Biology and palaeontology differ, among other things, in that the former studies living beings that perform all their vital functions and the latter studies vestiges of living beings that are currently integrated in the sedimentary record. However, there is an intermediate field relating to those beings that, although no longer alive, still retain some of their organic components unaltered. Apart from the exceptional preservation that occurs in environments with special conditions (freezing, mummification, etc.), the parts of animals that remain unaltered for the longest time after death are mainly mineralised tissues, such as bone or dentine.
Within this mineralised casing, biomolecules such as DNA or certain normally labile proteins can survive for thousands of years (Buckley & Collins Reference Buckley and Collins2011). Collagen is one such protein, and in recent decades it has been the subject of molecular palaeontology, the branch of palaeontology that studies the molecules contained in fossils in order to reconstruct various biological aspects of organisms from the past.
As the protein that it is, fossil collagen contains genetic information that in the last decade is being used for taxonomic purposes using the technique called zooarchaeology by mass spectrometry (ZooMS, Buckley et al. Reference Buckley, Collins, Thomas-Oates and Wilson2009). In this paper we will review the structure and composition of collagen and how it is applied to the identification of Pleistocene and Holocene European ursid remains.
1. Bone collagen as a subject of palaeontological study
1.1. On the bone collagen structure and composition
Bone is a living tissue made up of cells and extracellular matrix. It also has blood and lymphatics vessels, and nerve endings. The cells are found in spaces called lacunae and are responsible for secreting the components of bone, maintaining and remodelling it (Davies & Hosseini Reference Davies, Hosseini and Davies2000). The bone extracellular matrix is made up of two main components: the mineral fraction; and the organic matrix. The mineral fraction is mainly calcium phosphate or hydroxyapatite. The organic matrix, which accounts for one-third of the bone weight, consists mainly of proteins, including collagen (90%) and other non-collagenous proteins (3 to 5%) to which mineral crystals are bound (Davies & Hosseini Reference Davies, Hosseini and Davies2000). Dentine is similar in composition to the extracellular matrix of bone but lacks cells and does not undergo remodelling.
Type 1 collagen is the major insoluble fibrous protein of the organic matrix of bone or dentine (Henriksen & Karsdal Reference Henriksen, Karsdal and Karsdal2019). Like other proteins, collagen has a primary structure determined by the amino acid sequence. It is based on a repeat of the sequence glycine–X–Y, where X and Y can be any of the other amino acids except cysteine or tryptophan (Eastoe Reference Eastoe1955; Robinson & Rudd Reference Robinson and Rudd1974). Each molecule, or α-strand, acquires a three-dimensional secondary structure in the form of a left-handed helix. Individual strands are unstable and must aggregate for stability. Mature collagen or tropocollagen acquires a quaternary structure formed by three α-strands, twisted together into a right-handed triple helix. In tetrapods, the three strands of the triple helix of type 1 collagen are not the same: there are two α1-strands; and one α2-strand (Ricard-Blum Reference Ricard-Blum2011). Collagen synthesis takes place mainly in bone tissue cells, where α1 and α2 chains are synthesised separately (Henriksen & Karsdal Reference Henriksen, Karsdal and Karsdal2019). Both types of chains gradually mature losing the signal peptide and both terminal propeptides, leaving the helical region formed by more than 1000 residues (amino acids), flanked by two shorter telopeptides. The helical region of each chain undergoes post-translational modifications, which will eventually allow the three chains to wrap around each other to form the characteristic triple helix (Shoulders & Raines Reference Shoulders and Raines2009). The main modification is the hydroxylation of the amino acids, proline (Pro, P) and lysine (Lys, K). Hydroxylation is a chemical reaction in which a hydroxyl (OH) group is introduced to replace an H atom. Hydroxylation of proline and lysine can occur when they are in the third position of the sequence glycine–X–Y, and allows stable triple helices to be formed (Kuhn Reference Kuhn, Mayne and Burgeson1987). In type 1 collagen, approximately 50% of proline residues are hydroxylated, whereas this modification is more variable in lysine, between 15 and 90% of residues (Yamauchi & Sricholpech Reference Yamauchi and Sricholpech2012). Finally, each triple-helix associates into a right-handed super-super-coil referred to as the collagen microfibril. Each microfibril is interdigitated with its neighbouring microfibrils, forming the collagen fibre.
1.2. Fossil bones and fossil collagen
The bones of a present-day vertebrate contain on average 22% collagen (Crockett et al. Reference Crockett, Rogers, Coxon, Hocking and Helfrich2011). In fossils, the proportion decreases, depending on the age of the remains and the pH, humidity and temperature conditions to which they are subjected. Diagenetic degradation of bone can follow several pathways. The dissolution of the mineral fraction of the bone, due to the acidity of the soil, leads to the accelerated loss of collagen as it becomes accessible to microbial attack. Chemical degradation of collagen is dependent on moisture and temperature. It has been estimated that, under current European climatic conditions and with no other factors affecting the bone, it could survive for over a million years (Buckley & Collins Reference Buckley and Collins2011). Such ideal conditions do not usually occur in nature; even so, it is possible to recover sufficiently preserved collagen in bone remains tens of thousands or even several hundred thousand years old (Buckley & Collins, Reference Buckley and Collins2011; Britton et al. Reference Britton, Gaudzinski-Windheuser, Roebroeks, Kindler and Richards2012). In very rare cases, traces of collagen were found in dinosaur bones about one hundred million years old (Schweitzer et al. Reference Schweitzer, Zheng, Organ, Avci, Suo, Freimark, Lebleu, Duncan, Vander Heiden and Neveu2009; Lee et al. Reference Lee, Chiang, Huang, Chung, Huang, Wang, Chen, Chang, Liao and Reisz2017). Experimental studies showed that the composition of collagen remains virtually unchanged until only 1% of the initial amount remains in the bone (Dobberstein et al. Reference Dobberstein, Collins, Craig, Taylor, Penkman and Ritz-Timme2009).
