tested topology (ordered characters). •

Table S8
List of temporal shifts that are autapomorphic for the nodes in the best-supported hypothesis (unordered characters). •

Table S9
List of temporal shifts that are autapomorphic for the species in the best-supported hypothesis (unordered characters). •

Figure S1
All hypotheses for the position of turtles within Tetrapoda summarised (extension of Figure 1).

Figure S2
Alternative hypotheses for the relationships of turtle subgroups (modified from Figure 1).

Figure S3
Best-supported topology (modified from Figure 4) including the number of nodes that are referred in Table S8. •

Figure S4
Phylogentic reconstruction resulting from an event pair based PAUP* analysis using unordered characters. •

Figure S5
Phylogentic reconstruction resulting from an event pair based PAUP* analysis using ordered characters. •

References
References to supplementary information.
Online data files: • turtles.ep.nex Nexus data file containing the event paired matrix for all observed species (open i.e. with PAUP* or Mesquite software). Based on the rank table (Table S4), the event pairs were generated with eventpairsim.pl (see Table S5): The timing of two events (ranks) is compared among each other, generating an event pair: When event A occurs before B the event pair is coded as state "0". When event A and event B occur at the same time the event pair is coded as state "1" (simultaneity). When event A occurs after event B the event pair is coded as state "2".

Tachyglossus aculeatus
2 Generate an Excel file listing numbers of characters (events) in the first column and the name of these characters in the second column. Save as txt file (names.txt). 3 Based on stages.xls, rank the stages temporally -each species separately -by the following principle: In the first species: characters "A", "B", "C" occur first (rank 1), characters "D", "E", "F" occur second (rank 2), character "G", "H" occur third (rank 3). In a second species the character sequence is different: characters "A", "E", "F" occur first (rank 1), characters "B", "D", "G" occur second (rank 2), character "C", H" occur third (rank 3), etc.

4
Eliminate the first line (the event/character names) of ranks.xls because in the following steps only species/numbers lines are used by the programs. Information on event/character names does not get lost and in step 14 they will be replaced. Save this as a new file, turtles.txt (tabstop txt file), in a folder C:\analysis.    15 Now the numbers are replaced by the event names in the two files. Parsimov results and tree lengths (PAUP* analyses) for the tested hypotheses of the turtle origin within Tetrapoda (A-H) and the relationships of turtle crown taxa (a-c, compare Figure 1, Figure S1 and S2). Best-supported trees in bold type. Parsimov: Consensus (CON), Acctran (ACC), Deltran (DEL) of the ties-including analyses (ties = "incomplete" shifts from "0" to "1" or "2" to "1"). The first number in brackets is the count of consensus shifts in a ties-excluded analysis. The second number in brackets is the count of consensus shifts in both the ties-including and ties-excluding analyses. PAUP*: Consistency index (CI), homoplasy index (HI), CI excluding uninformative characters (CIe), HI excluding uninformative characters (HIe), Retention index (RI), Rescaled consistency index (RC) are listed. The Sauria + Testudines hypothesis ("C") combined to the cryptodire "b" hypothesis is the one with the highest Parsimov support. Note that the position of turtles next to all remaining amniotes (tree 1-3) could not be tested with Parsimov due to the use of only one outgroup (Ambystoma mexicanum): the direction of heterochronic shifts leading either to the outgroup or to the ingroup is not detectable. References for each hypothesis listed below.  Parsimov results and tree lengths (PAUP* analyses) for the tested hypotheses of the turtle origin within Tetrapoda (A-H) and the relationships of turtle crown taxa (a-c, see Figure 1, Figure S1 and S2). Best-supported tree in bold type. Parsimov: Consensus (CON), Acctran (ACC), Deltran (DEL) of the ties-including analyses (ties = "incomplete" shifts from "0" to "1" or "2" to "1"). PAUP*: Consistency index (CI), homoplasy index (HI), CI excluding uninformative characters (CIe), HI excluding uninformative characters (HIe), Retention index (RI), Rescaled consistency index (RC) are listed. The Sauria + Testudines hypothesis ("C") combined to the cryptodire "b" hypothesis is the one with the highest Parsimov support (tree No. 8). Note that the position of turtles next to all remaining amniotes (tree 1-3) could not be tested with Parsimov due to the use of only one outgroup (Ambystoma mexicanum): the direction of heterochronic shifts leading either to the outgroup or to the ingroup is not detectable. For references and illustrations of hypothesis tested, see Table S6.  List of temporal shifts that are autapomorphic for the nodes in the best-supported hypothesis (Sauria+Testudines/Cryptodira"a"). First character occurs L=late or E=early in relation to the second character(s). For shifts charactersing Testudines and Sauropsida (Testudines + Sauria) asterisks mark ties ("incomplete" shifts from "0" to "1" or "2" to "1"). For Sauropsida consensus shifts are marked by asterisks that occur both in the tie including analysis (as presented for all remaining taxa) and in the analysis excluding tie shifts (so called "complete shifts"). For referred nodes compare with Figure S3. For abbreviations, see Table S2.  List of temporal shifts that are autapomorphic for the terminal branches (species) in a topology assuming the best-supported hypothesis (Sauria + Testudines/Cryptodira"b" relationship, Figure 5). First character occurs L=late or E=early in relation to the second character(s). For referred nodes compare Figure S3. For abbreviations, see Table S2.  All hypotheses for the position of turtles within Tetrapoda summarised as tested in this study (A-H, modified from Figure 1). For references and results each hypothesis see Table S6 and S7. Images modified from different sources and representing used species.

Figure S2 -Alternative cryptodire arrangements
Alternative hypotheses for the relationships of turtle subgroups as tested in this study (modified from Figure 1). a-c show different cryptodire groups as basal taxa. For references and results each tree combination see Table S6 and S7. Images modified from different sources and representing used species.

Figure S3 -Best-supported topology
Best-supported topology (modified from Figure 4) including the numbers of nodes as referred in Table S8. Images modified from different sources and representing used species.