Putative fossil blood cells reinterpreted as diagenetic structures

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; Xu, Tang & Wang, 1999).

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 (⁴⁰Ar/³⁹Ar) 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) (Wang et al., 1999). 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).

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 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 Bi³⁺), as bismuth is not of biological interest and Bi³⁺ is very precise (Touboul et al., 2005).

We calibrated for mass using C, ¹⁸O, O₂, F, Na, C₂, Al, Si, P, Cl, ³⁷Cl, PO, PO₂, and PO₃. 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/mm²) 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 (FeS₂) 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).

c. Energy dispersive X-ray spectrometry

i. Hitachi TM-3000

The most abundant elements in the bone under EDS are phosphorus and oxygen (Figs. 5D, 5F). The signals from filled vessels differ across the thin section. Some, especially in regions noted under light microscopy to have many spheres, are dominated by aluminum and silicon, indicating the possible presence of clay minerals (Figs. 5B, 5E). At other locations, carbon is concentrated in the vessels and abundant in void spaces and fractures in the bone. Due to the differences in microscopy, we could not positively identify a sphere as identified on a light microscope under SEM. Figures 5B and 5E–5F show that vessel fills differ in composition from the surrounding bone.

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 ³⁵Cl and ³⁷Cl 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).