Turnover of non-polymeric leaf lipids in a loamy grassland soil

Leaves constitute a large input of lipids to soil, yet comparatively little is known about the fate of leaf lipids in soil. Our aim was to explore the initial stages of degradation of leaf lipids, both the loss of intact lipid and subsequent mineralisation. We focussed on intracellular lipids – triacylglycerols implicated in storage, membrane lipids such as phospholipids and galactolipids, and pigments – because they collectively constitute more than 1% of leaf mass. A mixture of U-13C lipids was extracted from leaves of wheat grown with 13CO2. The lipid mixture included the range of plant lipids soluble in organic solvent (e.g. free fatty acids, acylglycerols, pigments) but not polymeric lipids such as cutin and suberin. Mineralisation was deduced from 13CO2 efflux, while LC–MS examined degradation of intact 13C lipids. There was no delay before lipids were mineralised. Instead, mineralisation was significant within minutes and reached a maximum within three hours. There was rapid loss (i.e. degradation) of a broad range of intact lipids including phospholipids, galactolipids, pigments (chlorophylls), and triacylglycerols. Around two-thirds of added lipid-C was respired over the course of 15 days, with one-third of lipid-C persisting in soil. Our study indicates that non-polymeric leaf lipids degrade quickly in soil, yet a fraction of lipid-C likely persisted in degradation products and/or microbial biomass. Persistence of lipid-C probably also reflected the presence of lipids that are more resistant to degradation (e.g. phaeophytins), and a fraction of added lipid being protected (e.g. by interaction with clays).


Introduction
Lipids are a significant pool of ecosystem C comprising an average of 5% of the dry weight of plants, 9% of bacteria and 17% of fungi (Ohlrogge and Browse 1995;Horwath 2015). The dominant lipids in many living organisms are esters of one or more fatty acids with glycerol (acylglycerols) and predominantly occur within cells (i.e. are intracellular). Acylglycerols comprise two broad functional groups; the nonpolar triacylglycerols (TG) that are often associated 1 3 Vol:. (1234567890) with storage of C and energy (Murphy 1990;Bago et al. 2002;Bååth 2003) and polar lipids that are constituents of membranes (Ohlrogge and Browse 1995;Dörmann and Benning 2002;Sturt et al. 2004). Polar membrane lipids include phospholipids that are comprised of two fatty acids, a glycerol moiety, and a phosphate group that may be adorned with other molecules such as ethanolamine (phosphatidylethanolamine, PE), choline (phosphatidylcholine, PC), inositol (phosphatidylinositol, PI), or serine (phosphatidylserine, PS). In addition to phospholipids, membranes of photosynthetic plants and microbes can include large amounts of galactolipids (e.g. MGDG, DGDG) and sulfolipids (e.g. SQDG) (Wintermans 1960;Roughan and Batt 1969;Harwood 1980), while some microbes have ornithine lipids (Moore et al. 2015) or betaine lipids (Sato and Murata 1991). Other lipids that are abundant in membranes and can have taxonomic significance include photosynthetic pigments (e.g. chlorophylls), sterols and their derivatives. The polymeric lipids of plants (e.g. cutin and suberin) and microbes (e.g. polyhydroxyalkanoates), can account for a significant fraction of total lipid, but are beyond the scope of this study.
Lipids present in soil are assumed to reflect the major contribution of bacteria and fungi to the soil biomass (Brady 1990) and inputs of lipids from plants and (to a lesser extent) animals. However, the soil lipid profile is not a simple concatenation of soil organisms plus plant inputs. Galactolipids -the most abundant group of lipids in leaves (Wintermans 1960;Roughan and Batt 1969;Harwood 1980) -are notably absent from the soil lipid profile (e.g. Liu et al. 2010;Peterse et al. 2011;Warren 2019;Ding et al. 2020), suggesting galactolipids entering soil from leaf litter must be rapidly degraded. Differences among lipid classes in rates of cycling have been implied from changes in lipid profiles during early diagenesis (Bull et al. 2000;Jandl et al. 2005). For example, changes of leaf lipids during early diagenesis in soil suggested degradation was rapid for chlorophyll derivatives (quantified as phytyl esters) and fatty acid-based lipids (acylglycerols) but slower and more variable for triterpenoids (Nguyen Tu et al. 2017). Likewise for marine sediments changes during early diagenesis implied degradation was fastest for phaeopigments (degradation product of chlorophylls) > sterols > fatty acids > hydrocarbons (Colombo et al. 1997). Descriptive evidence based on lipid compositional changes during diagenesis suggest many acylglycerols turn over rapidly (Colombo et al. 1997;Nguyen Tu et al. 2017), but the experimental evidence is not so clear.
Previous experimental studies of acylglycerols focussing on phospholipids yielded contradictory estimates of turnover, and the scant evidence for nonpolar lipids is similarly contradictory. Experimental addition of phospholipids indicates turnover on a time scale of hours to days, and this is consistent with the typically low concentrations of phospholipid in soil (White 1993;Kuzyakov et al. 2014;Dippold and Kuzyakov 2016;Zhang et al. 2019). Nevertheless, there is approximately an order of magnitude variation among studies in the turnover of phospholipids. It is likely that much of this variation can be explained by the form in which phospholipids are added and whether turnover is gauged from the initial disappearance of intact lipid or its ultimate mineralisation. For example, degradation of free phospholipids is faster than degradation of phospholipids within microbial cells (Klamer and Bååth 1998;Kindler et al. 2009;Zhang et al. 2019). Some studies quantified the disappearance of intact lipid from soil, which is presumed to represent the first stage of phospholipid degradation in which intact lipids undergo enzymatic cleavage (Zhang et al. 2019); whereas other studies quantified mineralisation and thus encompassed the initial enzymatic cleavage step plus subsequent microbial uptake and use of lipid degradation products as carbon sources. The additional steps and extra time required for mineralisation is supported by a study with marine sediment showing added 14 C-phospholipid hydrolysed rapidly with a half-life of 12 h, whereas 14 CO 2 did not appear until 24 h had elapsed (Harvey et al. 1986). Less is known about degradation of non-polar acylglycerols such as the triacylglycerols (TG) often implicated in C and energy storage. One study indicated TG are degraded within weeks of addition to soil (Hita et al. 1996) whereas another suggested half-lives of TG in soil were 4-6 months (Soliman and Radwan 1981). However, both studies added TG to soil at orders of magnitude larger concentrations than occur in soil and thus it is possible that steady state TG turnover would be faster than reported.
Litterfall likely represents one of the largest fluxes of lipid input into soil given that lipids commonly comprise up to 5% of the dry mass of plants (Ohlrogge and Browse 1995), yet little is known about degradation of some of the most abundant lipid classes from leaves and some results are contradictory. For example, studies of soil and sediment diagenesis implied chlorophylls are degraded rapidly (Colombo et al. 1997;Nguyen Tu et al. 2017), whereas experimental evidence showed degradation is slow with some 90% of added chlorophylls remaining intact after 30 d (Simonart et al. 1959) and requiring two to four months to be degraded (Hoyt 1966). Anecdotal evidence suggests turnover of acylglycerols must be rapid. For example, rapid turnover of plant acylglycerols is suggested by the complement of acylglycerols in soil extracts more closely resembling soil microbes than plants (Ding et al. 2020). However, there is no experimental information on the cycling of the galactolipid and sulfolipids that can account for 50% of leaf lipids (Wintermans 1960;Roughan and Batt 1969;Harwood 1980). Uncertainty about the fate of plant-derived galactolipids and sulfolipids highlights the more general problem that little is known about the first hours of degradation when rates of loss and mineralisation should be fastest. For example, a recent study showed loss of 13 C-phospholipids was evident at two hours (Zhang et al. 2019), but no measurements were made at shorter intervals and inferences cannot be made about mineralisation because 13 CO 2 efflux was not quantified.
The broad aim of this study was to investigate the initial stages of degradation of non-polymeric leaf lipids added to soil. For some lipid classes there is already information on degradation, but existing estimates of degradation are not comparable among studies because lipids were added at different (and often unrealistically high) concentrations (e.g. Soliman and Radwan 1981). To provide new experimental evidence about turnover of a broad range of lipids, we added to soil a complex lipid mixture at tracer level concentrations to ensure steady-state processes were perturbed as little as possible. The lipid mixture was obtained by Bligh and Dyer (1959) extraction of wheat grown with 13 CO 2 , and thus included the broad range of leaf lipids soluble in methanol/ chloroform/water (e.g. free fatty acids, acylglycerols, pigments) but not the solvent-insoluble polymeric lipids (e.g. suberin and cutin). To examine the loss of intact lipids (e.g. due to enzymatic cleavage or microbial uptake) we used LC-MS measurement of acylglycerols and pigments, while the mineralisation of lipids was determined from tuneable diode laser measurements of 13 CO 2 efflux. These measurements enabled us to test whether mineralisation of added lipids is slow and lags 12 + hours behind loss of intact lipids (Harvey et al. 1986). Finally, we were able to bring direct experimental evidence to examine the contradiction between rapid loss of chlorophylls during diagenesis versus slow degradation of added chlorophylls (Simonart et al. 1959;Hoyt 1966;Colombo et al. 1997;Nguyen Tu et al. 2017).

