The first record of fossilized soft parts in ossified tendons and implications for the understanding of tendon mineralization

Abstract

Preservation of soft parts (collagen fibres, blood vessels and cells) in extinct vertebrates is rare and usually limited to fossilized bone and cartilage. Well-preserved coarse collagenous fibre bundles embedded in a mineralized matrix of tendons, as well as numerous hollow, tubular structures consistent morphologically with fibril bundles, blood vessels and associated cells, were identified in ossified tendons of Late Cretaceous ornithischians from North America and Central East Asia. Detailed, high-accuracy imaging, along with spectroscopic characterization of those fibrous structures and comparison with ossified tendons of modern-day turkeys, support the proposition that physiologically driven tendon ossification is common for avians and non-avian dinosaurs. The examined soft parts were preserved through the pathway of iron-induced crosslinking and alumino-silification, documenting a variety of pathways for the preservation of soft parts, depending on the burial environment. For the first time, the structure of dinosaur fossilized tendons is analysed in detail, revealing shared histogenetic principles with modern birds and the nature of preservation.

INTRODUCTION

Tendon mineralization has been recognized in all groups of non-avian dinosaurs and in birds (Supporting Information, Table S1). A number of clade-specific adaptive explanations for their development has been provided over time (Moodie, 1928; Brown & Russell, 2012; Klein et al., 2012; Hall, 2015). It is possible that, at least initially, the mineralization took as its basis an ancestral physiological propensity to mineralize the tendons, or that the process was a side-effect (spandrel) correlated with the physiology of mineral use or to selection for skeletal co-ossification (Organ & Adams, 2005). Intratendinous mineralization has also been reported in limbs, tails and possibly even the trunk of some Pterosauria, the sister-group of the Dinosauria (Owen, 1860; Williston, 1903; Bennett, 2001, 2003; Frey et al., 2006), so the potential for development of mineralized tendons appears to be widespread in Ornithodira. However, the ability to form bony sesamoids (intratendinuously developing skeletal structures) seems common to all vertebrates (Hall, 2015). Curiously, tendon mineralization shows a wide variety of clade-specific patterns among the Dinosauria (see Supporting Information, Data S1; Table S1). Dinosaur mineralized tendons have been the subject of numerous studies concerning their histology and microstructure (Adams & Organ, 2005; Lonardelli et al., 2005; Cerda, 2009; Zhou et al., 2010; Gallina, 2011; Klein et al., 2012; Preuschoft & Klein, 2013; Cerda et al., 2015, 2019; Horner et al., 2016; see also references in the Supporting Information, Data S1), but recently Horner et al. (2016) have questioned their osseous character and have noted a number of microstructural characters inconsistent with typical bone tissue, refuting the classical interpretation of mineralized metaplastic ossifications as true bones.

The reports about the presence of fossilized soft parts from mineralized tissues have so far been restricted to fossil bones (Schweitzer et al., 2005) and cartilage (Zheng et al., 2021), but thus far never documented in ossified tendons. While the potential for soft part preservation in mineralized tendons could have been expected based on their classically accepted identity as bone, recent questions about the true nature of their tissue (Horner et al., 2016) possibly impacting their vulnerability to taphonomic factors, as well as their small diameters, potentially increasing their susceptibility to permeation and diagenetic modification, remains to be tested. Our detailed studies of the tendons of two ornithischian dinosaurs from Central Mongolia (Homalocephale calathocercos Maryanska & Osmolska, 1974, Pinacosaurus grangeri Gilmore, 1933) and one from Alaska, USA (Edmontosaurus regalis Lambe, 1917) reveal for the first time the presence of fibrous structures. The aim of this study is to evaluate whether mineralized fibrous connective tissues exhibit a potential for preservation and, if so, whether they may have implications for palaeobiology or understanding of the biomineralogy of the skeletal connective tissue in avians (birds) and non-avian dinosaurs. Scanning electron microscopy, coupled with energy dispersive spectroscopy (SEM-EDS), were used for examination of tendons of these three dinosaurs. Atomic force microscopy (AFM) and a 3D microscope profiling system were applied to study the micro- and nanostructures of fossil tendons of H. calathocercos. SEM-EDS examination of ossified tendons of a recent turkey (Meleagris gallopavo Linnaeus, 1758) is provided for comparative purposes. Furthermore, selective demineralization was performed to remove the phosphate (apatite mineral) matrix and to reveal fibrous, vascular and cellular structures. Moreover, spectroscopic techniques were applied to determine their state of preservation and chemical composition. Our study is concerned with preservation of longitudinally oriented coarse collagenous fibre bundles in vascular canals, fossilized blood vessels and cells deeply embedded in the mineral matrix of ornithischian tendons. Finally, by documenting the ossified tendon micro- and nanostructures, we attempt to address the question of their homology with bone tissue [compare Horner et al. (2016)]. The presence of a functional lacunar–canalicular cellular network in the ossified tendons documents some physiological aspects of the mineralization mechanisms in dinosaur tendons, which share their histogenetic principles with those of modern birds, bringing us closer to the biology of these extinct animals. These observations indicate the burial environments in which the soft parts of tendons have the opportunity to be preserved.

