Soil organic carbon content and mineralization controlled by the composition, origin and molecular diversity of organic matter: A study in tropical alpine grasslands

The consensus for mechanisms controlling soil organic matter (SOM) persistence has shifted from traditional views based on SOM recalcitrance to a new paradigm based on SOM stabilization controlled by soil minerals and aggregates. Recent studies indicate that the origin, composition and molecular diversity of SOM are crucial to the decomposition and stabilization of SOM. However, it is not fully understood how the decomposition and stabilization of SOM are controlled at the molecular level. The objectives of this study were to investigate whether soil organic carbon (SOC) contents and mineralization are controlled by the composition, origin and molecular diversity of SOM. Soil samples were collected from contrasting bedrocks with different precipitation levels at tropical alpine grasslands of the Peruvian Andes. We applied a combination of a 76-day soil incubation exper-iment and pyrolysis-GC/MS assisted by thermochemolysis to investigate SOM decomposition and stabilization at the molecular level. The results indicated that soil samples with high SOC contents (92.6 ± 7.6 g kg (cid:0) 1 soil) and low SOC mineralization had abundant derivates of lignin, polysaccharides and n -alkanes. After the incubation, we observed neither a selective decomposition of any compound groups nor a decline of molecular diversity. In contrast, soil samples with low SOC contents (30.7 ± 2.8 g kg (cid:0) 1 soil) and higher SOC mineralization showed a depletion of plant-derived compounds, an accumulation of microbial-derived compounds and declined molecular diversity after the incubation. Furthermore, the SOC mineralization of these samples was positively correlated to the depletion of unsaturated fatty acids and the decrease in molecular diversity after the incubation. Therefore, we proposed that SOC contents and mineralization in our soils are (1) controlled by selective preservation of SOM molecular groups (e.g. plant-derived compounds), and (2) associated with changes in molecular diversity of SOM during microbial decomposition. Due to the selective preservation of organic compounds under different environmental conditions, we propose that environmental factors should be considered for the management of ecosystem services such as SOC sequestration in the studied region.


