Vol 463 | 11 February 2010 | doi:10.1038/nature08745 LETTERS Non-random decay of chordate characters causes bias in fossil interpretation Robert S. Sansom1, Sarah E. Gabbott1 & Mark A. Purnell1 Exceptional preservation of soft-bodied Cambrian chordates pro- anatomical interpretation and character coding in the fossils3,4. It is vides our only direct information on the origin of vertebrates1,2. also implicitly assumed that anatomical loss through decay is random Fossil chordates from this interval offer crucial insights into how with respect to the phylogenetic informativeness of a character and the distinctive body plan of vertebrates evolved, but reading this affects characters independently. We have lacked the data required to pre-biomineralization fossil record is fraught with difficulties, test the validity of this assumption, but deviation from random and leading to controversial and contradictory interpretations3,4. The independent loss will clearly bias and distort our interpretation and cause of these difficulties is taphonomic: we lack data on when and understanding of the fossil record. Furthermore, identification of how important characters change as they decompose, resulting in partially decayed remains requires knowledge not only of which a lack of constraint on anatomical interpretation and a failure to characters decay rapidly and which persist, but also of how they distinguish phylogenetic absence of characters from loss through decay. decay3. Here we show, from experimental decay of amphioxus and Current understanding of the early evolution of chordates and the ammocoetes, that loss of chordate characters during decay is origin of vertebrates provides one of the most notable examples of the non-random: the more phylogenetically informative are the most problems caused by taphonomic ambiguity. Fossil evidence from this labile, whereas plesiomorphic characters are decay resistant. The interval has the potential to transform our understanding of the taphonomic loss of synapomorphies and relatively higher pre- nature and timing of the fundamental episodes of initial body-plan servation potential of chordate plesiomorphies will thus result development, increasing complexity and genome duplications, as it in bias towards wrongly placing fossils on the chordate stem. has done for the jawed vertebrates2,5. Continuing disagreement over Application of these data to Cathaymyrus (Cambrian period of interpretations of the fossils, however, means that no clear patterns or China) and Metaspriggina (Cambrian period of Canada) high- processes are currently discernable. Only a few fossil taxa have been lights the difficulties: these fossils cannot be placed reliably in discussed in the context of early chordate evolution, largely because the chordate or vertebrate stem because they could represent the they pre-date the advent of biomineralized hard tissues; the early decayed remains of any non-biomineralized, total-group chordate. chordate fossil record is consequently limited to rare instances of soft- Preliminary data suggest that this decay filter also affects other tissue preservation. These soft-bodied fossils of purported chordates, groups of organisms and that ‘stem-ward slippage’ may be a wide- such as Metaspriggina, Pikaia, Cathaymyrus and yunnanozoans from spread but currently unrecognized bias in our understanding of the the Cambrian deposits of Canada and China (see, for example, early evolution of a number of phyla. refs 1, 6, 7), have proved to be highly contentious in terms of inter- The fossil record offers unique and crucial insight into many preting their anatomy. A broad range of phylogenetic affinities has important episodes in the evolution of life, and allows reconstruction been proposed for each taxon, and little consensus has been reached of the extinct relatives that comprise the stem lineages of major extant (Fig. 1a and Supplementary Fig. 1). clades and phyla (crown groups). Our view of the fossil record, To constrain interpretation of the anatomies and affinities of soft- however, is obscured by two persistent problems. First, most organisms bodied chordates, we undertook a series of decay experiments using are never preserved as fossils, and the patchy record provides only larval Lampetra (ammocoetes) and Branchiostoma (amphioxus). Of limited data from deep time. Second, even in cases in which exceptional the extant chordates, urochordates and jawed vertebrates have under- preservation provides a more complete picture, fossils themselves are gone considerable genetic and morphological specialization2,8 and the never anatomically complete or intact: they have been subject to the amphioxus (Cephalochordata) and the ammocoete (Vertebrata), with complex taphonomic processes and filters of