Bone collagen from subtropical Australia is preserved for more than 50,000 years

Ancient protein and ancient DNA (aDNA) analyses have become crucial tools in archaeology, with early work centered on cold Arctic contexts where decay is slowed. Research has expanded to lower latitudes and deeper time, but biomolecular studies in (sub)tropical regions remain limited. Compared to aDNA, proteins generally have higher preservation potential in harsh environments. Protein preservation varies by tissue and depositional environment; enamel and shell proteins are trapped in minerals, whereas bone proteins in mineralized composites are more exposed to diagenetic influences such as pH and hydrology. Preservation thus varies strongly between sites. The oldest peptides come from eggshell (6.5–9 Ma), while the oldest collagen peptides (from high Arctic material) date to 3.4 Ma. Predicting protein decay is more tractable in closed systems like eggshell than in open systems like bone, where pH, water, ions, and organics vary. In bone, alkaline pH preserves apatite but accelerates biomolecular hydrolysis; models predicting collagen survival for radiocarbon dating in temperate settings have been developed, yet cases of preservation exceeding predictions are especially informative. Most prior work on collagen degradation has focused on temperate Eurasia and North America, with few studies from warmer, wetter regions considered less favorable to biomolecule preservation. This study tests those assumptions for subtropical Australia by systematically assessing collagen survival across diverse Australian Quaternary archaeological and paleontological sites and depositional environments. The authors apply thermal age estimates, FTIR, ZooMS, and deamidation metrics to evaluate collagen preservation, compare results across sites and contexts, and explore mechanisms enabling unexpected long-term protein survival in subtropical settings.

Prior research emphasized high-latitude, temperate contexts for biomolecular preservation, reflecting reduced decay rates in cold environments. Ancient proteins often outperform aDNA in preservation, especially in mineral-protected tissues (enamel, shell). In bone, diagenesis is influenced by burial environment (pH, hydrology, ionic composition), leading to variable outcomes. The oldest mineral-bound peptides show deep-time survival in closed systems, while the oldest collagen peptides were reported from high Arctic Pliocene deposits. Models based on temperature-dependent gelatinization and thermal history have been used to predict collagen survival and radiocarbon dating success, but their applicability to open systems and warm climates is uncertain. Studies have documented both enhanced preservation in aridity and accelerated decay under alkaline conditions (e.g., bat guano). Previous collagen preservation studies largely focused on Eurasia and North America; tropical/subtropical contexts have been underrepresented, with some work indicating poor collagen survival in the tropics. Screening methods such as FTIR and NIR have been proposed to prescreen collagen preservation, but cut-offs are debated. Deamidation has been reconsidered as an index of preservation quality rather than a simple relative chronometer.

