Materials

The examined fossils include a fragment of dermal (acellular) bone from the heterostracan Psammosteus, two pectoral fin spines of the acanthodian Gyracanthus, a jaw fragment of the tristichopterid Hyneria, scales of the megalichthyid Megalichthys and porolepiform Holoptychius, and anteroventrolateral plates of the antiarch “placoderm” Bothriolepis and an indeterminate bothriolepid (Table 1 and Tables A and B in S1 File). Most of these fossils were recovered from sandy “red bed” fluviodeltaic siltstones at the famous Red Hill locality and other nearby sites in the Catskill Formation in Pennsylvania, but the indeterminate bothriolepid (NUFV 1586) and Psammosteus specimens were recovered from sandy fluvial overbank siltstones of the Fram and Nordstrand Point formations, respectively, in Canada [25,26].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary of cellular and soft tissue recovery from nine Devonian fish bones.

https://doi.org/10.1371/journal.pone.0335783.t001

The two Canadian Nunavut specimens examined in this study were collected in 2000 (NUFV 1587) and 2004 (NUFV 1586) by Nunavut Paleontological Expedition crews, and the seven specimens from Red Hill and other nearby sites in Pennsylvania were collected between 1994 and 2018 by Academy of Natural Sciences of Philadelphia (ANSP, now “of Drexel University”) crews led by Dr. Ted Daeschler. All necessary permits were obtained for those prior fossil collection campaigns, which complied with all relevant regulations [25–27]. For this study, fossil specimens were selected, in part, to achieve a diverse sampling of Devonian “placoderm”, stem chondrichthyan, and stem osteichthyan clades (Table A in S1 File). As negative controls, we also acquired and demineralized one rock matrix sample from the fossiliferous horizon at one of the source fossil localities in each of the three geologic formations (Catskill, Fram, Nordstrand Point) studied herein. Fragmented fossil specimens and fossil specimens coated with consolidants (e.g., vinac, butvar, or similar adhesives) were avoided during specimen selection to preclude glue casts as a potential source of any demineralization products.

Seven of the nine fossil bone samples examined in this study derive from the Famennian (~ 361 Ma) Catskill Formation (Table 1 and Table A in S1 File). The Catskill Formation primarily consists of grayish-red sandstone, siltstone, and mudstone deposited in prograding deltas [25]. The rock matrix at each of the three Catskill Formation sites from which the fossils examined herein were collected consists of pale red, very-fine grained sandstone [25]. Indeterminate bothriolepid sample NUFV 1586 was collected from an exposure of the Frasnian (~ 378 Ma) Fram Formation in Nunavut territory, Canada. The Fram Formation primarily consists of well-indurated sandstones and poorly-indurated siltstones deposited in meandering fluvial channels and adjacent floodplains [26]. Specimen NUFV 1586 was collected from site NV0403, which is one of several Late Devonian sites in the Fram Formation in the southwestern peninsula of Ellesmere Island. Fossiliferous rocks at this locality, and at nearby site NV2K17 from which a representative rock matrix sample was examined, consist of pale gray-red siltstone interpreted to represent a crevasse splay deposit [26]. The specimen of Psammosteus sp. examined in this study was collected from the Frasnian (~ 375 Ma) Nordstrand Point Formation in Nunavut territory, Canada. The Nordstrand Point Formation consists of green, pale-red, and gray siltstones interbedded with light-gray, very fine to fine sandstones [27]. The entombing rock matrix at site NV2K11, from which the Psammosteus sample was collected, consists of gray siltstone interpreted to have been deposited in a fluvial overbank setting [27].

Methods

Demineralization assays were conducted in a dedicated laminar flow hood within the former Molecular Paleontology Lab at Rowan University, which was dedicated to biomolecular analyses of fossils and in which modern tissue samples were prohibited (this facility has since been repurposed). Incubation in EDTA causes demineralization via chelation of calcium cations from bone, thus leaving behind iron oxides as well as any organic remains not possessing divalent cations [1].

Using a sterilized chisel, small fragments of cortical bone were collected for the demineralization assays. Disposable polypropylene pipettes (1/well) were used for EDTA solution exchanges. Nitrile gloves, a hair net, lab coat, and face mask were worn by all project personnel while conducting the demineralization assays in the Molecular Paleontology Laboratory. EDTA solution was refreshed every 48–72 hours. Most bone samples required ~ 10 weeks to breakdown into loose “debris”, which we collectively refer to in this study as demineralization products (in following of standard practice in this field; e.g., [2]). In contrast, two of the three representative rock matrix samples (specifically those from the NV2K11 and Red Hill localities) were found to be highly resilient to disaggregation in EDTA, so after 14 weeks they were gradually dissolved over a period of one week in 0.2 M HCl (as in [3]).