Soil
To examine lipid degradation, we used soil from a long-term mesocosm study that has been described previously (Warren 2014). Replicate 200-L mesocosms (572 mm diameter, 851 mm high) were filled in June 2009 with loam soil from Themeda triandra grassland in western Sydney. Climate is humid subtropical with average annual maximum 23℃, minimum 12℃, approximately 900 mm annual precipitation evenly spread through the year. The grassland was largely unmanaged, and the region has historically been open (< 10% tree cover) woodlands with scattered or clumped shrubs, and a dense ground cover of perennial grasses (principally T. triandra but also Microlaena stipoides) and perennial herbs. The intact soil was an abruptic lixisol derived from shale and alluvium. Mesocosms were planted with two perennial native grass species -Themeda triandra and Microlaena stipoides -and grown for 12 years with adequate water. At the time of sampling Themeda triandra accounted for > 90% of plant biomass in all replicate mesocosms. Experiments examining lipid turnover used freshly collected 0-15 cm mesocosm soil.

Pools of intracellular lipids in Themeda triandra leaves, roots and mesocosm soil
To examine if the signature of intracellular plant lipids is evident in soil, the lipid profile of leaves and roots of Themeda triandra was compared with the mesocosm soil it had grown in for 12 years. Lipids were extracted from fresh 0-15 cm soil, green leaves and fine roots based on the Bligh and Dyer extraction chemistry (Bligh and Dyer 1959) using citrate buffer (0.15 M, pH 4.0) (Frostegård et al. 1991) as described in more detail recently (Warren 2018). The extracts of leaves and roots would contain lipids of intracellular and extracellular origin, but only intracellular lipids were quantified owing to the specificity of the LC-MS protocol (Warren 2018). Intact lipids were analysed by reversed-phase LC-MS n using a nano-LC system (Ultimate 3000 RSLCnano, Thermo Scientific Dionex, Germering, Germany) interfaced to a mass spectrometer (AmaZon SL, Bruker Daltonics, Bremen, Germany). Samples of 50 nL were injected via partial loopfill onto a nano-scale analytical column (150-mm long,75 μm i.d., packed with Acclaim PepMap RSLC C18, 2 µm, 100 Å) and separated with mobile phases A (40% water, 60% acetonitrile, 10 mM ammonium formate, 0.1% formic acid), and mobile phase B (90% isopropanol, 10% acetonitrile, 10 mM ammonium formate, 0.1% formic acid). Total mobile phase flow was 500 nL min −1 , with 45% B maintained for one minute after injection, followed by a 45-min ramp to 99% B where it was held for 10 min before a one-minute ramp back to 45% B and a 12-min re-equilibration. Identification was based on MS 1 , class-specific neutral losses and/or product ions (determined via Auto-MS n ), and retention time (Warren 2018).

