Discuss And Critically Evaluate The Competing Hypotheses For The Origin Of Turtles

An important and largely unanswered question within palaeobiology is the sudden appearance of testudines in the fossil record and the identity of their ancestors. The unique chelonian anatomy, with shoulder and pelvic girdles inside the ribcage, which has expanded to form a shell, makes turtles a curious group to look at in terms of their derivation from traditional tetrapod anatomy. To define a clear ancestry, we must have agreed upon the systematic place of turtles so that we can trace their ancestry correctly. Unfortunately, their origin remains a highly contested issue, debate persists as to where turtles should be placed within amniota. According to Wang (2013), there are three major hypotheses which offer explanation for the evolutionary origin of turtles, all of which have been backed by varying evidence and have significant literature behind them, these are displayed in Figure 1. Firstly, 'A', the traditional view that they are part of the sister group to Diapsida, the largely extinct anapsida group, being classified as parareptiles (Laurin and Reisz, 1991; Lee, 2004; Lyson, 2010). Secondly, 'B' and 'C', that they are members of the lepidosauromorpha clade as either true lepidosaurs or sauropterygia (deBraga and Rieppel, 1997; Li et al., 2008) or finally, 'D', the hypothesis with the most popular support, that they are a sister group to archosaurs. (Kirsch and Mayer, 1998; Wang, 2013). Figure 1: Alternate hypotheses of testudine origin. Taken from Scheyer (2007); pp. 10.

Hypothesis 'A' is the traditional view on the evolutionary origins of turtles and is one broadly based on morphological evidence as molecular analysis only became available in the late 1990s. Outlined in Benton (2005), turtles were initially grouped with parareptilia by academics such as Lee (1993) and Laurin and Reisz (1991), making them the only extant anapsids. However, there was still debate between hypotheses even at this stage, with Lee arguing that they were most closely related to pareiasaurs, using Progantocheyls quenstedti as the stem turtle, whilst Laurin and Reisz argued that their closest relatives were procolophonids, utilising Owenetta as a transitional form. As mentioned, both of these theories were based on morphology, turtles share a number of unique characters with both groups, such as the lack of a tympanum with procolophonids and dermal amour with pareiasaurs. Ultimately, pareiasaur hypothesis had the better evidence behind it including a phylogenetic analysis of 128 osteological characters that strongly supported the hypothesis (Lee, 1997). In addition, the characters shared by turtles and procolophonids may easily have been a result of convergent evolution. However, the pareiasaur hypothesis suffered with the discovery of Odontochelys semitestacea, in China (Li et al., 2008) which offered a novel insight on the origin of the turtle shell. The rib structure of Odontochelys suggested that the plastron had evolved prior the carapace, which corresponds with the ossification of the shell in extant turtle embryos. By contrast, pareiasaur shells are thought to have formed from development of scutes to dermal bone plates, obviously with the carapace developing before the plastron (Yates, 2008).

However, despite Lee's theory being refuted by the discovery of Odontochelys, we must consider that the origin of turtles may lie with parareptillians of another variety. Interestingly, the embryological data presented by Werneburg and Sanchez-Villagra (2009) supports the basal position of turtles as a sister group to diapsida, lending support to a parareptillian affiliation and offering alternative evidence to support it. Lyson also states the observed sequence of ossification in embryos support turtles being placed outside of diapsida and suggested Eunotosaurus africanus as a replacement transitional form, lending support to a newly revived parareptillian hypothesis (Lyson et al., 2010; 2013). He argues that the strongest signal in morphological data sets firmly places Testudines as a sister group to Diapsida. He identifies 'six unequivocal synapomorphies' congruent with parareptillian classification, stating that both Odontochelys and Proganochelys exhibit these and Eunotosaurus to a lesser degree (2010). Analysing the most recent parareptillian phylogeny including stem turtles Odontochelys and Proganochelys, Lyson finds that there is strongest support for an exclusive Turtles + Eunotosaurus clade, demonstrated in Figure 2, which lies as a sister to pareiasaurs and procolophonids; he found that adding Eunotosaurus to the dataset only increased this support. The argument in favour of parareptillian turtles highlights the importance of integrating morphological data with the molecular to get the most accurate analysis, as both data sets are sensitive to bias (Lee et al., 2004). Furthermore, it emphasizes the importance of utilising embryological data as it provides unique 'cross-illumination' that is invaluable when studying vertebrate evolution (Burke, 2009). Figure 2: Cladogram showing proposed relationship of stem turtles to extant forms, Lyson's hypothesised Eunotosaurus + turtle clade is highlighted in red. Taken from Lyson et al. (2010); pp. 2.

