Zuteilung anfragen Kontakt English简体中文

WissenschaftTechnische Berichte

JM-TR-2013-001 Technischer Bericht Bibliothek pre-1.0

Endogenous peptide fragments recovered from an isolated Kem Kem Group theropod tooth: extraction, authentication and preliminary phylogenetic placement

Dr. Hao Lin 1 Dr. Yuting Zhou 1 Dr. Lei Chen 1,*

1 Jurassic Meat Shenzhen Research Center, Shenzhen, China
* Korrespondierende Autorin / korrespondierender Autor: Dr. Lei Chen · regulatory@jurassicmeat.com

Inhalt
  1. Zusammenfassung
  2. 1. Summary
  3. 2. Introduction
  4. 3. Materials and methods
  5. 4. Results
  6. 5. Discussion
  7. 6. Limitations and known issues
  8. 7. Implications for a reference library
  9. 8. Data availability
  10. 9. Acknowledgements
  11. 10. References
  12. Supplementary Table S1. Campaign-level reproducibility

Zusammenfassung

We report the recovery and authentication of endogenous peptide fragments from an isolated theropod tooth crown (specimen JM-SP-0001) from the Cenomanian Kem Kem Group of southeastern Morocco. Material was extracted in three independent campaigns between December 2011 and December 2012, the third performed by a different operator in a separate laboratory. Extracts were analysed by nanoflow LC-MS/MS and searched against a reference database of seven extant archosaur and outgroup taxa together with a contamination database, at a 1% false discovery rate. Of 731 unique peptide sequences identified across crown extracts, 196 passed all authentication criteria: no match to the contamination database, absence from all procedural and injection blanks, absence from method-development material, observation in at least two independent campaigns, and manual spectral review. The authenticated set is dominated by collagen type I (163 peptides), with a minority from enamel matrix proteins (26) and other non-collagenous proteins (7), and covers 1,712 unique residues. Glutamine and asparagine deamidation in the authenticated set (0.74 and 0.89) is markedly higher than in contaminant-assigned peptides from the same extracts (0.12 and 0.19) and in modern controls. Concatenated maximum-likelihood and Bayesian placement on a fixed extant reference tree recovers the specimen on the stem lineage of Aves (posterior 0.91; bootstrap 87%), consistent with membership of Theropoda. Median per-fragment placement posterior is 0.38. Limitations of provenance, specimen history and reference coverage are described in full.

SchlagwörterpalaeoproteomicsKem Kem GroupTheropodacollagen type Ienamel proteinsdeamidationcontamination controlphylogenetic placement

JM-TR-2013-001 · Public summary of internal report · Shenzhen, September 2013

1. Summary

  • Endogenous peptides were recovered from a ~95-million-year-old theropod tooth crown. 196 peptide sequences from specimen JM-SP-0001 passed all authentication criteria across three independent extraction campaigns (2011–2012).
  • Collagen type I dominates. 163 of 196 authenticated peptides derive from COL1A1 or COL1A2; 26 derive from enamel matrix proteins (amelogenin, enamelin, ameloblastin).
  • The deamidation profile supports antiquity. Authenticated peptides show Q and N deamidation of 0.74 and 0.89, against 0.12 and 0.19 for contaminant-assigned peptides in the same extracts.
  • Placement is consistent with Theropoda, not diagnostic of it. The specimen places on the stem lineage of Aves (posterior 0.91). No non-avian dinosaur reference sequences exist; finer placement rests on tooth morphology, not on molecular data.
  • The specimen's documented history is incomplete. Locality, date of collection and the period prior to 2011 are not recorded. These limitations are listed in Section 6 and are not resolvable.

2. Introduction

Claims of original protein in Mesozoic vertebrate fossils have been made and contested for more than a decade. Soft-tissue-like structures and immunological evidence of collagen were reported from Tyrannosaurus rex (Schweitzer et al. 2007), followed by mass-spectrometric peptide sequences from the same specimen (Asara et al. 2007) and, subsequently, from the hadrosaur Brachylophosaurus canadensis (Schweitzer et al. 2009). These reports drew specific and reasonable criticism, including the observation that some recovered sequences could be explained by contamination or by misassignment of spectra (Buckley et al. 2008). The field has since treated contamination as the default explanation for any deep-time protein claim, to be excluded rather than assumed absent.

This report describes work that began informally. In November 2011, following a conversation in Tokyo between the two eventual cofounders of the company, a single isolated theropod tooth from the Kem Kem Group of southeastern Morocco was sent by post to Dr. Lei Chen at Tsinghua University with a request to determine whether anything inside it survived. The initial extraction was performed by Dr. Hao Lin in December 2011. A positive result in that first campaign was not regarded as a finding; it was regarded as a hypothesis to be tested under conditions designed to make it fail. Two further campaigns followed, the second performed by a different operator (Dr. Yuting Zhou) in a separate laboratory in Shenzhen.

The objectives of this report are therefore narrow:

  1. to document the extraction and analytical methods applied to specimen JM-SP-0001;
  2. to describe the authentication framework and report which peptide sequences survive it;
  3. to provide a preliminary phylogenetic placement of the authenticated set against extant archosaur references; and
  4. to record, without qualification, what is not known about the specimen.

We do not report DNA, and did not attempt to recover it. Nucleic acids do not persist across the time scales concerned. All sequence information reported here is protein sequence, determined by tandem mass spectrometry.