Development of a labelled lipid-water suspension
Given our interest in determining how rapidly lipids are mineralised we developed a rapid labelling strategy based on injecting labelled lipids as a suspension carried by water. We rejected the option of adsorbing lipids onto sand then mixing the dry lipid-sand into soil (e.g. as used by: Zhang et al. 2019) because mixing lipid-coated sand into soil would be slow with the 1.25 kg soil aliquots required for measuring 12 CO 2 / 13 CO 2 efflux. We also rejected the alternative of adding lipids solubilised in organic solvent or surfactant because solvent or surfactant would artefactually solubilise membranes of soil microbes. The lipid-water suspension was created and characterised as described in Fig. S1. In brief, a mixture of 13 C-labelled leaf lipids was extracted from 13 C-labelled wheat leaves (> 97 atom% 13 C, IsoLife, Wageningen, Netherlands) using the Bligh and Dyer extraction chemistry (Bligh and Dyer 1959). Analogous lipid mixtures at natural isotopic abundance were obtained by extracting leaves of wheat grown in air at natural isotopic abundance. The organic (lipid) phase of Bligh and Dyer extracts was gently dried under a stream of N 2 gas. To suspend lipids, water was added to dry lipid extract at a ratio of 8.0 mL of water to 1.0 g of leaf extract, then lipids were brought into suspension by multiple cycles of ultrasonicating (40 kHz, 100 W) for 1 min then vortexing for 1 min. For LC-MS quantification of lipid loss we used suspensions of the 13 C lipid mixture, which was at 97.4 atom % 13 C (see below for details of how atom % 13 C was estimated). For tuneable diode laser quantification of mineralisation, we used a lipid mixture at 3.75 atom% 13 C obtained by mixing the 13 C lipid mix with the natural abundance wheat lipid mix.
To quantify the amount and chemical composition of the lipid-water suspension, four replicate aliquots of the lipid-water suspension were subject to Bligh & Dyer extraction to re-solubilise lipids (Fig. S1). Lipid mixtures were expected to contain the broad range of leaf lipids soluble in methanol/chloroform/water (e.g. free fatty acids, acylglycerols, pigments) but not solvent-insoluble polymeric lipids (e.g. suberin and cutin). Our study focussed on intracellular leaf lipids most of which can be analysed as intact compounds by LC-MS (Warren 2018); nevertheless, we sought to obtain a broader chemical characterisation of the mixture due to the likely presence of lipids not amenable to the LC-MS protocol. Hence the solubilised lipid mix was analysed by three complementary methods: LC-MS (to quantify acylglycerols, pigments, sterol glucosides), acid-catalysed hydrolysis/esterification and GC-MS (to quantify hydrolysable + free fatty acids, sterols, and alcohols), methoximationtrimethylsilylation and GC-MS (to quantify free fatty acids, and lipophilic metabolites). LC-MS also enabled verification of the 13 C enrichment of the lipid mixture based on mass spectra of five intact lipids using the FluxFix package (Trefely et al. 2016), as described in more detail recently (Warren 2022).
To quantify hydrolysable + free fatty acids, sterols and alcohols via acid-catalysed hydrolysis/esterification, an aliquot of the Bligh & Dyer extract was dried with N 2 gas, then 10 μL of internal standard (C17:0-d33 at 1.0 mg mL −1 ) along with 1.0 mL of 2.5% H 2 SO 4 in methanol, then samples were heated at 70 °C for 90 min. Cooled samples were phase separated by adding 500 µL of hexane and 1500 µL of 0.9% NaCl. An aliquot of the organic phase (containing methyl esters of lipophilic compounds) was analysed by GC-MS (GCMS-QP2010Plus, Shimadzu, Kyoto, Japan) fitted with an arylenemodified 5% diphenyl-95% dimethyl polysiloxane stationary phase (30 m long × 0.25 mm internal diameter × 0.25 μm film thickness; Rxi-5SilMS, Restek, Bellfonte, USA) and helium carrier gas. FAMES were quantified using a commercial FAME mixture (37 component fatty acid methyl ester mix from Sigma-Aldrich) plus retention indices (based on even numbered alkanes from C8 to C36) and EI mass spectra from the NIST library.
To analyse free fatty acids and lipophilic metabolites we used a generic methoxyamination-trimethysilylation derivatisation that gives good coverage of a broad range of lipophilic metabolites (Lisec et al. 2006;Warren 2014). Two internal standards were added (10 μg of C17:0 fatty acid and 1 μg of ribitol) then Bligh & Dyer extracts were dried with N 2 gas and derivatives prepared by adding 40 μL of methoxyamination reagent (20 mg mL −1 methoxyamine hydrochloride in pyridine), incubating at 37ºC for 90 min, then adding 70 μL of N-Methyl-N-trifluoroacetamide (MSTFA) and incubating at 37ºC for 30 min (Lisec et al. 2006). Derivatives were analysed by GC-MS and identified by reference to authentic standards and published mass spectra, as described previously (Warren 2014(Warren , 2020. Hydrolysable fatty acids were calculated as total fatty acids (determined via acid-catalysed hydrolysis/ esterification) minus free fatty acids (determined via methoxyamination-trimethylsilylation).
Quantification of soil CO 2 efflux and mineralisation of added lipid Mineralisation was determined from continuous measurements of soil efflux of 12 CO 2 and 13 CO 2 with a tuneable diode laser. Six soil respiration chambers (Ball Mason wide mouth preserving jar, Colorado, USA) were filled with 1.25 kg (fresh weight) of soil packed to field density ( Fig. 1). Soil respiration chambers were installed in the tuneable diode laser system and allowed to equilibrate for three days until rates of soil CO 2 efflux were steady. To determine mineralisation of added lipid to CO 2 , three replicate soil respiration chambers each received a 37.5 mL lipid-water suspension while three replicates served as controls and received 37.5 mL of water. Based on GC-MS quantification of total and free lipophilic compounds, the lipid-water suspension delivered 434 μmol lipid C per kg of soil (25 μmol of fatty acid per kg of soil) at 3.75 atom% 13 C. The lipid-water suspension was injected into the soil in 10 aliquots of approximately 3.75-mL using a 4-sideport Cass needle (18 gauge × 150 mm long, Victor-G & Company, Kanpur, India). To obtain a homogenous suspension the syringe was agitated vigorously prior to each injection, then the needle was pushed to near the bottom of the soil respiration chamber and thence slowly lifted up through the column of soil while dispensing the lipid-water suspension. Visual analysis suggested that this procedure led to the soil being evenly wet. Prior to injection the soil had volumetric water content 70% of field capacity, while after adding lipid-water suspension or water soil water content was around 85% of field capacity. 12 CO 2 and 13 CO 2 efflux were determined with a tuneable diode laser (TGA100a, Campbell Scientific, Logan, UT, USA) from continuous measurements of 12 CO 2 and 13 CO 2 from the soil headspace. Air from outside the laboratory was drawn at 2.5 L min −1 into a 10-L buffer volume and humidified to a dew-point of 15 °C (LI-610 dew-point generator, LI-COR, Lincoln, NE, USA). Air from the dew-point generator was split into eight streams: six served as inlet streams for the soil respiration chambers, one was directed through an empty soil respiration chamber and served as the reference gas, while the eighth stream was a vent to avoid over-pressurising the soil headspace. Samples of air from headspace of the six soil chambers, and reference gas were continuously drawn into the manifold system of the tuneable diode laser at 170 ± 10 mL min −1 . The tuneable diode laser was programmed to measure 13 CO 2 and 12 CO 2 of, in turn, the two calibration gases (416 μmol mol −1 , δ 13 C 8.5‰ and 612 μmol mol −1 , δ 13 C 35.0‰) the six soil chambers and empty reference chamber. Each intake was measured for 45 s, with the first 15 s ignored to minimize carryover and enable stabilization between intakes. Calibration of the tuneable diode laser involved linear interpolation between the two calibration gases, as described previously (Douthe et al. 2011).
Rates of soil respiration were calculated separately for 12 CO 2 and 13 CO 2 based on tuneable diode laser measurements of (dry) concentrations of 12 CO 2 and 13 CO 2 , and flow rate measured within the manifold system: where flow is the molar flow of air through the soil headspace, C out and C in are concentrations of 12 CO 2 or 13 CO 2 measured in the air exiting the chamber (C out ) or entering the chamber (C in , the reference gas). The carbon isotope composition of CO 2 , expressed as δ 13 C relative to the VPDB standard in permille units, (2) 13 CO 2 ef f lux = Flow 13 C out − 13 C in , was calculated from absolute concentrations of 12 CO 2 and 13 CO 2 : where R is 13 C/ 12 C of the VPDB standard. The δ 13 C of soil respiration (δ 13 C SR ) was calculated from isotopic mass balance:  C lipid (upper half) and loss of intact intracellular lipid from soil (lower half). To quantify mineralisation of added lipid, soil respiration chambers that contained 1.25 kg of soil received either 37.5 mL of a lipid-water suspension or 37.5 mL of water. 12 CO 2 efflux and 13 CO 2 efflux from the soil headspace were measured continuously with a tuneable diode laser system. To quantify loss of intact intracellular 13 C lipids from soil, a parallel experiment was carried out with 2.0 g subsamples of soil receiving 100 μL of lipid mix suspended in water, or 100 μL of water. Complementary LC-MS and GC-MS analyses ( Fig. S1) indicated > 85% of the lipid mixture was comprised of intracellular intact lipids amenable to LC-MS. The lipidwater suspension injected into respiration chambers delivered 25 μmol of fatty acid per kg of soil at 3.75 atom% 13 C, while the small 2.0 g soil aliquots received a comparable 25 μmol of fatty acid per kg soil but at 97.4 atom% 13 C. To disentangle biotic from abiotic loss of intact lipids, measurements made on soil that had been autoclaved four times over three days were contrasted with measurements on live soil. Arrows on the timeline indicate when soil samples were extracted by the Bligh & Dyer protocol: immediately prior to adding water or lipid (T0), then 15 min, 3 h and 1 day after water or lipid addition. For the live soil an additional sample was collected after five days, but autoclaved samples were not collected at 5 days day due to the likelihood of microbial recolonisation. Lipid extracts were subsequently analysed by LC-MS. Three biological replicates were used for all measurements where δ 13 C in and δ 13 C out are carbon isotopic compositions of CO 2 entering and leaving the chamber. The amount of respiration arising from added lipid (f lipid ) was calculated using a two-member mixing model from δ 13 C SR of lipid chambers (δ 13 C SR,lipid ), carbon isotope composition of the added lipid ( 13 C lipid ) and δ 13 C SR of control chambers δ. 13 C SR,control )

