1 Introduction

Glycerol dialkyl glycerol tetraethers (GDGTs), which were initially believed to be synthesized mainly by extremophilic archaea, have their carbon skeletons connected with glycerol by four ether bonds. The monolayer structure of the ether bond is highly stable compared with traditional ester bonds (Sinninghe Damsté et al., 2007). Other studies have demonstrated that these ether membrane lipids are widespread in various environments and are not only in harsh environments (Schouten et al., 2013). Two main kinds of GDGTs groups have been distinguished according to the difference in carbon skeleton structure: iGDGTs and branched GDGTs. The former is composed of isoprenoid carbon skeleton (biphytane with 0 – 4 cyclopentane moieties), while the latter contains two n-alkyl skeletons containing up to four methyl branches and one cyclopentyl moiety substitute (Hoefs et al., 1997; Sinninghe Damsté et al., 2000; De Jonge et al., 2014). Generally, iGDGTs are believed to be synthesized by archaea, while brGDGTs are produced from soil bacteria, although there is some debate regarding possible additional sources (DeLong et al., 1998; Macalady et al., 2004; Schouten et al., 2007). In addition to their differences in carbon skeleton, the two GDGTs also feature differences in the spatial configuration of the head glycerol: 2, 3-di-O-alkylsn-glycerol has been found in iGDGTs, while 1, 2-di-O- alkyl-sn-glycerol in brGDGTs (Becker et al., 2013). The differences in the structure indicate both the biological diversity and complexity of the GDGTs.

Studies have shown the GDGTs composition varies systematically in response to changes in environmental conditions. Based on these, many climate proxies have been established, for example, seawater temperature index – TEX₈₆ (Schouten et al., 2002), land source input index – BIT (Hopmans et al., 2004), methylation index of branched tetraethers – MBT (Weijers et al., 2007), cyclization ratio of branched tetraethers – CBT (Weijers et al., 2007), and methane indictor – MI (Zhang et al., 2011). Many studies have shown that brGDGTs are mainly produced by anaerobic soil bacteria and that a good correlation exists between MBT/CBT and mean annual air temperature (MAAT), as well as between MBT/CBT and soil pH (Weijers et al., 2007). These indices play increasingly important roles in terrestrial paleoenvironment reconstruction and contribute to GDGTs being the most promising biomarker.

However, further studies have shown that the GDGTs based indices, especially brGDGTs, have a particular feature: investigators found a clearly different distribution feature of brGDGTs in aquatic sediment and catchment soil, such as in the Norwegian strait (Peterse et al., 2009) and the lower Yangtze River – East China shelf area (Zhu et al., 2011). These phenomena have also been found in some lakes and coastal sediments (Tierney and Russell, 2009; Blaga et al., 2010; Tierney et al., 2010, 2012; Loomis et al., 2011; Wang et al., 2012; Buckles et al., 2014; Zell et al., 2014; De Jonge et al., 2015; Weber et al., 2015). Moreover, in the past 180 kyr years, the rebuilt MAAT was significantly lower than the actual observed value in a Scotland lake, and the MAAT in the glacial period was significantly higher than that in the interglacial period in the South China Sea; these imply the constraints of the proxy (Tyler et al., 2010; Dong et al., 2015).

