A heterocyte glycolipid-based calibration to reconstruct past continental climate change

The study addresses the challenge of reconstructing past continental climate change due to the limited applicability of existing organic temperature proxies in lakes. While proxies like UK′37, TEX86, and LDI have been invaluable in marine settings, their use in lakes is restricted: TEX86 is applicable mainly in select large lakes, alkenones for UK′37 are often absent in low-latitude lakes, and terrestrial MBT′/CBT-type proxies are complicated by mixed terrestrial and aquatic sources with differing temperature responses. Heterocyte glycolipids (HGs), unique to N2-fixing heterocytous cyanobacteria common in freshwater systems, exhibit temperature-dependent distributions, particularly between HG26 diols and keto-ols. Prior observations indicated HG composition tracks temperature in cultures and a single-lake time series. This study’s goal is to systematically evaluate HG distributions across lakes spanning broad environmental gradients to develop and calibrate an HG-based proxy (HDI26) for surface water temperature and apply it to reconstruct late Pleistocene–Holocene temperature history from Lake Tanganyika.

Marine organic palaeothermometers (UK′37, TEX86, LDI) have enabled reconstructions of past sea surface temperatures and ocean climate evolution, including deep-time records. Their lacustrine application is constrained: alkenones are often absent in tropical lakes; TEX86 is only suitable in some large systems; MBT′/CBT-derived temperature estimates can be confounded by mixed sources of branched GDGTs. HGs have been identified as biomarkers specific to heterocytous cyanobacteria, with distributions varying by taxonomic groups (e.g., Nostocaceae, Rivulariaceae, Scytonemataceae) and showing systematic temperature responses in culture, with HG26 diols increasing and HG26 keto-ols decreasing at higher growth temperatures. Environmental studies observed similar patterns seasonally and across climate zones, but a comprehensive, multi-lake calibration relating HG distributions to SWT had not been established.

Study design: The authors analyzed HG distributions in surface sediments from 46 East African lakes spanning 615–4504 m a.s.l. with wide gradients in depth (0.3–1470 m), pH (3.8–9.8), and SWT (5.7–27.9 °C). They complemented this with surface sediments or mats from eight globally distributed freshwater systems across tropical (Indonesia), temperate (Germany), subpolar (Argentina), and polar (Antarctica) climates. They also analyzed a sediment core from Lake Tanganyika (NP04-KH04-4A-1K; ~330 m water depth) spanning ~37,000 years to reconstruct SWT over time. Sampling and environmental data: East African lake SWTs were measured within the top 10 cm of the water column during 1996–2010. Additional SWT records were compiled or measured for Lake Towuti, Lakes Klakah and Lading (Indonesia), Lake Constance and Lake Schreventeich (Germany), Laguna Potrok Aike (Argentina), and Antarctic meltwater ponds. The Lake Tanganyika core was sampled at ~5 mm slices, yielding an average temporal resolution of ~1500 years; chronology is based on 14C AMS dating and stratigraphic correlation. Lipid extraction and analysis: Between 0.5 and 2 g of dried sediments were extracted via a modified Bligh and Dyer protocol using MeOH/DCM/phosphate buffer, followed by phase separation, solvent removal, and re-dissolution in n-hexane:2-propanol:H2O (71:27:1). HGs were analyzed by HPLC/ESI-MS (Waters Alliance 2690, Phenomenex Luna NH2 column, 150×2 mm, 3 μm; MRM mode on Micromass Quattro LC) with specified solvent gradients and monitored transitions for HG26–HG32 diols, keto-ols, triols, and keto-diols. Peak integration provided abundances, expressed as fractional abundances relative to total HGs. Proxy definition and computation: The heterocyte diol index (HDI26) was defined as HDI26 = HG26 diol / (HG26 diol + HG26 keto-ol), using, for each sample, the most abundant structural isomer of HG26 diol and of HG26 keto-ol, due to strongest correlations with SWT. Fractional abundances of individual isomers were computed relative to summed HG26 diol and keto-ol isomers. Statistics and calibration: Bivariate and partial correlation analyses were performed (SPSS v27) to assess relationships between HG metrics (HG26 diol, HG26 keto-ol fractional abundances, HDI26) and environmental variables (SWT, elevation, MAAT, BWT, depth, conductivity, pH, DO). Hierarchical clustering (Past v4.03) characterized cyanobacterial community biozones along altitude. Linear regressions related HDI26 to SWT for East African lakes and for the global set; ANCOVA tested slope/intercept differences. Analytical precision of HDI26 was assessed by replicate analyses. Downcore reconstruction: The Lake Tanganyika core’s HG distributions and HDI26 values were converted to SWT using the East African calibration. Temporal trends were compared to global CO2 and regional temperature records (e.g., TEX86 from Lake Tanganyika and Lake Malawi).

