Integumentary structure and composition in an exceptionally well-preserved hadrosaur (Dinosauria: Ornithischia)

General strategy

As an overall strategy, and following an initial petrographic study of the sample, Scanning Electron Microscopy-Energy Dispersive X-ray spectroscopy (SEM-EDS) elemental maps of the skin were first derived to investigate possible markers that could (a) discriminate the skin from the sedimentary matrix, and (b) identify potential regions for the presence of organic contents. Subsequently, the chemical states of a series of elements were identified using X-ray methods, along with analyses using Fourier Transform Infrared Spectroscopy (FTIR).

SEM studies of a 20 µm thick cross section of the hadrosaur skin were conducted with the intentions listed. The thin section covers the first 2 cm of the sample starting from the outer surface of the scales (see section ‘Sample Preparation’). Characteristic X-ray lines, generated in processes involving the interactions between the electron beam and atoms in the sample, were used to produce elemental maps covering a ∼100 µm thick region spanning from the outer surface to the deepest regions of the sample, and single-point spectral measurements (see section ‘Results’ for details). To better understand the morphology of the hadrosaur skin, histological skin samples were prepared for a chicken (Gallus gallus domesticus), saltwater crocodile (Crocodylus porosus) and a rat (Rattus norvenicus) and compared with the hadrosaur skin.

Based on these observations, some samples of the skin (at least 100 µm thick and incorporating the outer surface of the scales) were tested at different beamline endstations at the Canadian Light Source (CLS), Saskatoon, Canada, using synchrotron radiation techniques. Only the data from two techniques are included in this paper, corresponding to Fourier Transform Infrared (FTIR) spectroscopy and X-ray Near Edge Spectroscopy (XANES). FTIR and XANES were used to probe for complex compounds and the chemical state of certain elements of interest, respectively. An elemental map using X-ray fluorescence techniques is included in the Supplemental Information (Section 1).

Sample preparation

Histological skin samples were prepared from skin that was dissected from the thigh (Gallus, Rattus) or foot (Crocodylus). Samples were preserved in 4% paraformaldehyde (PFA) for a three day storage period. Following storage, the samples were slowly introduced to 30% sucrose. Samples were then embedded in freezing media and sectioned in 20 µm thick slices along the transverse plane using a cryostat, allowing for identification of all skin layers. To better improve cell layer identification and contrast, skin samples were stained with eosin. This step was not required for that of the preserved hadrosaur skin, as sufficient contrast was already present in the sample.

Small (∼1 cm²) sections of pristine skin were sampled from the hadrosaur forearm during excavation to avoid contact with potential contaminants (anthropogenic or otherwise). Gloves were worn to collect the sample which were then wrapped in previously unused aluminum foils and placed in zipped plastic bags. The samples were kept in a storage room inside a clean drawer. Gloves were always used while manipulating the sample. All analytical samples were prepared from a single pristine skin sample.

For measurements using scanning electron microscopy and for the histological analysis, a 20 µm polished thin section was manufactured (Vancouver Petrographics, Langley, British Columbia, Canada) using standard petrographic thin section preparation techniques. Carbon was used to coat the sample (under high vacuum conditions) to improve conductivity before the SEM-EDS measurements. A small carbon signal is, then, expected in all spectra collected with this technique. Any excess carbon observed is assumed to be endogenous to the sample.

The CLS Mid-Infrared (MidIR) beamline endstation (May, Ellis & Reininger, 2007; Canadian Light Source, 2019a) was used for the FTIR measurements. Three different samples were prepared. Two of these were sampled from a pristine piece of the dinosaur skin still attached to the sedimentary matrix. A microtome (Microtome, Leica EM UC7) with a diamond blade was used to “scrape” one of the scales on two different surfaces (Fig. 1D): 1. on the outer (superficial) surface of the scale, producing a light-coloured powder (hereafter called “light-coloured powder”), and; 2. on the broken cross-section of the scale and incorporating some of the underlying sedimentary matrix, producing a dark-coloured powder (hereafter called “dark-coloured powder”). The third sample, also a small pristine piece of skin still embedded in sedimentary matrix, was immersed in HCl for two days to remove most of its mineral contents. The remaining sample was then washed with ethanol and air-dried at room temperature.