Institutional abbreviations

• GIUS, Institute of Earth Sciences, Faculty of Natural Sciences, University of Silesia in Katowice, Sosnowiec, Poland

• MPC, Mongolian Paleontological Center, Ulaanbaatar, Mongolia

• UAMES, The University of Alaska Museum Earth Sciences Collection, Fairbanks, Alaska, USA

• ZPAL, Institute of Paleobiology, Polish Academy of Sciences, Warsaw, Poland

MATERIAL AND METHODS

The analysed material includes tendons of ornithischian dinosaurs: the ornithopod Edmontosaurus regalis (‘Ugrunaaluk kuukpikensis’ of Mori et al., 2016), marginocephalian Homalocephale calathocercos and thyreophoran Pinacosaurus grangeri. The control sample of a recent turkey, Meleagris gallopavo (Saurischia: Theropoda: Aves: Galliformes) was also used. The sample of the North American ‘Ugrunaaluk kuukpikensis’ consisted of three specimens UAMES 52615. The isolated tendons from an unknown location in the body were found at the Colville River, Alaska, where the lower Maastrichtian Price Creek Formation is exposed; the type horizon and locality for ‘Ugrunaaluk kuukpikensis’ was described by Mori et al.(2016). Note that Xing et al. (2017), Wosik et al. (2019) and Takasaki et al. (2019, 2020) consider this name a nomen dubium and the Alaskan specimens as belonging to Edmontosaurus regalis, and this recognition is used in our study. Two caudal tendons of the juvenile holotype specimen of Homalocephale calathocercos MPC-D 100/1201 came from Western Sayr, Nemegt, the Nemegt Formation (Maryanska et al., 1974; Evans et al., 2011). Initially, this specimen was a part of the ZPAL collection under the number ZPAL MgD-I/37 (Maryanska et al., 1974), but at the moment of description it was already assigned the number of the MPC-D collection: formerly GI SPS 100/51, recently changed into MPC-D 100/1201 (Maryanska et al., 1974; Evans et al., 2011). Two caudal tendons were sectioned longitudinally and transversally. Finally, the study includes two cross-sectioned samples of the caudal tendons of the representative of the Ankylosauridae Pinacosaurus grangeri ZPAL MgD-II/32 found in the Djadokhta Formation, Bayn Dzak locality (Maryańska, 1977; Hill et al., 2003). The body size of the individual is unknown, due to the incompleteness of the preserved caudal part of the specimen. For comparative purposes, the recent galliform ossified tendons from the lower leg of a turkey (Meleagris gallopavo, GIUS-12-3741) was used.

The extended results are also presented in the Supporting Information, Data S2–S5. The methods described below were applied to all the listed specimens in Table 1.

Preparation of thin sections

The fossil tendons were thin-sectioned using standard methods for microstructural studies (Padian & Lamm, 2013). The samples of the mineralized tendons were embedded in Araldite 2020 epoxy resin (Huntsman Advanced Materials, USA).

Microscopy

Thin sections were polished with silicon carbide and aluminium oxide papers, and finally cleaned using diamond paste with no cover slips. The photographs of the thin sections were obtained using an Olympus BX-51 microscope under transmitted light with Olympus microscope camera software in the Institute of Earth Sciences, University of Silesia in Katowice (Poland) and a Nikon Eclipse LV100 POL polarizing microscope with DS-Fil camera and NIS-Elements v.4.20.01 64-bit microscope imaging software at the Faculty of Geology, University of Warsaw, Poland.

The tendon surface micromorphology and ultrastructure were examined using two scanning electron microscopes (SEM): Phenom XL SEM, PhenomWorld (ThermoFisher Scientific, Eindhoven, Netherlands) equipped with an energy-dispersive X-ray spectroscope (EDS), and FEI Quanta 250 field emission scanning electron microscope equipped with EDS, as well as a secondary electron detector (SED) and circular backscatter detector (CBD), both installed in the Institute of Earth Sciences, the University of Silesia in Katowice. Measurements were made on non-coated samples under low vacuum conditions with an accelerating voltage of 15 kV. From each sample presented in this study, multiple EDS spectra were collected. The composite images were generated using the Phenom SEM ProSuite software, then stitched and processed using the Image Composite Editor v.2.0.3.0, 64-bit (Microsoft Research).

The 3D images and depth profiling data were captured using a Keyence VHX-7000 Series high-accuracy digital microscope during the presentation of the instrument by a Keyence Corporation sales representative in the Silesian Centre for Education and Interdisciplinary Research (see Acknowledgements).

Atomic force microscopy (AFM) was applied to characterize the nanostructure of the fibrous structures observed in the vascular canals of the tendons. The AFM surface maps and measurements performed on surfaces broken along the transverse-plane of the samples, and extract obtained in demineralization procedure (see Introduction), were collected using a JPK NanoWizard3 AFM equipped with BudgetSensors AFM tips (300 kHz resonant frequency, 40 N/m force constant, 125 µm length). The instrument is installed in the Laboratory of Scanning Probe Microscopy, A. Chełkowski Institute of Physics, Faculty of Science and Technology, University of Silesia in Katowice. The data were collected using JPK NanoWizard Control Station v.6.1.163 and processed in GWYDDION v.2.55, 64-bit (Nečas & Klapetek, 2012). The surface topography was measured using the standard intermittent contact mode.

Preparation of recent Meleagrisgallopavo tendon sample

A turkey lower leg originating from a food-breeding poultry farm was bought in a local butcher’s shop. Ossified tendons were separated from the surrounding muscles and ligaments using a sterile surgical lancet and scissors. The ossified tendon (tibialis cranialis) was isolated from surrounding muscles and fascia, rinsed in distilled water and desiccated. One part of the specimen, intended for SEM-EDS examination, was dried at room temperature and cut. Another part, intended for the FTIR study, was deeply frozen in liquid nitrogen, temperature about –195.8 °C and triturated in a mortar to an analytical fraction.

Geochemistry

Total carbon (TC), total inorganic carbon (TIC) and total sulphur (TS) were measured with an Eltra CS-500 IR-analyser equipped with TIC module, installed in the Laboratory of Organic Geochemistry, Faculty of Earth Sciences, University of Silesia, Poland (see Acknowledgements). Total organic carbon (TOC) was calculated as the difference between TC and TIC (for more details, see: Goryl et al., 2018).

X-ray powder diffraction (XRPD)

Powdered samples of fossilized tendons, a surrounding rock matrix and recent turkey ossified tendons were analysed by X-ray powder diffraction (XRPD) to investigate the mineral composition of the samples (compare Table 1). The PANalytical X’Pert PROMPD PW 3040/60 diffractometer at the Laboratory of X-ray Diffraction, Faculty of Earth Science, University of Silesia was applied in this study. Quantitative phase content was calculated using the RietveldModule in HighScore Plus software with the ICDD PDF-4 + pattern database.