The long preservation of collagen in bones and teeth allows biological and evolutionary data to be obtained from organisms that ceased to live long ago. Biomolecular analyses of collagen are increasingly used to reconstruct past life, whether in humans or animals. Stable isotope analyses of fossil bone collagen began to be applied in the last quarter of the 20th century, although initially only in an exploratory manner. However, their use has grown exponentially to the present day (Katzenberg & Waters-Rist Reference Katzenberg, Waters-Rist, Katzenberg and Grauer2019). As for taxonomic identification by mass spectrometry or peptide fingerprinting, it is a very recently implemented technique and is still under development. In zooarchaeology it has been specifically termed ZooMS (Buckley et al. Reference Buckley, Collins, Thomas-Oates and Wilson2009).
1.3. The ZooMS technique
The identification of bone remains by means of their collagen peptide fingerprint, or ZooMS, is proving to be a powerful tool in palaeontology. In Pleistocene sites, and even more so if they are of anthropogenic origin, the taxonomic identification of faunal remains is not always easy due to the high degree of fragmentation that bones usually present. To identify these small remains, one possibility would be to sequence their DNA. However, this is an expensive and laborious technique, which does not always give good results because the DNA of ancient remains is usually degraded and contaminated.
Another possibility is the sequencing of bone proteins, for example, collagen. The advantage over the study of DNA is that collagen is much more abundant in bone remains than DNA, is better preserved and is easier to extract. The disadvantage is that collagen is less specific. Within each major taxonomic group (e.g., mammals), the proteins that perform the same function are usually very similar, although they have small differences related to the evolution of the different lineages separately and the accumulation of mutations over time. Some very specific proteins, such as collagen, cannot accumulate too many mutations because they would lose their structure and function. Even so, small differences in amino acid sequence occur between collagen molecules of different taxa that allow differentiation of these taxa (Buckley Reference Buckley, Giovas and LeFebvre2018). In order to identify the taxon from which the bone remnant from which the collagen is obtained comes, it would be necessary to sequence the complete protein, that is, to identify the entire sequence of amino acids of which it is composed. This is a complex technique and only represents a small advantage over DNA sequencing, which is the greater ease of extraction of collagen compared to DNA, due to its abundance.
Peptide mass fingerprinting (PMF) (James et al. Reference James, Quadroni, Carafoli and Gönnet1993; Pappin et al. Reference Pappin, Hojrup and Bleasby1993) is a technique for protein identification that does not require sequencing and is based simply on the differences in the molecular weight of the various peptides. Since each amino acid has a specific molecular weight, due to its chemical composition, small differences in the amino acid sequence will produce proteins of different molecular weight. If we consider the entire collagen molecule, in which each α-strand contains more than 1000 amino acids, the difference in molecular weight will be very small and it is not possible to know in which position the substitution(s) are located. The basis of the PMF technique consists of cleaving the protein at specific sites by enzymatic digestion. This yields a series of peptides of different sizes, each of which has a specific amino acid sequence and thus a characteristic mass. In this way it is easier to know at which points in the sequence (in which of the peptides) the substitutions are found.
Trypsin is an enzyme produced in the pancreas and acts in the digestion of proteins, breaking them into smaller fragments called peptides. This cleavage is specific, as it only acts by breaking the peptide bonds at the C-terminal end of the amino acids, lysine (Lys, K) and arginine (Arg, R), except when the following residue is a proline (Olsen et al. Reference Olsen, Ong and Mann2004). Digestion of a protein with trypsin in vitro produces the same cleavage effect on the protein under study. This results in a series of peptides (which, because they are obtained by trypsin, are called tryptic peptides) that will always end in a lysine or an arginine.
The substitution of a single amino acid produces a peptide of different molecular weight. These small differences cause homologous peptides in different taxa to have different masses. By measuring these masses, it is possible to recognise the different taxa without the need for sequencing or reading of the amino acid sequence. The application of ZooMS to identify bone proteins from ancient remains began to be developed at the beginning of the 21st century (Ostrom et al. Reference Ostrom, Schall, Gandhi, Shen, Hauschka, Strahler and Gage2000) and is still under development. Given the highly conserved amino acid sequence of collagen, most of the peptides obtained by tryptic digestion are identical across taxa and only a few of them are useful for differentiation at the taxonomic level. Collagen type 1 alpha 1 (col1α1) is more conservative among taxa (Buckley Reference Buckley, Giovas and LeFebvre2018) and only two of its peptides are used as markers for taxonomic purposes. The remaining peptide markers are found on alpha chain 2 (col1α2).