3. Materials and methods

3.1 Specimen

Specimen JM-SP-0001 is an isolated, labiolingually compressed, serrated (ziphodont) tooth crown with an attached root (Table 1). It was purchased by a cofounder of the company, before the company existed, from a dealer in the Sunday market at Rissani, Tafilalt region, Morocco, and was sold as Carcharodontosaurus. Commercially traded theropod teeth from this region derive overwhelmingly from the Kem Kem Group (Cenomanian, early Late Cretaceous). No locality more precise than "Kem Kem" was provided at the point of sale, and none has been recoverable since.

Table 1. Specimen JM-SP-0001: description and recorded history.
Attribute Value
Specimen ID JM-SP-0001
Element Isolated tooth crown with attached root
Formation (attributed) Kem Kem Group, SE Morocco; Cenomanian (~95 Ma)
Locality Not recorded
Taxonomic attribution at sale Carcharodontosaurus
Morphological attribution (this report) Theropoda; Carcharodontosauridae indet. (preferred); Abelisauridae not excluded
Crown height / crown base length / crown base width 71.4 mm / 31.2 mm / 13.9 mm
Mesial / distal denticle density 9 / 8 per 5 mm
Total mass on receipt 63.8 g
Date of purchase Not recorded (before 2011)
Storage before receipt Residential display cabinet, ambient temperature; duration not recorded
Transport to laboratory International post, padded envelope; received 2011-11-21
Consolidants / repairs Cyanoacrylate at a transverse crown fracture (mid-height); root bonded to crown (see §3.2)
Crown mass consumed by sampling (all campaigns) 0.249 g (approx. 0.6% of crown mass)

3.2 Pre-sampling examination

Before any destructive sampling the specimen was photographed, weighed, and examined by micro-computed tomography (isotropic voxel 42 µm). CT showed that the crown carries a transverse fracture at approximately mid-height, repaired with a cyanoacrylate adhesive, and that the root is a separate element bonded to the crown base with adhesive and a pigmented mineral filler. The root is therefore not part of the same tooth and cannot be assumed to be part of the same individual. This is common in commercially traded Kem Kem material and is noted here without inference about the intent of any party.

All crown samples were taken more than 5 mm from the repaired fracture and from the crown–root junction. The root element was sampled once (Campaign A) as an internal comparison and is otherwise excluded from this report.

3.3 Laboratory environment

Campaigns A and B were performed at Tsinghua University in a dedicated bench space that had not previously been used for vertebrate tissue extraction, under a laminar-flow hood, with nitrile gloves, face masks and hair covers, single-use plasticware, and reagents opened for the campaign. Campaign C was performed in a newly fitted clean bench at the company's temporary Shenzhen laboratory, with reagents from different lots and suppliers. No modern archosaur tissue had been handled in the Shenzhen laboratory before Campaign C.

3.4 Sampling and extraction

The crown surface was removed to a depth of approximately 0.5 mm using a sterile tungsten-carbide burr at low speed, and the abraded powder was discarded. Dentine was then drilled from the exposed interior; enamel was sampled separately by abrasion of the outer crown surface after initial surface removal. Theropod enamel is thin, and enamel yields were correspondingly small.

Powder was demineralised in 0.6 M HCl at 4 °C for 18 h. The acid-soluble fraction and the acid-insoluble residue were processed separately. Acid-insoluble residue was gelatinised in 50 mM ammonium bicarbonate at 65 °C for 1 h. Half of each fraction was digested with sequencing-grade porcine trypsin (enzyme-to-substrate ratio estimated at 1:50) at 37 °C for 16 h; the other half was left undigested, on the basis that ancient protein is already extensively fragmented and that undigested extracts avoid the introduction of enzyme-derived peptides. Peptides were desalted on C18 solid-phase tips, dried, and resuspended in 0.1% formic acid. Enamel powder was demineralised separately in 5% trifluoroacetic acid, without enzymatic digestion.

Table 2. Sampling campaigns and controls.
Campaign Dates Site Operator Crown dentine (mg) Enamel (mg) Extraction blanks Other controls
A 2011-12 Tsinghua H. Lin 52 11 3 Root element (40 mg)
B 2012-05 Tsinghua H. Lin 47 — 3 Sediment from crown surface (2 mg)
C 2012-10 to 2012-12 Shenzhen Y. Zhou 49 10 3 Modern bovine bone (run last, separate batch)
Total 148 21 9

Extraction blanks were carried through every step, including demineralisation tubes, burr contact (the burr was run against a blank substrate) and digestion, in parallel with the specimen.

3.5 LC-MS/MS

All extracts were analysed on the same Orbitrap-class hybrid mass spectrometer coupled to a nanoflow liquid chromatography system. Campaign C extracts were transported to Beijing frozen on dry ice in sealed, labelled secondary containers with a transfer log. Instrument settings are summarised in Table 3.

Table 3. LC-MS/MS acquisition parameters (all campaigns).
Parameter Setting
Column 75 µm × 25 cm, C18, 3 µm particles; 2 cm trap column
Gradient 2–35% acetonitrile in 0.1% formic acid over 120 min; 300 nL/min
MS1 Resolution 60,000 at m/z 400; scan range m/z 350–1,600; AGC target 1×10⁶
MS2 Top-10 data-dependent; CID in linear ion trap (Campaigns A, B); HCD at resolution 7,500 (Campaign C replicate injections)
Isolation window 2.0 m/z
Normalised collision energy 35 (CID); 28 (HCD)
Charge states 2+ to 5+; singly charged and unassigned excluded
Dynamic exclusion 30 s
Lock mass Polysiloxane, m/z 445.120025
Injection order Solvent blank → extraction blank → specimen → solvent blank ×2
Solvent (injection) blanks 42 in total, including 12 after the 48 h column wash preceding Campaign C

Each extract was injected at least twice. The instrument had been used in 2011, before Campaign A, for method development on modern chicken (Gallus gallus) bone collagen. This material is treated as a named contamination source (§3.7).