• HGs, especially HG26 diols and HG26 keto-ols, are ubiquitous in surface sediments of all studied freshwater environments from tropics to polar regions.
• Along East Africa’s altitudinal/SWT gradient, HG26 diols increase and HG26 keto-ols decrease with increasing SWT; similar trends observed across global sites.
• Strong correlations in East African lakes: HG26 diol (most dominant isomer) vs SWT r = 0.953 (p < 0.0001; n = 42); HG26 keto-ol (most dominant per sample) vs SWT r = −0.979 (p < 0.0001; n = 42); HDI26 vs SWT r = 0.975 (p < 0.0001; n = 42). Partial correlations indicate minor or indirect influence from other environmental variables once SWT is controlled.
• East African calibration: HDI26 = 0.0155 × SWT + 0.5619 (r^2 = 0.95; RMSE = 1.8 °C; n = 42; excluding four >4300 m tarns for correlation assessment). Global calibration (polar to tropical sites): HDI26 = 0.0167 × SWT + 0.5041 (r^2 = 0.99; RMSE = 1.7 °C; n = 8). ANCOVA shows no significant slope difference (p = 0.17) but a significant intercept difference (p = 0.007), consistent with regional/species effects.
• Analytical precision of HDI26 is ±0.002 (~±0.2 °C). Proxy uncertainty dominated by calibration RMSE (~1.7–1.8 °C) without deviation from linearity across ~1–28 °C.
• Seasonality: In temperate/polar lakes, HDI26 reflects summer-biased SWT (e.g., Lake Constance HDI26-based SWT 19.1 °C vs annual mean 12.3 °C), aligning with peak cyanobacterial activity; in equatorial lakes seasonal bias is smaller.
• High-elevation tarns (>4300 m) show anomalously high HDI26 relative to measured SWT, likely due to species-specific lipid responses (dominance of Scytonema) and/or local cell-surface warming during high irradiance, with possible contributions from UV effects or terrestrial cyanobacteria.
• Lake Tanganyika application (~37 ka record): HDI26-inferred SWT ~23.8 °C at ~37 ka BP, cooling to ~22.5 °C during LGM, then warming through deglaciation and early Holocene to a mid-Holocene maximum of 26.5 °C (~4.2 ka BP), followed by ~1.8 °C cooling and late increase to modern 27.2 °C. Net warming from LGM to preindustrial is ~4.1 °C. A ~0.4 °C cooling near ~12.8 ka suggests a Younger Dryas signal.
• Comparison with TEX86: Broad agreement in deglacial warming magnitude and trends; mid–late Holocene differences likely reflect residence depth biases of GDGT producers (oxycline ~100 m) versus surface-dwelling cyanobacteria and changing stratification.

The study establishes that the HDI26, derived from the relative abundance of HG26 diols and keto-ols, is primarily temperature controlled and robustly reflects SWT across diverse freshwater settings. Strong correlations and low analytical error indicate suitability for quantitative reconstructions with uncertainties comparable to other proxies. The proxy’s biological basis—heterocytous cyanobacteria inhabiting the surface mixed layer—provides a well-constrained habitat depth and a direct link to SWT. However, regional offsets (intercept differences) imply species/community composition effects that warrant regional calibrations for highest accuracy. Seasonal biases are expected in temperate/polar lakes, where HDI26 reflects peak summer SWT associated with cyanobacterial blooms; in tropical lakes with weak seasonality, this bias is smaller but can still skew towards warmer conditions. Application to Lake Tanganyika demonstrates that HDI26 captures long-term deglacial to Holocene warming consistent with independent regional records and global CO2 trends, and may detect abrupt events like the Younger Dryas. Differences with TEX86 during the Holocene likely arise from differing source depths and stratification dynamics. Overall, HDI26 adds an independent and widely applicable continental temperature proxy, improving our ability to reconstruct and compare terrestrial climate histories.

This work develops and calibrates a heterocyte glycolipid-based temperature proxy (HDI26) that correlates strongly with surface water temperature across equatorial to polar freshwater environments. The proxy exhibits analytical precision of ±0.2 °C and calibration uncertainties of ~1.7–1.8 °C, comparable to established geochemical thermometers. The first downcore application in Lake Tanganyika reveals ~4.1 °C warming from the Last Glacial Maximum to the onset of the industrial period, consistent with other East African records and global climate forcings, and indicates potential detection of abrupt events such as the Younger Dryas. Given the widespread occurrence and preservation of HGs, including in Cretaceous sediments, HDI26 has potential to reconstruct continental climates across the Cenozoic and earlier. Future work should focus on expanding regional calibrations to account for community composition effects, refining understanding of seasonality and irradiance impacts (e.g., in high-elevation tarns), and conducting additional culture and environmental studies to constrain species-specific lipid responses.

• Regional/species effects lead to intercept differences between calibrations, necessitating regional calibrations for optimal accuracy.
• In temperate to polar lakes, HDI26 reflects a summer-biased SWT rather than annual mean, due to seasonal cyanobacterial productivity.
• High-elevation tarns (>4300 m) show anomalously high HDI26 relative to measured SWT, potentially due to Scytonema dominance, local surface warming under high irradiance, increased UV, or minor terrestrial inputs.
• Calibration uncertainty (RMSE ~1.7–1.8 °C) dominates reconstruction error; while linear across tested temperatures, uncertainty remains.
• Downcore resolution (~1500 years) limits detection of millennial-scale variability in detail; potential community shifts and depositional changes (e.g., lower lake levels, littoral inputs) may influence HG distributions.
• Requires careful selection of dominant isomers per sample for HDI26 computation; variations in isomer dominance across biozones may introduce additional variability.