There is little current support for a parareptillian origin, with most academics coming to the conclusion that turtles are a sister group to archosaurs, hypothesis 'D'. Molecular gene analysis, both mitochondrial and nuclear, has become a central technique in these studies and it unanimously supports an archosaur affinity. Nearly all early molecular analyses (Kirsch and Mayer, 1998; Hedges and Poling, 1999) placed turtles firmly within Diapsida and associated closely with archosaurs. Some of these studies can be regarded as questionable, with Hedges and Poling finding tuataras to be more closely related to archosaurs than to squamates (1999). However, it seems that many, if not all, of the more recent and comprehensive genome studies continue to support this hypothesis. One of the papers that most strongly supports the consensus of molecular analyses is a study done by Wang et al. (2013). They performed the first genome-wide phylogenetic analysis on the genomes of C. mydas and P. sinensis, which 'robustly indicated' that turtles are likely to be a sister group of archosaurs, having found over 1110 orthologous sets of genes. Their analysis indicates that the divergence from archosaurs took place 257.4 million years ago, well encompassing Odonotocheyls (220 Million years ago). Interestingly, this also places the divergence at approximately the same time as the Permian-Triassic mass extinction event, which suggests that perhaps the emergence of turtles is related to the extinction (Wang et al., 2013).

Whilst some argue that there is little morphological support for a turtle + archosaur relationship (Lyson et al., 2010), there are been several morphological characters in favour of this affinity. A comparison of basal turtle shell histology with several other 'osteoderm-bearing' taxa including pareiasaurs, placodonts, archosauromorphs and lepidosaurs, suggests a close relationship with archosaurs (Scheyer, 2007). Secondly, the laterosphenoid bone, shown in Figure 3, is nearly identical in basal turtles, such as Proganochelys and Kayentachelys aprix, and archosaurs, although modern turtles do not possess the bone at all (Bhullar and Bever, 2009). When including the laterosphenoid in phylogenetic tests, it places testudines as a sister taxa to archosaurs, in accordance with the molecular data. An important morphological implication for a turtle + archosaur clade is that it would mean that the temporal fenestrae in the turtle skull were most likely secondarily lost in the turtle lineage (Wang et al., 2013). This is not necessarily a problem, as temporal fenestrae have proved to be highly variable within diapsids, particularly in squamates, and turtles have emarginations which provide the same functionality. However, it would imply there is an older fossil turtle we have yet to find with a clearly diapsid skull configuration. Figure 3: Laterosphenoid in Proganochelys (left) and an archosaur, Crocodylus porosus (right). Taken from Bhullar and Bever (2009); pp. 3.

Gilbert and Corfe (2013) review Wang's findings alongside those of Shaffer et al. (2013), a study which also supported a turtle + archosaurs clade, but question whether these two studies provide the definitive answer to the question of turtle origins. The paper discusses at length that despite the convincing argument of the new data, there are still many unanswered questions and that there may be biases within molecular studies. They suggest that including other non-molecular sources of data would enable 'truly simultaneous analysis of fossil and molecular data', such as phenomic data sets for fossil and neontological turtle species which have been used by O'Leary et al. (2013) in a study on mammalian ancestors. Additionally, long branch attraction has been suggested by many academics (Lee, 2004; Lu et al., 2013) to artificially pair turtles and archosaurs. This is a common issue when working with molecular phylogenies due to the inherent limitations of working with DNA sequences. With this bias in mind, Lu et al. analysed 4500+ genes and concluded an uncertain phylogenetic placement of turtles. Indeed, given that there are constant conflicts between morphological and molecular data, not only with turtles but with many other aminote clades, it can be considered premature to firmly conclude any placement. (Field et al., 2014).