For XANES analysis, a piece of hadrosaur skin and the attached sedimentary matrix was sectioned at 100 nm into a water bath containing de-ionized water at room temperature (Microtome, Leica EM UC7). Due to the hardness of the sample, only flakes of the sample, containing skin and debris from the sediment, were obtained. The flakes and sediment were removed from the water bath using a wire loop and deposited onto a silicon nitride membrane, air-dried at room temperature and examined using X-ray spectromicroscopy with the Spectromicroscopy (SM) (Kaznacheyev et al., 2004; Kaznatcheev et al., 2007; Canadian Light Source, 2019b) beamline endstation.

Instrumentation

A Jeol JSM-6360 electron microscope (Department of Geology, University of Regina) and a Zeiss Axio Observer Z1 phase-contrast optical microscope (Department of Biology, University of Regina) were used to investigate the histological thin sections of the hadrosaur, chicken, rat, and crocodile skin. Linear measurements of various histological features were taken using the publicly available ImageJ 1.52k software.

A Nikon eclipse E600 Pol microscope (Department of Geology, University of Regina) was used for petrographic analysis of the hadrosaur skin and associated sedimentary matrix. Mineral identification were made utilising standard petrological microscope techniques; compositions were confirmed by reference to SEM-EDS analysis.

For the SEM measurements, the sample was subjected to high vacuum. To determine the best operational conditions, several areas were imaged at a range of beam currents (varied by increasing or decreasing the beam spot size), utilizing a smaller beam aperture and variable kV. Optimum conditions that provided the clearest and sharpest images at high magnification, but that did not lead to sample charging or sample degradation, were found to be 10–15 kV and a spot size of 65–70 µm. Examinations of samples were undertaken with both the secondary electron (SEI), and back-scattered electron (BSE) detectors fitted to the Jeol JSM-6360 instrument. A Noran 7 energy dispersive system attached to the Jeol SEM completed the X-ray spectroscopy analysis and mapping.

For FTIR measurements, the MidIR beamline provides a beam in the mid infrared region produced by a bending magnet, covering an energy range of 0.070–0.744 eV. A Hyperion 3000 IR microscope, cooled with liquid nitrogen, was used with the top objective set to 0.45 and the bottom objective set to 0.30, resulting in a beam spot size of 20 µm. 128 scans in transmission mode were performed per measured point. The data was processed using the Bruker commercial software package OPUS (Bruker, 2019) (baseline corrections were applied as provided with the software).

The scanning transmission X-ray microscope (STXM) endstation on the SM beamline (10ID-1) was used to collect all X-ray imagery and spectroscopy. Image sequences (i.e., stacks) (Dynes et al., 2006b) were collected at the C, Mg, Al and Si K-edges and at the K, Ca and Fe L-edges. The as-measured transmitted signatures (I) were converted to optical density (absorbance = OD = −In (I I0-1)) using the incident flux (I0) measured through areas where there was no sample present. The stacks (dimensions 8 × 8 µm) were collected at a spatial resolution of 60 nm and a dwell time of 1 ms/pixel. Principal component and cluster analysis (PCA-CA) was used to determine the number of chemical components present and/or derive spectra in selected image sequences (Lerotic et al., 2004). Quantitative maps from the image sequences were obtained using singular value decomposition (SVD) to fit the spectrum at each pixel to a linear combination of reference spectra of the components suspected to be present, or using spectra derived from the image sequence (Dynes et al., 2006a). Representative spectra were derived from the component maps using threshold masking (Dynes et al., 2006b; Dynes et al., 2006a). Axis2000 (Hitchock, 1997) was used for all STXM data processing.

The Very Sensitive Elemental and Structural Probe Employing Radiation from a Synchrotron (VESPERS) CLS beamline endstation (McIntyre et al., 2010; Canadian Light Source, 2019c) was also used to investigate possible chemical markers differentiating the skin from its associated sedimentary matrix. X-ray fluorescence spectroscopy using a polychromatic beam with energy ranging from 6 keV to 30 keV was employed, covering elements such as Ti, Ca, Mn, Fe, Zn, Sr, Y and Zr, among several others. The results of these measurements for Fe, Ca, Cu and Mn are presented with the Supplemental Information.