LC-MS quantification of loss of acylglycerols
To quantify loss of intact 13 C acylglycerols, three replicate 2.0 g FW aliquots of soil in glass centrifuge tubes received 100 μL of a 13 C lipid-water suspension then were extracted and analysed by LC-MS 15 min, 3 h, 1 day and 5 days after lipid addition (Fig. 1). The lipid-water suspension added to soil 25 μmol of fatty acid (at 97.4 atom % 13 C) per kg. The lipid was added by injecting the suspension onto the soil surface then gently mixing with a spatula. To disentangle the relative roles of abiotic versus biotic processes in the loss of intact lipids, an additional three replicate soils were autoclaved four times over the course of three days and then 13 C lipids were added as for the "live" soils except that samples were collected only at 15 min, 3 h and 1 day. The 5-day sample was not collected because it was reasoned microbial recolonisation of autoclaved soils was likely after 5 days. For the live and autoclaved soils, replicate samples were also extracted immediately prior to label addition (T zero). Lipids were extracted from soil based on the Bligh and Dyer extraction chemistry (Bligh and Dyer 1959) using citrate buffer (0.15 M, pH 4.0) (Frostegård et al. 1991) as described in more detail recently (Warren 2018). Blanks were extracted alongside samples.
For LC-MS analysis we optimised an existing method (Warren 2018) to improve detection limits for 13 C acylglycerols by analysing in positive mode via MS 2 using a scheduled precursor list. All 13 C lipids were quantified based on peak area of a (4) 13 C SR = ( 13 C out ⋅ ( 12 C out + 13 C out ) − 13 C in ⋅ ( 12 C in + 13 C in )) 12 C out + 13 C out − ( 12 C in + 13 C in ) (5) 13 C SR,lipid = f lipid 13 C lipid + f control 13 C SR,control (6) 1 = f lipid + f control carefully chosen MS 2 ion relative to MS 1 peak area of the instrumental internal standard. In addition to this relative quantification, phaeophytins A and B and their corresponding chlorophylls, were subject to absolute quantification. Standards of chlorophylls and phaeophytins were not available, so LC-MS data were cross-calibrated against spectrophotometric estimates. To cross-calibrate, lipids were extracted with methanol from wheat leaves, subject to a 3-step serial dilution, then analysed by LC-MS and spectrophotometrically using the HCl acidification approach (Lichtenthaler 1987) adjusted to work in microplate format (Warren 2008).