Introduction
Soil organic carbon (SOC) is one of the largest terrestrial carbon pools and plays an important role in global carbon dynamics. The storage and stability of SOC are largely dependent on the persistence of soil organic matter (SOM). In recent decades, there has been a shift from traditional views regarding SOM persistence based on SOM recalcitrance to a new paradigm based on SOM stabilization as an ecosystem property with an important role for interactions of SOM with the soil matrix that limit its accessibility for microorganisms (Lützow et al., 2006;Schmidt et al., 2011). Models explaining SOM persistence also have changed from the "humification" model based on aqueous extractions to a consolidated continuum model. In the continuum model, SOM persistence is controlled by a progressive SOM decomposition and SOM stabilization regulated by interaction with soil mineral surfaces and aggregates (Han et al., 2016;Lehmann and Kleber, 2015). Although molecular recalcitrance is now generally deemed less important as a SOM stabilization mechanism than previously assumed (Dungait et al., 2012), the effects of SOM molecular composition on SOM persistence are not fully understood. For example, a consensus is still absent on the extent to which the composition and origin of SOM control the stabilization of SOM via governing interaction with minerals and aggregates.
To unravel underlying mechanisms of SOM persistence, it is crucial to have insight at the molecular level (Kögel-Knabner, 2017, 2002. Carbohydrates are usually distributed in non-stabilized soil fractions and act as ready-to-use priming sources for microorganisms (Gunina and Kuzyakov, 2015). Lignin has been proved not as persistent as previously believed because it may disappear quickly in favorable environmental conditions (Schmidt et al., 2011;Thevenot et al., 2010). In contrast, compounds such as long-chain alkanes and amino sugars are reported to have a long residence time and can dominate stabilized OM under the right environmental circumstances (Barré et al., 2018;Jansen and Wiesenberg, 2017;Joergensen, 2018). Lehmann et al. (2020) proposed a conceptual framework to show the importance of molecular diversity in controlling SOM persistence. In general, SOM decomposition is alleviated at high molecular diversity because the microbial utilization of substrate is restricted by the complexity of compound composition (Cotrufo et al., 2013). However, it is not sufficiently understood how to quantify SOM molecular diversity to predict the decomposition and persistence of SOM (Lehmann et al., 2020).
Microbial transformation of plant-derived materials is considered as a key process for the formation of stable SOM (Cotrufo et al., 2013;Liang et al., 2017). Kallenbach et al. (2016) gave the first evidence that soil microorganisms produce chemically diverse SOM from limited types of organic compounds. Sokol et al. (2019) and Zhu et al. (2020) reported that microorganisms utilize labile plant compounds (e.g. simple molecules from the rhizosphere) to produce microbial compounds that accumulate as stabilized SOM in the form of microbial necromass (e.g. amino sugars). Recent studies indicated that a large proportion of SOM is microbial-derived (Liang et al., 2019;Miltner et al., 2012). However, the relative contributions of plant-versus microbial-derived compounds are largely dependent on environmental and pedogenic factors (Angst et al., 2021). It is not well understood how these factors control the preservation of plant-and microbial-derived compounds, as well as the relevant microbial processes (Angst et al., 2021;Lehmann et al., 2020). Therefore, more studies are required to unravel the roles of plant-and microbial-derived organic matter (OM) in the stabilization of SOM, especially at the molecular level and under various environmental conditions.
Tropical alpine ecosystems preserve abundant SOC and act as hot spots under the stress of global change, although they cover limited areas of the terrestrial surface (Buytaert et al., 2011). Alpine grassland soils of the Peruvian Andes are characterized by high SOC stocks and contribute to important ecosystem services such as food production, water provision and carbon sequestration (Rolando et al., 2017;Yang et al., 2018). Therefore, it is important to gain insights into the stabilization of SOC (e.g. at the molecular level) in this region. Ecosystems of the Peruvian Andes are characterized by heterogeneous environmental conditions such as contrasting bedrocks and a gradient of precipitation decreasing from north to south (Geo GPS Perú, 2014;Rolando et al., 2017). Our previous studies showed that these environmental factors had clear effects on not only soil aggregation and mineralogy but also SOC stocks and stabilization mechanisms in the studied regions (Yang et al., 2020b(Yang et al., , 2020c. Therefore, the environmental heterogeneity will provide sufficient variation in soil properties to investigate the effects of SOM molecular composition on the persistence of SOC. Meanwhile, the outcome of this study will contribute to the management of SOC storage and ecosystem services in the studied region. The objectives of this study were to investigate whether SOC contents and mineralization are controlled by the composition, origin and diversity of SOM, using (1) individual compound groups, (2) overall molecular diversity, and (3) plant-versus microbial-derived OM. To gain insights into SOM decomposition and stabilization at the molecular level, we applied a combination of soil incubation experiments and pyrolysis-GC/MS assisted with thermochemolysis using soil samples collected from the grassland of Peruvian Andes.

Soil sampling
Soil samples were collected from high-altitude grasslands of the Peruvian Andes. A detailed description of the sampling sites is provided by Yang et al. (2020b). Briefly, soil samples were collected from contrasting bedrocks (limestone and acid igneous rocks) and two precipitation levels (wet and dry sites), controlled with similar altitude and average temperature (Table 1). Limestone soils (LSs) had A and B horizons overlying the parent material and belong to the soil group of Phaeozems. In contrast, acid igneous rock soils (ASs) only had A horizons directly overlying parent materials and can be classified as Andosols or Umbrisols (WRB, 2014). Soil samples (18 samples) were collected in three replicates in each combination of precipitation levels and bedrocks from A horizons (LSs and ASs) and B horizons (only LSs). Two of the three replicates (12 samples) were used for analyses after drying at 40 • C.

