Lipid extraction and GDGT analysis

Aliquots of the soil samples were prepared for GDGT analysis by freeze-dried at −18 °C in a refrigerator. The soils were ground into less than 200 mesh size. An organic solvent (9:1 dichloromethane: methanol) was added to each sample to extract organic compounds using ultrasonic extraction at least three times. n-Hexane was added to obtain the neutral extracts (at least three times). The neutral extracts were then purified and separated by silica-gel chromatography using hexane/DCM (9:1) and DCM/methanol (1:1) as subsequent eluents to separate into non-polar and polar fractions. The polar fraction containing the GDGTs was dried under nitrogen gas and then re-dissolved in hexane/isopropanol (99:1, v/v). The resulting samples were passed through a 0.45 µm polytetrafluoroethene filter before analysis.

The GDGT were analyzed at Tongji University by HPLC-atmospheric pressure chemical ionization-mass spectrometry (HPLC-APCI-MS), performed with a LC-MS. The 5-methyl and 6-methyl brGDGTs (Supplementary Fig. ¹) were separated with an improved liquid chromatography method. Comparing with the new method with only one silica column, new method uses two coupled silica columns (each 150 mm × 2.1 mm, 1.9 μm; Thermo Finnigan; USA) at 40 °C using n-hexane and EtOAc (84:16 v/v) as elutes for pump A and pump B, respectively (modified from De Jonge²⁴ and Yang³²). Selected ion monitoring (SIM) was used to target specific [M + H]⁺, including those for the 15 brGDGTs ([M + H]⁺ 1050 IIIa III′a, 1048 IIIb III′b, 1046 IIIc III′c, 1036 IIa II′a, 1034 IIb II′b, 1032 IIc II′c, 1022 Ia, 1020 Ib and 1018 Ic). Whilst in old method, only 9 brGDGTs ([M + H]⁺ 1050 IIIa, 1048 IIIb, 1046 IIIc, 1036 IIa, 1034 IIb, 1032 IIc, 1022 Ia, 1020 Ib and 1018 Ic) can be detected.

The relative abundances of individual brGDGTs were calculated according to the integrated peak areas. MBT′_5ME, CBT′ were calculated based on the specific brGDGT group (5- and 6-methyl brGDGTs), where MBT′_5ME = (Ia + Ib + Ic)/(Ia + Ib + Ic + IIa + IIb + IIc + IIIa) and CBT′ = ¹⁰log[(Ic + IIa′ + IIb′ + IIc′ + IIIa′ + IIIb′ + IIIc′)/(Ia + IIa + IIIa)]¹⁸. Replicate HPLC/APCI-MS analysis of samples showed the reproducibility of MBT′_5ME and CBT′ (Supplementary Table ¹) brGDGTs to be ± 0.0019 and ± 0.018, respectively. For old method, the MBT′ and CBT calibrations are: MBT′ = (Ia + Ib + Ic)/(Ia + Ib + Ic + IIa + IIb + IIc + IIIa) and CBT = −¹⁰log[(Ib + IIb)/(Ia + IIa)]²².

The soil transfer functions of MBT′_5ME-MAT and CBT′-pH are: MAT = −8.57 + 31.45 × MBT′_5ME (r² = 0.66, RMSE = 4.8 °C) and pH = 7.15 + 1.59 × CBT′ (r² = 0.85, RMSE = 0.52)¹⁸. The peat transfer functions of MBT′_5ME-MAT_peat and CBT_peat-pH are MAT_peat = 23.05 + 52.18 × MBT′_5ME (r² = 0.76, RMSE = 4.7 °C) and pH = 8.07 + 2.49 × CBT_peat (r² = 0.58, RMSE = 0.8)¹⁷. For the old method, the transfer functions are: MAT = 0.81–5.67 × CBT + 31.0 × MBT′ (r² = 0.59, RMSE = 5.0 °C) and pH = 7.90–1.97 × CBT (r² = 0.70, RMSE = 0.8)²².