3.6 Database search and de novo sequencing

Raw files were searched with MaxQuant (Cox and Mann 2008) against a reference database comprising collagen, enamel matrix and selected non-collagenous bone protein sequences from seven extant taxa (Table 6), concatenated with a contamination database (§3.7) and reversed decoy sequences. Searches were non-specific (no enzyme constraint) for undigested extracts and semi-specific for tryptic extracts. Variable modifications: deamidation (N, Q), oxidation (M), hydroxylation (P, K), pyroglutamate formation from N-terminal Q and E, and N-terminal acetylation. No fixed modifications were set, as extracts were not reduced or alkylated. Precursor tolerance was 4.5 ppm after recalibration; fragment tolerance 0.5 Da (ion trap) or 20 ppm (HCD). Peptide-spectrum matches (PSMs) and peptides were filtered at 1% false discovery rate (FDR) by target-decoy estimation; minimum peptide length was six residues.

Because extinct sequences will differ from all references, spectra not matched in the database search were additionally subjected to de novo sequencing. De novo candidates with an average local confidence of at least 80% were aligned to the reference proteins by homology and re-searched as an expanded database; only peptides subsequently identified at 1% FDR in that second search were retained. Isoleucine and leucine are isobaric and are not distinguished; they are reported as "J" in internal records and resolved to the reference residue for alignment purposes.

3.7 Contamination database and exclusion list

The contamination database contained 247 sequences: human keratins and keratin-associated proteins; porcine trypsin; bovine, porcine and ovine collagens types I–III; bovine serum albumin; common laboratory proteins; and the full set of sequences detected during the 2011 Gallus method-development runs. A peptide was excluded if it matched any contamination sequence, if its sequence (allowing for I/L equivalence) was observed in any extraction blank or solvent blank, or if it was identical to a sequence observed in the method-development material.

The last criterion is conservative: because collagen is highly conserved, some genuine theropod sequences are expected to be identical to their chicken orthologues and are lost by this rule. We accepted that loss.

3.8 Authentication criteria

A peptide was classified as authenticated only if it satisfied all of the following:

  1. identified at 1% PSM and peptide FDR in a crown (dentine or enamel) extract;
  2. no match to the contamination database or exclusion list (§3.7);
  3. observed in at least two of the three independent campaigns;
  4. supported by at least two PSMs in total; and
  5. passed manual spectral review by two reviewers (H.L. and Y.Z.) for fragment-ion continuity, unexplained dominant peaks and plausibility of modification assignments.

Deamidation was used as an aggregate authenticity marker rather than a per-peptide filter. The deamidated fraction of N and Q was estimated from extracted-ion intensities of deamidated and non-deamidated forms, aggregated by peptide class.

3.9 Phylogenetic placement

Authenticated peptides were aligned to the reference proteins. Fragments identical across all reference taxa were classified as non-diagnostic. Diagnostic fragments were placed on a fixed reference tree of the seven extant taxa (topology from established vertebrate phylogeny) by maximum likelihood under a JTT+Γ model, and by Bayesian placement (two runs, four chains, 2×10⁶ generations, 25% burn-in, average standard deviation of split frequencies <0.01). Per-fragment placement posteriors are the posterior probability of the highest-supported branch. A concatenated placement of all diagnostic fragments was also computed, with 1,000 non-parametric bootstrap replicates for the ML analysis.

4. Results

4.1 Identifications and the authentication cascade

Across all crown extracts, 231,485 MS2 spectra were acquired, yielding 5,109 PSMs at 1% FDR and 731 unique peptide sequences. The extraction blanks (pooled) yielded 418 PSMs, all of which were assigned to the contamination database, predominantly human keratin and porcine trypsin autolysis products (Fig. 1).