Demineralization of fossil samples

Fragments of fossilized tendons were incubated in 0.5 M EDTA solution buffered to pH 8.0 with daily changes, as described in detail by Surmik et al. (2016). Samples were mounted on top of a vacuum filtration kit (Sartorius AG, Germany). First, the EDTA solution was poured on top of the column and, after reaction, it was drained by a manual vacuum pump to the bottom of the column. Subsequently, deionized water was loaded on top of the filtration column to rinse the sample and then drained down. The operation was repeated multiple times until the phosphate and EDTA salts were completely drained off. The obtained extract consisted of a reddish brown residuum and was a subject of further analysis.

Spectroscopy and mass spectrometry

The structural properties of the samples were determined using an Agilent Cary 640 Fourier transform infrared (FTIR) spectrometer equipped with a standard source and DTGS Peltier-cooled detector. The spectra were collected using a GladiATR diamond accessory in the 400–4000 cm^–1 range. All spectra were recorded by accumulation of 16 scans with a spectral resolution of 4 cm^–1. Post-processing analysis, including baseline correction, carbon dioxide and water vapour removal, was done using the GRAMS v.9.2 software package. Finally, a band fitting analysis using a Lorentz–Gauss function with the minimum number of components was performed using GRAMS. The samples (both fossils and recent) were analysed as powdered complete tendon fragments (bulk analysis), as well as an extract obtained by removal of the apatite matrix (demineralization).

The mass spectra were collected using a Time-of-Flight secondary ion mass spectrometry (ToF-SIMS) 5 spectrometer (IONTOF GmbH, Münster, Germany). Data were collected from the surfaces of two prepared specimens after cleaning with a bismuth ion beam (DC mode, the estimated thickness of material removed was within the order of 0.5 µm) in the ultra-high vacuum conditions (~2-5 × 10^– 9 mbar). Due to the three-dimensional structure of the characterized specimens, the analysis areas were selected individually for each sample (indicated in the Supporting Information, Data S4). High-resolution mass spectra were obtained using the Bi₃⁺primary beam (0.3 pA, 30 kV). The measurements were performed in positive polarity. The use of the flood gun was necessary to eliminate the effect of surface charging. The SurfaceLab6 software (IONTOF GmbH, Münster, Germany) was used for data analysis. The mass spectra were calibrated using the H⁺, CH⁺, CH₃⁺, C₂H₂⁺, C₂H₃⁺, C₃H₃⁺, C₃H₅⁺, C₄H₄⁺, C₅H₅⁺, C₆H₆⁺peaks.

RESULTS

Histological sampling

Edmontosaurus regalis (UAMES 52615)

Both cross-sectioned tendons show a highly porous structure built of fibrous primary matrix (Fig. 1A–C). The general structure of the sectioned tendons is homogenous. The matrix is composed of coarse collagenous fibre bundles containing numerous bone cell lacunae (Fig. 1A, C). The vascular canals forming the porous structure of the tendons are young Haversian canals (Fig. 1A–C). The cross-sectioned larger tendon shows few scattered secondary osteons in the centre of the tendon surrounded by collagen-fibre bundles (Fig. 1D). The tendon of a slightly smaller cross-section (3.5 mm diameter; Fig. 1A, B) is more porous than the larger one (6 mm; Fig. 1C, D). The different structure between the sampled tendons may result from a different location in the body.

Homalocephale calathocercos (MPC-D 100/1201)

The cross-section of the tendon shows dispersed primary osteons embedded in mineralized fibrous primary matrix (Fig. 2A–F). In some cases, the structural remnant of a primary osteon encapsulates a smaller concentric ring of lamellar bone apparently separated with a cement line, appearing to be a small secondary osteon deposited in the centre of a primary osteon (Fig. 2D). These structures probably represent a first generation of secondary osteons (a short cycle of resorption of primary tissue and subsequent deposition of secondary tissue along blood vessels). The osteons are bigger and denser in the innermost part of the tendon, and become less abundant and smaller closer to the periphery (Fig. 2A, B, E). Under normal light, in both the longitudinal and cross-section, visible bone cell lacunae are numerous in the centre of the tendon, but rare to absent close to the external margin of the tendon (Fig. 2A, E). Under polarized light, the longitudinal section shows bundles of fibres running in a herringbone-like pattern along the tendon (Fig. 2F), in the cross-section the fibres form a lattice-like pattern (Fig. 2B, C). Thus, the osteons are surrounded by coarse collagenous fibre bundles. The width of the longitudinally sectioned tendon varies from 1.1 to 1.4 mm, its total length is 7 mm.

Pinacosaurus grangeri (ZPAL MgD-II/32)

The caudal tendons reveal extensive secondary remodelling, with at least three generations of secondary osteons (Fig. 3A–E). The cross-section of the flattened, elliptical tendon shows a densely secondary remodelled inner part (Fig. 3B, C) in the thickest place and more loosely arranged secondary osteons in the thinner parts of the cross-section and the periphery of the tendon (Fig. 3D). The latter is mainly composed of bundles of mineralized fibres and few vascular canals. Although, the secondary osteons mainly overlap each other in some places, between them patches of mineralized fibres are present (interfascicular spaces; Fig. 3E, F). The cross-section of the second sample from the same individual shows three circular tendons. The two bigger tendons (2.1 and 1.7 mm in diameter) are extensively secondarily remodelled (Fig. 3A) and almost nonvascular mineralized fibres are present on their periphery. In contrast, the smallest tendon (0.6 mm in diameter) is exclusively composed of mineralized fibres, weakly vascularized. The difference in the structure between the smallest and two larger tendons result from different places of sectioning: the smallest is from the terminal part of the tendon, and the bigger ones closer to the middle part.