Therefore, the hypothesis that brGDGTs are derived from an in situ aquatic column or sediment has been proposed. An isotopic study will be a possible solution to confirm this hypothesis (Schouten et al., 2013). Owing to the high molecular weight and strong polarity of the GDGTs, their isotopic composition cannot be detected directly by gas chromatography-isotope ratio mass spectrometry (GC-IRMS). The chemical degradation of the larger GDGTs molecules into GC-amenable detectable molecules followed by GC-IRMS, and offline isolation by preparative liquid chromatography followed by elemental analysis-isotope ratio mass spectrometry (EA-IRMS) are the traditional methods for the isotope study of GDGTs. Thus far, most studies on stable carbon isotope of GDGTs employed the chemical degradation methods to detect the isotope; for example, Weijers et al. (2010) used semi-preparative high-performance liquid chromatography (HPLC) to obtain purified GDGTs; then, released the alkyl moieties by HI/LiAlH₄ treatment. In a previous study, the stable carbon isotope relationship between total organic matter and brGDGTs was attributed to a heterotrophic mechanism (Weijers et al., 2010). Research by Lu et al. (2018) on the Lantian paleosol also showed similar results by BBr₃/LiEt₃BH treatment. In addition to the chemical degradation method, high-performance liquid chromatography-isotope ratio mass spectrometry (HPLC-IRMS), developed in the past years, can directly detect the stable carbon isotope of individual GDGTs (Pearson et al., 2016; Colcord et al., 2017). Compared with the HPLC-IRMS method, GC-IRMS is equipped with well-established technology. Besides, most published studies on GDGTs stable carbon isotope are based on the traditional method. The development of a similar method has made results comparison feasible, so that the chemical degradation method is still relevant.

At present, continuous and long-timescale paleoenvironment research requires the processing of numerous samples. In addition, the mixture compounds in the chemical degradation and the large volume material required for the EA-IRMS limit related research. Regarding the batch analysis of samples, semi-preparative HPLC is inappropriate as its mechanism allows the analysis of only one sample at a time. Thus, developing an operational laboratory manipulation that would allow the separation of many samples at the same time is necessary.

In this study, a new method based on silica column chromatography was built to isolate the GDGTs into three groups (iGDGTs, brGDGTs, and other membrane lipids). Ether cleavage was further applied to estimate the purification of the separation. Meanwhile, the traditional semipreparative HPLC was also employed for comparison.

2 Materials and Methods

The sample used in this research was peat soil collected from the surface to the 5 cm below surface in Xinjiang Province, China. The samples were stored in dry ice in the field and in a refrigerator at −40℃ in the laboratory.

2.1 Lipid Extraction

The sample was freeze-dried, homogenized, and then extracted ultrasonically according to the procedure modified from Li et al. (2013). Each sample (about 5 g) was first spiked with the internal standard of C₄₆ tetraether (Huguet et al., 2006) and then extracted in sequence using methanol (MeOH), dichloromethane (DCM)/MeOH (1:1), and DCM, twice for each extraction solvent. The supernatant liquid was collected after centrifugation for 15 min. The extracted concentrated product was dissolved in 6% KOH in MeOH under a water bath of 50℃ for 12 h for saponification. Then, neutral components were recovered by nhexane. The neutral lipids were separated into three subfractions for the following steps: GDGTs analysis, in which samples must be filtered through a 0.45 mm-pore-size, 4 mm-diameter PTFE filter, and then HPLC analysis.

2.2 GDGTs Fractionated by Silica Gel
Column Chromatography

The silica gel (100 – 200 mesh) was successively washed with DCM and n-hexane for three times. Then, it was dried at 120℃ for 12 h, deactivated under room temperature for 12 h, and soaked in n-hexane solution for storage. About 2 g silica gel was loaded on a glass column with a diameter of 4 mm and length of 8 cm. The top subfraction of the total neutral lipids was dried under N₂ stream and dissolved in a small aliquot of n-hexane; it was then applied on the top of the silica gel column. Five fractions were successively collected using 20 mL of the following solvent systems: n-hexane (aliphatics), DCM (ketones, alkanols, and sterols), n-hexane/isopropyl alcohol (IPA) (96: 3) (iGDGTs), n-hexane/IPA (96:4) (brGDGTs), and n-hexane/IPA (9:1) (other ether membrane lipids). The first two fractions were aliphatics, alkanols, and sterols. The last three fractions were dried and dissolved with n-hexane/EtOAc (84:16) for HPLC in conjunction with atmospheric pressure chemical ionization mass spectrometry (HPLC-APCI-MS).

