The whole-soil warming experiment at University of California Blodgett Experimental Forest is located on the foothills of the Sierra Nevada near Georgetown, CA (120°39^′40^′′ W; 38°54^′43^′′ N) at 1370 m a.s.l. (above sea level) (Hicks Pries et al., 2018). The site has a Mediterranean climate with a mean annual air temperature of 12.5 °C and a mean annual precipitation of 1774 mm (Bird and Torn, 2006). The experiment is situated in a thinned 80-year-old mixed coniferous temperate forest, dominated by ponderosa pine (Pinus ponderosa), sugar pine (Pinus lambertiana), incense cedar (Calodefrus decurrens), white fir (Abies concolor), and douglas fir (Pseudotsuga menziesii; Hicks Pries et al., 2017). The soils are Holland series and classified as fine-loamy, mixed superactive, mesic ultic Alfisol of granitic origin (mean pH 5.5), which is equivalent to Dystric Cambisols (IUSS Working Group WRB, 2022).

Briefly, the whole-soil warming experiment consists of 6 plots in total arranged in three replicated blocks, each having a pair of warmed and controlled circular plots 3 m in diameter. To maintain warming down to 1 m depth, twenty-two 2.4 m-long resistance heating cables (BriskHeat, Ohio, USA) were vertically installed in metal conduits at a radius of 1.75 m, surrounding each plot. Two concentric rings of surface heater cable were installed at 1 and 2 m in diameter, 5 cm below the soil surface, to compensate for surface heat loss. The average soil temperature was elevated by +4 °C in warmed plots relative to ambient plots (except for at 0–20 cm only elevated by +2.6 °C due to surface heat loss), while preserving seasonality and natural temperature gradient with soil depth (Hicks Pries et al., 2017; Pegoraro et al., 2024). The setup of the control plots is identical to the warmed plots but without heating cables placed inside the metal conduits (Hicks Pries et al., 2017).

Common wild oat (Avena fatua) is an annual grass whose fast growth enabled uniform ¹³C-labelling in this experiment compared in contrast to the slower-growing native coniferous trees. This made its roots a suitable model substrate for decomposition studies (Hicks Pries et al., 2018). Avena fatua seedlings were grown for 12 weeks in a greenhouse within an airtight chamber at the University of California, Berkeley. Every 4 d the source of CO₂ was switched between ambient CO₂ and 10 atom % ¹³CO₂ (Cambridge Isotope Laboratories, Inc., Massachusetts, USA). After this labeling phase, roots were excavated, dried and cut in 1–2 cm pieces (<2 mm diameter). The root substrate was enriched by 5.6 % atom % ¹³C and had a carbon concentration of 0.463 g C g root⁻¹ (Castanha et al., 2018; Hicks Pries et al., 2018).

A total of six soil cores were extracted in each plot by using a perforated custom coring system made of polycarbonate tube (5.04 cm outer diameter and 4.41 cm inner diameter). Each tube consisted of four sections (10, 35, 40, and 10 cm in length, respectively) that were threaded on each end (male threads) and connected with female-to-female threaded polycarbonate connectors. Each section was threaded onto a sharpened aluminum cutting edge at the bottom for coring, and an aluminum tube with a weighted pounding head on the top. We cored the mineral soil based on the length of each polycarbonate section at the following depths: 0–10, 10–45, 45–85, and 85–95 cm. The top 4 cm of soil in each section (target experiment depth: 10–14, 45–49, and 85–89 cm) was marked and scooped into aluminum tins. Then a pre-weighed, 0.14 g aliquot of Avena fatua fine roots were added. The labeled Avena root litter was added to four out of the six cores and these cores will be termed as root litter treatment in this paper. The same disturbance was applied to the rest of the two cores without addition of labeled root litter and will be denoted as disturbance control treatment. Mesh disks with 1 mm aperture were positioned at the top and bottom of each target 4 cm soil increment to delineate the physical boundaries of the inserted root substrates. This setup was intended to retain the added root litter and prevent ingrowth of living roots into the cores and was also done in disturbance control treatment. The soil mixture was then added back to each polycarbonate section. The polycarbonate core sections were consequently connected with the female-to-female connectors and the resulting 95 cm long core was placed back in the soil. In July 2019, i.e., after 3 years of in-situ field incubation, two cores with root litter treatment and one core with disturbance control treatment were retrieved from each plot. The retrieved cores were wrapped in foil and transported to Lawrence Berkeley National Laboratory and stored in a -20 °C freezer before further processing. The more detailed description of experimental setup and incubation methods is demonstrated in Pegoraro et al. (2024). For brevity, DC will be used to refer to the disturbance control treatment in the rest of this paper.