Nanostructure of the tendons

Detailed scanning electron microscopy (SEM) observation of vascular canals of fossilized tendons revealed the presence of mostly randomly oriented fibrous matrix in all three studied ornithischian samples (Fig. 4A–G). Comparative SEM images of modern-day turkey tibialis cranialis tendons (Fig. 4I–H) reveal similar fibrous structures. The observed structures exhibit a hierarchic pattern – individual fibres are gathered in bundles (or fascicles). The diameters of the bundles range from 0.7 to 2 µm (Supporting Information, Fig. S1) and the bundles are located in the walls of vascular canals of diameters varying between a dozen and 50 µm. The fibrous structures are visible only in the vascular walls exposed on the surfaces of freshly broken tendons.

The EDS spectra collected from the fibrous structures and surrounding mineral matrix of all studied tendons reveal oxygen, calcium and phosphorus as the main components in all studied taxa (Fig. 4K, L), which is a typical elemental signature of modern and fossil bones. To visualize vascular canal shape and geometry, topographic microscope imaging (Fig. 5A, B) of a Homalocephale calathocercos tendon sample was applied. The results indicate that the observed fibrous structures are located deep inside the vascular canals exclusively and incorporated in their walls along their length. In addition to the surface morphology imaging, information of vascular canal topography in the tendon was collected using an atomic force microscope (AFM) in contact mode. The generated images (lock-in amplitude and lock-in-phase, Fig. 5B–G) show the nanostructure of the fibres. Furthermore, AFM imaging on an individual fibre (Fig. 5G, H) reveals its striped pattern and the performed measurements allow estimation of a periodicity measuring about 24 nm (Fig. 5I, J; Supporting Information, Data S2).

Preservation of soft parts

All fossil tendon samples were demineralized (see Material and methods) to remove the apatite matrix and to assess whether any soft parts were preserved. No fossilized soft parts were obtained from the extract of Pinacosaurus grangeri. The demineralization of the tendons of Homalocephale calathocercos (Fig. 6A) revealed an extract consisting of individual brown- to orange-coloured tubular structures, as well as a dense network of vascular canals still attached to a translucent layer loosening from the tendon (Fig. 6C, D; Supporting Information, Fig. S2). Furthermore, the Edmontosaurus regalis sample revealed the presence of numerous rust- to brown, dark to translucent, elongated, bifurcated and H-shaped structures (Fig. 6B). Additionally, over a dozen tubular, branched structures (Fig. 6C, D) seems to be morphologically consistent with blood vessels or fibril bundle layers lining the vascular canals observed with SEM. Among them, ovate to tent-shaped cell-like structures attached to the surfaces of vessels or bundle layers were found (Fig. 6F–K). Other cell-like structures, more spindle-shaped, and projecting branching structures (Fig. 6L–N, black asterisks) were usually observed in isolation (Fig. 6L, N), sometimes attached to the fibrous matrix. The EDS spectra collected from the demineralized parts of the tendons show that their elemental composition is different from the mineral matrix of the tendons. In the case of the external sheath of a partially demineralized tendon of H. calathocercos, which detached from the tendon, a dense network of vessel-like tubes was observed (Fig. 6C, D; compare also: Supporting Information, Fig. S2A, B). In this sample, oxygen, silicon and aluminium dominate (Fig. 4E), while remains of phosphates are still present (Supporting Information, Fig. S2C). The fossilized vascular-like tubular structures with attached cell-like structures (Fig. 6F–K), as well as individual cells (Fig. 6L–N) extracted from the samples taken from E. regalis and H. calathocercos, are composed mainly of iron and sulphur with smaller addition of oxygen and silicon (Fig. 6P).

Organic matter residues

Total carbon (TC), total organic carbon (TOC) and total inorganic carbon (TIC), as well as total sulphur (TS), contents were measured (see Material and methods) to document the preservation of organic matter in all studied fossil samples. Obtained results demonstrate TOC values equal to 1.28 wt% for Homalocephale calathocercos, 2.57 wt% for Pinacosaurus grangeri, and 2.81 wt% for Edmontosaurus regalis tendon samples. An elevated level of total sulphur (4.54 wt%) was also reported for the latter. Furthermore, the control sample of sediment associated with E. regalis tendons reveals the highest level of TOC, up to 6 wt% with sulphur content below 0.5 wt%.

Spectroscopic studies and mass spectrometry

FTIR spectroscopic studies, intended mainly to identify organic residues in the fossil samples, were performed only on the selected tendon samples that were never glued or treated with any organic-based materials (e.g. consolidants). Therefore, the Pinacosaurus grangeri tendon sample was not included (Supporting Information, Data S3). A modern-day turkey (Meleagris gallopavo, GIUS-12-3741) was a control sample. FTIR studies of H. calathocercos, E. regalis and M. gallopavo show a band arrangement pointing to co-association of various minerals, including phosphate (PO₄)^3–, carbonate (CO₃)^2– and iron oxide (FeO₄)^5–, as well as silicon dioxide (SiO₂) moieties, suggesting the presence of carbonate apatite, goethite and silicates (Fig. 7). The results are thus consistent with the EDS spectra (Fig. 4H, L, P) for the same samples. The FTIR measurements of demineralized parts of tendon samples from H. calathocercos and E. regalis revealed also signals associated with organic residues (Fig. 5B, D), especially well visible due to several bands in the 1550–1800 cm^–1 corresponding to the amide I (Fig. 7F, G). A percentage proportion of each component in the amide I band was computed as a ratio of a fractional area of the suitable peak after the band fitting and the sum of the areas of the peaks belonging to the amide I band (Byler & Susi, 1986; Jackson & Mantsch, 1995; Litvinov et al., 2012). As a result, two intensive bands were observed at 1633 and 1668 cm^–1 with 64 and 36% proportion between β-sheet structures and turns (Litvinov et al., 2012) for H. calathocercos, as well as three bands at 1630, 1663 and 1690 cm^–1 with 56, 25 and 19% for E. regalis. Two other marginal bands (at 1594 and 1714 cm^–1) were also detected in both samples and correspond to the lipid signal, including fatty acids. The comparison of the fossilized tendons with the turkey tendon (Fig. 7E, H) strongly support the possibility of organic (proteinaceous) preservation in the fossil samples, especially based on the comparison of the amide I fingerprint region (Fig. 7F–H).