decay and preservation. their relatively simple and comparable morphologies9, are widely This degradation creates difficulties when comparing the morphology thought to be the best anatomical proxies for early chordates8,10. of fossils with the anatomy of extant relatives to identify the shared Both organisms have been used to make direct comparisons with characters (synapomorphies) on which evolutionary analysis ulti- purported fossil chordates11,12. mately rests. Further complications arise because placement of fossils Previous work in the field of experimental taphonomy (see, for in the stems of extant clades reflects the absence of certain features; our example, refs 13–15), including some on Branchiostoma16, has generally failure to find a particular character in a fossil, however, may reflect focused on the transformation of organisms as a whole rather than on post-mortem degradation of the original anatomical signal (tapho- individual character change, but we have developed a new approach nomic loss) rather than true evolutionary absence (where the fossil based on analysis of the rate at which and sequence in which individual precedes the origin of the character). Failure to distinguish between anatomical characters are transformed and lost during decay. The the underlying causes of character absence will lead to erroneous evolu- apomorphies of Branchiostoma and larval Lampetra were identified tionary conclusions3. and categorised a priori in terms of the synapomorphies of the nested The main cause of these phylogenetic problems is not the patchi- clades to which each belongs (for example Chordata, Cephalochordata ness of the fossil record, but our lack of taphonomic data to constrain or Vertebrata, Fig. 1b). We recorded how, and when, each of the 1Department of Geology, University of Leicester, Leicester LE1 7RH, UK. 797 ©2010 Macmillan Publishers Limited. All rights reserved LETTERS NATURE | Vol 463 | 11 February 2010 a Figure 1 | Phylogeny and anatomy of chordates. a, Deuterostome 7,29,30 Echino- Hemi- phylogeny based on data from extant forms . Asterisks mark the extant dermata chordata Chordata proxies used in the experiments. Each fossil taxon (coloured) has been Cephalochordata Urochordata Vertebrata proposed to have a wide range of affinities, with little consensus existing. Myxinoidea Petromyzontida Gnathostomata References for each placement are available in Supplementary Information. b, Cephalochordate anatomy (top) and ammocoete anatomy (bottom). C, chordate synapomorphy; V, vertebrate synapomorphy. * * synapomorphies decayed, allowing recognition of partially decayed characters, distinction between phylogenetic absence and taphonomic loss, and identification of patterns of loss. For Branchiostoma, the general changes in pH, mass and state of organismal integrity through time (Supplementary Fig. 2), compare Stem vertebrates well with those documented in ref. 16, thus validating experimental Pikaia taphonomy as replicable. Our quantitative data show that the rate of Myllokunmingiids cumulative morphological character decay in both Branchiostoma and Metaspriggina Lampetra follows the same sigmoidal pattern (Fig. 2c). Figure 2a, b Nectocaris Stem chordates Cathaymyrus shows the chronological sequences of character decay. Analysis of these Yunnanozoans sequences reveals a striking pattern: in both taxa, the first characters to Stem deuterostomes decay are synapomorphies of relatively more derived clades (that is, cephalochordate synapomorphies for Branchiostoma and petromy- b Myomeres Dorsal nerve chord Notochord Anterior Dorsal storage organ C C zontid, cyclostome and [petromyzontid 1 gnathostome] synapomor- rostral extension bulb phies for Lampetra). The more plesiomorphic characters (for example chordate notochord and myomeres) are the last to decay. Although Buccal cirri there are some exceptions to this pattern, there is a very strong positive Anus C Caudal !n Velum/wheel organ Notochord correlation between the rank of a character in the decay sequence and C C Gonads Midgut Atriopore Pharynx Endostyle caecum the phylogenetic rank at which a character is synapomorphic (RS 5 0.75 (Spearman’s rank correlation coefficient) and P 5 0.0014 Nasohypo- 5 5 physial Sensory Myomeres Dorsal nerve chord for Branchiostoma (15 d.f.); RS 0.71 and P 0.000019 for Lampetra complex organs Brain C Median !ns C (28 d.f.)). Skull V V V Branchial cartilage V Our results reveal two important patterns. First, homologous Buccal cirri Oral hood chordate characters in different taxa have comparable decay resis- Lower lip Anus Caudal !n tances, despite some fundamental structural differences (for example V V Gut C Velum C C Heart Liver Notochord cephalochordate and vertebrate notochord and pharyngeal-arch Pharynx Endostyle