Histology

Optical microscopy of histological samples of the skin from three extant representatives (Gallus, Crocodylus, Rattus) reveals a thin but characteristically multi-layered epidermis and a deeper, thicker dermis. The outermost epidermal layer corresponds to the stratum corneum, a keratinous layer that aids in protection of the internal organs (including the underlying epidermal and dermal layers) from desiccation. The stratum corneum, which is composed of stratified layers of β-keratin, is the thickest component of the epidermis in Crocodylus owing to the presence of keratinised scales (Fig. 2D). As a result of this cornified layer, the epidermis is also the thickest in Crocodylus in both absolute and relative terms. The stratum corneum is comparatively thin in both Rattus and Gallus, but where epidermal scales are present on the avian podotheca, such as Gallus, they are similarly covered by a thick stratum corneum (Pass, 1989), although these were not sampled in our specimen. In birds, the relative thickness of the epidermal layers differs between locations (Lucas & Stettenheim, 1972); however, the epidermis is consistently thinner in Gallus than it is in Rattus and Crocodylus, a feature that has been linked to the progressive lightening of the avian body and the evolution of flight (Stettenheim, 2000).

In Crocodylus, the epidermis is invaginated to form the hinge area between scales (Fig. 2C). Much deeper invaginations, invading both the epidermis and dermis, are formed by hair and feather follicles in Rattus and Gallus, respectively (Fig. 2F).

Underlying the epidermis is the dermis, which contains openings for the blood vessels, fat deposits, and abundant pigment cells, the latter of which are more diverse in Crocodylus and other reptiles due to their naked skin (e.g., Bagnara & Hadley (1973); Fig. 2D). Glands are relatively scarce and/or small in avian and reptilian skin (e.g., Quay, 1972; Stettenheim, 2000; Jacobson, 2007), but are a salient feature of mammalian dermis (Quay, 1972; Fig. 2F). Thick subcutaneous hypodermis dominated by fat stores and blood vessels underlies the dermis in both Gallus and Rattus, whereas it is relatively thin in reptiles and not sampled in our specimen of Crocodylus.

Phase-contrast optical microscopy of the hadrosaur skin reveals an outermost (superficial) dark-coloured layer ∼35–75 µm in thickness, which overlies the sedimentary matrix (Figs. 2A and 2B). This outer layer is composed of clearly-defined, alternating dark and lighter-coloured layers, which typically range from ∼5 µm to ∼15 µm in thickness (Fig. 2B). Individual layers are typically laminar or undulatory giving the entire outer layer a stratified appearance. These finer layers may deviate around sedimentary particles that are occasionally found embedded within the outer dark layer (Fig. 2I). Aside from these occasional particles, sedimentary grains are typically restricted to the sedimentary matrix underlying the dark outer layer. In places, the dark layers appear to be composed of oval substructures measuring a few micrometers in maximum dimension (Fig. 2I). The entire dark stratified layer is, in places, capped by a pale-coloured, faintly laminated region identified as barite (see SEM and petrographic analyses below). No other evidence of integumentary features that could be interpreted as hair follicles, feathers or glandular structures could be identified anywhere in the sample. Other epidermal/dermal features such as osteoderms and melanosomes are also absent.

Petrography

The sample associated with the Hadrosaur skin is chiefly a fine to very fine grained detrital sedimentary rock (or subarkosic arenite), with local clay partings. The bulk of the grains are angular to subangular equant quartz (55–60%; 50–250 µm), and angular to subrounded clasts of K-feldspar/alkali-feldspar (10%; 25–100 µm), albite and plagioclase feldspar (5–10%; 50–150 µm). Chert/chalcedony is a minor component, with grains ranging from angular to subrounded (2–3%; 50–100 µm). Zircon (2–3%, 20–40 µm) is an important accessory phase, with many scattered angular and, more rarely, rounded grains occurring throughout the rock. Apatite, rutile and pyrite are also rare accessory minerals. Locally, euhedral grains of dolomite (2–3%, ∼100–150 µm) and barite (1%, up to 100 µm) occur, and may represent diagenetic phases. Calcite forms rare discrete grains (up to 50 µm), and is a common cement to the detrital grains (∼5%). Additional matrix minerals (<5%) include kaolinite and intermixed clays (potentially chlorite, smectite and illite), that commonly form local thin clay rich laminae, chiefly adjacent to the “skin” layer. The albite feldspar is commonly altered to sericite/muscovite and clay minerals.