Mass spectra (Supporting Information, Data S4) were collected both from the demineralized sample of H. calathocercos and E. regalis tendons across the regions of interest with a mass-to-charge ratio (m/z) range of 1–150 Da, the fingerprint region for lower mass amino acid anions (Surmik et al., 2017). There, various non-organic (Si-containing and Fe-containing), as well as organic (N-containing), species were identified. The detected ions CH₄N⁺ (at m/z 30.03 Da), C₂H₃N⁺ (at m/z 41.03 Da), C₂H₆N⁺ (at m/z 44.05 Da), C₃H₃N⁺ (m/z 53.02 Da), C₃H₆N⁺ (at m/z 56.04 Da), C₄H₆N⁺ (at m/z 68.05 Da), C₄H₈N⁺ (at m/z 70.07 Da), C₄H₁₀N⁺ (at m/z 72.08 Da) and C₅H₁₀N⁺ (at m/z 84.08 Da) are related to amino-acids (Surmik et al., 2017) (see also Supporting Information, Data S4). The presence of the Si- and Fe-containing species correlates with the results of the EDS surveys (Fig. 6E, P) and confirms the mineralogical composition of the studied samples.

DISCUSSION

Samples and geological background

The remains of the sampled ornithischians were found in formations indicating different burial environments. Homalocephale calathocercos (MPC-D 100/1201) was described from the Nemegt Formation, consisting of alluvial plain, paludal, lacustrine and fluvial deposits (Jerzykiewicz, 2000). The fossils of E. regalis (UAMES 52615) came from the alluvial, coastal plain environment of the Prince Creek Formation, which also includes plant material suggesting a polar woodland environment (Flaig et al., 2013) while deposited in organic-rich volcanoclastic sediments (Gangloff & Fiorillo, 2010). Finally, the remains of P. grangeri (ZPAL MgD-II/32) were found in the Djadochta Formation, which is interpreted as deposited in a semi-arid climate with sand dunes and heavy rainfall events (Jerzykiewicz, 2000; Dingus et al., 2008). The examined tendons in each case accompanied findings of skeletons.

Fibrous structures

The sampled tendons reveal different stages of secondary remodelling, as shown by histological sections. The lack of secondary remodelling and porous structure, entirely built of coarse collagenous fibre bundles in the studied Edmontosaurus regalis tendons (Fig. 1), indicates their ontogenetically early ossification stage (Adams & Organ, 2005). Together with the small size of the sampled tendons, this suggests that they belonged to a relatively young individual. Dispersed primary and secondary osteons within the whole section of the Homalocephale calathocercos (Fig. 2) tendon differ from the tendon microstructure organization reported from other pachycephalosaurids (Organ & Adams, 2005) and juvenile ornithischian dinosaurs (Adams & Organ, 2005). The sections from the middle part of the tendon show denser remodelling, and close to the terminal parts they are filled with more fibres and the secondary osteons are less extensive (Adams & Organ, 2005). However, the presence of only the first generation of secondary osteons (Fig. 2D) harmonizes with the earlier identification of the individual as a juvenile, based on its skeleton (Evans et al., 2011). Pinacosaurus grangeri (Fig. 3) shows extensive secondary remodelling within the tendon, indicating an advanced ossification stage similar to other adult ornithischian dinosaurs (Adams & Organ, 2005). Thus, the sampled tendon plausibly belonged to a non-juvenile individual.

Fibrous structures were recognized in all analysed tendon samples using SEM. The highest abundance of fibres (observed to occupy most of the vascular canals) was found in the comparatively least secondarily remodelled tendons of Edmontosaurus regalis (compare also Supporting Information, Fig. S1E–I). In Homalocephale calathocercos, the occurrence of fibres in the vascular canals is less frequent (fibres are observed in about half of all examined vascular canals, the other half showing smoother internal surfaces not different from the surrounding bone tissue; compare also Supporting Information, Fig. S1A–D). In the strongly secondarily remodelled tendons of Pinacosaurus grangeri, only two vascular canals exhibit fibrous structures in their spaces. Therefore, our observations suggest that the exposed fibres are abundant in the tendons that are not secondarily remodelled, i.e. at early stages of histogenesis. With an increasing level of secondary remodelling, the exposed fibres are less common, but still present. Several lines of evidence support an endogenous character of the fibres, indicating that they are not a contamination from environment: (1) they are visible (embedded) in vascular canals deep in the mineralized matrix of tendon (confirmed by SEM-EDS observations) in all three sampled specimens; (2) the fibrous structures similar to those observed in vascular canals are also visible on the surfaces of tubular structures comparable with fossilized blood vessels released from the mineral matrix of the tendons by phosphate removal; (3) these fibres occur in the mineralized tendons of different ornithischian taxa from different sedimentary environments; (4) SEM images of non-demineralized tendons indicate that the fibres arise in continuity with the mineral phase of the bone, without any separation; and (5) similar structures were also observed in the ossified tendon of an extant turkey.