2.3 GDGTs Fractionated by Semi-Preparative HPLC

One neutral lipids subfraction was separated into three fractions by 20 mL n-hexane (F1, aliphatics), 20 mL DCM (F2, ketones, alkanols, and sterols), and 20 mL MeOH (F3, total GDGTs) by silica gel chromatography. The third part was collected for further processing. After blow-drying under N₂, F3 was dissolved in 50 μL n-hexane/EtOAc (84: 16) for the semi-preparative HPLC-APCI-MS (Agilent 6130); the instrument was equipped with a semi-preparative NH₂ column (Econosphere, 10 mm×250 mm×5 μm; Alltech Prevail, USA) with 50 μL injection volume. The column was eluted isocratically with 10% EtOAc in n-hexane for 90 min, followed by a linear gradient to 50% EtOAc in 100 min, and then back to 10% EtOAc in 120 min, with a flow rate of 1 mL min⁻¹. The system was cleaned with 10% EtOAc and re-equilibrated to initial conditions (10 min each). The target fraction was collected under the Agilent G1364C collector, depending on the time window detected by the mass spectrometer (Agilent 6130), which corresponded to 27 – 45 min for iGDGTs, 50 – 80 min for br- GDGTs, and 80 – 105 min for other ether membrane lipids.

2.4 Ether Cleavage

The ether bond cleavage procedure of the GDGTs followed the methods modified from Weijers et al. (2010) and Lu et al. (2018). After these methods were carried out, aliquots of the separated GDGTs groups were subjected to excess 56% HI in H₂O for 4 h at 150℃ to cleave the ether bonds linking the alkyl skeleton and the glycerol side chain. The resulting alkyl iodides were extracted by n-hexane for four times and then reduced to corresponding hydrocarbons by LiEt₃BH for 3 h at 75℃ in tetrahydrofuran and cooled to room temperature. Deionized water was used to quench the excess oxidizers and reductant in the solvent (five times). Finally, the products were extracted with n-hexane and purified using a silica gel column, eluting with n-hexane. The extract was dried under a gentle N₂ stream for gas chromatography-mass spectrometry (GC-MS) analysis.

2.5 HPLC-APCI-MS for GDGT Analysis
and Quantification

The GDGT analysis was conducted following a method described elsewhere (Hopmans et al., 2004; Dong et al., 2015). The analysis using HPLC-APCI-MS included the total GDGTs extracted from the peat, three separated fractions by silica column chromatography, and semi-preparative HPLC. Samples were dissolved in 500 μL n-hexane/EtOAc (84:16) and 20 μL of the sample was injected. Separation was achieved using an Agilent 1200 liquid chromatograph equipped with an automatic injector and a Prevail Cyano column (2.1 mm×150 mm×3 μm; Alltech, Deerfield, IL) maintained at 40℃. The GDGTs were first eluted isocratically with n-hexane and EtOAc as follows: 90% n-hexane and 10% EtOAc for 90 min, followed by a linear gradient to 50% EtOAc in 100 min, and back to 10% EtOAc in 100 min (1 mL min⁻¹) till 120 min. The system was cleaned with 10% EtOAc and re-equilibrated to initial conditions (10 min each). The flow rate was 0.2 mL min⁻¹. After each analysis, the column was cleaned by back-flushing with n-hexane/EtOAc (84:16) at 0.2 mL min⁻¹ for 10 min. Detection was performed on an Agilent 6460 triple-quadrupole mass spectrometer with an atomspheric pressure chemical ionization (APCI) ion source. The conditions for APCI mass spectroscopy were as follows: nebulizer pressure, 60 psi; vaporizer temperature, 400℃; drying gas (N₂) flow of 5 L min⁻¹; temperature, 200℃; capillary voltage, −3.5 kV; corona, 5 μA (3.2 kV). The singleion monitoring mode was set to detect the five isoprenoid GDGTs and nine branched GDGTs (m/z 1302, 1300, 1298, 1296, 1292, 1050, 1048, 1046, 1032, 1034, 1036, 1018, 1020, 1022) and the C₄₆ GDGT internal standard (m/z 744), with a dwell time of 237 ms per ion. The GDGTs were quantified by the integration of peak areas of both target GDGTs and a known concentration of internal standard. Here, the liquid chromatography–mass spectroscopy detection limit for the C₄₆ GDGT was 30 pg with a signal to noise ratio of > 10.