Scanning Electron Microscopy (SEM) analysis

Back-scattered electron (BSE) SEM energy-dispersive X-ray imagery of a thin section comprising both skin and associated sedimentary matrix was used to identify concentrations of lower atomic number (low Z) elements (dark regions in Figs. 3C and 3D) and higher-Z elements (brighter areas in Figs. 3C and 3D). Thus, carbon-based structures (low Z), if present, are expected in the darker areas, while sedimentary material (high Z) will be mostly localized in brighter areas. SEM images show that, in general, the entire sample is dominated by angular particles of a sedimentary nature (white-to-grey in Figs. 3C and 3D, mostly detrital grains of quartz and feldspar). In contrast, the outermost 50–75 µm differs in both texture and concentration of elements. BSE SEM shows an outermost layer of white (high-Z; ∼25 µm thick) material identified as barite (Supplemental Information, Section 2) that caps a region of numerous thin, dark bands (low Z) up to a total thickness of ∼35–75 µm. The dark bands are interrupted by discontinuous laminae, or lenses, of higher Z material (grey in Fig. 3C).

To partially assess the elemental compositions of the light and dark bands, a number of spectral distributions (using SEM-EDS) were collected at different points on the sample as depicted in Fig. 3C. Most of these points (points 1 and 3–6 in Fig. 3C) were from grey or white areas (all high Z) identified by BSE SEM. Unsurprisingly, they showed no particular characteristic indicating contribution of organic components to each respective spectrum (Supplemental Information, Section 2). The most interesting spectrum comes from a region corresponding to lower-Z elements (point 2 in Fig. 3C and darker region indicated by an arrow in Fig. 3D). Figure 4 depicts this spectrum and is typical of all similar low Z areas distributed in the sample. These dark, carbon-rich areas are restricted to the outer parts of the sample and, in most cases, are covered with layers of kaolinite (among other minerals including quartz; point 5 in Fig. 3C; see also Supplemental Information). Interestingly, kaolinite appears to be mostly concentrated around these darker areas, which might indicate a correlation between carbon and kaolinite during diagenesis.

An elemental map of an area containing both high and low-Z regions (sedimentary and carbon-rich regions, respectively) was achieved using SEM X-ray in emission mode at a 5 µm scale. BSE imaging (Fig. 5A) of the carbon-rich (low Z) region confirms the identification of discrete layers or bands (dark structures in Fig. 5A). However, the SEM images alone are not sufficient to resolve those structures due to the two-dimensional nature of the measurement technique. Carbon distribution is directly correlated to the darker area in the sample (Fig. 5B), whereas oxygen, aluminum and silica are mostly concentrated outside the carbon-rich region (Figs. 5C, 5D and 5E, respectively).

Measurements with the MidIR beamline endstation

The following references were used for the identification of the different compounds in this section: Silverstein et al. (2014) for organic material; Derrick, Stulik & Landry (1999) for inorganic material. Extra references are provided in the text.

The first measurements using the MidIR beamline were carried out using the sample cleaned with HCl. This technique served to remove the hardest mineral parts from the sample leaving only a small amount of material associated with the epidermal scale. Most of this remaining material is identified as kaolinite (in agreement with the SEM results; Supplemental Information, Section 2), which is characterized by peaks at 3,690, 3,651, 3,619, 1,113, 1,026, 1,005, 936, and 910 cm⁻¹ (Fig. 6) (Vahur, 2015; Madejová & Kamodel, 2001).