The first studies using ZooMS were focused on the identification of large terrestrial mammals (Buckley & Collins Reference Buckley and Collins2011; Buckley & Kansa Reference Buckley and Kansa2011; Buckley et al. Reference Buckley, Cosgrove, Garvey and Prideaux2017a, among others), but their use soon expanded to marine mammals (Kirby et al. Reference Kirby, Buckley, Promise, Trauger and Holdcraft2013; Buckley et al. Reference Buckley, Fraser, Herman, Melton, Mulville and Pálsdóttir2014), micromammals (Buckley et al. Reference Buckley, Gu, Shameer, Patel and Chamberlain2016) or marsupials (Buckley et al. Reference Buckley, Harvey and Chamberlain2017b; Peters et al. Reference Peters, Richter, Manne, Dortch, Paterson, Travouillon, Louys, Price, Petraglia, Crowther and Boivin2021). Significant progress is also being made in the identification of fish (Richter et al. Reference Richter, Wilson, Jones, Buckley, van Doorn and Collins2011; Harvey et al. Reference Harvey, Daugnora and Buckley2018), bird bone (Horn et al. Reference Horn, Kenens, Palmblad, van der Plas-Duivesteijn, Langeveld, Meijer, Dalebout, Marissen, Fischer, Vincent Florens, Niemann, Rijsdijk, Schulp, Laros and Gravendeel2019) and eggshell (Stewart et al. Reference Stewart, Allen, Jones, Penkman and Collins2013; Presslee et al. Reference Presslee, Wilson, Woolley, Best, Russell, Radini, Fischer, Kessler, Boano, Collins and Demarchi2017), amphibians (Buckley & Cheylan Reference Buckley and Cheylan2020) or sea turtles (Harvey et al. Reference Harvey, LeFebvre, Defrance, Toftgaard, Drosou, Kitchener and Buckley2019), among others, which reveals the great potential of the use of ZooMS.
1.4. Identification of Ursids by ZooMS
Although the cave bear (Ursus spelaeus sensu lato) is a common component in many European Pleistocene cave sites, no peptide fingerprint taxonomy study has paid special attention to this species up to now. The first work devoted to the application of peptide mass fingerprinting to fossil mammals (Buckley et al. Reference Buckley, Collins, Thomas-Oates and Wilson2009) does not include any ursid specimens. Subsequently, bear peptide markers were offered in some works (Buckley & Collins Reference Buckley and Collins2011; Kirby et al. Reference Kirby, Buckley, Promise, Trauger and Holdcraft2013; Welker et al. Reference Welker, Hajdinjak, Talamo, Jaouen, Dannemann, David, Julien, Meyer, Kelso, Barnes, Brace, Kamminga, Fischer, Kessler, Stewart, Pääbo, Collins and Jean-Jacques Hublin2016; Buckley et al. Reference Buckley, Cosgrove, Garvey and Prideaux2017a). However, the recent implementation of this technique and the fact that it is still under development causes some disparity in the definition of peptides, either in their validity as markers, or in their position in the molecule, or simply in their nomenclature (Richter et al. Reference Richter, Codlin, Seabrook and Warinner2022). For peptide markers in Ursidae, the most recent proposal is that of Welker et al. (Reference Welker, Hajdinjak, Talamo, Jaouen, Dannemann, David, Julien, Meyer, Kelso, Barnes, Brace, Kamminga, Fischer, Kessler, Stewart, Pääbo, Collins and Jean-Jacques Hublin2016), which is based on collagen obtained from a modern brown bear sample. The m/z values of the marker peptides (Table 1) do not differ from those proposed in previous or subsequent works.
List of peptide markers identified so far for brown bear and polar bear.
The markers identified as A', F' and G' are the same as their namesakes, but with an extra hydroxylation that adds 16 Da to the peptide.
Initially the peptide markers were identified as correlative letters of the alphabet (Buckley et al. Reference Buckley, Collins, Thomas-Oates and Wilson2009). Subsequently, a system was adopted that identifies peptides by indicating which chain they come from (α1 or α2) and their order in the molecule (Buckley et al. Reference Buckley, Collins, Thomas-Oates and Wilson2009). A recent proposal for standardisation of peptide nomenclature (Brown et al. Reference Brown, Douka, Collins and Richter2021a) in addition to indicating the chain, identifies each peptide by the position of the amino acid with which it begins and ends starting from the beginning of the helical region, which facilitates the task of identifying the peptides.
Since none of the works identifying peptide markers included specifically cave bear collagen sequences, in this work we will attempt to cover this gap by studying cave bear samples from different origins and different chronology. The purpose of this study is twofold: firstly, we will check whether the marker peptides proposed in the literature for brown bears coincide with those obtained in cave bears; and secondly, we will try to find out if there is any difference between both cave and brown bears from distant geographical regions or of different chronology. For our purpose, we will rely on the direct study of collagen samples, but also on the in-silico study of the sequences available in protein databases. This allows, through the use of bioinformatics tools, to obtain the theoretical tryptic peptide spectrum and compare it with those obtained from bone samples.
2. Material and methods
2.1. In silico study of the sequences of ursid bone collagen
One way to know the sequence of amino acids that make up the a and b chains of ursid collagen is to resort to the databases available in UniProt Knowledgebase (UniProtKB). This is a central hub of protein knowledge by providing a unified view of protein sequence and functional information, made freely available by The UniProt Consortium in https://www.uniprot.org/uniprot/ (Magrane Reference Magrane2011).