Materials: Seventeen Australian localities spanning northeastern to southwestern regions and ages from recent to Middle–Late Pleistocene were studied, representing cave, fluvial, lacustrine, and swamp depositional settings. In total, 765 bone fragments were analyzed. Thermal age estimates: Present-day mean annual temperature, mean monthly temperature, and precipitation were compiled from nearby weather stations. Palaeoclimate variables (BIO01, BIO05, BIO06, BIO07, BIO12, altitude) were extracted at 1 kyr intervals up to 120 kyr using the Pastclim R package (v0.9.0) in R 4.0.1, validated against modern station data. Altitude corrections were applied. Two burial-depth models were used: a shallow model assuming near-surface deposition with annual temperature fluctuation, and a deep model assuming buffered, stable temperatures at depth. Thermal age estimates were calculated for each locality under both models. Taphonomic assessment: All bone fragments underwent visual documentation, color recording, and assignment of weathering stages following Behrensmeyer (0–5 scale). FTIR: Minimally destructive ATR-FTIR analyses were performed with an Agilent 4300 Handheld FTIR using a diamond ATR. About 1–2 mg of bone powder was analyzed (2000–650 cm−1, 4 cm−1 resolution, 32 scans). Backgrounds were run between samples. Peak heights for phosphate ν3 (~1035/1010 cm−1), carbonate (~1415 cm−1), Amide I (~1650 cm−1), and Amide II (~1550 cm−1) were measured to compute Amide I/phosphate (Am1/P), Amide II/phosphate (Am2/P), and carbonate/phosphate (C/P) ratios. ZooMS: Collagen was extracted using an acid-soluble approach with blanks. 50–70 mg bone powder was demineralized in 400 μl 0.6 M HCl for 5 days, heated 30 min at 65 °C, and the supernatant ultrafiltered (10 kDa, Microcon). Filters were washed twice with 300 μl 50 mM ammonium bicarbonate (AmBic). Retentates were resuspended in 100 μl 50 mM AmBic and digested on-filter with 1 μl of 0.4 μg μl−1 trypsin for 18 h at 37 °C. Peptides were purified/concentrated using C18 ZipTips. MALDI-TOF-MS (Autoflex Speed LRF, Bruker, smartbeam-II laser) was used; samples were spotted in duplicate with α-cyano-4-hydroxycinnamic acid matrix (10 mg ml−1 in 50% ACN/0.1% TFA). SNAP averaging parameters (C: 4.9384, N: 1.3577, O: 1.4773, S: 0.00417, H: 7.7583) were applied. Spectra were inspected in FlexAnalysis 3.4 and compared to published Australian ZooMS peptide marker databases. Deamidation: Glutamine deamidation indices (%Gln, 0–1) were computed from MALDI spectra using the Q2E R package. A value of 0 indicates fully deamidated peptides; 1 indicates no deamidation. Deamidation was calculated only for samples taxonomically identified to family level or higher; samples without sufficient collagen for identification were assigned 0 for statistical analyses. Statistics: Analyses were performed in R 4.0.1 (RStudio 1.2.1717). Violin plots and Student’s t-tests evaluated differences in FTIR metrics (Am1/P, Am2/P) between ZooMS-successful and -failed samples; significance at p < 0.01 unless noted. Pearson correlations assessed relationships between ZooMS success rates and average deamidation rates. A cut-off was estimated via the third quantile of failed samples. Clustered heatmaps visualized meteorological variation; variables were normalized and clustered using Euclidean distance. Figures were generated with ggplot2.

• Thermal age estimates: Most localities had thermal ages >200,000 years under even the more optimistic deep model; some (e.g., Robert Broom Cave, Tripot Cave) exceeded 1 million years, predicting minimal collagen preservation. Sites with thermal ages <200,000 years (e.g., Beehive, Devil’s Lair, Kudjal Yolgah Cave, Lake Victoria, Lancefield Swamp, Mammoth Cave, Tight Entrance Cave, Yellabidde Cave) were expected to preserve collagen better.
• ZooMS outcomes: 167 samples yielded sufficient collagen for taxonomic identification. ZooMS success rates varied widely by locality. No collagen was recovered at Boodie Cave, Darling Downs, Lake Victoria, Morwell, South Walker Creek, and Strathdownie. Highest success was observed at Millennium Cave (Broken River) at 100% and Devil’s Lair at 85%.
• Deamidation: Deamidation values ranged 0.16–0.99, with site averages 0.17–0.95. Spring Creek showed highest deamidation (lower preservation), McEachern Cave lowest (better preservation). Average deamidation strongly correlated with ZooMS success (Pearson r = 0.97), but not with absolute age, thermal age, or FTIR values. Deamidation is supported as an index of preservation quality rather than a relative chronometer.
• Method comparisons: No correlation between thermal age estimates and ZooMS success; relatively high ZooMS success (>35%) was achieved even at sites with thermal ages predicting no collagen, and low success (<10%) occurred across a broad range of thermal ages (52.3–6923.5 ka thermal ages). Bone weathering stage was a weak predictor of ZooMS success; slightly more successful outcomes occurred in mildly weathered bones, but sample size was limited.
• FTIR predictors: Significant differences between successful and failed samples were found for Am1/P (t = −13.0, df = 191.5, p < 0.001) and Am2/P (t = −12.2, df = 198.2, p < 0.001). Previously suggested Am1/P cut-offs (>0.02 or 0.04) would wrongly exclude many successful samples in this dataset; an Am1/P > 0.05 aligned better but would still exclude 20.6% of successful samples, indicating cut-offs must be applied cautiously.
• Depositional environment effects: Caves consistently showed the highest ZooMS success across regions, particularly protected cave deposits with capping flowstones or distal to large entrances and minimal biotic activity. Open systems with large entrances tended to fail. At open-air sites, carbonate-rich contexts (springs, tufa) performed best. Limestone caves favored collagen survival compared to acidic environments; limited water movement and carbonate saturation may enhance preservation.
• Exceptional subtropical preservation: Tripot Cave (Broken River) exhibited collagen preservation far exceeding chemical predictions based on thermal age, with U-Th constrained Pleistocene ages of associated speleothems indicating the deposits are not Holocene. High C/P ratios at high-success sites suggest carbonate enrichment, though within Tripot Cave, C/P did not correlate specimen-by-specimen with collagen success.
• Mechanistic insights: Proposed mechanisms include the polymer-in-a-box stabilization of collagen fibrils when constrained by bone mineral and potential secondary carbonate or other ions (e.g., Na, F, Sr; possibly Cr(III) analogies) promoting remineralization and stabilization of bone apatite and collagen. Thermal-age models, reliant on temperature and age, failed to predict observed preservation, underscoring missing microenvironmental variables.