The peak at ∼1,635 cm⁻¹ is likely related to the presence of montmorillonite (a type of smectite) and arises from the bending vibration of the OH group. The spectrum of montmorillonite has several peaks that superimpose with some others from kaolinite, although the peak at ∼1,635 cm⁻¹ is strictly characteristic of the smectite group. On the other hand, the broad peak centered at ∼3,400 cm⁻¹ is likely due to O-H stretch vibrations, which may be indicative of water impurities in the sample (Sigma-Aldrich, 2019). If that is true, O-H scissoring vibrational excitations could also explain the peak at ∼1,635 cm⁻¹. Another possibility could be contamination by alcohol (used to clean the glass slides that hold the samples); however, the absence of other peaks characteristic of the alcohol group (Sigma-Aldrich, 2019), such as those at 1,420–1,330 cm⁻¹, refutes this supposition.

Interestingly, the peaks located at ∼2,922 cm⁻¹ and ∼2,849 cm⁻¹ are likely the result of C-H vibrations and could provide evidence for the presence of organic material in the sample. The peak at ∼1,635 cm⁻¹ could support this statement as it can be associated to the presence of amide-1 (indicative of carboxylic acid in the sample) (Sigma-Aldrich, 2019); however, as discussed above, this peak is likely of inorganic nature. Thus, the absence of other peaks that could support the presence of, for example, some carbonyl groups does not support the interpretation of an organic origin for those two peaks around 2,900 cm⁻¹. Another plausible explanation for the appearance of those peaks is that they arise from C-O vibrations in carbonates. However, the complete absence of any significant peak near 1,400 cm⁻¹ (characteristic of carbonates) makes this a weak proposition (Sigma-Aldrich, 2019).

The potential contamination by water absorption makes the above sample unsuitable for searching for organic material. The peaks arising from vibrational excitations of water molecules can superimpose those produced by vibrational modes in organic groups, making identification of the latter difficult.

Several maps and spectra were also collected from both the light and dark-coloured powder samples with the MidIR beamline using FTIR in transmission mode. The sample preparation in this case was less invasive (from a chemical perspective) and required less manipulation. While the light-coloured powder sample yielded spectra rich in clay mineral signatures (Supplemental Information, Section 4), such as kaolinite and montmorillonite, the dark-coloured powder provided a more suggestive selection of spectra.

Four spectra were selected from the map shown in Fig. 7 based on the features present in one or more of the areas of interest (Fig. 8) (the complete set of spectra can be found in the Fig. S7). While some minerals can be associated with some of the observed peaks, there are other peaks that likely indicate the presence of organic traces (Table 1).

In particular, the appearance of the 1,789 cm⁻¹ and 1869 cm⁻¹ peaks in a FTIR spectrum is indicative of the presence of anhydrides acid (Coates, 2000). Interestingly, this acid is produced from the dehydration of carboxylic acids (Bruckner, 2010). The peak at 1,789 cm⁻¹ corresponds to C-O symmetric stretch while that at 1,869 cm⁻¹ corresponds to C-O asymmetric stretch.

Measurements using the SM beamline endstation

To further constrain the possible explanations for the observations described here, data were collected using the SM beamline from samples produced using a microtome. Optical images of the sample examined using the STXM are depicted in Fig. 9.

Carbon K-edge

The C K-edge image sequence was fit with protein, lipid and polysaccharide spectra (representing the major biomacromolecules commonly found in environmental samples), as well as carbonate and K (Dynes et al., 2006a). The spectra derived by threshold masking of these biomacromolecule component maps showed the same spectrum for each component map (Fig. 9A), which did not contain the characteristic peaks/spectral shape of protein (288.2 eV peak attributed to peptide band of proteins), polysaccharide (289.2 eV peak attributed to aliphatic C-OH) or lipids (Lawrence et al., 2003). The fact that only one spectrum could be derived indicates that the organic material was consistent throughout the skin. The main peaks in the spectrum are due to carbonyl (288.5 eV), ketone carbonyl (286.7 eV), carbonate (290.3 eV) and K (297.2 and 299.9 eV), C functional groups commonly found in soils (Gillespie et al., 2014).