The UniProtKB consists of two sections: UniProtKB/Swiss-Prot; and UniProtKB/TrEMBL. UniProtKB/Swiss-Prot is manually curated, so that the information in each entry is annotated and reviewed by a curator. The records in UniProtKB/TrEMBL are automatically generated and the records await full manual annotation. This means that not all the entries are fully reliable. In the case of the Ursidae, none of the sequences of collagen 1 are manually revised. To choose the most reliable entries, we used the sequence alignment function and chose those sequences that share the most common positions.
For col1α1 there are three sequences, one from Ursus maritimus, the polar bear (A0A384BX56 in UniprotBK) and two from Ursus arctos horribilis (the North American grizzly bear), of which only one (A0A3Q7X3Q3) preserves the complete helical region (that which constitutes the mature collagen fibrils). The alignment shows that, although there are some differences at the ends of the molecules, the helical region, which is used to identify marker peptides, is almost completely coincident (only two substitutions in a molecule of more than 1000 residues).
For col1α2, there are 10 sequences from U. maritimus and only one from U, arctos horribilis. Only one of the U. maritimus sequences (A0A384BPF6_URSMA) is almost complete. Performing an alignment of this entry with the U. arctos horribilis one, the coincidence in the amino acid sequence is 100% in the helical region. Since the sequence is identical in both species, we consider it to be valid. In addition, for comparative purposes, we have used the sequences of other carnivore species (Table 2). Of these, the dog is the only one that is manually checked. For the carnivores we added punctually the comparison with the collagen sequences of human, cow, sheep and horse.
Sequences of col1α1 and col1α2 in UniProtKB used in this work.
The sequences obtained were analysed using the Peptide Mass tool available on Expasy, (Swiss Bioinformatics Resource Portal, https://www.expasy.org/). This tool allows performing a tryptic digestion simulation and obtaining the peptide spectrum of each type of collagen, with its amino acid sequence and the theoretical m/z value. To this value it will be necessary to add the difference for each possible hydroxylation of the P or K residues. In addition, the deamidation of glutamine (Gln, Q), a frequent alteration in ancient collagen (Van Doom et al. Reference Van Doorn, Wilson, Hollund, Soressi and Collins2012; Wilson et al. Reference Wilson, van Doorn and Collins2012), would add +0.984 Da (practically one unit) for each altered glutamine (Robinson & Rudd Reference Robinson and Rudd1974).
2.2. Cave and brown bear bone samples
For this study we have selected 20 samples of cave bear, U. spelaeus Rosenmüller, Reference Rosenmüller1794, identified morphologically and, in most cases, genetically (Table 3). The samples come from several sites in the Iberian Peninsula, Austria, Italy, Slovenia and Russia (Fig. 1) and cover most of the cave bear varieties described according to their mitochondrial lineages: U. spelaeus; Ursus ingressus; Ursus ladinicus; Ursus rossicus; and Ursus kanivetz (Barlow et al. Reference Barlow, Paijmans, Alberti, Gasparyan, Bar-Oz, Pinhasi, Foronova, Puzachenko, Pacher, Dalén, Baryshnikov and Hofreiter2021). Direct carbon-14 dating is available for five of them, while the ages of the remaining samples are known from their stratigraphic position.
Map showing the location of the sites from which the studied samples originated. Below, an enlargement of the inset of the map above.
Data on the bone samples used in this study: location, chronology, taxonomy and collagen quality parameters. %Ccol: percentage of C in collagen. %Ncol, percentage of N in collagen. C:Nat, carbon-to-nitrogen atomic weight ratio in collagen.
The taxon column refers, for cave bears, to the genetic identification of the cave bear variety according to its mitochondrial DNA: (1) according to González-Fortes et al. (Reference González-Fortes, Grandal-d'Anglade, Kolbe, Fernandes, Meleg, García-Vázquez, Pinto-Llona, Constantin, Torres, Ortiz, Frischauf, Rabeder, Hofreiter and Barlow2016); (2) according to Barlow et al. (Reference Barlow, Paijmans, Alberti, Gasparyan, Bar-Oz, Pinhasi, Foronova, Puzachenko, Pacher, Dalén, Baryshnikov and Hofreiter2021); (3) González-Fortes et al. (Reference González-Fortes, García-Vázquez, Pinto Llona and Grandal-d'Anglade2017); (3*) González-Fortes et al. (Reference González-Fortes, García-Vázquez, Pinto Llona and Grandal-d'Anglade2017) but no DNA was recovered; and (4) no direct DNA study. (?) indicates the lack of genetic study in the site. The age of the samples is given in carbon-14 dating ages BP and their error when the sample is directly dated; in ka BP when it is a stratigraphic age (obtained by dating other coeval bones). (*) dates obtained by amino acid racemisation. Abbreviations: IP = Iberian Peninsula.
Additionally, we included in the study 10 samples of brown bear (Ursus arctos Linnaeus, Reference Linnaeus1758), all identified morphologically and most of them also by their mitochondrial DNA. The samples come from the Cantabrian region in the Iberian Peninsula. Two are of Pleistocene age, seven are Holocene and the last one is a modern specimen, deceased in 2015 in the vicinity of the town of Belmonte de Miranda (Asturias, Spain).