2.6 GC-MS Detection

The alkyl moieties released from the GDGTs were detected on the GC-MS instrument (TSQ8000 Thermo Quest) to confirm the structural features. The capillary column (CP Sil-5CB, 30 m×0.32 mm×0.25 μm, J&W) was equipped in a Trace GC system. Helium was the carrier gas with a flow rate of 1 mL min⁻¹. For the GC procedure, the temperature was first 90℃, then increased by 15℃ min⁻¹ to 170℃ (2 min), increased by 10℃ min⁻¹ to 250℃, maintained at this temperature for 2 min, increased by 5℃ min⁻¹ to 300℃, and then maintained at this temperature for 25 min. The TSQ 8000 Evo MS operated with an electron ionization ion source at 70 ev and a scan range of m/z 50 – 600 with a 0.24 s cycle. The temperatures of transfer tube and ion source were both 280℃.

3 Results and Discussion 3.1 Separation Effect of GDGTs by Silica Column Chromatography and Semi-Preparative HPLC

Before separation, GDGTs compositions in the studied peat sample were detected by HPLC-APCI-MS. Five isoprenoid GDGTs (Fig. 1, iGDGTs 0 – 4 and relevant composed alkyl moieties e – h), nine branched GDGTs (Fig. 1, Ia, Ib, Ic, IIa, IIb, IIc, IIIa, IIIb, IIIc and relevant composed alkyl moieties a – d), and other ether membrane lipids (possible degradation products from GDGTs, i.e., glycerol monobiphytanyl glycerol diether, glycerol monobiphytanyl monoether, glycerol dibiphytanol diether (GDD) in Fig. 1; Liu et al., 2012, 2018; Becker et al., 2013) are shown in Fig. 2. The obtained GDGTs are divided into three kingdoms: iGDGTs, brGDGTs, and other ether membrane lipids. The brGDGTs exhibited a higher proportion in the peat, containing about 80% of the total GDGTs, a finding similar to that of a previous study on peat sediments (Weijers et al., 2011). Two methods were applied for the purification of GDGTs in this study: silica gel column chromatography fractionation and semi-preparative HPLC fractionation. The separation effects of these two methods on the peat sample are demonstrated in Fig. 2. Generally, both approaches could achieve the fundamental purpose of isolating the total GDGTs into three groups. The semi-preparative HPLC based on the retention time exhibited a good separation of the three kinds GDGTs owing to the high sensitivity of HPLC.

Moreover, silica gel column chromatography illustrated a similar isolation effect. The elution of the silica gel column with solvents of increasing polarity allowed the sequenced withdrawal of iGDGTs, brGDGTs, and other membrane lipids. The results of the elution behavior of the extracted lipids under different solvent mixtures of n-hexane and IPA were tested. The ratios of 96:3 for iGDGTs, 96:4 for brGDGTs, and 9:1 for other ether membrane lipids were the appropriate proportions for the GDGTs separation (Fig. 2). However, because of the extremely close polarities and chemical properties between the three kinds of GDGTs, avoiding crossover in the chromatography is quite difficult. Although the isolation fractionation of GDGTs was acceptable, little tiny brGDGTs (< 3%) were still found in the iGDGTs fraction, and little tiny GDDs (< 4%) in the brGDGTs fraction.

Furthermore, GDGTs with different structures would not perform similarly under the fixed mobile phase. While the solvent collection in the semi-preparative HPLC depended on the specific time window, the more sensitivity and accuracy of the HPLC system guarantees a possibility of obtaining a relatively stable and high yield.

3.2 Recovery Comparison

To evaluate the efficiency of the silica column chromatography, nine brGDGTs and GDGT-0 were used as representatives in the following calculation regarding the studied peat sample.