Table 1:

Associations between peaks in Fig. 8 and possible compounds (Silverstein et al., 2014; Derrick, Stulik & Landry, 1999).

Peak position	Possible association
1,610 cm−1	Carboxylic acid, COO anti-symmetric stretch.
1,680 cm−1	Carbonyl, C-O stretch α, β-unsaturated aldehydes and ketones; alkene, C-C stretch.
1,789 cm−1 and 1,869 cm−1	Anhydrides acid, C-O symmetric stretch and C-O asymmetric stretch, respectively.
2,514 cm−1	Calcite; Dolomite with calcite phase.
2,849 cm−1 and 2,964 cm−1	Aliphatic compounds (alkenes), C-H vibrations.

DOI: 10.7717/peerj.7875/table-1

In order to map the organic carbon, carbonate and K separately from each other, a modified carbon spectrum (referred to as “carbonyl”) was made by subtracting a carbonate spectrum (i.e., calcite) from it and removing the K peaks (Fig. 10B) (Dynes et al., 2006a). Two fittings were performed: one consisted of using calcium carbonate (i.e., calcite) and K (K₂CO₃ with carbonate signal subtracted) spectra, while the other used only a K₂CO₃ spectrum. Both fittings used the carbonyl spectrum and a slowly varying featureless signal (FS) (i.e., slightly decreasing straight line) (Dynes et al., 2006b), which represents the inorganic components present in the sample. The calcite and K spectra were published previously (Dynes et al., 2006a). This is the first time that the K₂CO₃ spectrum has been published to our knowledge. It exhibits the spectral shape of carbonates (Brandes, Wirick & Jacobsen, 2010). The component maps from the fittings are shown in Fig. 11. The carbonyl and FS maps are similar for both fittings, thus, only those from the first fitting are shown. It is apparent that the K₂CO₃ component map (Fig. 11D) is predominantly the sum of the carbonate and K component maps. Overlay of the carbonate and K component maps revealed that there are at least two different carbonate particles present: K₂CO₃ and another carbonate in the skin. In the sediment, only carbonates were identified (both calcite and dolomite were observed in the host rock, see section ‘Petrography’); neither organic carbon nor K was present Brandes, Wirick & Jacobsen (2010).

Ca L-edge

Previous work (Benzerara et al., 2004; Fleet & Liu, 2009) has shown that the energy positions of the main Ca L-edge features are the same for the calcium carbonate polymorphs (e.g., calcite, aragonite, vaterite) and dolomite (CaMg(CO₃)₂ occurring at 249.3 eV (Ca LIII-edge) and 252.6 eV (Ca LII-edge). The number, energy positions, and relative intensities of the small peaks at lower energy relative to the main features are dependent on the coordination, reflecting electronic and crystal-chemical details within the first coordination sphere (Benzerara et al., 2004). Ca spectra derived from the image sequence from the skin and the sediment are shown in Fig. 12E. The ratio of the peak at 351.4 eV and adjacent to the main peak at 352.6 eV in the skin is 1.88. The ratio for the LII-edge peaks from the Ca in the sediment is 1.18. Hence, it appears that the Ca in the sedimentary areas occurs as calcite and in the skin as vaterite (Benzerara et al., 2004). However, potential spectral distortions associated with absorption saturation for optical density greater than two are possible, and may have decreased the ratio of these peaks, particularly for the Ca in the inorganic spectrum (Obst et al., 2009). Thus, it is possible that the CaCO₃ in the sediment is vaterite and not calcite. Nevertheless, there is a small energy shift of the small peaks (Fig. 12F) from the skin and sediment indicating that their coordination of Ca is slightly different. Two types of Ca compounds are apparent when the Mg component map is overlaid with the Ca component maps (not shown); calcium carbonate and either dolomite or calcium carbonate containing significant amounts of Mg.

