Putative fossil blood cells reinterpreted as diagenetic structures

Red to red-orange spheres in the vascular canals of fossil bone thin sections have been repeatedly reported using light microscopy. Some of these have been interpreted as the fossilized remains of blood cells or, alternatively, pyrite framboids. Here, we assess claims of blood cell preservation within bones of the therizinosauroid theropod Beipiaosaurus inexpectus from the Jehol Lagerstätte. Using Raman spectroscopy, Energy Dispersive X-ray Spectrometry, and Time of Flight Secondary Ion Mass Spectroscopy, we found evidence of high taphonomic alteration of the bone. We also found that the vascular canals in the bone, once purported to contain fossil red blood cell, are filled with a mix of clay minerals and carbonaceous compounds. The spheres could not be analyzed in isolation, but we did not find any evidence of pyrite or heme compounds in the vessels, surrounding bone, or matrix. However, we did observe similar spheres under light microscopy in petrified wood found in proximity to the dinosaur. Consequently, we conclude that the red spheres are most likely diagenetic structures replicated by the clay minerals present throughout the vascular canals.

Here, we analyze the preservation and vascular canal contents of the holotype of B. inexpectus. We aim to determine the composition of this specimen's putative blood cells and the degree of diagenetic alteration of the specimen using the following combination of analytical tools: light microscopy, Scanning Electron Microscopy (SEM), Energy Dispersive X-ray Spectroscopy (EDS or EDX), Raman spectroscopy (Raman), and Time of Flight Secondary Ion Mass Spectrometry (TOF-SIMS). Although each analytical tool has limitations, the combined strengths of these tools offer new insights into intra-vascular preservation.

Geological setting
Beipiaosaurus inexpectus (IVPP V11559) comes from the Sihetun locality of the Yixian Formation. The locality is a protected site within the Beipiao Bird Fossil National Nature Reserve of Liaoning Province, China (Figs. 1A, 1B) (Wang, 1998;. In the Sihetun area, the Yixian is composed of two sedimentary and two volcanic units. The Lujiatun Unit is the lowest and consists of coarse-grained sediments. This is succeeded by the lower volcanic unit, the shales and siltstones of the Jianshangou Unit, and then the upper volcanic unit (Jiang et al., 2011;Jiang & Sha, 2007). The tuffaceous claystones and shales of the Jianshangou Unit host B. inexpectus as well as many other vertebrate and invertebrate fossils (Chang, 2011;Jiang & Sha, 2007;Pan et al., 2013;Zhou, 2006b). Radiometrically dated ( 40 Ar/ 39 Ar) tuffs near the fossil-bearing layer constrain this unit to 124 Ma (Chang et al., 2009)-125 Ma (Fig. 1C) (Swisher et al., 2002, consistent with palynological data (Li & Batten, 2007). The Jianshangou Unit is bracketed by the volcanic units, dated at 125-128 Ma (lower volcanic unit) (Chang et al., 2017) and approximately 122 Ma (upper volcanic unit) (Jiang et al., 2011;Wang et al., 2001). At the Sihetun locality, the layer bearing B. inexpectus (beds 25-29 in Wang et al., 1999) underlies a tuff dated between 124.35 and 126.1 Ma by less than 3.5 m (Fig. 1C) . Based on these dates, the specimen of focus in this study is from the late Barremian or early Aptian age of the Early Cretaceous Period (Chang et al., 2017;Chang et al., 2009;Cohen et al., 2013 (updated);He et al., 2006;Jiang & Sha, 2007;Wang et al., 2001).  Zhang & Tang (2002) and field observations. This section is approximately 15 m from the original collection site and is now covered by new construction. Tuff thicknesses vary laterally. (D) Detailed stratigraphy of the fossil horizon of the holotype specimen of Beipiaosaurus inexpectus. Drawn from field observations. Full-size DOI: 10.7717/peerj.12651/ fig-1

MATERIALS & METHODS
All samples examined in this study, as well as slide/sample number, element (if applicable), and analyses performed, are listed in Table 1. For histological analysis of fossil bone under light microscopy, we re-analyzed the petrographic thin sections of the left humerus of the B. inexpectus holotype (IVPP V11559) described in Yao, Zhang & Tang (2002) (see Fig. 2A for sampling locations). The original thin sections by Yao, Zhang & Tang (2002) were covered in a highly fluorescent epoxy as well as a glass coverslip, and thus were not useful for Raman, EDS, SEM, or TOF-SIMS analysis.
For SEM, Raman, and EDS analysis (as well as additional light microscopy examination of histology), we prepared new petrographic thin sections from materials collected during the initial excavation of IVPP V11559 (Yao, Zhang & Tang, 2002). These include a fragmentary gastralium and additional fragments of the left humerus from the holotype, as well as petrified wood found in association with this individual. A hand sample containing gastralia and associated matrix from the holotype specimen was sampled directly for TOF-SIMS analysis. Finally, we also prepared thin sections of modern bone for comparison of Raman spectra.