Because prior work at this site concentrated on three depths (Hicks Pries et al., 2017; Soong et al., 2021), we selected the same depths to facilitate cross-study, temporal and spatial comparisons. These depths also capture major pedogenic zones: topsoil, a transitional mid-depth between topsoil and subsoil, and the deep subsoil horizon.

In the results and discussion for this paper, the topsoil denotes the soil depth at surface (10–14 cm), and the subsoil describes the soil depths at 45–49 cm (mid-depth) and 85–89 cm (deep soil).

In the laboratory, we opened the polycarbonate cores to retrieve the 4 cm section where labeled roots were added (same for disturbance control). We also sampled the 4 cm section above and below the target depths. The soil samples were sieved <2 mm. Roots were picked off the top of the sieve by tweezers. Bulk soil < 2 mm was freeze-dried and re-weighed. A subsample of bulk soil samples was ground by a ball mill (MM400, Retsch, Haan, Germany) and analyzed for carbon and nitrogen concentrations, as well as stable carbon isotope composition (δ13C) using an elemental analyzer-isotope ratio mass spectrometer (EA-IRMS; Flash 2000-HT Plus, linked by Conflo IV to Delta V Plus isotope ratio mass spectrometer, Thermo Fisher Scientific, Bremen, Germany). The results are reported in the δ notation: 1δ13C=RsampleRstandard-1×1000 where Rsample and Rstandard are the ¹³C / ¹²C ratios of the sample and the international standard, Vienna Pee Dee Belemnite (VPDB, 0.01118), respectively. At least two analytical replicates were measured for all samples. Calibration was carried out using IAEA-certified primary standards (e.g., N600 caffeine) and caffeine (Merck) as secondary standard.

All soil samples (<2 mm) were pre-extracted by Soxhlet following an established protocol (Wiesenberg and Gocke, 2017) to remove solvent-extractable lipids with dichloromethane : methanol (93:7; v/v) for 48 h. The extraction residues were dried until constant weight.

The extraction residues were homogenized with a ball mill (MM400, Retsch, Haan, Germany) and then hydrolyzed according to Zosso et al. (2023). Therefore, an aliquot of each residue equivalent to >20 mg carbon was weighed in a 250 mL round bottom flask. The sample was mixed with the extraction solution methanol: deionized water (9:1; v/v) with 6 % potassium hydroxide (KOH) and then saponified for 20 h at around 85–88 °C in a water bath under reflux. Subsequently, the solution was filtered and transferred to a separation funnel for phase separation. The solution was acidified to pH 2.0 using 6 M hydrochloric acid (HCl) and then extracted with dichloromethane. The collected fractions were volume-reduced and remaining water was removed by water-free sodium sulfate (Na₂SO₄).

For quantification of hydrolysable lipids, deuterated eicosanoic acid (D₃₉C₂₀; Cambridge Isotope Laboratories, Inc.) was added to the samples as an internal standard. The samples were silylated at 80 °C for 1 h with bis(trimethylsilyl)acetamide (BSA; Wiesenberg and Gocke, 2017). Individual compounds were quantified on an Agilent 7890B gas chromatograph (GC) equipped with a multi-mode inlet and a flame ionization detector (FID). Compound identification was performed on an Agilent 6890N GC equipped with split/splitless inlet and coupled to an Agilent 5973 mass selective detector (MS). Compounds were identified by comparison of mass spectra with those of external standards and from the NIST and Wiley mass spectra library. Both instruments were equipped with DB-5MS columns (50 m × 0.2 mm × 0.33 µm) and 1.5 m de-activated pre-columns, with helium as the carrier gas (1 mL min⁻¹). Silylated fractions were injected in splitless mode at an initial GC oven temperature of 50 °C that was kept isothermal for 4 min, then increased to 150 °C at a rate of 4 °C min⁻¹. Thereafter, the temperature ramped up to 320 °C at a rate of 3 °C min⁻¹ and held for 40 min. The GC-MS was operated in electron ionization mode at 70 eV and scanned from m/z 60–650. The analysis of the data was processed with Agilent Chemstation software. The concentration of each compound was finally normalized to the organic carbon concentration of the respective sample (stated as µg g⁻¹ OC). The weight of samples weighed in for hydrolyzation is always corrected by accounting for the mass loss due to free lipid extraction: 2Mcorrected=Mweighed1-pfreelipids where Mcorrected is the corrected weight of soil samples, Mweighed is the weight of soil samples weighed in for hydrolyzation, and pfreelipids is the proportion of free extractable lipids to the mass of soil samples weighed in for Soxhlet extraction.