2.3 Pretreatment of the samples and extraction of bone collagen
A sample of approximately 1 g was cut from each bone with a hand tool equipped with a diamond disc. Cancellous tissue and superficial concretions, if present, were mechanically removed. The fragment was repeatedly rinsed in an ultrasonic bath, successively in deionised water and acetone (a minimum of five rinses in acetone and six rinses in water, or more if necessary, until no turbidity was observed) and then left to dry in glass Petri dishes that protect it from dust and other possible contaminations, at room temperature, for at least 48 h.
The collagen extraction protocol follows a modified method of Longin (Reference Longin1971) described in Bocherens et al. (Reference Bocherens, Billiou, Patou-Mathis, Bonjean, Otte and Mariotti1997), with further modifications implemented in the Laboratory of Molecular Palaeontology of the University Institute of Geology, University of A Coruña (Spain), where the treatment was carried out. Our purification protocol is based on successive filtrations, which eliminate collagen fragments and retain only the large collagen strands. For each specimen studied, bone fragments (about 500 mg) were manually ground with agate mortar and pestle. The bone powder was sieved to obtain the fraction less than or equal to 0.5 mm. The use of powdered bone shortens the demineralisation time and therefore reduces the possibility of collagen degradation.
From each sample, a portion of between 250 and 300 mg of bone powder was taken, demineralised in about 30 mL of 1 M hydrochloric acid for 20 min, washed in deionised water until reaching a neutral pH and filtered through nitrocellulose filters (Sartorius Stedim®) of 5 μ pore size. The solid residue was incubated for 21 h at room temperature in 30 mL of 0.125 M sodium hydroxide to remove possible organic contaminants, such as fats or humic acids. After washing at neutral pH and further purification by filtration, the solid fraction containing collagen was solubilised in 20 mL of 0.1 M hydrochloric acid for 17 h at 90°, filtered a third time to remove insoluble mineral particles, frozen at −80 °C and freeze-dried for the analysis.
2.4. Collagen analysis
For peptide fingerprinting or ZooMS analysis, an aliquot of the collagen isolated from each bone was digested with trypsin, which breaks the molecular bonds between specific amino acids (after a lysine, K or an arginine, R, if not followed by proline, P). Thus, a set of peptides of different mass and charge (m/z) was obtained, identified by matrix-assisted laser desorption/ionisation, time-of-flight (MALDI-TOF). This analysis was performed at the Mass Spectrometry and Proteomics Unit of the Research and Technological Development Support Infrastructures Network, University of Santiago de Compostela (Spain).
For each sample, an aliquot of 1 to 5 mg of lyophilised collagen was dissolved in ammonium hydrogen carbonate buffer. After addition of Promega Trypsin Gold, mass spectrometry grade, the samples were digested at 37 °C overnight. The sample solution was mixed with a matrix of α-cyano-4-hydroxycinnamic acid by applying 1 μl of the mixture onto the MALDI plate in a Bruker Ultraflex® III MALDI-TOF/TOF mass spectrometer equipped with a smart beam laser. The principle of this type of analyser is simple. Once the collagen-matrix mixture is introduced into the plate, laser shots cause the mixture to be gently ionised. Next, an extraction voltage induces the mobilisation of all peptides simultaneously. These will pass through an accelerating electrostatic field, acquiring a high kinetic energy that propels them in the direction of the flight tube facing the detector. The travel time of the flight tube length will be directly proportional to the mass to charge ratio (m/z) of the ionised peptides. The spectrum produced is compared with published reference spectra (Welker et al. Reference Welker, Hajdinjak, Talamo, Jaouen, Dannemann, David, Julien, Meyer, Kelso, Barnes, Brace, Kamminga, Fischer, Kessler, Stewart, Pääbo, Collins and Jean-Jacques Hublin2016, the most complete and recent database) and with those obtained from in silico tryptic digestion, to identify the peptide markers and their m/z values.
3. Results
Collagen had been extracted from all the samples included in this work for stable isotope studies. Some results have been published (Pérez-Rama et al. Reference Pérez-Rama, Fernández-Mosquera and Grandal-d'Anglade2011; García-Vázquez et al. Reference García-Vázquez, Pinto-Llona and Grandal-d'Anglade2018; Grandal-d'Anglade et al. Reference Grandal-d'Anglade, Pérez-Rama, García-Vázquez and González-Fortes2019), while others are currently in preparation. Collagen from all samples yielded good results in terms of the usual quality criteria: yield; % carbon (C) and (N); and C:N atomic ratio (DeNiro Reference DeNiro1985; Ambrose Reference Ambrose1990; Van Klinken Reference Van Klinken1999).
The results obtained from tryptic digestion and MALDI-TOF analysis are shown in Table 4, where only the marker peptides are listed following Welker et al. (Reference Welker, Hajdinjak, Talamo, Jaouen, Dannemann, David, Julien, Meyer, Kelso, Barnes, Brace, Kamminga, Fischer, Kessler, Stewart, Pääbo, Collins and Jean-Jacques Hublin2016). The full spectra can be seen in the Online Supplementary Material (Table S1) available at https://doi.org/10.1017/S1755691023000038. Despite the absence of some markers in some samples, this result allows us to identify all specimens as members of the genus Ursus.
Results of the identification of peptide markers in the analysed samples.
4. Discussion
4.1. Identification of peptides in Ursidae
The peptide spectra obtained from the collagen analysis of 30 ursid specimens show in general a good coincidence in some of the markers, such as P1, B, D and G, while in others the presence is variable, or even does not appear in any of the samples.