Statistics
Univariate statistics were computed with Minitab 17.3.1 (Minitab Inc., State College PA, USA) and multivariate statistics with PAST 4.03 (Hammer et al. 2001). Univariate lipid concentrations were compared via T-tests. To compare the intracellular lipid composition of leaves and roots with soil we ordinated relative lipid concentrations with non-metric multidimensional scaling (NMDS). The high dissimilarity in lipid profiles and consequent many zeros in the data matrix necessitated use of the Bray-Curtis similarity index (Legendre and Legendre 2012). The significance of multivariate difference in lipid profiles among leaves, roots and soil was examined with permutational ANOVA (PERMANOVA) (Anderson 2001) with Bray-Curtis similarities and 9,999 permutations of group membership. To examine if incubation duration and autoclaving affect quantitative U-13 C-lipid profiles we ordinated (via NMDS) logtransformed absolute concentrations using a Euclidean distance measure and determined statistical significance with PERMANOVA.

Pools of intracellular lipids in Themeda triandra leaves, roots and soil
Both multivariate and univariate statistics showed lipid profiles differed among Themeda triandra leaves, roots and the soil it had grown in for 12 years. For example, NMDS ordination at the level of lipid sum composition showed tight clustering of replicate samples with leaves, roots and soil clearly separated from each other (Fig. 2). The lipid profile of soil was closer in 2-D multivariate space to roots than leaves. Much of the difference among leaves, roots and soil was due to relative abundance of different lipid classes (Fig. 3), and there was the same general trend (as for NMDS) for the lipid profile of soil to be less dissimilar to roots and most dissimilar to leaves. The largest differences were galactolipids (i.e. DGDG, MGDG) that comprised a total of around 55% of the lipid profile of leaves, around 3% of roots and were below detection limits in soil extracts. Two phospholipid classes -PA and PG -accounted for 2-10% of leaf lipids, around 1% of root lipids but were below detection limits in soil. PC and PE were present in leaves, roots and soil. In leaves TG was less than 2% of quantified lipids, whereas TG was 50-80% of quantified lipid in roots and soil. The betaine lipid DGTS was detected in soil but not leaves or roots. In addition to fatty acid-based lipids (acylglycerols), chlorophylls and their oxidation products phaeophytin a and b were detected in leaves and soil but not roots, while leaves but not roots or soil contained a small pool of acylated sterol glucosides (ASG). Lipid composition of mixture added to soil Multivariate (Fig. 2) and univariate (Fig. 3) analysis indicated lipids extracted from wheat were a reasonable proxy for leaf lipids of T. triandra with both quantitatively dominated by MGDG (36:6) and DGDG (36:6). The modest difference in lipid profiles between wheat and T. triandra was borne out by NMDS ordination showing wheat leaves and T. triandra were separated not by the first NMDS axis, but instead by the second NMDS axis that explained only a minor proportion of variation in the multivariate dataset (R 2 = 0.066). Lipids that were 13 C-labelled could be identified because a) their mass was shifted to higher mass relative to the same lipid at natural isotopic abundance, and b) they had an isotope pattern tailing to lower masses (reflecting decreasing probability of isotopologues with multiple 12 C) versus natural abundance lipids for which the isotope pattern tailed towards higher masses (reflecting decreasing probability of isotopologues with multiple 13 C) (Fig. S2). For five lipids the total 13 C enrichment of the intact lipid (calculated from LC-MS mass spectra) varied from 96.3 to 98.1 atom %, with a mean of 97.4 atom%.
Intracellular intact lipids analysable by LC-MS (i.e., acylglycerols + pigments) accounted for > 85% of the lipid mixture (Fig. S3). The combination of LC-MS with two GC-MS methods revealed hydrolysable fatty acids (primarily acylglycerols) accounted for > 80% of the molar concentration of the lipid mixture, with much of the remainder chlorophylls + phaeophytins (~ 7% of total), free fatty acids (3.5% of total) and degradation products of lipids (e.g. glycerol 3-phosphate, glycerol, ethanolamine) (Fig. S3). There were also small amounts of sterols, alcohols and some aromatic organic acids.  Loss of U-13 C intracellular leaf lipids after addition to soil Absolute concentrations of U-13 C phaeophytin a and b and chlorophyll b were 0.03 nmol g −1 or lower (Fig. 4), confirming the tracer-level addition of intact lipid. In soil that had been autoclaved repeatedly, the concentrations of 22 out of 26 intact U-13 C lipids did not change between samples collected at 15 min and 3 h or 24 h (T-test, P > 0.05). Two lipids had significantly lower concentrations at 24 h than 15 min, while 2 lipids had higher concentrations at 24 h than 15 min. Compared with 13 C lipid concentrations measured after 15 min, the average relative concentration of all 26 lipids was 44% greater at 3 h and 32% greater at 1 day. In the live soil, 18 out of 26 intact 13 C lipids had significantly lower concentrations at 24 h than 15 min (T-test, P < 0.05), while all 26 lipids had significantly lower concentrations at 5 days than 15 min (Fig. 4). The strong decrease in lipid concentrations in live soil was borne out by the average relative concentration of the 26 lipids being greatest at 15 min, 30% smaller after 3 h, 52% smaller after 1 day and 68% smaller after 5 days.   C-labelled lipid mixture to soil. A 13 C-lipid mixture (at 97.4 atom % 13 C) was injected into soil that had been pre-treated by 4 cycles of autoclaving over a 4-day period (autoclaved soil, a), or soil that was not subject to any pre-treatment (live soil, b). Relative quantification was based on MS 2 peak area divided by MS 1 area of the internal standard, with subsequent normalisation to the T15 min measurement. The inset graph shows absolute quantification (± standard deviation) for phaeophytin a and b and chlorophyll b in live soil. Unidentified lipids are annotated with retention time_analysis mass, while lipids identified to class level are annotated with retention time_lipid class, a handful of lipids were fully identified and are annotated by their name. For the live soil 18 out of 26 lipids had significantly lower concentrations at 24 h than 15 min (T-test, P < 0.05), while all 26 lipids had significantly lower concentrations at 5 days than 15 min. For the autoclaved soil, 2 lipids had lower concentrations at 24 h than 15 min and 2 had higher concentrations at 24 h than 15 min. Data are means of at least three replicates Vol.: (0123456789) Ordination via NMDS also indicated U-13 C lipid composition of autoclaved soil differed little over time, whereas U-13 C lipid composition of live soil changed dramatically over time (Fig. S4). This was confirmed by PERMANOVA indicating the multivariate U-13 C-lipid composition of autoclaved soil did not differ between 15 min and 24 h (pairwise comparison, P = 0.13), whereas lipid composition of live soil differed significantly between 15 min and 24 h (pairwise comparison, P = 0.0157) (2-way PER-MANOVA, interaction term P = 0.014, Fig. S4).