In this paper, mid-length and long-chain monomers are the compounds with a carbon chain length n between 14 and 20 (14≤n<20) and length ≥ 20, respectively. For each compound class, fatty acids are the synonymously used for n-carboxylic acids, ω-hydroxy acids are short for ω-hydroxy carboxylic acids, diacids stand for α, ω-alkanedioic acids, alcohols are abbreviated for n-alcohols, and mid-chain acids stand for mid-chain hydroxylated fatty acids, referring to fatty acids with functional groups or structural modifications located in the middle of their carbon chain, typically at the C-9 and C-10 carbon positions, such as x, ω-dihydroxyhexadecanoic acid (x=9 or 10; Graça, 2015).

Different monomers can be used as markers for leaf and needle (cutin) or woody and root (suberin) biomass. However, there are no universal markers across a variety of studies since the relative proportions of cutin and suberin markers could vary among plant taxa, plant functional type or plant organ (Jansen and Wiesenberg, 2017; Mueller et al., 2012). Here, we selected ω-hydroxy alkanoic acids and diacids as suberin markers since these monomers exist in substantial amount in the roots of the dominating plant species around the experiment plots and are 10 times higher in concentration compared to the same monomers in leaves or needles in the same species (Fig. S2). Cutin markers could not be distinguished since all the mid-chain acids which were traditionally considered as cutin markers were present in considerable amounts in added ¹³C-labeled roots (Table S1 in the Supplement).

To the best of our knowledge, there are no studies reporting the composition of mid-chain fatty acids, ω-hydroxy or diacids from microorganisms in soil, although it was reported that microorganisms can synthesize the compound classes mentioned above (Huf et al., 2011; Kim and Park, 2019). These compound classes, however, are more regio-specific (the enzymes will target only one or a few defined carbon positions) and differ from those in plants and animals (Kim and Oh, 2013) and usually have a carbon chain-length < 20 (Zhang et al., 2024). Therefore, we hypothesize that all the compounds with a carbon chain length ≥ 20 and mid-chain fatty acids are exclusively plant-derived.

To determine the δ13C of individual compounds, a Trace GC Ultra, coupled via GC Isolink II and Conflo IV to Delta V Plus isotope mass spectrometer (Thermo Fisher Scientific) was used to perform compound-specific δ13C analysis of individual hydrolysable lipids. The settings of the instrument and temperature program used here was the same as mentioned above. Reproducibility and stability (<0.6 ‰) of δ13C values were checked with pulses of CO₂ reference gas and n-alkane standard mixture (C_20–30; Sigma Aldrich) of known isotope composition. The δ13C values were presented in per mil (‰) relative to the Vienna-Pee Dee Belemnite (V-PDB) reference standard. Every sample was measured with three analytical replicates and the difference between measurements typically did not exceed 1.0 ‰ for natural abundance samples and 10 % of the measured isotope value for ¹³C labeled samples.

The isotope composition of individual hydrolysable lipids was corrected for the value of the δ13C value of each trimethylsilyl group that was added during silylation as: 3δUD=(n+3×a)×δD-3×a×δMn where n is the number of carbon atoms in the underivatized hydrolysable lipids and δUD and δD are isotope ratios of the underivatized and the derivatized hydrolysable lipids, respectively, a is the number of functional groups in individual compounds that were derivatized by BSA. δM is the carbon isotope ratio of the added trimethylsilyl group (-44.3 ‰). δM was determined by repeated measurement (n=8, with 3 analytical replicates of each) of derivatized standard FAs (C₁₀ and C₁₂ FAs with known δ13C isotope composition).