The absence of peptide A or COL1α2 978–990 is not unusual, as we found it mainly as the variety with an extra hydroxylation (A'). In contrast, the peptide COL1α2 767–799 is preferably found in its version with one less hydroxylation than all the others (G instead of G'). In most cases it shows one extra unit, the result of the deamidation of the single glutamine (Q) residue it contains.
The absence of some of the peptides in some of the samples may be due to the collagen molecule being broken during the diagenetic phase, so that digestion with trypsin produces peptides of a smaller size than expected. This tends to occur if the bones are badly damaged and is most noticeable in the larger peptides (Buckley et al. 2011). However, the low presence or even complete absence of some peptides such as P2, E or F, which we will discuss in more detail below, is noteworthy.
4.1.1. P2 peptide, COL1α2 292–309
This peptide was initially proposed as useful for cetacean identification (Buckley et al. Reference Buckley, Fraser, Herman, Melton, Mulville and Pálsdóttir2014). In terrestrial mammals, and according to previous studies, it is identified by a peak of m/z 1609.8, common to all canids, felids and mustelids for which there are ZooMS data (Welker et al. Reference Welker, Hajdinjak, Talamo, Jaouen, Dannemann, David, Julien, Meyer, Kelso, Barnes, Brace, Kamminga, Fischer, Kessler, Stewart, Pääbo, Collins and Jean-Jacques Hublin2016).
However, our in-silico study on bear and other carnivores’ col1α2 sequences shows at that position the sequence GPNGEAGSAGPSGPPGLR whose m/z is 1577.7. It contains a proline (P) susceptible to hydroxylation (before glycine, G), so the peptide could reach a m/z of 1593.7. The same sequence is found in felids and canids. In addition to this discrepancy, the only taxonomic study that includes a significant amount of ursid samples, from Denisova Cave (Brown et al. Reference Brown, Wang, Oertle, Kozlikin, Shunkov, Derevianko, Comeskey, Jope-Street, Harvey, Chowdhury, Buckley, Higham and Douka2021b) does not identify this peptide in any of the 175 samples identified as ursids. Nor does it appear in any of the carnivores of that site.
In our cave bear set, a 1609 peak does not appear in any of the 20 cases, but a 1592 mass peptide appear in 18 of them, and 1593 in another one. The same is true for brown bears, where the peak 1609 is absent in all the samples, with a peak at 1592 in seven of the samples and one with 1593. This peptide could correspond to COL1α2 292–309, but the systematic difference of one unit less than the theoretical mass in most of the samples does not allow us to state this before applying a technique capable of identifying each residue, and not just the mass of the whole peptide. This lower-than-expected value may be due to cave bears showing low δ 15N and δ 13C bone collagen values (Grandal et al. Reference Grandal-d'Anglade, Pérez-Rama, García-Vázquez and González-Fortes2019) due to their plant-based diet. If the isotopic ratios of the samples are different from those calculated in silico, based on a homogeneous isotopic mixture, it would not be impossible to obtain a slightly lower m/z value. It could be argued that this reason is not valid for the more omnivorous brown bear, but all our samples come from the Cantabrian area, where brown bears showed a diet mainly based on plant foods (García-Vázquez et al. Reference García-Vázquez, Pinto-Llona and Grandal-d'Anglade2018). We also found this possible effect in the peptide Col1α2 502–519 (C) where five of the cave bears also show one unit less than expected, while there is only one in the brown bear set. In any case, all results point to the fact that this peptide was misidentified in carnivores, or at least in ursids, in the databases published so far.
4.1.2. Peptide E, Col1α2 454–483
The validity of this peptide as a taxonomic marker was initially proposed by Buckley et al. (2011) but was later excluded from further studies by those authors since it was not present in many ancient samples (Buckley et al. Reference Buckley, Cosgrove, Garvey and Prideaux2017a). Recent studies, however, keep considering it among the useful markers. Welker et al. (Reference Welker, Hajdinjak, Talamo, Jaouen, Dannemann, David, Julien, Meyer, Kelso, Barnes, Brace, Kamminga, Fischer, Kessler, Stewart, Pääbo, Collins and Jean-Jacques Hublin2016) assign this peptide a m/z of 2808 for the polar bear, whereas for brown bear and American black bear it is unresolved. These authors rely on sequences obtained from modern samples, but their sequences are not complete (Welker et al. Reference Welker, Hajdinjak, Talamo, Jaouen, Dannemann, David, Julien, Meyer, Kelso, Barnes, Brace, Kamminga, Fischer, Kessler, Stewart, Pääbo, Collins and Jean-Jacques Hublin2016, supplement). The antiquity, however, does not seem to be the reason for the absence of this peptide in our samples, as discussed below.
In the analysis of the Col1α2 sequences of UniProtKB, both the brown bear (A0A3Q7VKW6) and the polar bear (A0A384BPF6) show that positions 454 to 483 are occupied by the sequence GEQGPAGPPGFQGLPGPAGT AGEAGKPGER, with m/z = 2744.3. The sequence contains three proline residues susceptible to hydroxylation and two glutamine (Q) residues susceptible to deamidation. This offers a variety of possibilities as to the final m/z of the peptide. With all three prolines hydroxylated, the value would be 2792. If we add the possibility of Q deamidation, the final value could be 2793 or 2794.