Figure 1. Peptide-spectrum matches by extract and assignment classPSMs at 1% FDR in each extract, split into contaminant-assigned and candidate-endogenous classes. Candidate-endogenous PSMs are those not matching the contamination database; they are not yet authenticated. No candidate-endogenous PSMs were observed in any extraction blank.PSMs (1% FDR)02004006008001,0001,2001,4001,6001,800A dentineA dentine Candidate endogenous: 584 Total: 1,486A dentine Contaminant-assigned: 902 Total: 1,4861,486A enamelA enamel Candidate endogenous: 103 Total: 311A enamel Contaminant-assigned: 208 Total: 311311B dentineB dentine Candidate endogenous: 565 Total: 1,352B dentine Contaminant-assigned: 787 Total: 1,3521,352C dentineC dentine Candidate endogenous: 667 Total: 1,618C dentine Contaminant-assigned: 951 Total: 1,6181,618C enamelC enamel Candidate endogenous: 123 Total: 342C enamel Contaminant-assigned: 219 Total: 342342Root elementRoot element Candidate endogenous: 46 Total: 197Root element Contaminant-assigned: 151 Total: 197197Extractionblanks (n=9)Extraction blanks (n=9) Contaminant-assigned: 418 Total: 418418ExtractFigure 1. Peptide-spectrum matches by extract and assignment classPSMs at 1% FDR in each extract, split into contaminant-assigned and candidate-endogenous classes. Candidate-endogenous PSMs are those not matching the contamination database; they are not yet authenticated. No candidate-endogenous PSMs were observed in any extraction blank.05001,0001,5002,000A dentineA dentine Candidate endogenous: 584 Total: 1,486A dentine Contaminant-assigned: 902 Total: 1,4861,486A enamelA enamel Candidate endogenous: 103 Total: 311A enamel Contaminant-assigned: 208 Total: 311311B dentineB dentine Candidate endogenous: 565 Total: 1,352B dentine Contaminant-assigned: 787 Total: 1,3521,352C dentineC dentine Candidate endogenous: 667 Total: 1,618C dentine Contaminant-assigned: 951 Total: 1,6181,618C enamelC enamel Candidate endogenous: 123 Total: 342C enamel Contaminant-assigned: 219 Total: 342342Root elementRoot element Candidate endogenous: 46 Total: 197Root element Contaminant-assigned: 151 Total: 197197Extraction blanks(n=9)Extraction blanks (n=9) Contaminant-assigned: 418 Total: 418418PSMs (1% FDR)
Figure 1.Peptide-spectrum matches by extract and assignment class. PSMs at 1% FDR in each extract, split into contaminant-assigned and candidate-endogenous classes. Candidate-endogenous PSMs are those not matching the contamination database; they are not yet authenticated. No candidate-endogenous PSMs were observed in any extraction blank.Source: JM Shenzhen, internal dataset PP-2013-01.
Source data
ExtractCandidate endogenousContaminant-assigned
A dentine584902
A enamel103208
B dentine565787
C dentine667951
C enamel123219
Root element46151
Extraction blanks (n=9)0418

Table 4 shows how the 731 unique crown-derived sequences were reduced to the authenticated set. The largest single exclusion was the contamination database, which removed 55% of sequences. The replication criterion removed a further 51 sequences that were observed in one campaign only; these are retained in a separate unvalidated file and are not used further.

Table 4. Authentication cascade for unique peptide sequences from crown extracts.
Step Criterion Excluded Remaining
0 Unique sequences at 1% FDR, all crown extracts — 731
1 Match to contamination database 402 329
2 Observed in any extraction or solvent blank 38 291
3 Identical to Gallus method-development sequence 27 264
4 Observed in fewer than two independent campaigns 51 213
5 Fewer than two PSMs, or failed manual spectral review 17 196

Of the 196 authenticated peptides, 131 were observed in all three campaigns and 65 in two. Campaign C, performed in a different laboratory by a different operator with different reagents, independently recovered 173 of the 196. The operator of Campaign C was aware of the results of Campaigns A and B.

The root element yielded 46 candidate-endogenous PSMs and 9 unique candidate sequences, none observed in more than one campaign by design. It was not analysed further.

4.2 Protein assignment and coverage

Collagen type I accounts for 83% of authenticated peptides (Table 5), consistent with its dominance in dentine and with the general expectation that collagen is the most durable bone and tooth protein. Enamel-derived peptides were assigned to amelogenin, enamelin and ameloblastin. Seven peptides were assigned to non-collagenous proteins, of which five map to osteocalcin and two to secreted phosphoprotein 1; these are reported for completeness and not used in placement.

Table 5. Authenticated peptides and residue coverage by protein.
Protein Gene Authenticated peptides Unique residues covered Reference length (residues) Coverage
Collagen α1(I) COL1A1 118 744 1,057 70.4%
Collagen α2(I) COL1A2 45 571 1,040 54.9%
Amelogenin AMELX 14 121 198 61.1%
Enamelin ENAM 7 112 1,098 10.2%
Ameloblastin AMBN 5 88 421 20.9%
Other non-collagenous BGLAP, SPP1 7 76 — —
Total 196 1,712

Reference lengths for collagens refer to the mature chain including telopeptides. Coverage within collagen is concentrated in the triple-helical domain (Fig. 2); the N- and C-terminal telopeptides are almost entirely absent, which is expected given their susceptibility to early degradation.