Soil analyses and incubation experiment
Soil pH was measured using demi-water (w:v = 1:5). Soil bulk densities were calculated using samples collected using Kopecky rings (100 cm 3 ) by weighing them after freeze-drying. Total carbon (C) and nitrogen (N) contents were determined using a VarioEL Elementar analyzer (Elementar, Germany). For samples with pH values > 5.5, carbonate contents were measured using a TIC module in the Elementar analyzer. As carbonate contents were found to be negligible in all samples, SOC contents were equal to total C contents.
Bulk soil samples were separated by aggregate size using a drysieving method. Between 170 g and 230 g air-dried samples were placed on two mesh sieves (diameter 2 and 0.25 mm, top and bottom) and shaken with a horizontal shaker (30 Hz for 20 s). The sieving produced 3 fractions: large macroaggregates (LM, >2 mm), small macroaggregates (SM, 0.25− 2 mm) and microaggregates (Mi, <0.25 mm), respectively. Total C and N contents of these fractions were also measured by the Vario EL Elementar Analyzer (Elementar, Germany). The LM and SM fractions (24 samples) were used for incubation and molecular analyses because they contained more than 60% of the carbon of the bulk soils (Table 1).
The incubation experiment was reported in detail by Yang et al. (2020b). Briefly, samples were rewetted on a pF tray at pF = 2 (− 100 mbar) for 10 days to reach favorable moisture contents and to activate microorganisms. The moistened samples were incubated at 20 • C in sealed jars (120 cm 3 ) and the headspaces were sampled at Day 1, 2, 6, 9, 13, 20, 28, 48 and 76. After each sampling, synthetic air (CO 2 free) was injected into jars to maintain the internal pressure and avoid the depletion of O 2 . The CO 2 concentrations of the headspaces were measured using a gas chromatography flame ionization detector (GC-FID with a methanizer, Thermo Scientific, Trace GC Ultra). Based on the CO 2 concentrations for all sampling days, cumulative SOC mineralization (g CO 2 -C kg − 1 SOC) during the 76-day incubation was calculated for each sample and used for further statistical analyses. We used the unit of g CO 2 -C kg − 1 SOC rather than g CO 2 -C kg − 1 soil mass because the former is generally used to estimate the decomposability or the stability of the SOC while the latter is applied to evaluate CO 2 emission from soils and implies information on SOC stability and contents. The incubation was terminated at Day 76 and the incubated soil samples were dried at 40 • C for pyrolysis-GC/MS analyses.

Pyrolysis-GC/MS analyses
Before pyrolysis-GC/MS analyses, homogenized samples (20 -50 mg) before and after the incubation (48 samples) were mixed with 20-60 µL tetramethylammonium hydroxide (TMAH) solution (25% by weight in H 2 O) for derivatization (He et al., 2020). The purpose of derivatization (methylation) was to improve the identification of polar and non-volatile compounds, such as fatty acids, lignin and monomers of cutin and suberin. These compounds provide information on SOM origin and microbial processes (Section 2.4). After the TMAH addition, a small amount of the mixture for each sample was placed on a ferromagnetic wire and dried under a halogen lamp for the pyrolysis. A Curie-point pyrolyzer (Horizon Instruments; 600 • C; 5 s) was applied for pyrolysis and it was coupled to a ThermoQuest Trace GC gas chromatograph (Waltham, USA). Helium was used as the carrier gas. The initial temperature of the oven was 40 • C for 1 min and heated to 320 • C with a rate of 7.0 • C min − 1 and a hold time of 10 min. The separation of analytes was performed on a ZB1-MS column (Phenomenex: 30 m, 0.25 mm i.d, 0.50 µm df). The column was further connected to a Finnigan Trace MS mass spectrometer (Waltham, USA; m/z: 47-500, ionization energy: 70 eV, cycle time: 0.45 s).

Peak identification and calculation
Peak identification and quantification were conducted using the Xcalibur software (Version 3.1, Thermo Fisher Scientific Inc.). Peak identification was based on quantifying ions using the NIST library (Gaithersburg, USA) and information from other publications (e.g. Brock et al., 2019;Nierop et al., 2007;Schulten and Schnitzer, 1997). A peak was deemed valid only when the signal of the peak was at least 3 times larger than that of the background noise and when the spectra pattern of the compound could be clearly identified. We identified 104 compounds in total and these compounds were classified into 12 chemical groups. The groups were short-chain saturated fatty acids (FASS, less than 20 C), short-chain unsaturated fatty acids (FASU, less than 20 C), long-chain fatty acids (FAL, at least 20 C), α,ω-dioic acids (DA), ω-hydroxyl alkanoic acids (HA), n-alkanes (Alka), n-alcohols (Alch), esters (Ester), lignin and phenols (Lg), polysaccharides (Ps), nitrogen-containing compounds (N) and other compounds (Other). We classified fatty acids into three sub-groups because: (1) the carbon chain length reveals the origin of the fatty acids (i.e. microorganisms or plants, Wiesenberg et al., 2010); and (2) previous studies for these samples indicated that the carbon chain length and the presence of double bonds are important for the stabilization of fatty acids (Yang et al., 2020a). Peak quantification was conducted by manual peak integration. The sum of peak areas of each sample was set to 100% and the relative amount of each compound to the total peak area was calculated as relative abundance (RA). The RA of each compound group was the sum of RAs for all compounds in this group.
To estimate the molecular diversity of SOM, we calculated the Shannon index, the Simspon Index and the Pielou Index, based on the RA of each compound (n = 104). We decide to use the Shannon index to estimate the molecular diversity of SOM because (1) it is widely used to estimate diversities in ecology (Spellerberg and Fedor, 2003), and (2) it had strong correlations with the Simpson Index and the Pielou Index (Supplementary Table S1). The Shannon index was calculated by the following equation: In which Shannon RA is the Shannon index, RA i is the RA of compound i, and RA k is the total relative abundance.
The sums of the relative abundances of DAs and HAs were used to estimate plant-derived compounds because they are widely used as biomarkers of cutin and suberin in soils (Otto and Simpson, 2006).