For imaging of each sample, the tip was cut off a disposable polypropylene pipette, which was then used to transfer two to three drops (~ 50 µl) of concentrated demineralization products to standard glass microscope slides. A glass coverslip was then placed over each slide prior to imaging with the Z-stack feature on a Zeiss AxioLab A1 transmitted light microscope with an integrated AxioCam 506 camera. Slides were initially examined under 10X to find larger, potentially-organic remains (i.e., ‘vessels’), then 20X and 40X objectives were primarily used to search for and image small potentially-organic remains (i.e., ‘osteocytes’ and ‘fibrous matrix’ fragments). In the absence of biochemical verification of their identities, such purported cells and soft tissue microstructures isolated from the fossils will hereafter be referred to using single quotation marks, in following of common practice in similar prior studies [cf., 8,28,29]. In especially productive samples, a simple counting method was used to estimate the approximate number of microstructures present in each slide. This procedure involved tallying the number of each type of microstructure within each field-of-view area under 20X magnification until the entire area of the covered-slipped sample had been searched.

In addition to transmitted light microscopy, demineralization products from select samples were further examined (without gold or carbon coating) using a variable pressure FEI Apreo field-emission gun scanning electron microscope (FEG-SEM) with an Everhart Thornley Detector and an integrated Oxford AZtec energy dispersive x-ray spectroscopy (EDX) probe for elemental analyses. To concentrate and stabilize the isolated demineralization products for imaging by SEM-EDX, they were rinsed three times in Type I (18.2 Ω) water, transferred as above onto ~ 1 cm² silicon chips (as in [8]), and then left to dry for 48 hrs. Structures morphologically consistent with blood vessels required a magnification of 120–1,500X for imaging, whereas those morphologically consistent with osteocytes and fibrous matrix were best viewed around 2,000–6,500X. Select “close up” images were acquired at 10,000–35,000X magnification. SEM images were collected in secondary electron imaging mode at a working distance of ~10 mm, typically using an accelerating voltage of 10 kV; in a few instances, a lower kV setting was used (down to 2 kV). EDX analyses were performed under these operating conditions at an incident angle of 35° to the stage and over reading times ranging from 25–50 s.

Extending the record

Microstructures that are morphologically consistent with ‘osteocytes’, intricately-branching ‘blood vessels’, and mesh-like fragments of ‘fibrous matrix’ have each previously been recovered from demineralization products of bones from a wide variety of extinct vertebrates, including nonavian dinosaurs, mammals, birds, turtles, fish, and marine reptiles dating as far back as the Permian [21]. Here, we have identified this style of preservation in a diverse suite of Late Devonian fish, none of which had previously been examined by demineralization, namely an acanthodian, heterostracan, antiarch “placoderm”, and several basal osteichthyans. These discoveries extend the temporal record of this type of exceptional preservation of cellular and soft-tissue micromorphology [sensu 41] back by approximately 100 million years, supporting the inference [22] that this style of cellular/soft-tissue preservation may be independent of geologic age. Our findings also complement those of Ullmann and Schweitzer [21] who found negligible statistical support for preferential preservation of these cellular and soft tissue microstructures across geologic time.

This study also extends the record of this style of cellular and soft-tissue preservation to several bioapatitic tissue types beyond endochondral/perichondral bone, namely osteodentine (in Gyracanthus), orthodentine (in Psammosteus, Gyracanthus, and Megalichthys), and aspidin (in Psammosteus). Of these three tissue types, only orthodentine had previously been tested for soft-tissue preservation by demineralization, and neither study that previously examined this question [8,42] recovered any microstructures morphologically consistent with an organic source from teeth (comprised primarily of orthodentine, which lacks osteocytes during life). Consequently, our successful recovery of ‘osteocytes’, ‘vessels’, and ‘fibrous matrix’ from diverse bioapatitic tissue types demonstrates that this style of preservation is also independent of bone/dentine microstructure, which in turn implies that a much larger portion of the vertebrate fossil record may contain informative cellular, soft tissue, and (potentially also) biomolecular data than is currently appreciated.