Figure 2. Residue coverage along each protein, by alignment decileFraction of reference residues covered by at least one authenticated peptide, in ten equal segments of each aligned protein (N- to C-terminus). Collagen coverage is concentrated in the triple-helical domain. n = 196 authenticated peptides.Alignment segment (N → C)12345678910COL1A1COL1A1 × 1: 0.060.06COL1A1 × 2: 0.710.71COL1A1 × 3: 0.880.88COL1A1 × 4: 0.920.92COL1A1 × 5: 0.850.85COL1A1 × 6: 0.940.94COL1A1 × 7: 0.890.89COL1A1 × 8: 0.830.83COL1A1 × 9: 0.790.79COL1A1 × 10: 0.070.07COL1A2COL1A2 × 1: 0.030.03COL1A2 × 2: 0.420.42COL1A2 × 3: 0.660.66COL1A2 × 4: 0.710.71COL1A2 × 5: 0.580.58COL1A2 × 6: 0.740.74COL1A2 × 7: 0.690.69COL1A2 × 8: 0.610.61COL1A2 × 9: 0.440.44COL1A2 × 10: 0.040.04AMELXAMELX × 1: 0.810.81AMELX × 2: 0.920.92AMELX × 3: 0.770.77AMELX × 4: 0.680.68AMELX × 5: 0.550.55AMELX × 6: 0.620.62AMELX × 7: 0.710.71AMELX × 8: 0.480.48AMELX × 9: 0.360.36AMELX × 10: 0.110.11ENAMENAM × 1: 00.00ENAM × 2: 0.310.31ENAM × 3: 0.220.22ENAM × 4: 0.090.09ENAM × 5: 0.040.04ENAM × 6: 0.120.12ENAM × 7: 0.070.07ENAM × 8: 00.00ENAM × 9: 0.170.17ENAM × 10: 00.00AMBNAMBN × 1: 0.440.44AMBN × 2: 0.380.38AMBN × 3: 0.290.29AMBN × 4: 0.120.12AMBN × 5: 0.080.08AMBN × 6: 0.210.21AMBN × 7: 0.310.31AMBN × 8: 0.060.06AMBN × 9: 00.00AMBN × 10: 00.00Figure 2. Residue coverage along each protein, by alignment decileFraction of reference residues covered by at least one authenticated peptide, in ten equal segments of each aligned protein (N- to C-terminus). Collagen coverage is concentrated in the triple-helical domain. n = 196 authenticated peptides.Alignment segment (N → C)12345678910COL1A1COL1A1 × 1: 0.06COL1A1 × 2: 0.71COL1A1 × 3: 0.88COL1A1 × 4: 0.92COL1A1 × 5: 0.85COL1A1 × 6: 0.94COL1A1 × 7: 0.89COL1A1 × 8: 0.83COL1A1 × 9: 0.79COL1A1 × 10: 0.07COL1A2COL1A2 × 1: 0.03COL1A2 × 2: 0.42COL1A2 × 3: 0.66COL1A2 × 4: 0.71COL1A2 × 5: 0.58COL1A2 × 6: 0.74COL1A2 × 7: 0.69COL1A2 × 8: 0.61COL1A2 × 9: 0.44COL1A2 × 10: 0.04AMELXAMELX × 1: 0.81AMELX × 2: 0.92AMELX × 3: 0.77AMELX × 4: 0.68AMELX × 5: 0.55AMELX × 6: 0.62AMELX × 7: 0.71AMELX × 8: 0.48AMELX × 9: 0.36AMELX × 10: 0.11ENAMENAM × 1: 0ENAM × 2: 0.31ENAM × 3: 0.22ENAM × 4: 0.09ENAM × 5: 0.04ENAM × 6: 0.12ENAM × 7: 0.07ENAM × 8: 0ENAM × 9: 0.17ENAM × 10: 0AMBNAMBN × 1: 0.44AMBN × 2: 0.38AMBN × 3: 0.29AMBN × 4: 0.12AMBN × 5: 0.08AMBN × 6: 0.21AMBN × 7: 0.31AMBN × 8: 0.06AMBN × 9: 0AMBN × 10: 0
Figure 2.Residue coverage along each protein, by alignment decile. Fraction of reference residues covered by at least one authenticated peptide, in ten equal segments of each aligned protein (N- to C-terminus). Collagen coverage is concentrated in the triple-helical domain. n = 196 authenticated peptides.Source: JM Shenzhen, internal dataset PP-2013-01.
Source data
Protein12345678910
COL1A10.060.710.880.920.850.940.890.830.790.07
COL1A20.030.420.660.710.580.740.690.610.440.04
AMELX0.810.920.770.680.550.620.710.480.360.11
ENAM00.310.220.090.040.120.0700.170
AMBN0.440.380.290.120.080.210.310.0600

4.3 Fragment length

Authenticated peptides range from 6 to 31 residues, with a median of 11 (Fig. 3). The distribution is shifted towards shorter lengths relative to tryptic peptides from modern bovine bone analysed on the same instrument (median 15), as expected for material subject to extensive backbone cleavage. Undigested extracts contributed 58% of authenticated peptides, indicating that much of the recovered material was already fragmented before any enzyme was applied.

Figure 3. Length distribution of authenticated peptidesPeptide length in residues for the 196 authenticated peptides from JM-SP-0001. Median 11; range 6–31. Peptides of 6–7 residues were retained when all authentication criteria were met.Peptides010203040506810121416182024326–8: 238–10: 4110–12: 4412–14: 3314–16: 2116–18: 1318–20: 820–24: 824–32: 5Peptide length (residues)Median 11Figure 3. Length distribution of authenticated peptidesPeptide length in residues for the 196 authenticated peptides from JM-SP-0001. Median 11; range 6–31. Peptides of 6–7 residues were retained when all authentication criteria were met.Peptides0102030405010152025306–8: 238–10: 4110–12: 4412–14: 3314–16: 2116–18: 1318–20: 820–24: 824–32: 5Peptide length (residues)Median 11
Figure 3.Length distribution of authenticated peptides. Peptide length in residues for the 196 authenticated peptides from JM-SP-0001. Median 11; range 6–31. Peptides of 6–7 residues were retained when all authentication criteria were met.Source: JM Shenzhen, internal dataset PP-2013-01.
Source data
Peptide length (residues)Peptides
6–823
8–1041
10–1244
12–1433
14–1621
16–1813
18–208
20–248
24–325

4.4 Deamidation

Authenticated peptides show high levels of glutamine and asparagine deamidation in both dentine and enamel (Fig. 4). Contaminant-assigned peptides from the same specimen extracts, which share the specimen's entire burial and post-excavation history only to the extent that they were introduced during it, show low deamidation, as do human keratin peptides in the blanks and the modern bovine bone control. The root element shows deamidation comparable to the crown, which is consistent with its being genuinely ancient material, from a tooth other than this one.