a. Petrographic sectioning
To prepare thin sections of B. inexpectus material and petrified wood, we embedded the fragments in Castolite-AC (EP4101polyester casting resin; Eager Polymers, Chicago, IL) catalyzed with MEKP, and then cut 0.5 mm thick slices on an IsoMet 1000 Precision Sectioning Saw (Buehler, Lake Bluff, IL) using tap water to fill the basin (for lubrication) and approximately 5 mL of Cool 2 Cutting Fluid (for cooling; Buehler). Before processing each new specimen, we cleaned and dressed the blade using a priming block. We air-dried slices for a minimum of 12 h. Before mounting, we roughened slice-sized areas on mounting side of plexiglass slides with a 120-grit sanding sponge to improve specimen adhesion. We mounted each slice on a roughened plexiglass slide using Loctite ULTRA Gel Control commercial superglue (Henkel Corp., Bridgewater, NJ). Mounted slices were then ground on a Metaserv 2000 Grinder/Polisher (Buehler) until they were thin enough to view details in transmitted light (approximately 100 µm thick). We first ground the specimens using 240 grit until they were ∼300 µm thick, then 400 grit to ∼150 µm, and finally using 800 grit until the specimens were transparent enough that their histological features were visible when dry. All specimens prepared for this study were rinsed clean, but their surfaces received no further chemical treatment after grinding (unless otherwise noted), and they were not coverslipped.

b. Raman spectroscopy
Raman spectroscopy is a well-established method in taphonomy (Bernard et al., 2007;Jehlička, Jorge Villar & Edwards, 2004;Marshall et al., 2012;Thomas et al., 2007;Thomas et al., 2011;Wiemann, Yang & Norell, 2018;Witke et al., 2004). It is useful for assessing the degree of diagenetic alteration to bone (Thomas et al., 2007;Thomas et al., 2011) and Table 1 Materials used in this study. All dinosaur material is from IVPP V11559 (holotype specimen of Beipiaosaurus inexpectus). Each specimen listed here was examined during this study, and its anatomical and preparatory context are listed in addition to the types of analyses used. Some were not part of the final results.

Figure 2 Photograph of sampled specimen (Beipiaosaurus inexpectus, IVPP V11559) (A) and transmitted light micrographs of representative thin sections (B-D).
In thin section images, black arrows indicate spheres (putative red blood cells), white arrows indicate osteocyte lacunae, and grey arrows indicate non-spherical vessel fills (A) Initially published portion of IVPP V11559, with sampled areas marked ''1'' and ''2''. Samples from these areas include previously published thin sections (Yao, Zhang & Tang, 2002) as well as thin sections produced by the authors for this study. We also sampled fragments of associated gastralia (not pictured here) (B) Vessel containing spheres as well as non-spherical amorphous vessel fill (grey bracket) in thin section LJ98B-1 (sample area 1; originally prepared for (Yao, Zhang & Tang, 2002)). evaluating the origin of compounds and structures in fossil material (Marshall et al., 2012;Thomas et al., 2014;Wiemann, Yang & Norell, 2018). We employed Raman spectroscopy to test the alternative interpretation of the putative red blood cells as framboids of iron minerals (Martill & Unwin, 1997), and to assess the quality of bone preservation from a diagenetic perspective. All Raman spectra were obtained from thin sectioned material, and collected on a high-resolution, 800 mm focal length spectrometer (LabRAM HR800; Horiba Scientific, Kyoto, Japan) with a 785 nm diode laser at the Virginia Tech Vibrational Spectroscopy Lab. The laser beam was focused on an approximately 8 µm diameter spot to reduce heat on the sample and points for analysis were chosen using a 50x objective lens. Raman spectral data were collected at 1/10 power (maximum power = 150 mW) for 5 s per collection. Decreased laser power reduces the intensity of Raman peaks, but also lessens laser damage of the specimen. As blank tests, we also took Raman spectra of the plexiglass, epoxy, and superglue used in slide preparation. The relatively thick (100 µm) sections and the weak laser power used in the analysis ensured that the Raman data were not contaminated by epoxy or plexiglass signals; this was confirmed by a comparison between sample data and blank data. The spectral resolution is 0.69 wave numbers. Raman spectra were baselined using the Gaussian baseline correction algorithm in Fityk (Wojdyr, 2010) and CrystalSleuth (Laetsch & Downs, 2006). Reference peak positions (e.g., for pyrite and apatite) were obtained from the CrystalSleuth database.

