Biomolecular Characterization and Protein Sequences of the Campanian Hadrosaur B. canadensis

See allHide authors and affiliations

Science  01 May 2009:
Vol. 324, Issue 5927, pp. 626-631
DOI: 10.1126/science.1165069


Molecular preservation in non-avian dinosaurs is controversial. We present multiple lines of evidence that endogenous proteinaceous material is preserved in bone fragments and soft tissues from an 80-million-year-old Campanian hadrosaur, Brachylophosaurus canadensis [Museum of the Rockies (MOR) 2598]. Microstructural and immunological data are consistent with preservation of multiple bone matrix and vessel proteins, and phylogenetic analyses of Brachylophosaurus collagen sequenced by mass spectrometry robustly support the bird-dinosaur clade, consistent with an endogenous source for these collagen peptides. These data complement earlier results from Tyrannosaurus rex (MOR 1125) and confirm that molecular preservation in Cretaceous dinosaurs is not a unique event.

Soft tissues and cell-like structures observed in a variety of vertebrate fossils (1, 2) raise the question of whether, and to what extent, original molecular components remain associated with these structures and the bones containing them. A suite of experiments supported the presence of original collagen molecules in the bony matrix of Tyrannosaurus rex [Museum of the Rockies (MOR) 1125] (3), and phylogenetic analysis of tryptic fragments sequenced by mass spectrometry (4, 5) placed T. rex within Archosauria, closer to birds than other vertebrate taxa in the sampled database (6, but see also 710).

Deep burial in sandstone seems to favor exceptional preservation (2). We recovered a single femur from an articulated hind limb of Brachylophosaurus canadensis (MOR 2598). The pes elements, tibia, and fibula were collected in 2006; the femur was untouched and protected under ~7 m of Judith River Formation sandstones, Eastern Montana, USA, until recovery in 2007. The femur was not exposed in the field but jacketed with ~10 to 12 cm of sediments to maintain equilibrium. Immediately after opening the jacket, bone and sediment samples were collected with sterile instruments, wrapped in layers of foil, and placed in sealed jars with desiccation crystals until laboratory analyses were performed. Bone and/or processed samples were dispersed to multiple institutions for independent analyses. Some bone fragments were subjected to chemical extraction (3, 10), and others were demineralized, showing marked preservation of microstructures resembling soft, transparent vessels, cells, and fibrous matrix in this pristine bone (Fig. 1 and fig. S1) (10).

Fig. 1

FESEM and transmitted light micrographs of demineralized B. canadensis (MOR 2598) bone matrix, vessels, and cells, compared with extant ostrich components imaged under similar conditions. (A) Low-magnification of FESEM demineralized MOR 2598 shows fibers in a plywood-like array characteristic of bone. Arrows indicate osteocytes in lacunae within fibers. (B) Demineralized matrix showing detail of fibers. (C) Demineralized ostrich bone matrix imaged under the same parameter as in (B). (D) Demineralized bone matrix from MOR 5928 is white under transmitted light. (E) MOR 2598 vessels and matrix after EDTA demineralization show rounded red inclusions (10). (F) Higher magnification of isolated MOR 5923 vessel, showing amorphous red intravascular contents. (G) Ostrich vessel with amorphous contents and associated osteocyte (arrow). (H) FESEM shows hollow B. canadensis vessels, often associated with osteocytes (arrow). (I) FESEM image of collapsed ostrich vessel and associated matrix with osteocytes (arrows). (J) Osteocytes attached by interconnected filipodia, with filipodia embedded in fibrous matrix. (K) FESEM of single B. canadensis osteocyte with long filipodia. (L) Single ostrich osteocyte isolated from demineralized, digested bone.