Table 1
General information of sampling sites and soil samples.

SOC distributed in SM (%)
Wet-LS ( SOC: soil organic carbon, C/N: C/N ratio, BD: bulk density, LM: weight percentage of large macroaggregates (>2 mm), SM: weight percentage of small macroaggregates (0.25-2 mm). 1 bulk density was measured every 10 cm and the averaged value was calculated for each horizon. Original data and detailed information are from Yang et al. (2020b).
Short-chain saturated fatty acids have a large contribution from microorganisms (Wiesenberg et al., 2010) and were reported to be accumulated during the incubation in our previous study (Yang et al., 2020a). Therefore, we use short-chain saturated fatty as an indicator for microbial-derived compounds in this study.

Statistics
Linear regressions were applied to identify significant factors controlling the molecular composition of SOM. The results showed that the molecular composition of SOM was controlled by precipitation, lithology and horizon but was poorly predicted by aggregate size (Supplementary Table S2). As a result, our analyses only focused on precipitation, lithology and horizon. As the LM and SM fractions from the same soil samples were not strictly independent samples, a linear mixed model was applied to compare SOC properties and SOM molecular composition between different samples. A Fisher's least significant difference (LSD) test was used for post hoc analyses. In addition, A principal component analysis (PCA) based on the correlation matrix was performed to explore the shift in RAs for different soil samples after the 76-day incubation.
A partial least square regression (PLSR) was used to identify significant chemical groups controlling SOC contents and SOC mineralization. A multiple linear regression could not be used because of the collinearity between chemical groups. The PLSR solved this problem as it is a combination of PCA and linear regression that allows for: (i) the decomposition of multiple predictors into several latent variables despite collinearity, and (ii) the prediction of dependent variation using predictors via the transformation into latent variables (Abdi, 2003). The data was normalized (mean = 0, variance = 1) prior to the PLSR analysis, whereas the cross-validation was conducted using the "leave-one-out" method. Significant predictors were selected using the variable importance for projection (VIP) method, in which predictors characterized by VIP scores > 0.8 were considered significant (Chong and Jun, 2005). Predictors and dependent variables were visualized in figures by their correlation coefficients at the levels of latent variable 1 (LV1) and latent variable 2 (LV2).

Soil properties
For A horizons, Wet-LS samples had the highest SOC contents (92.6 ± 7.6 g kg − 1 ) and the lowest SOC mineralization (7.77 ± 2.44 g kg − 1 C, Day 76). In contrast, Dry-LS samples had the lowest SOC contents (30.7 ± 2.8 g kg − 1 ) and the highest SOC mineralization (22.08 ± 2.23 g kg − 1 C, Day 76). Medium levels of SOC contents and mineralization were found in both Wet-AS and Dry-AS samples ( Fig. 1-A and 1-C). Soil C/N ratios were the highest in Dry-LS samples, followed by Wet-AS and Wet-LS samples, and were the lowest in Dry-LS samples ( Fig. 1-B). Only LS samples had B horizons, which had lower SOC contents and C/N ratios compared to A horizons ( Fig. 1-A and 1-B).  (n ¼ 4). The linear mixed model is used to identify significant differences between soil samples, as indicated by different letters (P < 0.05). In graph C, letters on the right indicate differences on Day 76. Wet: the wet site, Dry: the dry site, LS: limestone soils, AS: acid igneous soils, A: A horizons, B: B horizons.
In addition, Wet-LS samples had higher pH values than other samples, whereas LS samples were characterized by larger aggregate sizes than AS samples (Table 1).