c. Energy dispersive X-ray spectrometry
To locate vessels exposed on the surface of thin sections and to obtain element maps of vessel-filling material and near-vessel bone, we used a Hitachi TM-3000 Tabletop SEM coupled with a Quantax 70 Energy Dispersive X-ray Spectrophotometer (EDS) system (Hitachi High-Tech America, Inc., Schaumburg, IL). In initial analyses, we left the samples uncoated, but surrounded them with aluminum tape to improve conductivity, and used the Quantax70 to interpret and visualize the EDS data. Based on these preliminary analyses, we identified specimen 2018-L2 (longitudinal section of a gastralium fragment, see Table 1) as the best candidate for quantitative EDS spot analysis due to the abundance of vessels exposed on the surface of the section.
We then collected spot analyses from 2018-L2 on an FEI Quanta 600FEG environmental SEM (FEI Company, Hillsboro, OR) with both back scattered electron (BSE) and secondary electron (SE) capability, operating at a voltage of 5-20 kV. The sample was coated in a mixture of gold and palladium and surrounded by aluminum tape to improve conductivity. We used a Bruker EDX to carry out spot analyses of elemental concentrations and to generate additional elemental maps on vessel fills and lacuna fills.

d. Time of flight -secondary ion mass spectrometry
Relative to other mass spectrometers for fossil analysis, TOF-SIMS is minimally destructive; it removes only about 1 nm of surface material and charges material up to 30 nm deep in the specimen, a depth easily removed with sputtering if further analysis in a spot is desired (Cheng, Wucher & Winograd, 2006;Debois, Brunelle & Laprévote, 2007;Touboul et al., 2005). TOF-SIMS creates a map of relative ion abundance across an area of a specimen, can analyze both organic and inorganic molecules simultaneously, and can detect molecules as well as individual elements with superb spatial precision (Thiel & Sjövall, 2014).
To minimize potential contamination from epoxy, we did not embed TOF-SIMS specimens. Instead, we used a hand sample of gastralia from IVPP V11559, which is surrounded by finely laminated mudstone matrix. We ground a fresh edge of a transverse break through one gastralium (section 2018-X3; see Table 1) using 240 grit and then 800 grit sandpaper on the Metaserv 2000 Grinder/Polisher described above. Just prior to TOF-SIMS analysis, a slice approximately 1 mm thick (including the ground edge) was cut off with a Dremel tool and rinsed with isopropyl alcohol in order to remove any surface contamination from handling the specimen during collection, storage, and processing. Although isopropyl may remove carbonaceous compounds, this step was necessary because IVPP V11559 was not originally collected and handled with chemical analyses in mind. Section 2018-X3 was then mounted on a silicon stub and placed into the vacuum chamber of an IONTOF TOF.SIMS 5 (Iontof GmbH, Munster, Germany). Areas of interest were sputtered with a 2 kV cesium beam for a minimum of 5 min to reveal a clean surface at depth (Graham & Gamble, 2018;Taylor, Graham & Castner, 2015). Spectra were acquired from the sputtered surfaces using a 30 kV bismuth ion beam (primary species Bi 3+ ), as bismuth is not of biological interest and Bi 3+ is very precise (Touboul et al., 2005).
We calibrated for mass using C, 18 O, O 2 , F, Na, C 2 , Al, Si, P, Cl, 37 Cl, PO, PO 2 , and PO 3 . After identifying calibration peaks, we defined the width of each peak and then assigned non-calibration peaks starting at low masses. We aimed for deviation measures (the difference of an observed peak from the expected location for a given secondary ion) below 200 ppm to improve peak assignment accuracy. Components of shouldered peaks (see (Green, Gilmore & Seah, 2006) for comparable peak morphology) were distinguished from one another whenever the IONTOF TOF.SIMS 5 software was able to separate them, and we placed the boundary at the nadir of the valley or inflection point of the shoulder.