Figure 4. Deamidated fraction of glutamine and asparagine by peptide classIntensity-weighted deamidated fraction of Q and N residues. Authenticated peptides from crown dentine and enamel are compared with contaminant-assigned peptides from the same extracts, the root element, human keratin from extraction blanks, and a modern bovine bone control.Deamidated fraction0.000.200.400.600.801.00Crown dentine(authenticated)Crown dentine (authenticated) Gln (Q): 0.74Crown dentine (authenticated) Asn (N): 0.89Crown enamel(authenticated)Crown enamel (authenticated) Gln (Q): 0.81Crown enamel (authenticated) Asn (N): 0.93Root elementRoot element Gln (Q): 0.69Root element Asn (N): 0.84Contaminant-assigned,crown extractsContaminant-assigned, crown extracts Gln (Q): 0.12Contaminant-assigned, crown extracts Asn (N): 0.19Human keratin,blanksHuman keratin, blanks Gln (Q): 0.06Human keratin, blanks Asn (N): 0.11Modern bovineboneModern bovine bone Gln (Q): 0.04Modern bovine bone Asn (N): 0.09Peptide classFigure 4. Deamidated fraction of glutamine and asparagine by peptide classIntensity-weighted deamidated fraction of Q and N residues. Authenticated peptides from crown dentine and enamel are compared with contaminant-assigned peptides from the same extracts, the root element, human keratin from extraction blanks, and a modern bovine bone control.0.000.400.80Crown dentine(authenticated)Crown dentine (authenticated) Gln (Q): 0.74Crown dentine (authenticated) Asn (N): 0.89Crown enamel(authenticated)Crown enamel (authenticated) Gln (Q): 0.81Crown enamel (authenticated) Asn (N): 0.93Root elementRoot element Gln (Q): 0.69Root element Asn (N): 0.84Contaminant-assigned,crown extractsContaminant-assigned, crown extracts Gln (Q): 0.12Contaminant-assigned, crown extracts Asn (N): 0.19Human keratin,blanksHuman keratin, blanks Gln (Q): 0.06Human keratin, blanks Asn (N): 0.11Modern bovine boneModern bovine bone Gln (Q): 0.04Modern bovine bone Asn (N): 0.09Deamidated fraction
Figure 4.Deamidated fraction of glutamine and asparagine by peptide class. Intensity-weighted deamidated fraction of Q and N residues. Authenticated peptides from crown dentine and enamel are compared with contaminant-assigned peptides from the same extracts, the root element, human keratin from extraction blanks, and a modern bovine bone control.Source: JM Shenzhen, internal dataset PP-2013-01.
Source data
Peptide classGln (Q)Asn (N)
Crown dentine (authenticated)0.740.89
Crown enamel (authenticated)0.810.93
Root element0.690.84
Contaminant-assigned, crown extracts0.120.19
Human keratin, blanks0.060.11
Modern bovine bone0.040.09

We regard the deamidation contrast between authenticated and contaminant-assigned peptides within the same extracts as the single most important piece of evidence in this report. It does not prove antiquity, but it is difficult to reconcile with recent introduction.

4.5 Preliminary phylogenetic placement

Of the 196 authenticated peptides, 131 contain at least one position that varies among the reference taxa and are classified as diagnostic; 65 are identical across all references and carry no placement information. Best-supported branches for the 131 diagnostic fragments are summarised in Table 6. The modal placement is the stem lineage of Aves (71 fragments). The 17 fragments placing on the crocodilian branch are disproportionately enamel-derived (11 of 17). Modern birds are edentulous, and their amelogenin, enamelin and ameloblastin genes are absent or degraded; enamel peptides can therefore be compared meaningfully only with the crocodilian references (the turtle outgroup is also edentulous), and their placement is biased toward the crocodilian branch by construction.

Table 6. Reference taxa and per-fragment placement of diagnostic peptides (n = 131).
Branch (best-supported) Fragments Of which enamel-derived Median posterior
Stem lineage of Aves 71 3 0.44
Within crown Aves (Gallus, Anas, Taeniopygia, Struthio) 22 0 0.31
Crocodylia (Alligator, Crocodylus) 17 11 0.36
Stem Archosauria 14 4 0.29
Unresolved (root, Chrysemys outgroup) 7 2 0.21
All diagnostic fragments 131 20 0.38

Individually, fragments are weakly informative: the median per-fragment posterior is 0.38, and placement confidence rises with fragment length (Fig. 5). Fragments shorter than 8 residues have a median posterior of 0.17 and an interquartile range spanning most of the tree.

