Fossilization transforms vertebrate hard tissue proteins into N-heterocyclic polymers

Vertebrate hard tissues consist of mineral crystallites within a proteinaceous scaffold that normally degrades post-mortem. Here we show, however, that decalcification of Mesozoic hard tissues preserved in oxidative settings releases brownish stained extracellular matrix, cells, blood vessels, and nerve projections. Raman Microspectroscopy shows that these fossil soft tissues are a product of diagenetic transformation to Advanced Glycoxidation and Lipoxidation End Products, a class of N-heterocyclic polymers generated via oxidative crosslinking of proteinaceous scaffolds. Hard tissues in reducing environments, in contrast, lack soft tissue preservation. Comparison of fossil soft tissues with modern and experimentally matured samples reveals how proteinaceous tissues undergo diagenesis and explains biases in their preservation in the rock record. This provides a target, focused on oxidative depositional environments, for finding cellular-to-subcellular soft tissue morphology in fossils and validates its use in phylogenetic and other evolutionary studies.


Supplementary Notes 1-4 1. Supplementary Note 1: Decalcification process
To confirm that the decalcification routine does not affect the chemistry of the soft tissues we investigated how the incubation of hard tissue with hydrochloric acid solution affects molecular preservation or AGE/ALE generation using Allosaurus bones to obtain high-resolution Raman spectra of in situ undecalcified and extracted decalcified samples.
If incubation in hydrochloric acid solution induced crosslinks in the samples, the AGE/ALE band would be present in the decalcified sample but absent in the undecalcified sample. However, the AGE/ALE band heights were identical in the normalized spectra of undecalcified and decalcified samples ( Supplementary Fig. 4).
Peak labels shown in Supplementary Fig. 4 are obtained through the automatic peak label function (threshold 5.00 %, prominence 3) in SpectraGryph 1.2 spectroscopic software.
The AGE/Amide I ratios were also identical, suggesting that hydrochloric acid incubation has no effect on the extracted proteins/peptides and their crosslinks, perhaps predicted by the fact that protein crosslinking is favored under alkaline, not acid, conditions 1,2 . The only Raman band induced by hydrochloric acid incubation is a C-Cl band at Raman shift 750-790 cm -1 . Generally, decalcification results in greater resolution of the Raman bands in a soft tissue sample.

An additional implication of the comparison between undecalcified and decalcified
Allosaurus bone is the prominence of organic AGE/ALE and amide I bands in the undecalcified spectrum. Such an analysis offers a nondestructive assessment of proteinaceous/peptide soft tissue preservation in a fossil hard tissue. spectra in the RRuff database 14 . We focused on pyrite, hematite and related oxo-hydroxoiron compounds (e.g., goethite and fluorapatite, which occurs in different colors due to trace metal incorporation and crystal lattice defects), to investigate hard tissue discoloration due to staining minerals ( Supplementary Fig. 7). Peak labels shown in Supplementary Fig. 7 were obtained through the automatic peak label function (threshold 5.00 %, prominence 3) in SpectraGryph 1.2 spectroscopic software. Pyrite was absent in all modern, matured, and fossil soft-tissue yielding samples (thus, samples from oxidative environments). Fluorapatite, the diagenetic product of hydroxyapatite, was present only in the undecalcified extant, matured, and fossil samples from both reducing and oxidative settings, and there were no significant band shifts indicating trace metal substitutions in the crystal lattice. Hematite and related iron compounds were found as traces in the undecalcified fossil samples from oxidative settings but were absent in the decalcified samples from oxidative settings yielding soft tissues. The extracted fossil soft tissues exhibited a significant dark brown stain but yielded only organic Raman signatures and no evidence of staining mineral compounds. The correlation between discoloration intensity and AGE/ALE concentration ( Fig. 1) indicates that the observed discoloration in matured and fossil soft tissues is organic in origin.