If then, it is premature of us to definitively rule turtles as a sister taxon to archosaurs, we must consider another alternative, hypotheses 'B' and 'C'. It has been argued, initially by deBraga and Rieppel (1997) that morphological evidence supports the placement of turtles among the lepidosauromorph clade. Their results refute earlier hypotheses of parareptillians affinity, and suggest a close relationship with sauropterygia and placodonts. There are several important morphological characters shared by turtles and lizards, although these are mostly post cranial. In particular, the fusion of astragalus and calcaneum in the ankle is noted as one of the most significant for validating the turtle + lepidosauromorph connection (Lyson et al., 2012). With cladistics work, Rieppel and Reisz admit that the nature of this work is very dependent on which taxa are included in the analysis, as removal of sauropterygians from their dataset results in turtles being sister group to pareiasaurs (1999). However, they believe that the most accurate and appropriate osteological data available indicates a sister relationship with sauropterygia (Rieppel and Reisz, 1999). The presence of morphological evidence is very important for strengthening the hypothesis as by contrast, there is very little morphological support for the molecular supported pairing of turtles and archosaurs (Rieppel, 2000). The inclusion of Odontochelys into phylogenetic analyses supported and provided further credence to the Sauropterygia origin theory (Li et al., 2008).

In another potential advance, Lyson et al. (2012) introduced a new molecular dataset utilising presence or absence of specific microRNAs. The conclusion is that this dataset unambiguously supports a turtle + lepidosaur grouping. However, the study was revisited by Field et al. (2014) and they found that by expanding the dataset and employing more rigorous criteria, there is support for an archosaur sister relationship rather than lepidosaur. In addition to the amendment of this study, there are also many criticisms of the lepidosaur hypothesis, arguably the most important of which is that the lepidosaur argument lacks a supporting transitional form. A proposed transitional form Priscochelys hegnabrunnensis was suggested by Karl (2005), but was later disproved and shown to be a placodont by Scheyer (2008). Secondly, the idea of a placodont affiliation holds the same problems we encountered with pareiasaurs, that the shell develops from osteoderm and not from expanded ribs as can be seen in extant turtles. It can be argued that both morphological and molecular support for this arrangement is weak, lacking both the morphological similarities of parareptillians and the near unanimous molecular agreement of archosaurs. Indeed, Lee (2004) states that the cladistics that do support the sauropterygia origin can be dismissed as 'homoplasy and secondary signals' which 'push turtles artificially close to lepidosauromorphs'. These hypotheses have implications, aside from the conflict between academics. Proganochelys infers a terrestrial origin for turtles due to the sediments it was found in and its shell histology (Scheyer 2007). However, Odontochelys by contrast was clearly aquatic, due to its front limb structures and being found in marine deposits (Li et al., 2008). The older Eunotosaurus is only found in terrestrial sediments and lacks marine adaptation (Lyson et al., 2010) leading to suggestion that Odontochelys may be a paedomorphic radiation of terrestrial turtles returning to water (Reisz and Head, 2008). Ultimately, deciding on an origin point will help us know where to look for older turtle fossils to further the research. There is one thing that many academics do agree on - that a multidisciplinary approach is the only way to truly answer this question, by combining work from developmental biology, as well as genetics and palaeontology. It is hard to say which hypothesis has the strongest evidence behind it currently, due to the potential issues with the molecular work, but undoubtedly academics will eventually reach a consensus.