The main fibrous protein contributing to the connective tissues of the skin, tendons and bones is collagen (Sagi & Afratis, 2019). The nature and organization of the fibres differ between osteoderms, endoskeletal and intramembranous bone and ossified tendons. Both osteoderms and ossified tendons mineralize through metaplasia (although osteodermal histogenesis may involve, or mix in, other processes, mostly intramembranous ossification – see, e.g. Scheyer, 2007; Scheyer & Sander, 2009; Vickaryous & Sire, 2009; Buffrénil et al., 2010; Scheyer & Desojo, 2011; Scheyer et al., 2014; Vickaryous et al., 2015), but their structure differs between each other. The ossified tendons maintain their fibrillary organization, consisting of coarse collagenous fibre bundles oriented parallel to the main axis of the element. Typical dinosaur osteoderms, on the other hand, are built of packed bundles of mineralized collagen fibres of dermal origin, i.e. structural fibres. Endoskeletal and intramembranous bone has non-metaplastic origins and shows no structural fibres (e.g. Cerda et al., 2019). Collagen fibres are bundles of densely-packed fibrils in which triple-helical collagen molecules are mutually interconnected in a head-to-tail arrangement (Marino et al., 2017). The striated pattern of the collagen fibres is a result of their molecular structure, which incorporates a repeating Gly-X-Y amino acid sequence (Sagi & Afratis, 2019) to form the triple-helical structure of tropocollagen. The periodicity of the striations, also called the D-spacing, D-periodicity or banding, typically equals ~67 nm (Wallace et al., 2011; Fang et al., 2012) and is considered a diagnostic ultrastructure feature for quaternary structure of collagen I (Fang et al., 2012; Bertazzo et al., 2015). The preservation of collagen in fossil bones was believed to be rare (Schweitzer et al., 2005, 2008), but recent data have provided plentiful cases of such preservation (Lindgren et al., 2011; Bertazzo et al., 2015; Surmik et al., 2016; Lee et al. 2017). Some of the examined fibres resemble bundles (e.g. Fig. 4A, D–F; Supporting Information, Fig. S1) comparable to those described by Bertazzo et al. (2015: fig. 1c, d) and Lindgren et al. (2011: fig. 1h, i, m). However, one incompatible feature is observed. The detailed SEM observations and AFM measurements of H. calathocercos revealed a striped pattern perpendicular to the axis of the fibrous structures (Fig. 5C–I), but the pattern displays periodicity not consistent with that of native collagens (Vesentini et al., 2013; Orgel et al., 2014; Bertazzo et al., 2015). The measurement of the striations gave varied results, ranging from 16.9 to 36.5 nm, with an average of ~24 nm (Supporting Information, Data S2). The D-spacing distributions in various connective tissues may be altered by many factors, including changes at the bundle or at the individual fibril levels (Fang et al., 2012), drying (Wess & Orgel, 2000) or diseases (osteopenia, osteogenesis imperfecta) (Wallace et al., 2011; Fang et al., 2012), resulting in shorter values of the D-spacing; usually it is no less than 60 nm (Wess & Orgel, 2000; Fang et al., 2012), with one exception (~24 nm) documented from collagen I from dehydrated (air-drying) human cornea (Jastrzebska et al., 2017). Therefore, it seems likely that the observed periodicity is not the original pattern of collagen fibres, but an artefact caused by the alteration of the mineralized tissue during fossilization. Many extrinsic factors (like elevated temperature) may erase an original striped pattern (Sasaki et al., 2019). The shapes, sizes and location of the studied fibres in the mineralized tissue of the tendon are consistent with typical collagen-like pattern, supporting such recognition of the fibrous structures of collagen origin.

Are mineralized tendons bone?

The character of the mineralized tendons and their tissue is ambiguous. Although they have been commonly treated as a product of metaplastic ossification resulting in bone formation, Horner et al. (2016) questioned the homology of the two.

First, Horner et al. (2016) noted a scalloped nature of the structures typically called secondary osteons in mineralized tendons and instead they introduced the term ‘secondary reconstructions’. Interestingly, there is no indication of scalloped edges in any of our samples of Pinacosaurus grangeri, either in SEM or transmitted normal or polarized light, even when numerous generations of remodelling are involved. Although we notice that the edges of those structures can be interrupted by fibre bundles, resulting in minor unevenness, in most cases they do not differ significantly from secondary osteons observed in endochondral and intramembranous bones. Moreover, undulations along the margin of secondary osteons (Howship’s lacunae) can also be present in normal endochondral bone, and are caused by osteoclast activity in bone tissue undergoing reabsorption (Skedros et al., 2005). Thus, this cannot be used as a feature distinguishing between normal bone and metaplastic tissue mineralization. The well-marked cementing lines (Fig. 3E, F) indicate osteoclastic activity and support the interpretation that these structures are true secondary osteons [contrary to Horner et al. (2016)].

Furthermore, Horner et al. (2016) noted the presence of bone cell lacunae with canaliculi only in secondary mineralized tendon tissue and their absence in primary tissue. Therefore, they proposed that the spaces typically interpreted as bone cell lacunae are, in fact, empty spaces between fibre fascicles or tenocyte lacunae. Based on SEM images of our samples, we confirm that most of these spaces in primary mineralized tendon tissue lack canaliculi. However, they are mostly short, non-continuous and spindle-shaped, indicating that most of them are cell lacunae rather than interfascicular spaces. Moreover, we were able to extract cells from demineralized samples and, while it appears that many of those indeed represent tenocytes (tendon-specific fibroblasts comprising approximately 95% of tendon tissue; compare: Kannus, 2000), even in the non-remodelled Edmontosaurus regalis sample, we also found cells with incipient or even exuberant and highly branched projections identifiable as filopodia, which are best interpreted as osteocytes. Such a transforming character of cells at early (non-remodelled) stages of histogenesis is expected from metaplastic ossification, suggesting a homology between the mineralized tendon tissues and bone. While some interfascicular spaces are visible in our samples, they are much less abundant compared to what was proposed by Horner et al. (2016) and what we observe in Meleagris gallopavo. It is possible that this difference results from the stage of histogenesis, systematic position of the studied animals or even diagenetic modification of the mineral phase of the specimens.