The ¹³C-excess, which can be expressed as percent atom excess, presents the enrichment of ¹³C in individual hydrolysable lipids. The value is defined as the ratio of the relative abundance of the heavier stable isotope in a labeled sample to the natural isotope abundance in the identical unlabeled sample (Epron et al., 2012). It was calculated as followings (Speckert et al., 2023): 413C-excess[%]=10013C/12Cdistcontrol×13C/12Clabelled-100 where ¹³C / ¹²C_distcontrol is the atomic ratio of the stable isotopes in the compartments (bulk soil carbon, and individual monomers of hydrolysable lipids) of the disturbance control plots as natural abundance values, and ¹³C / ¹²C_labeled is the atomic ratio in the corresponding compartments in the plots with added labeled root litter.

As individual isotope values can vary a lot in between different homologues for each compound-class specifically in isotope labeling experiments, a more meaningful measure was chosen to express the δ13C values of the respective compound-classes that can be assigned to the same carbon source (Wiesenberg et al., 2008). The δ13C values and ¹³C-excess for the most abundant compound classes (n-alcohols, n-fatty acids, diacids, and ω-hydroxy acids) within the hydrolysable lipid fractions were calculated separately as weighted means of individual compounds within each compound class. This means within each compound class, weightings will be given to each monomer in this compound class (e.g., ω-hydroxy acids) based on the proportional contribution of individual monomer (e.g., C₂₄ ω-hydroxy acids) to the total concentration of this compound class. Then for individual monomers, their weightings will be multiplied by their δ13C values, and we sum up all the monomers identified to get the weighted mean δ13C values for this compound class: 5μc=∑i=abxci×wci where μ denotes the average value and subscript “c” represents different compound classes, x denotes the value of either δ13C or ¹³C-excess, a and b represent the lower and upper limits of the respective carbon number range, wi indicates the relative abundance of the individual compounds within compound class c.

The amount of root carbon that was recovered in bulk soil was calculated by dividing the amount of ¹³C-labeled root-derived carbon left in the soil by the amount of carbon added with the original labeled roots. The proportion of root-derived carbon (froot) was estimated by using a simple mixing model (Hicks Pries et al., 2018): 613Catom%sample=13Catom%DC×fsoil+13Catom%root×froot7froot+fsoil=18Recoveryroot=froot×Msoil×C%soil0.14×0.463 where ¹³C atom %_sample is the ¹³C atom % of soil samples where labeled root litter was added; ¹³C atom %_DC is the ¹³C atom % of soil samples in the disturbance control plots; ¹³C atom %_root is the ¹³C atom % of initial ¹³C labeled roots; fsoil and froot are the proportion of carbon originally derived from native soil and labeled root litter, respectively. Msoil is the mass of the soil sample and C %_soil is the carbon content of the corresponding soil sample. 0.14 is the mass of root litter (g) added at individual soil depth and 0.463 is the carbon content of the added root litter.

The decay rate, k, of initially added ¹³C-labeled roots and root-derived hydrolysable lipids was calculated based on the following model (Olson, 1963): 9-kt=ln⁡MtM0 where M0 is the mass of original roots or root-derived hydrolysable lipids, Mt denotes the mass of root-derived carbon or hydrolysable lipids at time t, and t is the duration of incubation which is 3 years in our study. The model assumes that M is a well-mixed carbon pool with first-order decay kinetics. The residence time is the reciprocal of k.

The priming effect of added ¹³C labeled root litter was calculated using the mass of carbon in DC cores at individual depths as background values. Then, based on the calculations shown in Eq. (6), the proportion of native SOM that is left in the labeled cores within the same plot after 3 years of incubation was calculated: 10Primingeffect=MSOMl-MSOMDCMSOMDC where MSOM_l and MSOM_DC denote native SOM left in labeled cores and SOM in DC cores, respectively. The priming effect for hydrolysable lipids was calculated with the same approach. One outlier is excluded in priming effect calculation at mid-depth in ambient plots. It is noteworthy that this is only an estimated proxy used for priming effect. The more classical and straightforward way to calculate priming effect is described in Schiedung et al. (2023).