a. Light microscopy
Sections 2018-L1 through 2018-L5 (a series of longitudinal thin sections from a fragment of a gastralium, see Table 1) are roughly rectangular (2C & 3A-3D). The bone growth pattern does not differ substantially in the exterior and interior of the bone, which is expected for gastralia (e.g., Klein, Canoville & Houssaye, 2019).
Sections 2018-L1 through 5 visually resemble sections LJ98B-1 & LJ98B-4 from Yao, Zhang & Tang (2002) in terms of their vascularization pattern, lacunae with preserved canaliculi, and a diversity of vessel fills. Most vascular canals are longitudinally oriented, but they anastomose laterally and through the depth of the section, connecting otherwise parallel longitudinal canals. Osteocyte lacunae are visible in each of these sections and are generally lenticular in shape (Figs. 2B-2D & 3A-3D). They are densely distributed throughout the sections (approximately 650/mm 2 ) except at the bone margin and directly adjacent to vascular canals. Canaliculi are often visible and the sections have a high visual quality, similar to a modern histological section (Hedges, Millard & Pike, 1995). Vessel fills have a variety of textures and colors, from massive and opaque black to granular and translucent orange (Figs. 3A-3D). These fills include translucent, red-orange spheres located throughout the vascular canals.
The spheres (i.e., purported red blood cells) are most abundant at points where two vascular canals connect (Fig. 3D) and near other shapes and textures of vessel fill material, especially translucent orange and red-orange material. The spheres range in color from pale orange to red-orange, overlapping the color range of but never appearing as deeply red as some other vessel fills. In some vessels, it is unclear whether the vessel is densely filled with spheres, filled with blocky grainy material, or a mixture of the two (Fig. 3A). The individual spheres range in size from 6 to 15 µm, but the majority are about 10 µm in diameter. In areas where the spheres are sparse and easily distinguished from one another, their external texture appears bumpy, and appears to be composed of 1-3 µm round subunits. These subunits are best visible when adjusting focus through the depth of the section. Sections 2018-X1 and 2018-X2 (see Table 1; not figured) were cut transversely in sequence from a 3.5 × 1 mm fragment provided by Dr. Yao. The edges of these slices are gently curved, as they are part of the circular transverse section of another gastralium or small rib fragment. The transverse sections reveal dark-filled osteocyte lacunae with distinct canaliculi, as well as primary osteons. The contents of these osteons' central canals are not as easy to discern in this view; the variation in textures and colors of vessel fill is not apparent and neither are the spheres.
The petrified wood thin section, 2018-1 (Table 1, Fig. 4), is a square slice with two zones of carbonate preservation surrounding a zone of silicious preservation. The carbonate zones do not preserve histological detail. Within the silicious zone, however, lines of boxy segments each about 20-40 µm long are visible; these are possibly tracheid cells of the xylem. Some of these segments contain dark round structures that range in size from <5 µm to 20 µm and are similar in appearance to the purported blood cells in the associated bone (Fig. 4, arrowheads).

b. Raman spectroscopy
Raman spectra acquired from across these specimens are characterized by a complex of peaks around 1,000-1,700 cm −1 . Apatite peaks are visible in most of the Raman spectra, with an average position of 962.4 cm −1 (range: 959.8-965 cm −1 ; Table S1). Spectra collected from the plexiglass slide, epoxy, and superglue around our specimen had no peak overlap with our sample spectra (see Fig. S1).
We did not observe peaks for pyrite (FeS 2 ) or its associated iron oxides, which would occur at wavenumbers below 700 cm −1 (De Faria, Venâncio Silva & De Oliveira, 1997;Vogt, Chattopadhyay & Stolz, 1983). We also did not observe any of the five marker peaks for heme compounds, which all occur above 1,340 cm −1 (Asher, 1981;Schweitzer et al., 1997).  Figures 5B and 5E-5F show that vessel fills differ in composition from the surrounding bone.

ii. FEI quanta 600FEG
Using area scans of various exposed vessels on 2018-L1, we found some fills with carbonaceous compositions and others composed of aluminosilicates. The low-chlorine carbon-rich fills were distinguishable from chlorine-rich epoxy infill by their chlorine content. By employing quantitative spot analyses of the aluminosilicate fills, we found aluminum to be, by far, the dominant element, with high magnesium and calcium contents (7.96, 2.85, and 1.25 atomic %, respectively). Based on a magnesium number of 100 (Mg was detected in the spot analyses, but not Fe) and a very low ratio of alkalis to aluminum ((Na+K)/Al = 0.121), we identified the aluminosilicates as kaolinite clay minerals (Page et al., 2008).