Mineralisation of added 13 C-labelled lipids
The rate of soil CO 2 efflux (sum of 12 CO 2 and 13 CO 2 ) differed among replicates from 17 to 25 μmol kg −1 h −1 , and slowly decreased over the 15-day measurement period (Fig. 5). Maintenance of the tuneable diode laser and diel changes in δ 13 C and [CO 2 ] of air entering the respiration chambers led to temporal variation in rates of CO 2 efflux and δ 13 C ( Fig. 5 and 6), but did not confound analysis because temporal variation affected similarly the control Fig. 5 Rate of efflux of CO 2 (sum of 12 CO 2 and 13 CO 2 ) from the soil headspace of a 1.89-L glass jar containing approximately 1.25 kg of soil. At time zero, soil respiration chambers received either lipids suspended in 37.5 mL of water (lipid) or were controls that received 37.5 mL of water only (water). Data are shown as absolute rate in the upper panel, while the lower panel shows normalised rates whereby the rate of respiration was normalised to the rate measured in the 18 to 6 h before adding lipid mix or water. Data are means and standard deviations of 3 replicate chambers Time (hours) 0 (water) and treatment (lipid mix suspended in water) (e.g. see close tracking of δ 13 C between treatments and among replicates, Fig. 6). For the procedural controls, water addition led to a modest 20% increase in respiration for approximately three hours. By contrast, addition of 13 C lipid mix led to a larger 55% increase in respiration. The rate of respiration increased within minutes of lipid addition, reached a maximum after three hours, and respiration was statistically different (T-Test, P < 0.05) to the procedural control from 2 h after injection until 7 h after injection. Over 15 days, soil receiving the 13 C lipid mix respired an average of 434 ± 376 μmol more 12 CO 2 and 17 ± 4 μmol more 13 CO 2 than the control. The δ 13 C of air exiting chambers increased within minutes of adding the 13 C lipid mix, but was unaffected by addition of water (Fig. 6). The δ 13 C of soil respiration (δ 13 C SR ) reached maxima of 450-650‰ (vs δ 13 C SR of -25‰ of control, and 2660‰ of added lipid) within 2 h of label addition, then slowly decreased. Fifteen days after adding the 13 C lipid mix, δ 13 C SR remained elevated at 35 to 70‰. In the first day after lipid addition, 87 μmol of Fig. 6 The carbon isotope composition of air exiting soil respiration chambers (δ 13 C, upper panel) and the calculated carbon isotope discrimination of soil respiration (δ 13 C SR , lower panel). At time zero, soil respiration chambers received either lipids suspended in 37.5 mL of water (lipid) or were controls that received 37.5 mL of water only (water). Carbon isotope composition was calculated from efflux of 12 CO 2 and 13 CO 2 from the soil headspace of 1.89-L glass jars containing approximately 1.25 kg of soil. To illustrate the modest variation among replicates, data are shown for individual soil respiration chambers lipid C was respired, which was equivalent to 16% of the lipid-C added (Fig. 7). Over the same 1-day interval there was a three times larger decrease in mean intact lipid concentrations (Fig. 4). After 5 days, the cumulative amount of lipid-C respired was 198 μmol or 37% of added lipid-C, while over the same interval the decrease in intact lipids was almost twice as large. After 15 days the cumulative lipid-C respired was 349 μmol (64% of added lipid-C).

Discussion
The degradation of intracellular leaf lipids such as acylglycerols and pigments has not been extensively examined, despite the quantitative significance of the flux of leaf lipids into soil. It is probable global lipid inputs from leaves are at least 351 Tg per annum given global scale leaf litterfall is around 35 Pg per annum (Meentemeyer et al. 1982) and intracellular lipids are commonly more than 1% of leaf mass. There is growing recognition that plant-derived lipids account for a significant fraction of soil organic C (Dai et al. 2022), despite common extraction protocols underestimating lipid content of soil (Angst et al. 2021). Studies to date have focussed on degradation of other compound classes such as amino acids and proteins (Wanek et al. 2010;Hu et al. 2020) and cellulose and lignin (Donnelly et al. 1990), while much of the research on lipids has focussed on their use as biomarkers of plant inputs. Studies have for example examined the similarity between plant and soil lipids with a focus on molecular biomarkers that can be analysed by gas chromatography (Wiesenberg et al. 2010a;Jansen and Wiesenberg 2017). Studies examining incorporation of intracellular lipids into microbial biomass and mineralisation are rare and have tended to focus on fatty acids hydrolysed from acylglycerols (Wiesenberg et al. 2010b) rather than the intact acylglycerols themselves. To build an integrated picture of the fate of leaf lipids added to soil we consider three independent measures. First, the large difference in intracellular lipid profile between leaves and soil; second, the loss of U-13 C intracellular lipids from soil; and third, the mineralisation of added lipid to CO 2 . To provide a window into how leaf lipids are degraded, we first extracted lipids from leaves of wheat grown with 13 CO 2 then subsequently  Fig. 7 The proportion of soil respiration attributable to added lipid. At time zero, soil respiration chambers received either lipids suspended in 37.5 mL of water or the same volume of water without added lipid (water control). Respiration attributed to added lipid was determined from δ 13 C SR of lipid and water control chambers (Fig. 6). Respiration and isotope com-position were measured at high temporal resolution (e.g. Figure 6), but data are presented here as total daily respiration. Data are means of three replicates. Error bars are standard error. Respiration attributed to added lipid was significantly different to zero at all time points (one-sample T-test ≠ 0, P < 0.05) added the lipid mix to soil at a tracer level such that the steady state acylglycerol pool (and by inference lipid fluxes) were affected by less than 20%. The lipid mixture extracted from wheat was a reasonable proxy for non-polymeric leaf lipids of T. triandra and was quantitatively dominated by lipids of intracellular origin (Figs. 2 and 3). Using this combination of approaches we established that mineralisation of a leaf lipid mixture begins more rapidly than previously reported; intact intracellular lipids are degraded more quickly than lipid-C is mineralised; and much lipid-C is lost within two weeks, yet a significant amount of lipid-C persists in soil.