The brGDGTs concentrations in the peat were generally higher than those in the soil and ocean sediment, and the difference within different structural brGDGTs was huge. Taking the Xinjiang peat as an example, some concentrations could be up to 15000 ng g⁻¹ (IIa, m/z 1036) while some were only 3 ng g⁻¹ (IIIc, m/z 1046). The brGDGTs concentrations in the total GDGTs before fractionation were considered as the original content. The concentration results obtained from silica gel column chromatography and semi-preparative HPLC are illustrated in Fig. 3. Comparing the collected brGDGTs with the total GDGTs, it is clear losses occurred in the collected brGDGTs by both methods.

The recoveries of each brGDGT are shown in Fig. 4. Original compounds before separation were used for the efficiency assessment during the overall analytical protocol. The performances of nine brGDGTs in the semi-preparative HPLC were averaged at 70%, with the highest value of 85% for brGDGTs IIIa and the lowest (55%) for brGDGTs IIa. Moreover, the brGDGTs recovery from silica gel column chromatography was slightly lower than that from semi-preparative HPLC, with the highest yield of 70% for brGDGTs IIIc and the lowest of < 20% for brGDGTs IIIb. The specific recoveries of each brGDGT are listed in the following Table 1.

Although the iGDGTs were less than the brGDGTs in the sample, we analyzed the performance of GDGT-0 as a representative. Its 69% recovery using the silica gel method and 81% using the semi-preparative HPLC column were similar to the recoveries of brGDGTs using the corresponding methods.

Partial irreversible adsorption of GDGTs might exist in both the semi-preparative HPLC NH₂ column and silica gel column. The small injection volume and high sensitivity of the semi-preparative NH₂ column content in semipreparative liquid chromatography were possibly the main causes of the relatively high recovery. Probably, the absorption characteristics of silica gel and the multiple transfer steps during the silica gel experiment would cause a high mass loss. This could account for the lower yield obtained using the silica gel method.

Although both methods could separate iGDGTs, br- GDGTs, and other membrane lipids from the total GDGTs, different recoveries were obtained for various brGDGTs. The recovery exhibited weak relationship with the concentration; similar recoveries were found in low concentrations of brGDGTs Ib, Ic, IIb, and IIc. However, explaining the special low recovery of brGDGTs IIIb in silica gel column chromatography is difficult.

3.3 Purification Comparison

The released alkyl moieties from GDGTs after ether cleavage were detected to assess the purification of the fractionated GDGTs from silica column chromatography and semi-preparative HPLC. Theoretically, four kinds of iGDGTs-derived isoprenoid hydrocarbons and four kinds of brGDGTs-derived branched alkanes should be released. In addition, the GDGTs purification was essential to the isotopic detection.

Three isolated GDGTs fractionations from the silica gel column chromatography and the semi-preparative HPLC, as well as the corresponding GC chromatographs of hydrocarbons released by ether cleavage, are shown in Figs. 5 and 6, respectively. In terms of products derived from iGDGTs collected by semi-preparative HPLC, only isoprenoid hydrocarbon 'e' derived from iGDGT-0 and iGDGT-1 was detected. The mass spectra of previous published data confirmed the structure of the compound (Fig. 7) (Sinninghe Damsté et al., 2000; Weijers et al., 2010; De Jonge et al., 2014). Other cleavage products derived from iGDGTs, i.e., 'f', 'g', and 'h' were not detected for their low contents. Additionally, four extra peaks existed in the spectra of the ether cleavage products. These materials were not compositionally related to GDGTs based on the GC-MS analysis. They may co-elute during the HPLC separation.

Coincidentally or not, indistinguishable peak mixtures also existed when the silica gel method was used. These mixtures may be caused by the insufficiency of the silica column chromatography, whose separation mainly depended on the polarity differences among organic matters. After the n-hexane and DCM, n-hexane/IPA 96:3 eluted not only iGDGTs but also many other compounds, and with similar polarities; these were all exhibited on the gas chromatogram and thus made the chromatogram complex. The absence of 'f' in the semi-preparative HPLC could be accounted for the instability of ether cleavage reaction, which happened accidentally.