All statistical analyses were performed in the RStudio interface 2024.12.1.563 (Posit team, 2025) with R version 4.4.2 (R Core Team, 2024). We used linear mixed effects models (LMEs) in the “nlme” package (v3.1.166; Pinheiro et al., 2024) to test the fixed effects on responsive variables. All the plots were created using the “ggplot2” package (v3.5.1; Wickham, 2016). Homoscedasticity and normality were visually checked by residual plots and Q–Q plots. If model assumptions were not met, the data was log-transformed.

Response of bulk soil organic carbon concentration was tested to warming, root treatment, depth, and their interaction. ¹³C-excess was tested in response to warming, depth, and their interaction. The hydrolysable lipids concentration normalized to soil organic carbon (SOC; mg g⁻¹ SOC) was tested in response to warming, root treatment, depth, and their interaction. We tested warming, depth, and their interaction on hydrolysable lipids mass change, and ¹³C-excess of each compound class separately. To test whether chain length increases suberin-derived monomers resistance to decomposition, we used warming, depth, carbon number (refers to the total number of carbon atoms in the backbone of the respective suberin monomers), and their interactions as fixed effects to evaluate their effects on mass change of individual ω-hydroxy acid and diacid monomers. For each statistical test, we regarded block as a random effect.

Model fits were evaluated by Akaike's Information Criterion (AIC; Akaike, 1998). We conducted a backward stepwise model selection to remove fixed effects that increased AIC values by 5 or more (Pegoraro et al., 2024). The alpha level was set to α=0.05 as significant in all statistical tests and a p between 0.05–0.1 was considered marginal.

In ambient plots, the normalized hydrolysable lipids concentration decreased from 18.4±4.4 mg g⁻¹ SOC in topsoil to 11.0±2.7 mg g⁻¹ SOC in the deep soil, whereas it decreased from 18.6±3.8 mg g⁻¹ SOC in topsoil to 12.5±2.1 mg g⁻¹ SOC in the deep soil in warmed plots (Fig. 1a). However, this average decrease of hydrolysable lipids concentration with depth was not significant (p=0.460; Table S10). Warming did not significantly alter the normalized hydrolysable lipids concentration (p=0.472; Table S10) across the soil profile.

In general, warming did not significantly alter the relative contribution of each compound class to the total hydrolysable lipids (p>0.1; Table S21) and depth only significantly decreased proportion of mid-chain acids (p=0.003; Table S21). In ambient topsoil, ω-hydroxy acids accounted for the largest proportion of total hydrolysable lipids by 31 %, followed by fatty acids (27 %), mid-chain acids (21 %), diacids (13 %), while alcohols only make up 9 % of the total (Fig. 1; Table S22). At the same depth, warming did not significantly affect the relative contribution of each compound class (p>0.1; Table S22). At mid-depth (45–49 cm), the proportion of each compound class to total hydrolysable lipids was consistent compared to topsoil although contribution of mid-chain acids marginally decreased by 7 % (p=0.072; Table S22). In the deep soil (85–89 cm), the proportion of mid-chain acids further significantly decreased by 10 % compared to its contribution in topsoil (p=0.011; Table S22). Albeit much higher average contribution of fatty acids (45 %) in ambient deep soil, it did not significantly differ relative to topsoil (p=0.360; Table S22). At this depth, warming did not significantly alter the proportions of all the compound classes. However, elevated temperature marginally decreased fatty acids and increased diacids proportion by 16 % (p=0.084; Table S23) and 3 % (p=0.064; Table S23), respectively.

¹³C-excess of bulk SOC did not significantly change with warming (p=0.527; Table S8) but increased significantly with depth (p=0.001; Table S8). At 10–14 cm, the ¹³C-excess of bulk SOC was on average 17.0 % lower in warmed than in ambient plots (Table S9), but this difference was not significant (p=0.527; Table S8). At 45–49 cm, ¹³C-excess of bulk SOC on average increased by 2.7 times from 3.8 % in topsoil to 10.5 % at this depth (p=0.016; Fig. 2; Table S9). In the deep soil, the ¹³C-excess on average furthered increased to 22.8 % (Fig. 2).