Lipid profiles of leaves, roots and soil
The lipid profile of soil from mesocosms grown with Themeda triandra for 12 years was different to the intracellular lipids of leaves and roots (Figs. 2 and 3).
The difference between T. triandra and soil reflected differences in the relative abundance of lipid classes (i.e. lipid headgroups) and to a smaller extent fatty acid inventories (data not shown). In accordance with previous studies, leaves were dominated by galactolipids and a variety of phospholipids (Wintermans 1960;Roughan and Batt 1969;Harwood 1980;Dörmann and Benning 2002;Boudière et al. 2014), roots contained little galactolipid and a similar assemblage of phospholipids as leaves, while soil was dominated by PC, PE and DGTS and contained no galactolipids above detection limits (Mangelsdorf et al. 2009;Liu et al. 2010;Peterse et al. 2011;Warren 2018Warren , 2019. The lipid profile of soil was closer to that of roots than leaves, but it is improbable that this broad similarity is because roots are a major pathway of lipid and C input into soil (Wiesenberg et al. 2010b;Liu et al. 2019;Gregory 2022). In the past circumstantial support for soil lipids originating directly from roots arose from the modest differences in fatty acid inventories between roots and soil (Wiesenberg et al. 2010a). However, we show here via analysis of intact lipids that several of the lipid classes present in roots were not detected in soil (i.e. PA, PG, MGDG, DGDG, see Fig. 3) and thus roots are unlikely the direct source of soil acylglycerols despite similarities in fatty acid inventories. Phaeophytins are degradation products of chlorophylls and were detected in substantial amounts in soil, which we suspect reflects input of the large amounts of chlorophyll or phaeophytins from leaves. Leaves are likely the main source of phaeophytins because photosynthetic soil organisms are unlikely abundant in mineral soil due to light limitations. That the soil lipid profile is missing many of the lipid classes present in plants (and see also : Bull et al. 2000;Ding et al. 2020), suggests that plant-derived acylglycerols such as PA, PG, MGDG, DGDG must be degraded rapidly when they enter soil whereas other plant-derived lipids such as phaeophytins are more persistent.

Rapid loss and mineralization of added lipids
The mineralization of 13 C lipids began within minutes of addition (Fig. 4). Previous studies noted loss of intact lipid was evident within a matter of hours (Harvey et al. 1986;Zhang et al. 2019), but there was a delay of up to one day before labelled CO 2 was evolved (Harvey et al. 1986). Mineralisation is thought to be delayed because it encompasses the initial loss of intact lipid that occurs due to enzymatic cleavage prior to uptake, plus subsequent intracellular metabolism. Nevertheless, by adding lipids rapidly and measuring 13 CO 2 efflux at high temporal resolution we determined that mineralisation of lipid-C began within ten minutes and reached a maximum within several hours of lipid addition (Fig. 6). These data do not support the idea that there is necessarily a delay before added lipids are mineralised. The rapid mineralisation of added lipid was likely fuelled primarily by direct uptake of the 10% of the lipid mix present as small lipid molecules such as free fatty acids, glycerol 3-phosphate, glycerol and ethanolamine (Fig. S3). We are unable to verify if small molecules were directly taken up because concentrations were too low to quantify 13 C abundance by GC-MS. Nevertheless, the pool of small lipids was sufficiently large to theoretically fuel around 60% of respiration measured in the first day and previous studies have provided strong support for rapid uptake of free fatty acids by soil microbes (Dippold and Kuzyakov 2016), while other studies have noted widespread capacity for bacterial uptake of glycerol and glycerol 3-phosphate (Blötz and Stülke 2017) and ethanolamine (Kaval and Garsin 2018). Moreover, the rapid microbial uptake of small lipids is supported by the observation G3-P, ethanolamine and glycerol are commonly below detection limits or at micromolar concentrations in soil solution (Warren 2013(Warren , 2014. We cannot discount the possibility that some of the rapid onset mineralisation was fuelled by the galactolipids and phospholipids that dominated the lipid mix (Fig. 3). However, galactolipids and phospholipids likely require extracellular degradation prior to uptake such that mineralisation would be slower than for smaller lipids that can be taken up without prior enzymatic degradation.
LC-MS measurement of intact 13 C lipids revealed loss of a broad range of 13 C lipids including phospholipids, galactolipids, ASG, pigments, and TG (Fig. 4). Rates of loss differed among lipids such that the multivariate profile of 13 C-lipids in soil differed between 15 min and 24 h (Fig. S4). As reported previously (e.g. Soliman and Radwan 1981), the loss of intact lipids was largely biotic given that loss of intact lipids was not significant in soil that had been repeatedly autoclaved to stop microbial and enzyme activity (Figs. 4 and S4). Losses of membrane lipids occurred on a time scale of hours to days as previously reported for tracer-level addition of phospholipids (e.g. White et al. 1979;Kindler et al. 2009;Zhang et al. 2019). The rate of degradation of TG was comparable to membrane lipids, with concentrations of the four quantified TG halving within one day of adding a tracer-level mix of lipids (approx 0.15 mg of TG added per kg of soil) (Fig. 4). In contrast significant loss of TG required weeks in a study adding TG at 2 g lipid per kg of soil (i.e. approx. 14,000 times the amount of TG added here) (Hita et al. 1996) and another study reported half-lives of TG and sterols in soil were 4-6 months after adding 1 g of lipid per kg of soil each week for nine successive weeks (i.e. approx. 60,000 times the amount of TG added here) (Soliman and Radwan 1981). When supplied at tracer-level concentrations it seems probable that the rate of loss of TG is comparably fast as that of a range of membrane lipids (phospholipids, galactolipids, sulfolipids), which perhaps belies their broad similarity as sources of highly reduced, C-rich fatty acyl residues and glycerol. The rapid degradation of all the most abundant acylglycerol classes helps mechanistically explain why the lipid profile of soil does not resemble that of leaves or roots (e.g. Figure 3 and Bull et al. 2000;Ding et al. 2020), but instead the lipids of living microbes. For example, the rapid degradation of added TG implies that the large amounts of TG in soil reflects microbes using TG as a store of C and energy (Alvarez and Steinbüchel 2002;Murphy 2012) rather than slow degradation of TG-rich roots.
Experimentally measuring the loss of added 13 C lipids from soil (Fig. 4) confirmed chlorophylls are degraded rather rapidly, but it is probable chlorophyll's immediate degradation product (phaeophytin) degrades substantially more slowly given the abundance of phaeophytin in soil (Fig. 3). Studies from 50 years ago identified trace amounts of chlorophylls in soil (Hoyt 1966;Cornforth 1969;Sanger 1971) but the rate of degradation has been unclear. Early studies often added 14 C-chlorophyll then subsequently implied degradation based on the amount of 14 C retained in soil. However, radioactivity retained in soil could reflect retention of unaltered 14 C-chlorophyll or retention of 14 C in other compounds (e.g. phaeophytins) or organisms (e.g. fungi) (Simonart et al. 1959). We found here that intact chlorophylls are rapidly degraded (Fig. 4), while the occurrence of substantial amounts of phaeophytin in soil (Fig. 3) is consistent with chlorophyll rapidly losing its central Mg atom with the resulting phaeophytin then degrading slowly. The rapid loss of chlorophylls then slow degradation of pheophytins is likely ubiquitous given pheophytins have been reported from marine sediments in which chlorophylls are at or below detection limits (Colombo et al. 1997).