d. Time of flight -secondary ion mass spectrometry
The analyzed material is a transversely cut one-mm-thick section of a gastralium fragment in situ in finely laminated mudstone that also contains a round, orange-colored concretion (2018-X3 in Table 1). The laminations of the mudstone dip between the concretion and the bone. This scan is a composite of 21 serial scan segments that together measure 1,500 by 3,500 µm (Figs. 6A-6C). This scan shows a clear difference in composition between the bone and its surrounding matrix, with silicates in the matrix and phosphate in bone, and reveals silicates in the pore spaces of the bone visible at this scale (Fig. 6, compare D to I).
Carbonaceous compounds are concentrated in the bone and at the contact between the laminated matrix and concretion but are not abundant within the concretion or matrix (Fig. 6J). Iron and manganese oxides are evenly dispersed through the matrix and nearby concretion. Aluminum is most densely concentrated in the concretion. Outside of the concretion, Al is densest where Si is also present, except in the matrix densest with Cl. Both 35 Cl and 37 Cl are concentrated in the bone, but also infiltrate the matrix in contact with the bone (Fig. 6F). The concentration of chlorine within the matrix decreases with distance from the bone. Fluorine is abundant throughout the sampled area but is especially concentrated in the bone (Fig. 6H).

DISCUSSION
The bone of Beipiaosaurus inexpectus (IVPP V11559) is highly chemically altered. Apatite in the holotype gastralia has incorporated both fluorite and carbonates, resulting in a Raman peak position at a relatively low wavenumber (ca. 962 cm −1 ; Fig. 3). This peak has a low wave number despite abundant fluorine in the bone (Fig. 6H), which drives apatite peaks towards higher wavenumbers. Raman peak positions for bioapatite shift higher with the addition of fluorine through diagenetic alteration. This results in a range of Raman shifts from 962 cm −1 in modern bone (Alligator mississippiensis) and fossil bone with little alteration, to 966 cm −1 in highly altered fossil bone and in fluorapatite (Thomas et al., 2011). The average position for the apatite peaks in our samples is 962.4 cm −1 , much closer to unaltered enamel apatite (962 cm −1 ) than to fluorapatite (966 cm −1 ). However, TOF-SIMS analysis shows high concentrations of both chlorine and fluorine in Beipiaosaurus bone. There is a gradient in Cl concentration from the bone into surrounding matrix observed with TOF-SIMS (Fig. 6F). The chlorine is likely a component of altered apatite (Keenan, 2016), and the diffusion gradient indicates that exchange between the bone and matrix has occurred, despite the low average wavenumber position of the apatite peak observed with Raman. Notably, carbonate in the bone has the opposite effect of fluoride, shifting the peak towards lower wavenumbers (Li & Pasteris, 2014;Thomas et al., 2011). Many of our apatite peaks are below 962 cm −1 , indicating a considerable carbonate influence on the peak position. Comparative analysis of modern and fossil bone specimens using integrated TOF-SIMS and Raman analysis is needed to further quantify the impact of fluorine on apatite Raman peak position.
The prevalence of diagenetic materials in the vascular canals also suggests that the bones of this fossil are chemically altered. Using EDS and Raman, we observed a high carbon content in the bone and some vessel (pore space) fills but could not distinguish specific carbonaceous compounds. Additionally, we observed carbon compounds both at the permeable contact between the concretion and matrix and in bone using TOF-SIMS analysis (Fig. 6J). This implies that at least some of the carbonaceous materials in the pores of the bone may be exogenous, either carried in by fluids during diagenesis or present due to post-burial microbial occupation of pore space (e.g., vascular canals) and fractures throughout the bone and matrix. We did not identify pyrite in our sample with Raman or EDS, even though 5-20 µm iron oxide pseudomorphs of framboidal pyrite have been observed in siliciclastic and carbonate precipitate laminations in sediments of the Jianshangou Unit in the Sihetun area (Hethke et al., 2013) and in unlaminated black mudstones at the Sihetun locality (Zhang & Sha, 2012). Additionally, pyrite framboids ranging in size from 5-31 µm are found in association with insect and plant fossils of the Jehol biota (Leng & Yang, 2003;Wang et al., 2012).
The complex of Raman peaks at 1,212, 1,296, 1,430, 1,570, and 1,690 cm −1 (Fig. 3) fall in a similar position to peaks identified in previous publications as Raman bands of carbonaceous compounds such as amides (Lee et al., 2017;Morris & Mandair, 2011;Puech et al., 1986) (Fig. 3). A similar peak complex has been observed in other spectra obtained from fossil dinosaur bone of various ages (e.g., Lee et al., 2017;Wang & Yang, 2007). The peaks in this region do not match any of the materials used in processing nor were these materials detected when targeting the specimen, indicating that our spectra are from the sample itself rather than contamination. The peaks do, however, match the expected positions of photoluminescent bands, an artefact in Raman spectroscopy that is produced by rare earth elements in fossil apatite and the surrounding shale matrix (Culka & Jehlička, 2018;Lenz et al., 2015) and we interpret them as such. We recommend that any future Raman spectroscopic analyses of fossil bone carbonaceous compounds or biomolecules target more recently collected specimens (per Pruvost et al., 2007) and employ corrective methods for rare earth element artefact bands (outlined in Culka & Jehlička, 2018).
We propose that the spheres in the vascular canals of B. inexpectus are botryoidal structures consisting of authigenic or diagenetic kaolinite clays (Fig. 7). Although we were unable to specifically target the spheres, the EDS and Raman analyses of the vessel fills recovered no evidence of heme or diagenetic products that would support an identity of fossilized red blood cells.
We did, however, observe similar spheres in a petrographic section of petrified wood (Fig. 4) that was collected with IVPP V11559 and stored with the bone fragments (see Table 1). The presence of these spheres in wood contradicts a blood cell hypothesis but is consistent with authigenic mineralization. The most common alternative hypothesis to other purported erythrocytes with a similar appearance to those studied herein is framboidal pyrite (Kaye, Gaugler & Sawlowicz, 2008;Martill & Unwin, 1997). However, we have found no chemical evidence indicative of pyrite or its oxidative products in our samples, aside from iron and manganese oxides (Mg 2 O − and FeO − ) scattered evenly in low abundance throughout the matrix (detected with TOF-SIMS; see Fig. S2).
The clays detected in IVPP V11559's vascular canals appear to have a variety of habits in addition to spheres. This does not rule out the possibility that former pyrite structures have been replaced by the clays we observed. In fact, kaolinite can form a pseudomorph of pyrite framboids, and the two minerals are often associated (Pollastro, 1981). Many other fossils bear similarly sized spheres that resemble those we observed in this study (Moodie, 1920;Schweitzer & Horner, 1999;Seitz, 1907;Wilby, 1993), though only some of these are hypothesized to be red blood cells. We are inclined to interpret these similar spherical structures as indicators of shared traits of diagenetic history across many fossil localities, reflecting similarities in the taphonomic histories of disparate fossil-bearing environments rather than red blood cell preservation. Individual analysis of these similarly appearing structures would be necessary to ascertain their compositions and thus identities.