Similar to bulk SOC, the weighted ¹³C-excess of each compound class was not significantly altered by warming (p>0.05; Table S12) but significantly increased with depth (p<0.05; Fig. 2b; Table S12). In topsoil, warming on average decreased weighted ¹³C-excess for all the compound classes between 3.2 % (diacids) and 8.5 % (alcohols) compared to ambient conditions, however, this decrease was not significant (p>0.1; Table S12). At 45–49 cm, warming non-significantly increased weighted ¹³C-excess for all the compound classes (p>0.1; Table S14), except for fatty acids, where warming on average decreased their weighted ¹³C-excess (p=0.920; Table S14). In the deep soil, warming on average largely reduced the weighted ¹³C-excess of fatty acids (28.0 %; p=0.356; Table S14), ω-hydroxy acids (17.1 %; p=0.728; Table S14), and diacids (8.2 %; p=0.760; Table S14), and slightly lower the weighted ¹³C-excess of alcohols (0.9 %; p=0.989; Table S14) and mid-chain acids (3.0 %; p=0.925; Table S14), although none of them was significant. The weighted ¹³C-excess of fatty acids increased significantly from 10–14  o 45–49 cm by 25.5 % (p=0.039; Table S13), whereas the weighted ¹³C-excess all the other compound classes non-significantly (p>0.1; Table S13) increased between 2.5 % (alcohols; Table S13) to 3.2 % (diacids; Table S13) at the same depth. At 85–89 cm, the weighted ¹³C-excess of fatty acids, ω-hydroxy acids, and diacids significantly increased compared to the values at 10–14 cm by 62.2 % (p=0.004; Table S13), 62.9 % (p=0.033; Table S13), and 38.1 % (p=0.015; Table S13), respectively. The weighted ¹³C-excess of alcohols and mid-chain acids marginally increased by 73.3 % and 36.7 % correspondingly (p=0.054 and 0.051, respectively; Table S13) at the same depth.

The mass change of each compound class represents the proportion of hydrolysable lipid-derived compound classes from the ¹³C-labeled root litter that remained after 3 years of incubation (Fig. 3). This metric allowed us to evaluate whether these compound classes experienced a net loss or accumulation during the incubation. Values above and below 100 % indicates accumulation and loss of corresponding compound classes, respectively.

The mass changes of different compound classes were affected differently by warming and soil depth. Both warming (p=0.035; Table S15) and depth (p=0.008; Table S15) had significant impacts on fatty acid mass change and the impacts of warming was independent from depth (p=0.409; Fig. 3a, Table S15). Fatty acids significantly accumulated across the soil profile, resulting in an increased their mass change by 67 % (p=0.044; Table S16) and 102 % (p=0.006; Table S16) at mid-depth and in the deep soil, respectively. Although non-significantly at 10–14 and 45–49 cm, warming lead to a consistent decreased fatty acid mass change across the soil profile, particularly at 85–89 cm, where there was a significant lower (74 %; p=0.029; Table S17) recovered fatty acids with elevated temperature. The accumulation of fatty acids at 45–49 and 85–89 cm derived primarily from Hexadecanoic acid (n-C_16:0), Octadecanoic acid (n-C_18:0), and Octadecenoic acid (n-C_18:1), where the mass change of these monomers exceeded 110 % or even 300 % (Table S2), e.g. Octadecanoic acid (n-C_18:0) at 85–89 cm in ambient and warmed plot (348±49 % and 181±24 %; Table S2).

The effects of warming on the mass change of mid-chain acids were marginally depth-dependent (p=0.057; Table S15). In the topsoil, warming significantly decreased the mid-chain acids recovered from hydrolysable lipids (p=0.029; Table S17), whereas warming non-significantly increased the mass change of this compound class at 45–49 cm (p=0.257; Table S17) and at 85–89 cm (p=0.729; Table S17). Neither warming nor depth had significant effects on the mass changes of all the other three compound classes (p>0.1; Table S15). There was a general trend that warming on average decreased the mass changes of alcohols, ω-hydroxy acids, and diacids in the topsoil, but increased them at mid-depth and in the deep soil (Table S17), however, none of these changes were significant (p>0.1; Table S15).