The study addresses the central question of whether collagen can persist in subtropical Australia beyond expectations informed by temperate-focused models. Results demonstrate that collagen can indeed survive >50 ka in subtropical contexts and that preservation quality varies more with site-specific depositional conditions than with thermal age alone. The strong correlation between deamidation and ZooMS success indicates deamidation is a robust proxy for collagen preservation and the likelihood of taxonomic identification, even though it is not a temporal proxy. Protected cave environments, carbonate-rich settings, and geochemically favorable microenvironments are key to collagen survival. The lack of correlation between thermal age and ZooMS success, and between macroscopic weathering stages and biochemical preservation, challenges reliance on these predictors. FTIR Amide/phosphate ratios differentiate preservation states statistically, yet universal cut-offs risk excluding viable samples, indicating screening thresholds must be context-sensitive. Tripot Cave exemplifies preservation exceeding thermal predictions, highlighting limitations of current models that omit microenvironmental parameters such as hydrology, pH, ionic composition, and mineral stabilization processes. Mechanistic hypotheses (polymer-in-a-box stabilization, persistent collagen-mineral association, secondary carbonate or alternative ion-mediated remineralization) provide plausible explanations. These findings broaden the geographical and environmental scope for palaeoproteomics, implying that subtropical and other previously deemed unfavorable environments can retain collagen suitable for analysis, and they motivate refinement of preservation models to incorporate geochemical and site-specific variables.

This work demonstrates robust collagen preservation in subtropical Australian bones exceeding 50,000 years, overturning assumptions derived from temperate-focused models and expanding the feasible scope of palaeoproteomics into warmer, wetter regions. The study provides a systematic, multi-proxy assessment (thermal age, FTIR, ZooMS, deamidation) across diverse depositional settings, identifying caves and carbonate-rich environments as most favorable. It reveals strong ties between deamidation and ZooMS success and cautions against rigid FTIR cut-offs. Future research should: (1) refine predictive models of collagen survival by incorporating microenvironmental and geochemical parameters; (2) conduct targeted geochemical characterization of exceptional sites like Tripot Cave to test stabilization mechanisms; (3) expand non-destructive screening (e.g., NIR) to triage samples; and (4) systematically survey (sub)tropical and other underexplored contexts. Given temperature-dependent kinetics suggesting collagen degradation may accelerate more than DNA at high temperatures, targeted aDNA searches in these contexts may be fruitful.

• Thermal age models used rely on temperature histories and age but do not incorporate key microenvironmental variables (hydrology, pH, ionic composition), limiting predictive power.
• Geochemical data for Tripot Cave are currently lacking, preventing identification of specific ions/minerals responsible for stabilization.
• Sample sizes for certain depositional environments (lacustrine and swamp) were limited, potentially biasing comparisons of preservation by environment.
• Bone weathering stage assessments had small sample sizes for statistical inference, and macroscopic appearance was a weak preservation predictor.
• FTIR screening thresholds are context-dependent; applying generic cut-offs risks excluding viable samples.
• Modern temperature datasets (e.g., 1961–1990) and palaeoclimate reconstructions may not capture full, site-specific thermal histories.