Figure 5. Per-fragment placement posterior by fragment lengthMedian posterior probability of the best-supported branch (line) and interquartile range (band), by peptide length, for the 131 diagnostic authenticated peptides. Seven-taxon extant reference tree.Posterior placement probability0.000.200.400.600.801.00681012141618202224Fragment length (residues)Below 8 residues: high varianceMedian posteriorFragment length (residues): 6 Median posterior: 0.17 (0.08–0.31)Fragment length (residues): 8 Median posterior: 0.26 (0.14–0.39)Fragment length (residues): 10 Median posterior: 0.34 (0.21–0.47)Fragment length (residues): 12 Median posterior: 0.40 (0.27–0.53)Fragment length (residues): 14 Median posterior: 0.46 (0.33–0.58)Fragment length (residues): 16 Median posterior: 0.51 (0.38–0.63)Fragment length (residues): 20 Median posterior: 0.57 (0.44–0.69)Fragment length (residues): 24 Median posterior: 0.62 (0.49–0.74)Figure 5. Per-fragment placement posterior by fragment lengthMedian posterior probability of the best-supported branch (line) and interquartile range (band), by peptide length, for the 131 diagnostic authenticated peptides. Seven-taxon extant reference tree.Posterior placement probability0.000.200.400.600.801.00101520Fragment length (residues)Below 8 residues:high varianceFragment length (residues): 6 Median posterior: 0.17 (0.08–0.31)Fragment length (residues): 8 Median posterior: 0.26 (0.14–0.39)Fragment length (residues): 10 Median posterior: 0.34 (0.21–0.47)Fragment length (residues): 12 Median posterior: 0.40 (0.27–0.53)Fragment length (residues): 14 Median posterior: 0.46 (0.33–0.58)Fragment length (residues): 16 Median posterior: 0.51 (0.38–0.63)Fragment length (residues): 20 Median posterior: 0.57 (0.44–0.69)Fragment length (residues): 24 Median posterior: 0.62 (0.49–0.74)
Figure 5.Per-fragment placement posterior by fragment length. Median posterior probability of the best-supported branch (line) and interquartile range (band), by peptide length, for the 131 diagnostic authenticated peptides. Seven-taxon extant reference tree.Source: JM Shenzhen, internal dataset PP-2013-01.
Source data
Fragment length (residues)Median posteriorMedian posterior interval
60.170.08–0.31
80.260.14–0.39
100.340.21–0.47
120.40.27–0.53
140.460.33–0.58
160.510.38–0.63
200.570.44–0.69
240.620.49–0.74

Concatenated, the diagnostic set places JM-SP-0001 on the stem lineage of Aves with a posterior of 0.91 (ML bootstrap 87%), with residual support for crown Aves (0.05), Crocodylia (0.03) and other branches (0.01) (Fig. 6). Excluding enamel-derived peptides raises the stem-Aves posterior to 0.95.

Figure 6. Concatenated placement of JM-SP-0001 on the extant reference treeFixed seven-taxon reference topology with the concatenated placement of 131 diagnostic peptides. JM-SP-0001 places on the stem lineage of Aves (Bayesian posterior 0.91; ML bootstrap 87%). The extant reference set contains no non-avian dinosaurs, so placement within Theropoda is not testable with these data.JM-SP-0001p = 0.91JM-SP-0001 — placement posterior 0.91StruthioStruthioGallusGallusAnasAnasTaeniopygiaTaeniopygiaAlligatorAlligatorCrocodylusCrocodylusChrysemysChrysemysFigure 6. Concatenated placement of JM-SP-0001 on the extant reference treeFixed seven-taxon reference topology with the concatenated placement of 131 diagnostic peptides. JM-SP-0001 places on the stem lineage of Aves (Bayesian posterior 0.91; ML bootstrap 87%). The extant reference set contains no non-avian dinosaurs, so placement within Theropoda is not testable with these data.JM-SP-0001p = 0.91JM-SP-0001 — placement posterior 0.91StruthioStruthioGallusGallusAnasAnasTaeniopygiaTaeniopygiaAlligatorAlligatorCrocodylusCrocodylusChrysemysChrysemys
Figure 6.Concatenated placement of JM-SP-0001 on the extant reference tree. Fixed seven-taxon reference topology with the concatenated placement of 131 diagnostic peptides. JM-SP-0001 places on the stem lineage of Aves (Bayesian posterior 0.91; ML bootstrap 87%). The extant reference set contains no non-avian dinosaurs, so placement within Theropoda is not testable with these data.Source: JM Shenzhen, internal dataset PP-2013-01.
Source data
TaxonHighlightedPlacement posterior
JM-SP-0001yes0.91
Struthio
Gallus
Anas
Taeniopygia
Alligator
Crocodylus
Chrysemys

(((JM-SP-0001,(Struthio,((Gallus,Anas),Taeniopygia))),(Alligator,Crocodylus)),Chrysemys);

5. Discussion

The central result of this report is that a single isolated theropod tooth crown yielded, across three independent extraction campaigns, a reproducible set of peptide sequences that are not explained by any contamination source we were able to identify, that are heavily deamidated relative to known contaminants in the same extracts, and that place phylogenetically where a non-avian dinosaur would be expected to place on a tree of living animals.

Each of these lines of evidence is individually insufficient. Reproducibility across campaigns excludes one-off procedural contamination but not a contaminant intrinsic to the specimen. Deamidation is an aggregate measure that depends on burial environment as well as age and is not a clock. Placement on the avian stem is what one would expect of genuine theropod sequence, but is also, in principle, what one would expect of a sufficiently degraded avian contaminant whose diagnostic positions have been misassigned. We excluded identical matches to our only documented avian laboratory material (Gallus); we cannot exclude undocumented avian material encountered elsewhere in the specimen's history.

Taken together, however, the evidence is more coherent than any single contamination scenario we have been able to construct. The authenticated peptides behave as a population with a common, old history, distinct from every recent protein source present in the same tubes.

The relationship to the published literature deserves direct comment. The authenticated set reported here is substantially larger than earlier mass-spectrometric reports from Mesozoic bone (Asara et al. 2007; Schweitzer et al. 2009). We do not regard this as evidence of a superior method. Possible contributing factors include the use of dentine rather than bone, the inclusion of enamel, the combination of undigested and tryptic extracts, and a de novo sequencing step that expands the searchable space. The last of these also expands the opportunity for error, which is why de novo candidates were admitted only after re-identification at 1% FDR in a conventional search.

On taxonomy: the specimen was sold as Carcharodontosaurus, and that attribution is morphologically reasonable. It is not tested by these data. The extant reference set contains no non-avian dinosaurs, and no reference set can, because no non-avian dinosaur sequence is available to compare against. Placement within Theropoda in this report therefore means: molecular placement on the avian stem, combined with ziphodont tooth morphology. Attribution below that level reflects the point of sale.