Brown et al. (Reference Brown, Wang, Oertle, Kozlikin, Shunkov, Derevianko, Comeskey, Jope-Street, Harvey, Chowdhury, Buckley, Higham and Douka2021b), in the Denisova Cave samples, record peaks of m/z 2792 in 10 samples of the 175 identified as ursids, and none with the value 2808. In our 20 cave bear samples, 15 yield a peak at 2793 and five at 2794. Additionally, only four of them show a peptide 2808. In the case of brown bears, seven of them show a peak at 2793, and none at 2808, including the present-day brown bear sample.
Based on our results and the absence of peptide 2808 in the Denisova spectra, we propose that the peptide value in cave bears and by extension in Ursidae is 2792, which could be 2793 or 2794 considering the possible deamidation of the Q residues in the ancient samples.
4.1.3. Pepide F or COL1α1 586–618
It is located in the α1 chain of bone collagen. In ursids it is identified by a peak at m/z 2853.4. However, in silico digestion of Col1α1 from any of the mammals we used in this study does not yield this peptide intact, but rather two contiguous peptides, at positions 586 to 603 and 604 to 618.
The first peptide, which we will call here F1 for short, is composed of the sequence GLTGPIGPPGPAGAPGDK which is invariant not only in ursids or carnivores, but in all the taxa reviewed. Its m/z is 1558.8, but as it presents two prolines in position suitable for hydroxylation, the m/z values could be increased to 1574.8 with one hydroxylation, and 1590.8 with both prolines hydroxylated. In addition, the existence of a K residue before a G (before tryptic digestion) may add an extra hydroxylation, which would yield an m/z value of 1606.8.
The second peptide, or F2 for short, has a variable sequence, depending on the taxa. In the review of in silico sequences we have found up to three variants, produced by the substitution of an amino acid in the third position of the peptide. In some taxa such as Ovis and Equus, the third amino acid is T and the mass of the peptide is 1311.6. In Homo sapiens the third place is occupied by S and the mass is 1297, as also occurs in other primates as well as in rhinoceros, hippopotamus and some seals. In most carnivores including bears, canids, felids and mustelids, the sequence is GEAGPSGPAGPTGAR and the m/z value is 1281.6. It does not present any amino acids susceptible to post-translational modifications. This same sequence is found also in Bos.
The sum of m/z of both peptides does not correspond to the value reported in the databases. This is due to the fact that the peptide bond established between the carboxyl group (–COOH) of an amino acid and the amino group (–NH2) of the immediately adjacent amino acid leads to the release of a water molecule (−18 Da). Therefore, the m/z value attributed in previous literature has a mass of 18 Da less than the sum of its two components.
According to this, in the peptide spectrum of a sample, these two peptides should be either found separately, or joined together if the action of trypsin failed to separate them. Whether or not the amino acid sequence is cleaved between residues 603 and 604 may be due to the performance of the trypsin used, or even to the digestion time. As the trypsin used for all our samples is the same, it can be estimated that it is the digestion time that determines the performance of tryptic cleavage of this fraction of the collagen molecule. In 11 of the cave bear samples studied in this work a peptide 1606 is detected, but only in four samples a peak 1281 appears. In four other samples there is a peptide 2853. In the brown bear set, the peak corresponding to the two peptides together appears in only two samples, while peak 1606 is in three (and in one more, the 1590 variant). Peak 1281, which would correspond to the second peptide fraction, appears in only one case. The scarce occurrence of this peak allows us to hypothesise that the theoretical sequence obtained from the in-silico analysis may not be accurate, which is not surprising since there is only one sequence of brown bear col1α1, recalling it was not curated.
Finally, Table 5 shows the presence of peaks in the analysed samples at the m/z values calculated for these three peptides from the in-silico analysis of the UniProtKB brown bear sequence. None of them is crucial for the identification of bear skeletal remains by ZooMS if the collagen is well preserved. However, the E marker (α2 454–483) may be useful for differentiating between ursids and felids when the larger peptide G (COLα2 767–799) is not present.
Alternative values for some peptide markers identified in the in-silico analysis of α1 and α2 collagen sequences and their presence in the analysed samples.
In any case, the final identification of the peptides must be carried out by other proteomic techniques, such as liquid chromatography–mass spectrometry. It is necessary to ensure that the peaks found in the spectra of the analysed bears really correspond to the sequences in the established position, and that they are not the result of m/z coincidence with other peptides or peptide fragments present in the collagen.
4.2. Comparison of sequences between brown bear and cave bear species
We found no differences between the peptide markers of brown bear and cave bear of any species, at least not for the commonly used peptide markers. This is not surprising, as the divergence between brown bears and all cave species has been set at 1.5 million years (based on nuclear DNA, Barlow et al. Reference Barlow, Paijmans, Alberti, Gasparyan, Bar-Oz, Pinhasi, Foronova, Puzachenko, Pacher, Dalén, Baryshnikov and Hofreiter2021). As we have already seen, the need to maintain the stability of the collagen molecule prevents major changes in its amino acid sequence. The rate of amino acid substitution is estimated to be one every 1–8 million years depending on the vertebrate class (Buckley Reference Buckley, Giovas and LeFebvre2018). Therefore, taxonomic identification based on ZooMS generally does not achieve more than genus rank. Also, genetic divergence between cave bear species occurred less than one million years ago (Barlow et al. Reference Barlow, Paijmans, Alberti, Gasparyan, Bar-Oz, Pinhasi, Foronova, Puzachenko, Pacher, Dalén, Baryshnikov and Hofreiter2021), which would not be sufficient to cause amino acid substitutions in collagen molecules, at least in the peptide markers used. For example, the oldest separation within the cave bear lineage was about 0.83 million years ago between U. rossicus (represented in our samples by those from Kizel in the Urals) and all others (Barlow et al. Reference Barlow, Paijmans, Alberti, Gasparyan, Bar-Oz, Pinhasi, Foronova, Puzachenko, Pacher, Dalén, Baryshnikov and Hofreiter2021) with no differences in the peptide markers commonly used. Nevertheless, there is a possibility that there are substitutions in other regions of the collagen molecules, but to detect these, the entire collagen molecules would have to be sequenced, as single substitutions cannot be detected simply by their m/z values.