Under field-emission scanning electron microscopy (FESEM) (10), demineralized bone matrix was fibrous, with fibers arranged either in distinct layers at ~30° (Fig. 1A) that retain the original plywoodlike orientation of collagen fibers in bone (11, 12) or arranged in parallel (Fig. 1B). Microstructures with filipodial extensions and internal contents, consistent in size and morphology to vertebrate osteocytes (13, 14), were visible at matrix layer boundaries (Fig. 1A, arrows). The matrix was virtually indistinguishable from recent demineralized ostrich bone (Fig. 1C) imaged under the same parameters. In transmitted light, demineralized matrix appeared white and fibrous (Fig. 1D) and autofluoresced when illuminated with a mercury lamp, similar to extant bone collagen (15).

Transparent, flexible vessels were observed; some contained spherical microstructures (Fig. 1E), whereas others contained an amorphous red substance (Fig. 1F) that is superficially similar to degraded blood products in vessels recovered from extant bone (Fig. 1G) (2). B. canadensis vessels were hollow (Fig. 1H), with walls of uniform thickness, and possessed a surface texture that differed from exterior to luminal surfaces, features not consistent with the relatively amorphous texture of biofilm (7). Vessel surface texture differed substantially from the fibrous matrix but was similar to that seen in extant ostrich vessels (Fig. 1I) (1) after demineralization and collagenase digestion. Osteocytes were closely associated with vessels in both extant and B. canadensis samples (Fig. 1, H and I, arrows). The variation in texture, microstructure, and color of dinosaur material is consistent with extant tissues and not plausibly explained by biofilm (7).

Ovoid red “cells” with long filipodia, similar in morphology to extant osteocytes, were embedded in or associated with white matrix (Fig. 1J and fig. S1) or vessels (Fig. 1H). In some cases, these were attached by their filipodia to adjacent cells (Fig. 1J, inset), forming an interconnecting network as in extant bone. The cells contain internal microstructures suggestive of nuclei. Red filipodia extend from cell bodies into the white fibrous matrix (Fig. 1J and fig. S1), reflecting original chemical differences at submicron levels between cells and matrix and inconsistent with recent microbial invasion (7). Under FESEM (10), B. canadensis osteocytes and filipodia (Fig. 1K) are similar in morphology, surface texture, and size to extant ostrich osteocytes isolated from bone digests (Fig. 1L) (1, 2, 13, 14).

MOR 2598 bone fragments were chemically extracted (2, 10) for use in multiple analyses. Electrophoretic separation revealed a smear of silver-stainable material, not present in either buffer controls or surrounding sediment extracted in tandem with the bone. A dark-staining high–molecular weight band was consistently seen in guanidine extracts, whereas a low–molecular weight smear was visible in trichloroacetic acid extractions (fig. S2) (10).

Both chemical extracts and demineralized dinosaur tissues showed binding of antibodies raised against collagen and other extant proteins. Enzyme-linked immunosorbent assays (fig. S3A) and immunoblots (figs. S3B and S9) of bone extracts showed positive reactivity to antibodies raised against avian collagen I and/or osteocalcin; controls of extraction buffers alone or coextracted sediments were nonreactive (figs. S3A and S9). Results were repeated in two separate laboratories (labs of R.K. and L.C.C.) on separate extractions of MOR 2598 bone fragments.

In situ immunohistochemical analyses conducted on demineralized dinosaur extracellular matrix showed positive reactivity to polyclonal antibodies raised against avian collagen I (Fig. 2A), osteocalcin (Fig. 2D), and ostrich whole bone extracts (fig. S4) (10), but monoclonal antibodies raised against a specific osteocalcin epitope (fig. S5) showed no binding. This may indicate that the single epitope targeted by the monoclonal antibody was either not preserved or not present originally in B. canadensis protein. Negative controls of secondary antibody only (no primary added) showed no binding (fig. S5), and specificity controls of inhibition (blocking antibody binding sites by first incubating with excess collagen, see Fig. 2B) and tissue digestion with collagenase before antibody exposure (Fig. 2C) resulted in the reduction of binding. Antibodies to ostrich whole bone extracts (fig. S4) bound dinosaur tissues slightly stronger than commercial chicken collagen antibodies (Fig. 2A), possibly indicating more shared epitopes between B. canadensis and ostrich bone proteins than this dinosaur and chicken. These results support the hypothesis that original epitopes of bone proteins are preserved in these 80-million-year-old dinosaur skeletal elements.