Molecular composition of soil organic matter before the incubation
For A horizons, most chemical groups were not significantly different between different samples, except for (1) higher ω-hydroxyl acids in the Wet-AS compared to Dry-LS and Dry-AS samples, (2) higher polysaccharide derivates in Wet-LS and Wet-ASs compared to Dry-AS samples, and (3) higher lignin derivates in Wet-LS than Dry-AS samples (Fig. 2). For B horizons, short-chain fatty acids were abundant, whereas lignin derivates were almost absent (Fig. 2-A and 2-I). In addition, Wet-LS samples had a larger differentiation between A and B horizons than Dry-LS samples (Fig. 2-A, 2-C, 2-D and 2-E).
In the graph of partial least square regression (Fig. 3-A and 3-B), LV1 and LV2 explained 34.6% and 18.8% of the total variation for predictors (i.e. RA of chemical groups), whereas 58.8% of the total variation for dependent variables (i.e. SOC contents and mineralization) was explained. Zone (1) showed that Wet-LS-A samples had high SOC contents, which were associated with abundant derivates of lignin, polysaccharides and n-alkanes (Fig. 3-A and 3-B). Zone (2) indicated that Dry-LS samples had high SOC mineralization, which was positively correlated to the presence of unsaturated fatty acids (Fig. 3-A and 3-B).

Shift of soil organic molecular composition after the incubation
In Fig. 3-C and 3-D, LV1 and LV2 explain 53.0% and 19.0% of the total variation for both predictors and the dependent variable. Zone (3) showed that Dry-LS samples were characterized by high SOC mineralization, which corresponds to the large depletion of unsaturated fatty acids and n-alcohols (Fig. 3-C and 3-D).
The shift of different chemical groups after the incubation is shown by the PCA that explained 54.6% of the total variation (Fig. 4). The principal component 1 (PC1) was positively loaded by derivatives of short-chain unsaturated fatty acids, long-chain fatty acids, α,ω-dioic acids, ω-hydroxyl acids, n-alcohols, lignin and polysaccharides, and was negatively loaded by short-chain saturated fatty acids and N compounds (Fig. 4-A). The PC2 had positive contributions from N-compounds and nalkanes (Fig. 4-A). After the incubation, Dry-LS samples showed a clear pattern moving towards the negative side of PC1, indicating the accumulation of short-chain saturated fatty acids and N-compounds, as well as depletions of other compounds (Fig. 4-A and 4-C). In contrast, other soil samples showed no clear pattern of chemical group changes after the incubation (Fig. 4-B to 4-D).
The molecular diversity before the incubation was not correlated with SOC contents or mineralization (P > 0.05, Fig. 5-A). After the incubation, the decline of molecular diversity was positively correlated with SOC mineralization (P < 0.05, Fig. 5-B). In addition, the Dry-LS samples showed a trend of consistent decrease of molecular diversity after the incubation, whereas other samples did not show a clear trend (Fig. 5-C). For plant-and microbial-derived compounds, the Dry-LS samples showed consistent depletion of plant-derived compounds (i.e. cutin and suberin biomarkers) and an accumulation of microbialderived compounds (i.e. short-chain saturated fatty acids) after the incubation ( Fig. 6-B). In contrast, the other samples had no clear trend ( Fig. 6-A and 6-C).

Overview
Pyrolysis-GC/MS assisted with TMAH is a useful analytical method to get an overview of SOM composition, but quantitative analyses are difficult due to the lack of analytical standards and the different response factors per compound (Derenne and Quéné, 2015;Shadkami and Helleur, 2010). In addition, the TMAH derivatization only improves the detection of certain compounds (e.g. fatty acids and lignin) in the subsequent GC/MS analysis (He et al., 2020). Therefore, the detected molecular groups of SOM do not necessarily correspond to the original composition of SOM in soil samples. To stay on the safe side, we focused on relative shifts in patterns only, by limiting ourselves to direct comparisons between our samples and with the results of other studies that used the same method.
For A horizons, significant differences in SOM composition between different soils were only found for three molecular groups (i.e. lignin, nalkanes and ω-hydroxyl acids). In contrast, the other nine groups were not significantly different (Fig. 2). The similarity in the molecular composition of SOM in the A horizons can be explained by the similar types of vegetation and land use (Table 1). Similarly, Nierop et al. (2007) also reported homogeneity in SOM composition in the Ecuadorian Andes. For B horizons, which are only present in LS samples, SOM molecular composition showed more differences compared to the A horizons. Lignin derivates were almost depleted in subsoils (B horizons) of the LS samples (Fig. 2). This is similar to the results of many studies as summarized by Thevenot et al. (2010) and is consistent with the situation in grassland soils of the Ecuadorian Andes as reported by Nierop et al. (2007). In addition, the B horizons were characterized by a large relative amount of short-chain saturated fatty acids (Fig. 2). This suggests that the B horizons contain more microbial-derived compounds than the A horizons, which is consistent with general views (Rumpel and Kögel-Knabner, 2010). In the following discussion, we only compare the A horizons between different soil plots because (1) B horizons were only presented in the LS samples, and (2) SOM properties and stabilization mechanisms can be different between A and B horizons (Rumpel and Kögel-Knabner, 2010).