Glues, epoxy resins, and consolidant
In order to test for contaminants which might produce soft tissue morphologies, or account for the observed amide bands, we analyzed Araldite epoxy resin, polyacrylamide glue, and acrylamide-based consolidant, all compounds used in the conservation of fossil bone. Polymeric acrylamide-based glues or consolidants are the only compounds used in vertebrate fossil conservation that might result in Raman peaks that could be confused with proteins (since they are polyacrylamides). The spectra obtained from glue, epoxy, and consolidant ( Supplementary Fig. 7), however, are very different. We also subjected all glue and epoxy samples to the same hydrochloride acid incubation as the tissue samples, and likewise obtained different signatures. We used these spectra to determine whether consolidant, glues, or epoxy were applied to our samples, complementing visual observations.
Raman spectra of these contaminants included many more narrow bands than the soft-tissue samples ( Supplementary Fig. 7). We identified three marker bands at 630, 830, and 1111 cm -1 which we used to identify glue, epoxy, or consolidant contaminated samples in our data set. This revealed glue in two samples, the Maiasaura bone and the ichthyosaur bone, both from reducing environments, which did not produce any soft tissues after decalcification, only small, transparent glue sheets of random 3-dimensional shape. The spectra from these samples included peaks for all three types of glue. We retained these "contaminated" specimens in our data set to ensure the reliability of our screening for consolidants (see Supplementary Fig. 7). None of our extracted soft tissues showed evidence of glue, epoxy, or consolidant. Thus, we can rule out the possibility that the extracted soft tissues represent polyacrylamide-based glue "casts" of voids such as osteocyte lacunae or Haversian canals. d.

Humic acid
The archeological literature provides many examples of bones buried for less than 100,000 years that exhibit a black stain due to humic acid impregnation 15,16 . Humic acids, as well as humic colloids, are abundant in coals, peats, and especially soils. They are polymeric carboxylic, carbonylic, and phenolic compounds which exhibit a brownish stain and have been extensively analysed using Raman spectroscopy 17  e.

Iron oxide/hydroxide as soft tissue composites
It has been suggested that iron oxides and hydroxides "coat" voids produced by soft tissue decay in fossil hard tissues 18 . This was suggested based on comparison of SEM images of a putative extracted, fossil vascular canal surface, and morphologically similar structures replicated by pyrite framboids 18 . Iron hydroxyoxides occur as colloids in soils and may percolate fossil bones. However, biofilm-associated iron oxides filling voids in hard tissue would not explain amide signals that are not associated with a biofilm (see g.

Systemic contamination
We analyzed extracted modern, matured, and fossil soft tissues, and undecalcified in situ samples ( Supplementary Fig. 9), as well as glue, consolidant, epoxy resin, and sediment samples. The restriction of proteinaceous/peptide AGEs/ALEs to matured modern soft tissues and fossil soft tissues from oxidative environments is clear evidence that our samples were not contaminated by a laboratory source of crosslinked proteins. h.

Sediment contamination
Dark sediment from reducing environments generally contains significant organics which are a potential source of contamination of contained fossil hard tissues. To test this possibility in our samples, we subjected sediment samples to in situ Raman Spectroscopy and compared the spectra with those of extracted soft tissues ( Supplementary Fig. 8).
Peak labels shown in the Supplementary Fig. 8

Supplementary Note 3: Samples from reducing environments
Undecalcified fossil hard tissues from reducing environments 1 , which did not produce soft tissues during decalcification, did not yield amide signals and do not contain proteins/peptides. Nevertheless, in situ analyses of some undecalcified fossil hard tissues (Supplementary Fig. 9) revealed lipids, which may be preserved via oxidative crosslinking 27

Supplementary Note 4: Chemospace PCA tests chemical similarity
We quantified the similarity among the spectra obtained for the fresh, matured and fossil material using Principal Component Analyses in PAST 3 (Fig. 4). We also included control samples: fossils preserved in reducing environments, and chemical compounds such as glues and epoxy, which may have been introduced during preparation and conservation ( Supplementary Figs. 7, 8). As previously proposed 26,27 , characteristic peaks can be extracted from the spectra to perform PCA. We selected 15 Raman shift ranges (Supplementary Tab. 7) representing the molecular composition of tissues using the software SpectraGryph 1.2. Before extraction of the peaks, spectra were subjected to baseline correction and normalization. We extracted normalized intensity values at these band positions and compiled them in a data matrix to be used for ChemoSpace analytics.
The data show a normal distribution. Overall, control samples do not overlap with modern, matured or fossil material (Fig. 3a).
The modern samples occupy a separate area of the chemospace with no overlap with fossils, whereas the artificially matured samples fall in an intermediate region (Fig. 3b).