Fig. 2

In situ immunohistochemistry of demineralized B. canadensis bone matrix and vessels. (A) Incubated with antibodies against avian collagen I . (B) Anti-collagen antibodies inhibited with avian collagen, then incubated with tissues as in (A). (C) Demineralized bone matrix digested with collagenase, then incubated with anti-collagen antibodies as in (A). (D) B. canadensis demineralized bone matrix incubated with anti-osteocalcin polyclonal antibodies (10). (E) B. canadensis vessels incubated with polyclonal antibodies against laminin (10) show weak binding above background levels. (F) Vessels incubated with antibodies against elastin proteins. (G) Vessels digested with elastase, then incubated with elastin antibodies as in (F). (H) B. canadensis vessels incubated with polyclonal antibodies against ostrich hemoglobin. (I) Ostrich hemoglobin antibodies incubated with excess hemoglobin to inhibit binding, then incubated with dinosaur vessels as in (H). Panels (A) to (D) were taken at 79-ms integration, 63× objective; (E) to (I) were taken at 120-ms integration, 63× objective, magnification as shown.

Because elastin, laminin, and hemoglobin are commonly associated with blood vessels in tetrapods, we used antibodies raised against these proteins to test for epitopes preserved in B. canadensis vessels. Elastin is a highly conserved, vertebrate-specific molecule generally resistant to degradation (16), and antibodies to this protein bound dinosaur vessels above background levels in all cases, both in immunohistochemistry (Fig. 2F) and immunoblot (fig. S6A). Antibodies to both laminin and elastin proteins showed specific binding, though they differed in binding strength in both assays. Three relatively high–molecular weight bands can be seen within a smear of stained material in immunoblot (fig. S6A, thin arrows), whereas two faint bands indicating laminin antibody binding were observed at different molecular weights when compared with elastin-reactive bands (fig. S6A, short arrows). Vessel extracts did not show binding to antibodies raised against ostrich hemoglobin, but hemoglobin antibodies bound specifically in in situ tests (Fig. 2H). A second set of experiments, conducted in a separate lab (by R.K.), confirm these results (fig. S6B). Laminin antibodies known to react to avian and reptilian proteins (10) demonstrate specific binding to B. canadensis extracts. Controls for spurious antibody binding were negative.

Attenuated total reflection infrared spectroscopy demonstrated clear amide I and amide II bands (fig. S7) in lyophilized vessels and surrounding demineralized matrix. This method cannot be used to identify which proteins are present or the source/endogeneity of these proteins, only characteristic bond vibration patterns. However, the data are consistent with other in situ data supporting the presence of proteinaceous material in B. canadensis tissues.

Posttranslational hydroxylation of proline is an identifying feature of collagen, and microbes cannot produce this modification (17, 18). Whole vessels and matrix from demineralized B. canadensis bone were subjected to in situ time-of-flight–secondary ion mass spectrometry (ToF-SIMS) with imaging capability (10) in a separate laboratory (the lab of R.A.). Focused Ga+ and Au+ ions generated molecular fragments from tissues that were detected with sufficient high mass resolution (m/Δm ~ 4000) to uniquely identify hydroxylated proline (Pro-OH), as well as Lys, Pro, Lys-OH, Ala, Gly, and Leu from demineralized dinosaur tissues (fig. S8). A number of organic fragments containing nitrogen were also detected, including unusual complexes between C, N, and Fe (Fig. 3A, inset, and fig. S8), suggesting alteration and modification to preserved organics, more consistent with an ancient source than recent contamination.