CONCLUSIONS
On the basis of comprehensive analyses conducted herein, we suggest that the purported blood cells in the bones of Beipiaosaurus inexpectus (IVPP V11559) are authigenic or diagenetic structures composed of kaolinite clay. Even though we were unable to sample the chemical composition of individual spheres directly, clay minerals are concentrated in the pore spaces of the bone, in the same areas of the thin sections where the spheres are most abundant. This suggests that the spheres and other vessel fills are some of the many habits clays can take. The presence of similar small, reddish spheres in the associated petrified wood also weakens the blood cell hypothesis. One factor limiting the certainty of our results is that these spheres, unmistakable when viewed under a light microscope, were not confidently identifiable with SEM.
To further investigate the nature and origin of these spheres, chemical analyses of extracted or etched specimens may allow better association between particular vessel fill structures and exogenous kerogen or clay. A more recently collected specimen may also be more informative for studying biomolecular preservation in fossil bone from the Yixian Formation.

Field Study Permissions
The following information was supplied relating to field study approvals (i.e., approving body and any reference numbers): Specimen observations and materials were provided by Dr. Xu Xing with permission from the Institute of Vertebrate Paleontology and Paleoanthropology in Beijing for destructive sampling.

Data Availability
The following information was supplied regarding data availability: All Raman spectra as well as the TOF-SIMS spectrum and peak assignment data are available in the Supplemental File.

Supplemental Information
Supplemental information for this article can be found online at http://dx.doi.org/10.7717/ peerj.12651#supplemental-information.