We used authentication tests developed for ancient DNA to evaluate claims by Asara et al. (Reports, 13 April 2007, p. 280) of collagen peptide sequences recovered from mastodon and Tyrannosaurus rex fossils. Although the mastodon samples pass these tests, absence of amino acid composition data, lack of evidence for peptide deamidation, and association of α1(I) collagen sequences with amphibians rather than birds suggest that T. rex does not.
Early reports of DNA preservation in multimillion-year-old bones (i.e., from dinosaurs) have been largely dismissed (1, 2) (table S1), but reports of protein recovery are persistent [see (3) for review]. Most of these studies used secondary methods of detection, but Asara et al.(2) recently reported the direct identification of protein sequences, arguably the gold standard for molecular palaeontology, from fossil bones of an extinct mastodon and Tyrannosaurus rex. After initial optimism generated by reports of dinosaur DNA, there has been increasing awareness of the problems and pitfalls that bedevil analysis of ancient samples (1), leading to a series of recommendations for future analysis (1, 4). As yet, there are no equivalent standards for fossil protein, so here we apply the recommended tests for DNA (4) to the authentication of the reported mastodon and T. rex protein sequences (2)(Table 1).
First, the likelihood of collagen survival needs to be considered. The extremely hierarchical structure of collagen results in unusual, catastrophic degradation (5) as a consequence of fibril collapse. The rate of collagen degradation in bone is slow because the mineral “locks” the components of the matrix together, preventing helical expansion, which is a prerequisite of fibril collapse (6). The packing that stabilizes collagen fibrils (6) also increases the temperature sensitivity of degradation (Ea 173 kJ mol–1) (Fig. 1). Collagen decomposition would be much faster in the T. rex buried in the then-megathermal (>20°C) (7) environment of the Hell Creek formation [collagen half-life (T½) = ∼ 2 thousand years (ky] than it would have been in the mastodon lying within the Doeden Gravel Beds (present-day mean temperature, 7.5°C; collagen T½ = 130 ky) (Fig. 1).
This risk of contamination also needs to be evaluated. Collagen is an ideal molecular target for this assessment because the protein has a highly characteristic motif that is also sufficiently variable to enable meaningful comparison between distant taxa if enough sequence is obtained (Fig. 2). Compared with ancient DNA amplification, contamination by collagen is inherently less likely. Furthermore, because the bones sampled in (2) were excavated by the authors, obvious contamination sources such as animal glue (used in conservation) can be excluded. However, concentrating protein from the large amounts of bone used (2.5 g) may have heightened the risk of extraneous proteins entering the sample during extraction, although there have been no systematic studies of this phenomenon. Independent extraction and analyses would have strengthened claims for the authenticity of the origin of the peptides (and potentially ameliorated the original problems of data interpretation) (4).
The remarkable soft-tissue preservation of the investigated T. rex specimen (MOR 1125) has been documented (8). However, microscopic preservation does not equate with molecular preservation (9). Immunohistochemistry provides support for collagen preservation, but Asara et al. (2) presented no data regarding inhibition assays with collagen from different species or cross-reactivity with likely contaminants [e.g., fungi (10)]. Curiously, no amino acid compositional analysis was conducted [see (11)], although immonium ions were identified by time-of-flight secondary ion mass spectrometry. In our experience, collagen-like amino acid profiles have been obtained in all bones from which we could obtain collagen sequence (Fig. 1, inset).
Regarding the proof of sequence authenticity, the spectra reported by Asara et al. (12) are inconsistent with some of the sequence assignments (13) (table S1). A common diagenetic modification, deamidation, not considered in (2), may shed light on authenticity. The facile succinimide-mediated deamidation (14) of asparagine occurred at N229G and N1156G in ostrich peptides (Ost 4 and Ost5) (see table S1 for nomenclature), presumably during sample preparation. Direct hydrolytic deamidation is slower (14), and an expectation of elevated levels of such products is reasonable for old samples. We agree with the most recent interpretation (13) of the spectrum illustrated in Fig. 2B as α1(I) G362SEGPEGVR370, the deamidated (Q→E367) form of the sequence found in most mammals (12). By way of contrast, none of the three glutamine residues in the reported T. rex peptides are deamidated (table S1). Only time will tell if Q→E is a useful marker for authentically old collagen, but from the evidence presented, the mastodon sequence looks more diagenetically altered than T. rex.
The unusual, fragmented nature of the reported T. rex sequence does not make it amenable to standard, model-based phylogenetic analysis. Instead, we examined the phylogenetic signal of the α1(I) fragments of mastodon and T. rex using Neighbor-Net analysis and uncorrected genetic distances. Using the sequences reported in (13), both the T. rex and mastodon signal display an affinity with amphibians (Fig. 2A). Our reinterpretation of the spectra (12) changes the affinity of mastodon but not of T. rex (Fig. 2B). In addition to the α1(I) peptides used in the Neighbor-Net analysis, Asara et al. reported two other peptides from T. rex (13); we question the interpretation of the α1(II) spectra (identical to frog) but not the α2(I) spectra (identical to chicken).
We require more data to be convinced of the authenticity of the T. rex collagen sequences reported by Asara et al. Nevertheless, the handful of spectra reported for the temperate Pleistocene mastodon fail neither phylogenetic nor diagenetic tests, thus highlighting the potential of protein mass spectrometry to bridge the present gulf in our understanding between the fate of archaeological and fossil proteins. To avoid past mistakes of ancient DNA research (1), we recommend that future fossil protein claims be considered in light of tests for authenticity such as those presented here.
Supporting Online Material