Fig. 3

Mass spectrometry analyses of B. canadensis tissues and extracts. (Inset to A) Collagen-specific Pro-OH functional group positively identified by TOF-SIMS in situ analysis of demineralized B. canadensis tissues by a preparation method distinct from tandem mass spectrometry analyses [see text and (10)]. (A to D) Product ion spectra recorded during tandem mass spectrometry (LC/MS/MS) analyses with CID of chemical extracts of bone. The four collagen peptides scoring lowest on Mascot search engine scores (but top-ranked versus the Swiss-Prot database) acquired from the LTQ-Orbitrap XL mass spectrometry were validated with high-confidence versions of the same sequences with a computational spectral comparison tool. (A) B. canadensis collagen a1(I) [M+3H]3+ peptide ion GLTGPIGPP(OH)GPAGAP(OH)GDKGEAGPSGPPGPTGAR using the MS Search 2.0 spectral comparison algorithm shows excellent alignment of fragment ion m/z values and relative intensities and was ranked first to a previously acquired high-confidence ostrich version of the same sequence [reverse match factor (RMF = 779)] in a database of >200,000 spectra. (B) Collagen a1(I) [M+2H]2+ peptide ion GATGAP(OH)GIAGAP(OH)GFP(OH)GAR was a top match to the synthetic peptide version of the same sequence (RMF = 494). (C) The collagen a2(I) [M+2H]2+ peptide ion GSN(deam)GEP(OH)GSAGPP(OH)GPAGLR versus a synthetically derived version (RMF = 482), and (D) the collagen a1(I) [M+2H]2+ peptide ion GVQGPP(OH)GPQGPR validated with a synthetic peptide (RMF = 562). (B, C, and D) Significant fragment ions attributed to water losses from the precursor ion are present in the spectra from Brachylophosaurus but are less apparent in synthetic or ostrich comparisons. The four other highest-scoring statistically significant MS/MS spectra are shown in fig. S9 in the SOM.

We used mass spectrometry (performed by J.M.A.) to analyze whole bone extracts of B. canadensis (MOR 2598) that were fibrous and lighter in color than those from T. rex (MOR 1125) (4, 5). Bone extracts (10) were proteolyzed with trypsin, purified, and concentrated by microreversed phase chromatography, then analyzed by reversed-phase microcapillary liquid chromatography tandem mass spectrometry (LC/MS/MS) using both linear ion trap and hybrid linear ion trap–orbitrap mass spectrometers (19). We searched fragmentation spectra from data-dependent acquisitions against the reversed Swiss-Prot protein database using Sequest (20) and Mascot (21) algorithms. We identified eight total collagen peptide sequences, six from collagen α1 type I and two from collagen α2 type I. The eight peptides totaled 149 amino acids from four different samples (six LC/MS/MS data sets), nearly double the number of amino acids (89) recovered from T. rex (MOR 1125) (4, 5). The eight peptides identified as collagen (Table 1) were top-ranked matches versus the Swiss-Prot target protein database. Four of the eight peptides, derived from three separate bone extracts, were sequenced from multiple scan events across multiple data files. We validated the sequences using scoring statistics, decoy databases, manual inspection, and spectral comparisons to high-confidence spectra obtained from synthetic peptides and peptides of identical sequence from extant organisms. One peptide sequence showed evidence of deamidation of asparagine, and all reported sequences contained at least one Pro-OH residue.

Table 1

Collagen α1(I) and α2(I) sequences acquired by ion trap and Orbitrap mass spectrometry for B. canadensis. m/z values of the peptide ion, molecular weight, mass spectrometer used, database search engine scores, expectation values, sequence validation method, and sequence identity based on BLAST searches versus the all-species NCBInr protein database and internally acquired ostrich and alligator sequences are shown. The interpretation of the sequence GLPGESGAVGPAGPP(OH)GSR was aided by the high mass accuracy of the Orbitrap, because hydroxyproline is more accurate than isoleucine/leucine at position 15 by 0.0364 daltons.