SOC contents and mineralization controlled by SOM molecular composition
The high SOC contents in the Wet-LS samples, which are associated with low SOC mineralization, are characterized by more abundant derivates of lignin, polysaccharides and n-alkanes than other samples (Figs. 1 and 3). After the incubation, these samples showed no selective and relative abundance changes of molecular groups after the incubation, D: scores of different soil samples corresponding to graph C. n = 24 in total. The PLSR model for graph A and B explains 53.4% and 58.8% of the total variation for predictors and respond variables (R 2 (X) = 0.534 and R 2 (Y) = 0.588), whereas the PLSR for graph C and D explains 72.0% of the total variation for both predictors and respond variables (R 2 (X) = R 2 (Y) = 0.720). VIP ≥ 0.8 indicates significant predictors. FASS: short-chain saturated fatty acid, FASU: short-chain unsaturated fatty acid, FAL: long-chain fatty acid, DA: α,ω-dioic acids, HA: ω-hydroxy acids, Alka: n-alkane, Alch: n-alcohol, Ester: ester, Lg: lignin, Ps: polysaccharides, N: N compounds, Other: other compounds, Wet: the wet site, Dry: the dry site, LS: limestone soils, AS: acid igneous soils, A: A horizons, B: B horizons. Zone (1): Wet-LS-A samples have high SOC contents, which are associated to abundant Lg, Ps and Alka. Zone (2): Dry-LS samples have high SOC mineralization, which are associated to abundant FASU. Zone (3): High SOC mineralization are correlated to large depletion of FASU and Alch, which is found in Dry-LS samples. In addition, FASU is a better predictor than Alch for SOC mineralization (Supplementary Table S3). accumulation or depletion of any compound groups, as indicated by no clear pattern of changes in compound groups (Figs. 4 and 6). Therefore, the abundant and stable SOC in the Wet-LS samples can be explained by the accumulation of lignin, n-alkanes and polysaccharides, as well as the fact that they were not clearly depleted after the incubation compared to the Dry-LS samples.
The low SOC contents and high SOC mineralization in the Dry-LS (Fig. 1) suggest that SOC is not well stabilized. The high SOC mineralization was further associated with a greater depletion of unsaturated fatty acids (Figs. 2 and 3). This is in line with previous studies showing that unsaturated fatty acids are more susceptible to decomposition in soils (Moucawi et al., 1981;Yang et al., 2020a). After the incubation, the selective depletion of compound groups, except for short-chain saturated fatty acids and N-compounds (Fig. 4), indicates the low capacity of the Dry-LS samples to stabilize certain OM compounds (e.g. monomers of cutin, suberin and lignin).  In our previous studies on these soils, we found that SOM stabilization was controlled by organo-mineral interactions, whereas SOM occluded in aggregates played a minor role (Yang et al., 2020b(Yang et al., , 2020c. The results of our present study suggest that SOC contents and mineralization are associated with the selective preservation of different OM molecules. The higher contents of lignin in the Wet-LS samples (high SOC content) can be explained by the higher pH values compared to other samples (Table 1), as the decomposition of lignin is closely related to white rot fungi that prefer acidic soils (Thevenot et al., 2010). This is consistent with results reported for instance by Brock et al. (2019), showing that lignin degradation was more pronounced in acidic soils. The selective preservation of polysaccharides and n-alkanes in the Wet-LS samples might be related to soil mineralogy, as the soils had high contents of pedogenic Fe and Al oxides and Ca 2+ bridges (Yang et al., 2020b(Yang et al., , 2020c. These oxides and Ca 2+ bridges can (1) promote stabilization of SOM compounds (e.g. polysaccharides and lignin) by association with mineral surfaces via ligand exchange and cation bridges (Lützow et al., 2006), and (2) improve soil f to protect compounds (e.g. n-alkanes) by occlusion in aggregates (Rowley et al., 2018). For the Dry-LS samples (low SOC content), the high SOC mineralization and the depletion of most compounds after the incubation are consistent with the previous findings that these soils had low capacity to stabilize organic molecules (Yang et al., 2020b). In addition, the abundant unsaturated fatty acids and other less stabilized SOM might promote microbial activities, through processes similar to priming effects (Kuzyakov, 2010), to decompose substrates with relatively high activation energy (see Section 4.4).