Inferred preservation pathways

We infer that our successful recovery of intact cellular and soft-tissue microstructures from fossils up to ~378 My in age is likely the product of initial postmortem circumstances which promoted natural fixation and partial mineralization, followed by chemical stability [cf., 2,11,17]. First, the fossils we examined were all collected from sandy, and therefore initially-permeable, fluvial sediments, which have been posited to favor soft tissue preservation as they can facilitate drainage of autolytic enzymes away from a carcass [14]. Moreover, the pale to deep maroon color of the “red bed” rocks entombing all but one of the fossils we examined (that from the Nordstrand Point Formation) typically signifies deposition under semi-arid conditions [43] in conjunction with, or followed by, early-diagenetic oxidation [44], the latter of which is known to facilitate preservation of soft tissues through free-radical-induced [20,39] and oxidative crosslinking (i.e., glycoxidation/lipoxidation) of their component biomolecules [42]. Alongside concurrent precipitation of iron oxides and/or hydroxides (e.g., goethite) onto and within cell membranes [i.e., 8,11,39] and tissue surface layers (i.e., the tunica externa of blood vessels), these processes can rapidly confer durable, lasting stability to organic microstructures, helping them to persist over geologic timescales [20]. Based on our observations, we infer that all three of these processes likely acted in concert during early diagenesis of the Devonian bones we examined. This conclusion is consistent with both the sedimentary context of the fossils and our EDX findings that many of the recovered microstructures are enriched in iron and oxygen (especially for ‘osteocytes’ and ‘fibrous matrix’; e.g., Fig 4C and 4I).

It is important to note that occasional permineralization of osteocyte lacunae by quartz in the Hyneria specimen is not surprising, as alternative forms of preservation of cellular features mere microns away from one another has been encountered previously [e.g., 11]. Further, that the permineralizing phase in this case was quartz indicates that the Hyneria specimen experienced greater diagenetic alteration than the others we examined, and that it was exposed to relatively more acidic and silica-rich pore fluids during early diagenesis [cf., 40]. Additionally, the greater abundance of Na, K, and Al in ‘vessels’ observed via EDX maps and point analyses (e.g., Fig 4D–4F) indicates that these microstructures may have been preserved through partial aluminosilicification [cf., 11] rather than permeation (or casting) with goethite. Cumulatively, then, our results are not only consistent with prior hypotheses about where this type of exceptional preservation may be encountered, but also support the use of the following taphonomic “search criteria” for future studies on soft tissue preservation within fossil bones: burial in initially sandy/permeable sediments [14], in oxidizing environments [42], and/or in association with abundant iron at the micro-scale [20,39].

Implications for taphonomy and molecular paleontology

Finally, it is important to note how, based on current knowledge, it is clear that ‘blood vessels’ and ‘osteocytes’ can potentially be recovered from a diverse array of bioapatitic tissues in fossil bones of any extinct vertebrate which possessed them, specifically since the origin of vascularized bone in pteraspidomorphs in the Early Ordovician [33] and cellular (osteocytic) bone in osteostracans in the early Silurian [45] respectively. Taphonomically, this implies that the early-diagenetic chemical reactions facilitating this type of exceptional preservation described above (i.e., free radical-induced crosslinking [20], sulfurization [46], aluminosilicification [11], and/or glycoxidation/lipoxidation [42]) can operate within most, if not all, bioapatitic tissues, and that they were likely operating as soon as these biomineralized tissues evolved.

These insights have implications for the value of the vertebrate fossil record, which paleontologists now recognize to extend to the molecular level. Although it is probable that the average severity of diagenetic alteration generally increases with geologic age [47,48], our findings corroborate those of numerous prior reports [e.g., 1–5,8,14,22,30,32] that have repeatedly found ‘osteocytes’, ‘blood vessels’, and ‘extracellular matrix’ to be remarkably resilient against the long-term rigors of diagenesis. Because each of these microstructures are, and always were, composed of a rich suite of proteins and other biomolecules during life, many of which also have high preservation potential in their own right (e.g., collagen I, actin, tubulin [15,49]), much can potentially be learned from biomolecular analyses of such fossil cells and soft-tissue microstructures. In fact, a growing number of studies have revealed how biochemical (and even proteomic) remnants occasionally identified within fossil ‘osteocytes’, ‘vessels’, and similar microstructures can shed light on diverse aspects of the physiology, phylogenetic relationships, and evolutionary history of extinct vertebrates. For example, collagen I and actin peptide sequences recovered from fossilized ‘extracellular matrix’ and ‘blood vessels’, respectively, have been used to evaluate the phylogenetic relationships of extinct taxa independent of skeletal morphology [13–16]. Similarly, fossil ‘osteocytes’ have been found to be reservoirs of information not only on an individual’s relative age at death [50], but also on bone growth rate [51] and genome size and evolution [e.g., 52,53]. Recognition that ‘osteocytes’ and ‘soft tissues’ can be recovered from fossil bones dating back to the middle Paleozoic implies that each of these types of histologic, paleogenomic, and molecular-phylogenetic analyses could potentially yield informative results from exceptionally-preserved fossils of the earliest vertebrates – an intriguing possibility that would have been considered unfathomable just a few short decades ago.