View this table:

The eight peptide sequences for collagen α1 type I and collagen α2 type I represent 7.8 and 2.5% of the full-length sequence for related organisms, respectively. In addition, three peptide sequences were confirmed in a separate laboratory (in the lab of W.S.L.) from two samples from a single bone extraction. Coextracted sediment and buffer controls were consistently negative for collagen peptidic material by all methods of analyses including mass spectrometry sequencing, although common contaminants including human keratins, ubiquitous in all labs, and some microbial peptides were sequenced from these bone extracts.

All eight sequences derived from B. canadensis extracts are high-confidence, and their spectra produced expectation values of <1 from Mascot searches. Four dinosaur spectra were high-confidence matches to existing protein databases, and statistical validation is sufficient for identification. The four lowest-scoring MS/MS spectra from Brachylophosaurus bone extracts, acquired with the use of collision-induced dissociation (CID) in the linear ion trap of an LTQ-Orbitrap XL mass spectrometer, are shown in Fig. 3, A to D. These MS/MS spectra were subjected to an extra level of validation with the MS Search 2.0 spectral comparison algorithm from the National Institute of Standards and Technology (NIST) against a database of >200,000 random peptide fragmentation spectra from various taxa (22, 23), including high-confidence versions of MS/MS spectra from four collagen sequences by adding previously acquired ostrich or synthetic peptides to the NIST spectral database. These four sequences derived from B. canadensis were top matches to high-confidence versions of the same sequences by fragment ion mass/charge ratio (m/z) values and relative intensities of fragment ions, providing additional validation of the data. All sequence data from Brachylophosaurus and validating sources are available in the supporting online material (SOM). The complete raw data from all 27,791 spectra, as well as previously reported sequences from T. rex, are available from the PRIDE database ( In addition, we have shown B. canadensis collagen sequences to be statistically significant, and false discovery rates for MOR 2598 sequence data have been calculated (10).

In a separate lab (that of L.C.C.), B. canadensis bone extracts showed positive reactivity by immunoblot to a mixture of collagen type I polyclonal antibodies at high molecular weights (~250 to 300 kD), whereas surrounding sediments showed negative reactivity (fig. S9). This supports the identification of collagen fragments in bone extracts and suggests modifications (for example, cross-linking) of collagen molecules (2). Although silver-stained material in gels (fig. S2) suggest concentrations within the low end of current mass spectrometry sensitivity thresholds (low nanogram), it is likely that most of the protein is present in unsequenceable form(s); that is, cross-linked and/or modified so as to make them incapable of resolution by current mass spectrometry technology and software. We hypothesize that processes contributing to preservation of these otherwise labile components make analyses difficult, but conversely, the low recovery and diagenetic alteration support an endogenous source, as these chemical modifications are not observed in modern proteins.

B. canadensis collagen sequences were aligned with collagen sequences from 21 extant taxa and two extinct organisms: Mammut americanum (MOR 605) and T. rex (MOR 1125) (see table S1 and SOM appendix) (6). The Anolis carolinensis amino acid sequence was inferred with the use of FGENESH+ (gene prediction on the basis of protein homology; Additional collagen α2(I) sequence data for five species were obtained from online public databases (table S1).

We used BayesPhylogenies (24) to infer phylogenetic relationships from collagen sequences with the Dayhoff amino acid substitution model (6). The resulting consensus tree of the posterior distribution (Fig. 4) was well resolved, as was a phylogeny inferred by maximum likelihood (fig. S11). Whereas the first molecular phylogeny containing a non-avian dinosaurian [T. rex, MOR 1125, (6)] resulted in a tree that misplaced A. carolinensis (6), additional sequence data from several extant species (fig. S11 and table S1) corrected this misplacement and improved clade support to 100% (number of times a clade was inferred in the posterior distribution of trees) for all groups except the Gallus/Struthio group (Aves), which had 36% support. Under a majority-rule criterion to building a consensus tree, Dinosauria (the group containing the two extinct dinosaurs and the two birds) collapsed into a three-way polytomy.