Stabilization of plant-versus microbial-derived OM
Lignin, n-alkanes, α,ω-dioic acids and ω-hydroxyl acids are widely used as biomarkers for plant-derived OM (Amelung et al., 2008). In contrast, short-chain fatty acids, especially branch-chain fatty acids identified in our samples (Supplementary Table S4), have a major origin of microorganisms as parts of the cell membrane (Kaneda, 1991;Zelles, 1999). In addition, compounds such as polysaccharides and N-compounds are ubiquitous with both plant and microbial origins (Barré et al., 2018).
We observed no selective degradation of plant-derived OM or accumulation of microbial-derived OM in the Wet-LS, Wet-AS and Dry-AS samples (high and medium SOC contents). This is indicated by (1) no clear trend in depletion of lignin, cutin and suberin biomarkers after the incubation, and (2) no clear trend in accumulation of short-chain saturated fatty acids after the incubation, compared to the Dry-LS samples (Figs. 4 and 6). A possible explanation is that plant-derived OM is well protected in these soils and receives less microbial transformation, as many monomers of plant-derived compounds (i.e. lignin, cutin and suberin) can be stabilized by association with mineral surfaces via hydroxyl and carboxyl groups (Angst et al., 2021;Lützow et al., 2006). In these soils, especially the Wet-LS samples, plant-derived OM can be stabilized by interacting with active Fe-and Al-oxides, as well as Ca 2+ bridges (see Section 4.2.). In addition, the higher pH values in the Wet-LS samples (Table 1) suggest a reduction of fungal activity, which can restrict the decomposition of plant-derived macromolecules (Koranda et al., 2014). In contrast, Dry-LS samples (low SOC contents) showed a strong transformation from plant-to microbial-derived compounds, as indicated by the depletion of most plant-derived OM and an accumulation of short-chain fatty acids after the incubation (Figs. 4 and 6). The strong microbial transformation might be attributed to a poor stabilization of the plant-derived compounds in these soils, as mentioned in Section 4.2. Moreover, it is also possible that these soils provide favorable pH and moisture to microorganisms for the intensive microbial transformation of plant-derived OM.
Recent studies found that a large proportion of SOM consists of stabilized compounds with a microbial origin (Liang et al., 2019;Miltner et al., 2012). However, the relative contributions of plant-versus microbial-derived SOM are largely dependent on environmental and pedogenic factors (Angst et al., 2021). Our results are consistent with this and suggest that the variation in plant-and microbial-derived SOM can be attributed to the degree of overall preservation of plant-derived OM. Similarly, Barré et al. (2018) found that persistent SOM is contributed by a substantial and variable fraction of plant-derived OM. Moreover, it is also possible that the persistence of microbial-derived OM is controlled by factors such as soil pH, texture or microbial communities (Hu et al., 2020).