The Ca speciation in the sediment surrounding (and away from) the skin was also examined. Spectral fitting of the sediment Ca L-edge image sequence with the two spectra used in the skin/inorganic material fitting, and subsequent threshold masking of these component maps indicated that results similar to those for the skin were obtained, with regards to the ratio of the peaks. Thus it appears that absorption saturation is likely the reason for some of the peak ratio difference. In any event, only one species of Ca is apparent in the sediment, as the positions of the small peaks are the same between the two spectra. The Ca was observed to coincide with the carbonate map from the C K-edge fitting (Pearson’s coefficient 0.89), confirming that the Ca occurs as calcium carbonate.

Fe L-edge

The oxidation state of Fe can be determined from the relative intensities of the double peaked Fe L3 signal and from the position and number of peaks at the Fe L2–edge (Dynes et al., 2006b). The Fe L-edge image sequence for the skin was fit using two Fe reference spectra (Fig. 13A), representing Fe(III) (goethite) and Fe(II) (siderite) oxidation states and a featureless signal (FS) spectrum. The goethite and siderite spectra were previously published (Dynes et al., 2015). The Fe(III) and Fe(II) component maps are shown in Figs. 13B and 13C, respectively. The Fe(III) and Fe(II) spectra (3-point smoothed) derived from the respective component maps are characteristic of Fe(III) and Fe(II) spectra. The Fe(III) spectrum from the sample also contains considerable Fe(II). The Fe L-edge image sequence of the sediment was fit with the goethite, siderite and FS spectra. Threshold masking of these component maps indicated that there are Fe(II) and Fe(III) species in the sediments. The Fe(II) species in the sediment is similar to that found in the skin, whereas the Fe(III) species in the sediment contained more Fe(II) compared to that in the skin as evident from the equal heights of the two peaks at the Fe L3-edge. In agricultural soils, both Fe(III) and Fe(II) can coexist, even in well aerated soils, attributed to mixed valence Fe(II)-Fe(III) minerals and/or from periodic low-redox conditions where Fe(III) is reductively transformed to Fe(II)-bearing materials (Chen et al., 2014).

Measurements of Al, Mg and Si K-edge were also performed and the data and discussions are provided with the Supplemental Information (Section 5).

Carbonyl, Ca and Fe maps of a carbon-rich layer and its substructures at the SM beamline

The skin flake was again measured at the SM beamline in order to identify the chemical state of the carbon associated with the optically-identified skin layers in Fig. 2. Measurements of the area containing the skin were performed with photon beam of 280 eV (corresponding to an energy lower than the carbon K-edge energy), and 288.5 eV (carbonyl K-edge excitation energy) using photon absorption mode. The image collected at 280 eV was used for background subtraction in the computation of the image at 288.5 eV. Red in Fig. 14A represents carbonyl after background subtraction, while cyan represents the measurements at 280 eV (non-carbonyl).

The yellow rectangle encapsulates a region of the sample with a layered structure similar to those identified in Figs. 2A, 2B, 2I and 2J. As Fig. 14 shows, this layer is rich in carbonyl, providing potential organic or organic-related material. The layers are clearly different both at chemical and morphological levels relative to other sedimentary regions of the sample.

A more detailed map of the area represented by the red rectangle in Fig. 14A was developed using a photon beam at 288.5 eV (carbonyl). This map is displayed in Fig. 14B. It shows potential carbonyl-rich substructures organized in a layer within the limits of the dinosaur skin. The existence of these substructures is reinforced by the evidence observed in both the optical and SEM images of the skin. The images captured using the three different techniques (X-ray SR, SEM and optical) all show similar geometrical features with apparent wave-like contours to the skin layers. These characteristics strongly suggest organized structures composed of subcircular substructures resembling cells. Artifacts introduced by the measurements can be discarded with high likelihood as it is unlikely that three completely different and independent techniques would randomly yield similar geometric features.

One of the substructures was mapped for carbon, Ca and Fe (Fig. 14C). As expected, the chemical states of these elements were found similar to those observed in section ‘Measurements using the SM Beamline Endstation’ for the carbon-rich area (predominantly carbonyl, vaterite/calcite and goethite). As depicted in Fig. 14C, carbonyl is relatively uniformly distributed, while vaterite/calcium appears to delineate the substructure. More remarkably, goethite is highly concentrated at the center of this form. We have not investigated the reasons behind these distributions, and can only speculate about them at this point.