Finally, Horner et al. (2016) noted the presence of numerous mineralized intratendinous collagen fibres both in primary and secondary mineralized tendon tissue. While the mineralized fibre bundles in primary tissue are unsurprising, because they are expected to be enclosed in the mineral matrix during the process of metaplastic ossification of already developed tissue, their observation in secondary tissue (i.e. deposited after the resorption of primary tissue and, supposedly, its fibres) would be more puzzling. As noted by Horner et al. (2016), this would indicate that the primary matrix of mineralized tendons is resorbed by some unknown cell type other than typical osteoclasts or, alternatively, the fibres of the primary tissue are either not resorbed alongside the matrix or new fibres are produced during the deposition of secondary matrix. Clear fibrous structures free from the matrix were observed by us only in the vascular canals in fossilized tendons, but consistently with previous studies (Horner et al. 2016), in the fresh mineralized tendon (modern-day turkey) they are also visible under the SEM in the mineral matrix (Fig. 5H, I, compare also Fig. 5L). Intratendinous fibres are clearly discernible embedded in the matrix in the thin sections of fossil specimens in transmitted light, but lack of their discernible characteristics under the SEM probably results from the hypermineralization of the tendons, either primary or due to diagenetic modification. Their exposure of the fibres from the matrix in the vascular canals of the fossilized samples may result either from the fact that (1) mineralized tissue at the time of the animals’ deaths was being resorbed or demineralized, the matrix was dissolved but the fibres remained or (2) when the mineralized tissue was deposited, the fibres were being deposited or incorporated into the front of mineralization before the matrix. Importantly, we were not able to identify these exposed fibres in secondary osteons. In contrast to Horner et al. (2016), we also note their absence in the secondary matrix, so fibres were resorbed during the process of bone remodelling. Based on the photographs published by Horner et al. (2016), we suspect that their identification of secondary fibrous tissue may be erroneous. While some structures observed by them (e.g. Horner et al. 2016: fig. 4A, E) indeed superficially resemble groups of secondary osteons surrounded by cement lines, it seems possible that they, in fact, represent co-ossified (or joined diagenetically) small tendons or subtendons (Cerda et al., 2019; Handsfield et al., 2020) consisting of primary fibrous tissue and central primary or secondary osteons. On the contrary, our samples show clear secondary osteons surrounded by cement lines with clearly distinguishable secondary and primary matrix (Fig. 2E, F). Based on these observations, the proposition of the lack of homology between the typical bone tissue and metaplastic bone seems premature.

The importance of understanding the mineralization process in dinosaur tendons

The coarse collagenous fibre bundles with cells aligned along them (Fig. 6G–O) resemble structures described by Zou et al. (2020) in turkey tendons. These authors, based on FIB-SEM images and corresponding 3D reconstruction of tenocytes, collagen fibrils and mineral deposits, made an attempt to understand the mineralization mechanisms in vertebrate fibrous connective tissue. In the proposed scenario, the spatial proximity between the cells, their structures, fibres and mineral phase suggest that substrates necessary for mineralization may be transported between the fibrils before crystallization. Visualization of such processes in biological tissues is difficult and their evolutionary history and broader phylogenetic distribution is obscure. Therefore, observation of fossilized soft tissues may provide important clues for deciphering various physiological aspects of organism functions elusive for typical biological (neontological) studies (Surmik et al., 2017). The studied tubular structures covered by fibrous surface are reminiscent of bundles aligned with cells (Fig. 6G–K), which, according to their shapes and location, could be interpreted as fossilized tenocytes or some unknown type of fibroblasts secreting minerals and building up the fibril bundles. These cells communicate directly with bundles via cytoplasmic processes (especially Fig. 6I), acting as an intermediary for transporting calcium and propagating the mineralization (ossification) of collagen fibrils (Fig. 8). The presence of highly diversified, spindle-shaped cells with long and ramified processes (Fig. 6L–N), which we interpret as osteocytes, documents fully a functional lacunal–canalicular cellular network in the ossified tendons of ornithischian dinosaurs.

Pathway of preservation

Both the elemental composition obtained from the EDS survey and results of FTIR confirm the presence of mineral apatite as a dominant component of fossilized tendons, as in fresh and fossilized bones (compare Surmik et al., 2016). The demineralization procedure (incubation in EDTA) results in the removal of phosphates from the sample and reveals the preservation of fossilized soft parts (Fig. 6). The obtained extract from E. regalis tendons contains fossilized cells (e.g. Fig. 6L), fibrous matrix (e.g. Fig. 6M), bundle-like tubes (e.g. Fig. 6O, yellow asterisk) and fossilized blood vessels. EDS spectra collected from these samples revealed the presence of iron and sulphur (Fig. 4O) as dominant elements with a lower input of oxygen and silicon. The elemental composition of the H. calathocercos sample shows that the dominant elements are oxygen, silicon and aluminium (Fig. 4E). Although the EDS survey gives general information about the elemental composition of the sample, FTIR is a more informative method to identify structural complexity and the diversity of chemical composition. FTIR examination shows that both in H. calathocercos and E. regalis samples of fossilized soft parts were preserved as alumino-silicates and iron sulphides or iron oxides. Here, FTIR spectra from H. calathocercos and E. regalis indicating bands between 650 and 550 cm^–1 (Fig. 7B, D), which correspond to the (FeO₄)^5– vibration of iron oxides (Fysh & Fredericks, 1983; Cambier, 1986). Additionally, bands between 3800–3400 cm^–1 to the hydroxyl groups and suggest the presence of hydroxylated iron oxides. Moreover, the mutual ratio between the (FeO₄)^5– and OH signals indicates a prevalence of more hydroxylated phases in the sample of E. regalis extracted from fossilized vessels compared to the H. calathocercos sample (Fig. 7B, D). The overlapping signal of the silica and phosphates in the H. calathocercos spectrum and silica or alumino-silicate phases for E. regalis is more visible by comparing two regions: 1200–950 cm^–1 and 550–350 cm^–1 (Fysh & Fredericks, 1983; Barth, 2007). Furthermore, the X-ray diffraction study of E. regalis extract points to the presence of volcanoclastic minerals (zeolite clinoptilolite, compare Supporting Information, Data S5). The sedimentation environment had a strong input into the sample and could have contributed to such a good state of preservation of soft parts (bundles, blood vessels, cells and fibrous matrix). Alumino-silicification as a mode of preservation for soft parts was documented in the Jehol fossil assemblage (Bailleul et al., 2020; Bailleul & Zhou, 2021; Zheng et al., 2021), where terrestrial biota was buried in falling volcanic ash (Zhou, 2014). Beside alumino-silicification, iron minerals played an important role in the preservation of soft parts (Schweitzer et al., 2014; Surmik et al., 2016, 2021). The iron-induced crosslinking and the Fe²⁺/³⁺ in-between interfacial effects (Surmik et al., 2021) is a recognized pathway for soft parts preservation with some potential mechanisms explained (Schweitzer et al. 2014). The detailed studies of the fossilization mechanisms on the pathways of alumino-silicification and iron-induced crosslinking in the fossilized dinosaur tendons, as well as their diagenetic alterations. are beyond the scope of this work and will be developed in the future.