For fatty acids with a carbon number ≥ 20, we did not observe significant warming effects on the mass change (p=0.255; Fig. 3b, Table S18), but marginal depth effects (p=0.076; Table S18). Warming on average reduced the mass change of fatty acids by 12 % at 10–14 cm, 7 % at 45–49 cm, and 16 % at 85–89 cm, but these effects were non-significant (p=0.260, 0.501 and 0.141, respectively; Table S20). Similar to fatty acids, there was no significant effects of neither warming nor depth on the long-chain ω-hydroxy acids and diacids mass change, and warming effects was not dependent on depth (p>0.1; Table S18). At 10–14 cm, warming led to non-significant decreases in the ω-hydroxy acids and diacids mass change by 16 % and 22 % (p=0.334 and 0.316, respectively; Table S20). In contrast, although warming on average enhanced the mass change of the two compound classes at both 45–49 and 85–89 cm, none of the changes were statistically significant (p>0.1; Table S20).

Neither depth nor warming had significant impacts on the mass differences of pre-existing SOC (p=0.558 and 0.422, respectively; Table S24). In topsoil, labeled root addition led to SOC accumulation on average by 28 % in ambient plots, but this accumulation was not significant (p=0.273; Table S25). The effects of root addition decreased on average by 37 % at mid-depth and 32 % at 85–89 cm, resulting in an on average loss of pre-existing SOC (Fig. 4a). However, these effects were non-significant (p=0.359 and 0.375, respectively; Table S25). Warming caused an average decline of the bulk SOC mass change at 10–14 and 85–89 cm by 17 % and 14 % (Fig. 4a), and no-change at mid-depth, but none of them were significant (p=0.629, 0.684 and 0.991 for 10–14, 85–89, and 45–49, respectively; Table S26).

The similar trend was observed for the mass change of pre-existing hydrolysable lipids as for pre-existing SOC, but the direction was opposite in the topsoil and the deep soil (Fig. 4b). Neither depth nor warming significantly altered the relative mass differences of pre-existing hydrolysable lipids (p=0.310, and 0.987 for depth and warming, respectively; Table S24). At 10–14 cm, labeled root addition led to a non-significant 19 % loss of pre-existing hydrolysable lipids (p=0.596; Table S25). In the subsoil, the responses of relative hydrolysable mass change were contrasting between 45–49 and 85–89 cm. At mid-depth, labeled root addition caused a further non-significant decrease of the mass change of hydrolysable lipids by 35 % (Fig. 4b; p=0.529; Table S25), whereas in the deep soil, it caused a non-significant increase of the hydrolysable lipids mass change by 45 % compared to topsoil (Fig. 4b; p=0.372; Table S25). In the topsoil, warming non-significantly decreased the mass change of hydrolysable lipids by 4 % (p=0.941; Table S25). In the subsoil, warming had also contrasting effects on the mass change of hydrolysable lipids (Fig. 4b) between mid-depth and deep soil. Warming on average increased the mass change of hydrolysable lipids at 45–49 cm by 30 % and decreased it by 14 % compared to corresponding ambient conditions, but none of them was significant (p=0.587 and 0.784, respectively; Table S26).

To further explore other factors governing the decomposition of suberin markers in our study, we examined the effects of chain length (number of carbon atoms in each monomer) on the mass change of ω-hydroxy acids and diacids after 3 years of incubation. Chain length had a significant effect on the mass change of suberin markers (p<0.001; Table S27) that they decomposed more slowly with longer chain length. This pattern is independent of warming, depth, and compound class. Our results are in line with multiple previous studies, which also reported decreased decomposability of hydrolysable lipids with increasing chain length by examining either single compound class (Kashi et al., 2023), or suberin- and/or cutin-derived polymers (Altmann et al., 2021; Angst et al., 2016a; Suseela et al., 2017). Generally, molecules with a low oxidation state or low O / C ratio (i.e., more reduced compounds such as lipids) provide less favorable energy yields for microbial oxidation, which often translates into slower decomposition rates (Kleber, 2010; LaRowe and Van Cappellen, 2011). As chain length increases, the carbon atoms in respective compound classes exhibit progressively lower oxidation states (Chakrawal et al., 2020). Therefore, the microbial utilization of long-chain lipids requires more energy investment than that of mid-chain lipids (Schönfeld and Wojtczak, 2016), which may explain the slower decomposition of long-chain ω-hydroxy acid and diacid observed in our study.