4.3. Sequence variability between ursids of different chronology
The samples studied range in age from a present-day brown bear, several Holocene brown bears to one brown bear more than 40,000 years BP, and several Pleistocene cave bears of different ages. It might be expected that the collagen from the older bears would be more degraded and have fewer marker peptides due to fragmentation of the molecule during diagenesis, but this is not a particularly visible effect (see Tables 4 and 5). Similarly, there is no pattern in the presence or modification of peptides according to their geographical origin. We might also expect more instances of Q deamidation in the older samples, as this is a known diagenetic process that was proposed as an indicator of the age of skeletal remains (Wilson et al. Reference Wilson, van Doorn and Collins2012). In peptide G or COL1α2 767–799 this type of degradation is visible in a +1 Da mass shift that seems to affect cave bear samples more than brown bear samples, although with little difference. Certainly, the extant brown bear sample did not show this increase in the mass of peak 2957. But in Eirós, whose two cave bears are separated by 20,000 years, the sample with deamidation is the most modern.
This is consistent with the observation that the extent of deamidation seems to be influenced more by burial conditions than chronological age (Van Doorn et al. Reference Van Doorn, Wilson, Hollund, Soressi and Collins2012; Schroeter & Cleland Reference Schroeter and Cleland2015; Welker et al. Reference Welker, Soressi, Roussel, van Riemsdijk, Hublin and Collins2017). We did not really expect much degradation in the extracted collagen, as all fossil bones come from cave deposits. Caves maintain fairly stable conditions of humidity and temperature, so the preservation of organic molecules can be sustained over time (Pinto-Llona et al. Reference Pinto-Llona, Andrews and Etxeberría2005; Torres et al. Reference Torres, Ortiz, Fernández, Arroyo-Pardo, Grün and Pérez-González2014; González-Fortes, et al. Reference González-Fortes, García-Vázquez, Pinto Llona and Grandal-d'Anglade2017).
5. Conclusions
The sequence of amino acids that make up the bone collagen molecule shows slight variations between taxa, that can be studied by peptide mass fingerprinting for taxonomic purposes. This requires reference databases that allow peptide identification. In this work we specifically review the identification of peptide markers in cave bears and brown bears, common components of the European Pleistocene fauna. We provide the peptide spectra of 20 cave bear and 10 brown bear samples, that have never been published before. We found no differences between the different cave bear species and brown bear spectra, at least not for the peptide markers described in the literature. We also found no evident correlation between the age of the samples and the post-depositional alterations of their collagen.
However, the in-silico study of the ursid collagen sequences published in UniProtKB revealed discrepancies in m/z values of some peptides, keeping in mind that the sequences are automatically generated and must be annotated. This means that they may contain errors. In particular, the peptides COLα2 292–309 (P2) and COLα2 454–483 (E) have different m/z values than those suggested for ursids in publications on the subject. These markers are also not found in many published fossil Ursidae spectra. Our analysis of the peptide spectra of 30 samples morphologically identified as ursids revealed a higher affinity for these markers to the values obtained from the in-silico analysis (in 27 of the 30 specimens studied, for both P2 and E peptide markers), suggesting that these two peptide markers are not well defined for ursids.
A third tryptic peptide raises problems, that are methodological in nature in this case. The peptide COL1α1 586–618 (F) is actually composed of two tryptic peptides that may or may not be cleaved, apparently depending either on the type of trypsin used or the digestion time, among other possible factors. In the samples analysed here, this peptide is barely detected; however, one of the peptides that we identified as one of its components appears in more than half of the samples. These types of limitations are common in a technique as young as ZooMS. It would be necessary to establish a standardised analysis protocol to avoid such discrepancies.
6. Supplementary material
Supplementary material is available online at https://doi.org/10.1017/S1755691023000038.
7. Acknowledgements
The authors are grateful to Dr G. Rabeder, Dr I. Martini and Dr G. Baryshnikov for providing cave bear samples from Austria, Italy and Russia for analysis, and to Dr JF. García Marín for the modern brown bear sample. ACPLL, JMG, TT, AGV and AGD obtained the samples from the excavation of the sites and made the morphological identification of the skeletal remains and its initial dating. AGV and AGD performed the collagen analysis and interpretation of the results. AGV and AGD wrote the text, with the participation of all authors in the final revision. The authors are grateful for the valuable comments of two anonymous reviewers. Funding for open access charge: Universidade da Coruña/CISUG.
8. Financial support
This study was carried out with the financial support of the project ED431B 2021/17 of the Autonomous Government of Galicia (Spain) awarded to AGD.
9. Competing interest
None.
Open access