Biomolecular preservation

As a result of the FTIR measurement of fossilized tendons, the fitting procedure of the 1800–1550 cm^–1 range (Fig. 7F–H), which is related to the amide I, revealed the presence of secondary protein structures (α-helix, β-sheet) (Litvinov et al., 2012), as well as polypeptide side chains (β-strands) (Byler & Susi, 1986; Litvinov et al., 2012) and amino-acid side chains (Byler & Susi, 1986; Jackson & Mantsch, 1995), accompanied by alumino-silicates and iron oxides (Fig. 7B, D). Recent research indicates an important role of the alumino-silicate in the protection of organic residues and soft parts on the cellular and nuclear level (Bailleul et al., 2020; Bailleul & Zhou, 2021; Zheng et al., 2021), as was previously known for iron oxides (Schweitzer et al. 2014; Surmik et al., 2016). However, the attribution of the infrared bands to specific amino-acids (even after using the band-fitting procedure) is not conclusive; several ion masses found in the mass spectra (compare Supporting Information, Data S4) corresponding to the collagen type I-associated amino-acids, such as glycine (CH₄N⁺) (Surmik et al., 2016), proline (C₄H₆N⁺, C₄H₈N⁺) (Surmik et al., 2016, 2017) as well as other peptide fragments (Surmik et al., 2016), were also identified.

CONCLUSIONS

The results of our study reveal that samples of ossified tendons of the Late Cretaceous ornithischians contain fibrous structures, embedded deep within their vascular canals. Based on their shape, size, presence of a striped pattern and location in the vascular canals, they are interpreted as fossilized collagen fibres or fibril bundles. The fibrous structures are more frequent in non- or weakly secondarily remodelled tendons. Various permineralization processes, including iron-induced crosslinking and alumino-silicification, play a role in the fixation of protein fibres, fibrous matrix, blood vessels and cells in the studied dinosaur tendon samples. Moreover, the presence of alumino-silicates (e.g. originating in organic-rich volcanoclastic sediments) provides a plausible environment for preserving soft parts. For the first time we have shown that the ossified tendons, alongside the fossilized bone and cartilage, are a precious source of molecular preservation in dinosaurs. Given the relative commonness and number of ossified tendons in the skeletons of numerous ornithodiran taxa, as well as their relatively low morphological informativeness and value for classic anatomical studies, making them a relatively good object for destructive analyses, they may thus potentially serve as a rich and convenient source of taphonomic data.

ACKNOWLEDGEMENTS

We are grateful to Nicole Klein for providing Edmontosaurus regalis tendons. We thank Mariusz Gardocki, Grzegorz Widlicki and Adam Zaremba for the preparation of histological thin sections. We also thank Aleksandra Gawęda and Krzysztof Szopa for taking microphotographs of tendons’ thin sections with the Olympus microscope and Michał Surowski who photographed the thin sections using the Nikon Eclipse polarizing microscope. Ewa Teper and Arkadiusz Krzątała are acknowledged for participation in the SEM microphotography of the studied samples. Leszek Marynowski and Dawid Balcer are acknowledged for performing TC/TOC/TIC/TS analyses. Keyence Corporation is acknowledged for access to the VHX-7000 digital microscope. We also thank Jakub Zalewski for his drawing of fossilized soft parts illustrated in Figure 6. Finally, we thank Ignacio Cerda, an anonymous reviewer and the editor Maarten Christenhusz for their important comments that improved our manuscript. This research project is supported by the National Science Centre, Poland (www.ncn.gov.pl), grant no. 2019/32/C/NZ4/00150.

AUTHOR CONTRIBUTIONS

DS conceived the study and secured the funding. DS, JS and TS prepared the manuscript. MW, DŚ, MD, KB and TK performed the experimental work and analysed the experimental data. DS, JS, TS and RP discussed and interpreted the data. All authors discussed the results and contributed to the final manuscript.

CONFLICT OF INTEREST

The authors declare no competing financial interests.

DATA AVAILABILITY

All data supporting these findings are included in this article and its online supporting information.

REFERENCES

SUPPORTING INFORMATION

Additional supporting information may be found in the online version of this article on the publisher’s website.

Data S1. Distribution of mineralized tendons among the Dinosauria.

Data S2. D-period measurement.

Data S3. Recent contamination of Pinacosaurus grangeri samples.

Data S4. Mass spectra.

Data S5. X‐ray structural characterization of UAMES samples.

Table S1. Selected occurrences of mineralized tendons and ligaments in Ornithodira.

Figure S1. Example sizes of fibrous structures in vascular canals of sample ornithischian dinosaurs

Figure S2: Extracted brown- to orange-coloured tubular structures and network of vascular canals attached to a translucent layer detached from the tendon of a Homalocephale calathocercos sample.