For the ether cleavage products of the second groups, the gas chromatograms were relatively pure (Figs. 5 and 6). Besides the expected alkyl moieties 'a' and 'b' from br- GDGTs, an extra hydrocarbon 'e' was also present in ether cleavage production in the silica column collection. This was considered as a released product from the iGDDs during the silica column collection, as shown in Fig. 2. Compared with the iGDGTs and their released alkyl moieties, the brGDGTs were much more concise, with little impurity, and the hydrocarbon was consistent with their parent compounds shown in the liquid chromatography.

Typical alkyl moieties 'a' and 'b' were present in the third fraction after HI/LiEt₃BH treatment (Figs. 5 and 6). Generally, the hydrocarbons 'a' and 'b' were believed to be from brGDGTs. Their presence in the third fraction implies that this fraction mainly comprised the natural degradation products of iGDGTs and brGDGTs, such as glycerol monobiphytanyl glycerol diether, GDD, and glycerol monobiphytanol monoether (structures shown in Fig. 1). Same carbon skeleton as that in the GDGTs produced the same hydrocarbon 'a', and 'b' after ether cleavage from these GDDs or other structural ether lipids. Hydrocarbon 'e' was also detected in the third fraction of the silica gel column. Although iGDGTs were less than the brGDGTs in the peat sample, iGDDs might still have existed in the third groups. In addition, the yields of the ether cleavage reactions of the iGDGTs and brGDGTs were different: on average, 70% for iGDGTs and 40% for brGDGTs. These could account for the high abundance of hydrocarbon 'e' in the third fraction.

The ether cleavage result indicates the GDGTs group separation was necessary, especially for the isotope study of GDGTs, because the same cleavage products were found in the third groups, especially the products 'a' and 'b'. Both iGDGTs fractionation and GDD fractionation could produce these hydrocarbons. Without GDGTs group separation, the isotopes of iGDGTs and brGDGTs would be mixture information. Remarkably, co-elution happened even in the HPLC separation, especially in the iGDGTs fractionation.

Although the semi-preparative HPLC exhibited a slightly better separation performance, its special equipment requirement and time consumption makes it impractical for the large-number batch experiments in paleoclimatic change studies. Considering the merits of rapidness, easy laboratory operation, no special equipment requirement, and acceptable separation results, silica column chromatography for the isolation of different groups of GDGTs is practicable. Although the semi-preparative HPLC featured automatic operation, repetitive transfer processes would take up too much time. In addition, the collection process was slow due to the limitation by the mobile phase. Therefore, the semi-preparative HPLC is more appropriate in highprecision analysis, i.e., specific compound study, while the silica column chromatography is more suitable for batch samples, i.e., paleoclimate and paleoenvironment studies.

4 Conclusions

In this study, GDGTs separation effect by silica column chromatography was compared with that by the semi-preparative HPLC method using a peat sample.

Using the silica column chromatography, 60% was obtained, a little lower than that obtained by the semi-preparative HPLC (70%). The alkyl moieties released after ether cleavage from the fractionated GDGTs indicated the separation efficiency. Some compositions unrelated to GDGTs were co-eluted in the iGDGTs collected by silica column chromatography. Meanwhile, the finding of same branched alkanes cleavage products (i.e., 'a', 'b') in the third groups imply caution should be taken in the isotopic study of the GDGTs in the sedimentary records. Overall, the HI/LiEt₃BH treatment results indicated similar purification effects between the silica column chromatography method and the semi-preparative HPLC approach. The merits of time saving, easy operation, and low cost make the silica column chromatography method a better choice for batch experiments in paleoclimate reconstruction studies.

Despite the good performance of the new method, further research covering various samples is needed to expand the application scope and test the accuracy under different circumstances.

Acknowledgements

We thank Dr. Liang Dong for providing the peat sample from Xinjiang. This research was supported by the National Natural Science Foundation of China (Nos. 416730 42, 41876042, and 41776049).