6. Limitations and known issues

The following limitations apply to all results in this report. They are listed in no particular order of severity.

  • Single specimen. All results derive from one tooth. No second specimen was analysed.
  • Locality not recorded. The specimen's collection locality is known only as "Kem Kem," as stated by the seller. Stratigraphic position, burial environment and excavation date are unknown.
  • Purchase not documented. No receipt or written record of purchase was retained. The seller is identified by first name. The date of purchase is not recorded.
  • Pre-laboratory storage. The specimen was kept in a residential display cabinet at ambient temperature for an unrecorded period before 2011. Handling by household members, visitors, and an unknown number of people at the point of sale is assumed. Exposure to household cleaning products cannot be excluded.
  • Composite specimen. The root is a separate element bonded to the crown. It is excluded from this report. Whether the crown itself has been reshaped or restored beyond the repaired fracture was not determined to a standard we would consider conclusive.
  • Consolidant. Cyanoacrylate is present at the crown fracture. Sampling avoided the fracture by more than 5 mm; diffusion of consolidant through microcracks was not assessed.
  • Transport. The specimen was sent by international post in a padded envelope, declared as a mineral specimen, and received after an in-transit period of 11 days. It was not bagged.
  • Shared instrument. Campaigns A–C used an instrument previously used for modern Gallus collagen. Identical sequences were excluded (27 peptides). Sequences differing from Gallus only at I/L positions are indistinguishable and were also excluded; sequences differing at one other position were retained.
  • Operator awareness. The Campaign C operator was not blind to earlier results.
  • Short fragments. Peptides of 6–7 residues (23 of 196) are authenticated but individually carry little placement information. Their placement is high-variance.
  • Enamel placement bias. Enamel peptides have no avian comparison and are biased toward the crocodilian branch.
  • Reference coverage. Seven extant reference taxa were used. Placement resolution is limited accordingly.
  • Funding and independence. Campaigns A and B were not performed under a funded research programme and were not reviewed by any external body. No results in this report have been submitted for external peer review.

7. Implications for a reference library

The work described here was conducted using working files: spreadsheets of identifications, alignment files maintained per campaign, and a placement output directory per analysis run. At the time of writing, the authenticated set exists in 17 such files across three formats, maintained by two people. This is adequate for a single specimen and a single report. It will not be adequate for anything that depends on these sequences in future.

We therefore recommend that the authenticated peptides, together with the extant reference sequences against which they are placed, be consolidated into a single, versioned reference library, with:

  • a stable identifier for each fragment and each specimen;
  • the authentication status and evidence for each fragment recorded alongside its sequence;
  • the reference taxon set expanded beyond the seven used here, with explicit treatment of enamel proteins for which birds provide no comparison;
  • a documented curation procedure, with any manual decision recorded and attributable; and
  • explicit retention of excluded and single-campaign sequences in a separate, non-production set, so that exclusion decisions can be revisited.

This work is scheduled for completion in the first quarter of 2014 and will be described in JM-TR-2014-002.

8. Data availability

Library contents are confidential. Raw mass spectrometry files, search outputs, the contamination database and the authenticated peptide list are retained by Jurassic Meat and are not available to third parties. Specimen JM-SP-0001 is retained at the Shenzhen laboratory. The root element is retained separately and is labelled as not belonging to the specimen.

9. Acknowledgements

We thank R. Williams for the loan, and subsequent donation, of specimen JM-SP-0001, and a dealer in the Rissani Sunday market for assistance with specimen provenance. We thank the staff of the shared mass spectrometry facility for instrument access and for the 48-hour column wash preceding Campaign C.

Funding. Campaigns A and B were funded personally by the authors and a cofounder. Campaign C and all subsequent analysis were funded internally.

Competing interests. All authors are employees of Jurassic Meat and may hold equity.

Ethics. This report predates the constitution of the Ethics & Bioresponsibility Council (2014). It was reviewed retrospectively by the Council following its constitution; no concerns raised.

10. References

  1. Asara JM, Schweitzer MH, Freimark LM, Phillips M, Cantley LC (2007). Protein sequences from mastodon and Tyrannosaurus rex revealed by mass spectrometry. Science 316:280–285.
  2. Buckley M, et al. (2008). Comment on "Protein sequences from mastodon and Tyrannosaurus rex revealed by mass spectrometry." Science 319:33.
  3. Cox J, Mann M (2008). MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification. Nature Biotechnology 26:1367–1372.
  4. Schweitzer MH, et al. (2007). Analyses of soft tissue from Tyrannosaurus rex suggest the presence of protein. Science 316:277–280.
  5. Schweitzer MH, et al. (2009). Biomolecular characterization and protein sequences of the Campanian hadrosaur B. canadensis. Science 324:626–631.
  6. Witmer LM (1995). The extant phylogenetic bracket and the importance of reconstructing soft tissues in fossils. In: Thomason JJ (ed.), Functional Morphology in Vertebrate Paleontology. Cambridge University Press, 19–33.

Supplementary Table S1. Campaign-level reproducibility

Table S1. Authenticated peptides by campaign of observation.
Observed in Peptides Of which collagen Of which enamel Of which other
A, B and C 131 125 0 6
A and B only 23 22 0 1
A and C only 30 4 26 0
B and C only 12 12 0 0
Total 196 163 26 7

Campaign B included no enamel extraction; enamel peptides can therefore be authenticated only by observation in Campaigns A and C, and all 26 were.