Molecular diversity
Our results highlighted the relationship between the molecular diversity and the decomposition of SOM, as indicated by the fact that the molecular diversity declined for the Dry-LS samples (low and unstable SOC) during the incubation but had no consistent trend for the other soil samples with medium to high SOC contents (Fig. 5). To our knowledge, this has not been reported before, but relevant studies showed that the molecular diversity of SOM was controlled by the type of OM input (Chen et al., 2021), soil mineralogy (Hall et al., 2020) and ecosystem characteristics (Tfaily et al., 2017). Also, the molecular diversity affects microbial decomposition and transformation of SOM and, therefore, controls the stabilization of SOM (Cotrufo et al., 2013;Lehmann et al., 2020). We propose two possible explanations for the significantly decreased molecular diversity in the Dry-LS samples. The first explanation is that most of the molecular groups were poorly stabilized and depleted under the detection limit of the instrument for the Dry-LS samples after the incubation. Therefore, we observed declined molecular diversity for these samples. Alternatively, the presence of abundant unsaturated fatty acids, which are easily degradable and act as potential priming substrates, could promote the activity and functional diversity of soil microorganisms (Kuzyakov, 2010). The activated microbial community would be able to utilize substrates with relatively higher activation energy (e.g. recalcitrant or weakly stabilized compounds) and, therefore, lead to decreased molecular diversity of SOM after the incubation (Blagodatskaya and Kuzyakov, 2008;Kuzyakov, 2010). To test the two explanations, further experiments on the priming effects of unsaturated fatty acids might be helpful.

Practical issues for soils in the Peruvian Andes
Tropical alpine grasslands, including Peruvian Andes, provide important ecosystem services including carbon sequestration and food provision (Chai et al., 2020;Rolando et al., 2017). To maintain SOC storage and soil fertility, local farmers perform land management such as a rotation in the order of cultivation, abandonment and grazing within a period of 2-5 years (Yang et al., 2018). However, the effectiveness of these measures depends on whether the OM from plant input can be stabilized in soils. Our results indicate that the decomposition of plant-derived OM in these soils is dependent on parent material and precipitation (Fig. 7). This suggests that traditional management to increase SOC storage and fertility (e.g. green manure addition) might have restricted effects in some soils (e.g. Dry-LS) because a part of plant-derived OM will be quickly mineralized rather than stabilized. As plant-derived compounds account for ≥ 50% of the SOM and contribute to stabilized SOC pools (Angst et al., 2021), we propose that environmental factors (e.g. lithology and precipitation) should be considered to maintain the ecosystem services such as carbon sequestration in the studied region.
Climate change can have crucial impacts on SOC sequestration in the Peruvian Andes. The shifting of forest lines to higher altitudes caused by warming in the Andes (Tovar et al., 2013) will potentially affect the SOM input and stabilization. In general, the shift from grassland to forest will increase the input of low-quality OM and the proportion of plant-to microbial-derived OM in soils (Angst et al., 2021). Because soils in our study area had different capacities to stabilize plant-derived OM, we expect contrasting feedbacks of SOC sequestration in these soils (e.g. Wet-LS vs. Dry-LS) to the shift from grassland to forest caused by climate change. The change in precipitation, in the short term, can alter the soil microenvironment and the microbial decomposition and transformation of SOM. For the long-term effects, precipitation controls the pedogenesis such as weathering processes and further controls soil mineralogy and SOM stabilization. As the effects of precipitation are stronger for LS samples compared to AS samples, the LS samples are more likely to be affected than the AS samples by climate change regarding the shift of precipitation.

Conclusions
Results of this study were summarized as an overview in Fig. 7. In the studied region in the Peruvian Andes, soil samples with high SOC contents and low SOC mineralization (Wet-LS) had abundant derivates of lignin, polysaccharides and n-alkanes. After the incubation, we observed neither selective decomposition of any compound groups nor decline of molecular diversity. In contrast, soil samples with low SOC contents and high SOC mineralization (Dry-LS) showed a strong transformation from plant-derived OM to microbial-derived OM and declined molecular diversity after the incubation. Overall, SOC mineralization was positively correlated to the decline of unsaturated fatty acids and that of molecular diversity after the incubation. Our results indicate that variation of SOC contents and mineralization are associated with selective preservation of SOM molecules and changes in molecular diversity under microbial decomposition.
The selective preservation and depletion of plant-derived OM in our samples help to understand the stabilization and the relative contribution of plant-versus microbial-derived OM in different environmental conditions. For SOM stabilization mechanisms, future studies could focus on specific stabilization mechanisms for different SOM molecules regarding their interaction with soil minerals and aggregates. Before the extrapolation of our conclusions to other soils, a crucial check is required for the applicability of our findings to soils with different environmental conditions. In addition, long-term experiments could be helpful to evaluate SOC stability and persistence as our 76-day incubation experiment could not fully represent the long-term stability of SOC. Finally, for practical issues, we propose that environmental factors should be considered for the management of ecosystem services such as SOC sequestration in the studied region of the Peruvian Andes.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.