It is noteworthy that when chain length was tested together with warming and depth on the mass change of ω-hydroxy acids and diacids monomers, warming emerged as a significant factor (p<0.001; Table S27), and its effect was depth-dependent (p<0.001; Table S27). Specifically, warming significantly enhanced the loss of suberin monomers in the topsoil (p=0.002; Table S29), but reduced it at 45–49 and 85–89 cm (p<0.001 and p=0.008, respectively; Table S29). This pattern was also observed at compound class level, although with limited statistical power due to limited number of replicates. The stronger statistical significance at the monomer level reflects the greater number of observations (i.e., higher degree of freedom) and lower variability within individual monomers compared to the aggregated variability among monomers when grouped at the compound class level. Therefore, our results can confirm that the effect of warming on suberin marker decomposition was depth-dependent.

Root-derived hydrolysable lipids, especially those suberin markers, degraded more slowly than bulk root carbon, indicated by higher proportions of hydrolysable lipids remaining in the soil than the bulk root recovery (Pegoraro et al., 2024). The depth-dependent effects of warming lead to depth-specific mean residence time (MRT) of these suberin markers (chain length ≥ 20). In the topsoil, warming reduced the MRT of ω-hydroxy acids and diacids due to accelerated decomposition from 5.8 and 5.6 years in ambient plots to 3.4 and 3.6 years. In the subsoil, slower decomposition of suberin markers caused by warming enhanced MRT compared to ambient conditions, particularly at 45–49 cm with warming, where the MRT for of ω-hydroxy acids and diacids was 11.9 and 15.0 years, respectively.The suberin markers had a shorter MRT in our study compared to a previous study, where Feng et al. (2010) reported a decadal MRT of hydrolysable lipids between 32 to 34 years. This discrepancy could be related to the fact that our model substrates are grass roots (Avena fatua) with less lignin and lower C / N ratios than local woody roots (Hicks Pries et al., 2018; Silver and Miya, 2001). Hydrolysable lipids without association to lignin could be steadily decomposed (Angst et al., 2016a). Additionally, an underestimation of MRT could exist in our study since we have a much shorter experimental duration compared (Feng et al., 2010).

In prior study, Suseela et al. (2017) reported that elevated temperature and CO₂ increased the chain length of ω-hydroxy acids in grass roots (Boutelou gracilis). Together with our finding that the decomposability of suberin markers decreases with chain length, this raises the possibility that plant physiological adaptations to future warming could enhance the persistence of root compartments against microbial decomposition and consequently mitigate potential soil carbon losses under warming.

Fresh biomass input can stimulate (prime) the decomposition of native, pre-existing SOM. This is especially relevant in subsoils, where SOM might have existed for a long time, from decades to millennia (Fontaine et al., 2007; Luo et al., 2019; Shahzad et al., 2018). Such priming could offset long-term carbon sequestration, especially in subsoil where there is usually substrate limitation (Bernard et al., 2022; Bingeman et al., 1953).

Three years after adding ¹³C-labeled root at three different soil depths, we found no evidence for significant priming across the whole soil profile, either for pre-existing bulk carbon or hydrolysable lipids. This likely reflects the fact that fresh input was added only at the beginning of the incubation. Priming is a transient response to fresh carbon input (Schiedung et al., 2023) and is commonly strongest at the beginning of the incubation (Fontaine et al., 2007; Tao et al., 2024), a pattern that has been also reported in lipid decomposition experiments (Angst et al., 2016a; Kashi et al., 2023). Therefore, any initial effect may have dissipated in the long-term. In addition, our calculations of priming effects showed large error bars, likely due to heterogeneous background SOC concentrations, particularly in the subsoil. For example, the mass of SOC in root litter added to the topsoil (10–14 m) represented only 4.0 %–8.5 % of pre-existing SOC, whereas at 85–89 cm, the same litter addition corresponded to 10.5 %–39.3 % of pre-existing SOC. This wide range in the proportion of added to native SOC, alongside reduced microbial abundance under warming may have biased our estimates and masked potential priming effects. Therefore, although our back-of-the-envelope calculation provides a preliminary estimate, the strength of the interpretation is limited. Short-term monitoring of CO₂ fluxes and their partitioning between ¹³C-labeled root- and native SOC-derived, would offer more robust insights into whether root litter addition primes native SOC decomposition, particularly in the subsoil.