Fig. 4

Consensus of the posterior distribution of phylogenetic trees including M. americanum (Mastodon, MOR 605) and the extinct dinosaurs B. canadensis (MOR 2598) and T. rex (MOR 1125, in bold). Colored backgrounds designate monophyletic groups. All nodes have 100% posterior probability (PP) support except the Gallus/Struthio group, which has 36% PP support. Branch lengths are reported as the mean of the expected number of changes per site.

Removing T. rex from the phylogeny resulted in a three-way polytomy as well. The amount of missing data in B. canadensis and T. rex sequences relative to extant samples resulted in relatively low resolution within Dinosauria, but even so, the phylogenetic relationship of recovered B. canadensis sequences supports the species’ placement within Archosauria, closer to birds than Alligator. However, on the basis of well-established morphological analyses (25), we predict that T. rex is more closely related to birds than it is to the ornithischian hadrosaur B. canadensis. Despite ambiguity within Dinosauria, obvious phylogenetic signal resides within recovered collagen sequences, supporting endogeneity (fig. S11) (10).

The hypothesis that endogenous proteins can persist across geological time, as first reported for T. rex (MOR 1125), was met with appropriate skepticism (79). However, the inclusion of additional sequence data from extant reptiles (6) and B. canadensis strengthens the hypothesis that the molecular signal is preserved at least to the Late Cretaceous.

The submicron differences in texture (Fig. 1 and fig. S1), elemental differentiation, sub-“cellular” inclusions in osteocytes and vessels, identification of the posttranslational Pro-OH modification not produced by microbes (18), differential binding of antibodies by both in situ and immunoblot studies, collagen protein sequences, and phylogenetic analyses do not support a microbial origin for either these microstructures or peptide fragments (7). Coupled with evidence for cross-linking and unusual chemical modifications, the congruence of evidence strongly supports an endogenous origin for this material. The most parsimonious explanation, thus far unfalsified, is that original molecules persist in some Cretaceous dinosaur fossils. Still unknown is the chemistry behind such preservation.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S12

Table S1


Appendix S1

Database S1

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank B. Harmon, N. Peterson, and the MOR crew responsible for recovery of MOR 2598; N. Equall and S. Brumfield for microscopy assistance; G. Fisher (Physical Electronics) for PHI nanoTOF assistance; T. Cleland for assistance with gel analyses; X. Yang for help with mass spectrometry sample preparation; S. Stein and Y. Mirokhin for MS Search 2.0 software enhancements; J. Starkey and R. Mecham for antibodies; B. Anderson for Fourier transform infrared spectroscopy access; T. Collier for sample preparation; E. Leithold for access to equipment; J. C. Fountain for support; and the Montana Department of Natural Resources and Conservation for land access. C.L.O. thanks S. V. Edwards for postdoctoral support. We acknowledge funding from NSF [awards EAR 0722702 and EAR 0634136 (J.M.A.) and EAR 0548847 and EAR-0541744 (M.H.S.)]; the David and Lucile Packard Foundation (M.H.S.); United Negro College Fund–Merck Postdoctoral Science Research Fellowship (M.B.D.); the Taplin Funds for Discovery (Harvard Medical School); partial support by NASA–Expermental Project to Simulate Competitive Research under grant NCC5-579 (R.A. and Z.S.); Damon Runyon Cancer Research Foundation (M.G.V.H.); NIH grants DK 55001, DK 62987, AA 13913, DK 61866, and CA 125550; and the Department of Medicine for the Division of Matrix Biology at the Beth Israel Deaconess Medical Center. Correspondence and requests for materials should be addressed to M.H.S. for biochemical and morphological analyses and to J.M.A. for mass spectrometry and molecular phylogenetics analyses.
View Abstract

Navigate This Article