Lipid-C persists in soil
Even though lipid mineralisation commenced within ten minutes of lipid addition, approximately 36% of lipid-C remained in soil after 15 days. Multiple mechanisms likely led to persistence of a fraction of lipid-C in soil. It is probable a significant fraction of lipid-C is retained first in microbes and second in necromass, consistent with data for other compound classes such as sugars (Gunina and Kuzyakov 2015). Retention of lipid-C in microbes and necromass is supported by the observation intact lipids were lost more rapidly than lipid-C was respired, and previous suggestions of lipid-C recycling by the microbial community (Kindler et al. 2009). Further support can be found in known microbial re-use of fatty acids (Dippold and Kuzyakov 2016), and likely re-use of other degradation products including G3-P, glycerol and headgroups such as choline and ethanolamine (Blötz and Stülke 2017;Kaval and Garsin 2018). The lipid-C persisting in soil may also reflect a fraction of the mixture comprising lipids that are chemically recalcitrant (e.g. phaeophytins) or protected by sorption to the organic matrix or clays or within aggregates (Theng 1974;Harvey et al. 1986;von Lutzow et al. 2006). The existence of a protected fraction of lipid is consistent with a previous study showing a fraction of added phospholipid is protected from degradation (Zhang et al. 2019) and the independent observation a fraction of lipid is not freely extractable (Lin and Simpson 2016). All of chemical recalcitrance, protection and retention within microbes and necromass are feasibly responsible for retaining lipid-C within soil, but additional experiments are required to determine the relative importance of the three mechanisms.

Limitations
The broad aim of this experiment was to assess the fate of intracellular leaf lipids in soil by adding a realistic mixture of lipids at tracer-level concentrations, but this has some drawbacks. Addition of a complex mixture of lipids (rather than an individual lipid) meant we were unable to trace flux of 13 C from individual lipids into degradation products (e.g. glycerol, G3-P, ethanolamine) or 13 CO 2 efflux. An additional limitation of using a complex natural mixture is that concentrations of individual lipids differed, and thus differences in lipid concentration confounded comparison of rates of loss of intact lipid (e.g. Figure 4). Nevertheless, finding the multivariate profile of 13 C-lipids detected in soil changed between 15 min and 24 h provides strong support for lipids differing in rate of degradation (Fig. S4). Finally, interpretation of lipid degradation was hamstrung by the lack of basic information about microbial utilisation of lipids. For example, it is not known which lipids can be directly taken up by microbes versus those that first require extracellular breakdown. The extracellular breakdown is itself subject to uncertainty because lipids can be cleaved at different sites by different enzymes, which leads to a variety of possible degradation products (e.g. phospholipids can be cleaved to lysolipid, glycerophosphodiesters, G3-P, phosphate, glycerol, headgroups, MG, DG, fatty acids). Together these limitations highlight the need for studies tracking the fate of individual lipids to determine degradation and utilisation pathways.
The degradation of intracellular lipids observed here may be different for leaf litter for several reasons. First, the leaf lipid extract contained a broad suite of intracellular, solvent-soluble lipids, but was not comprehensive because it did not include polymeric lipids (e.g. suberin and cutin). Second, lipids were added as a suspension in water and it is likely that lipid degradation would be slower for leaf litter due to physical protection and lower surface area:volume. Hence, lipids added in free form, whether adsorbed onto sand (Zhang et al. 2019) or as a suspension in water (here), probably turnover more rapidly than lipids added in intact microbial cells (e.g. Kindler et al. 2009). Finally, leaf litter contains a variety of additional compound classes (e.g. protein and amino acids, nucleic acids, poly-and mono-saccharides, lignin) that might affect microbial utilisation of lipids, but in ways that are not easy to predict and may differ among soils. For example, microbes might prefer to degrade monosaccharides than lipids (Gunina and Kuzyakov 2015), while in an N-limited soil the provision of N from proteins might ameliorate N limitations and accelerate use of C-rich lipids. At a finer scale, one might predict differential utilisation of specific lipid classes as a function of C and nutrient availability because all lipids are C-rich but some also contain N and/or P. For example, C-limited microbes might preferentially take up TG because they are N-and P-free, whereas P-limited microbes might preferentially degrade phospholipids and take up the liberated P. Taken together these potentially complex interactions highlight the need to consider degradation of leaf lipids in the broader context of leaf degradation and microbial nutrient economics.

Conclusions
Our study confirmed that only a subset of the nonpolymeric intracellular lipids of leaves and roots are detected in soil, despite leaves and roots likely representing a large input of lipids. That the soil lipid profile does not more strongly reflect that of plants is mechanistically explained by the lipid profile of leaves being dominated by acylglycerols (i.e. fattyacid based membrane and storage lipids) that are rapidly degraded upon addition to soil. Contrary to previous reports there was no evidence of a delay before lipids were mineralised, but the more rapid loss of intact lipid than respiration of lipid-C supports the idea that many lipids are degraded extracellularly prior to uptake and mineralisation. Around twothirds of added lipid-C was respired over the course of 15 days, meaning around one-third of lipid-C persisted in soil. It is not known to what extent the persistent lipid-C reflects presence of a fraction of lipids that are more resistant to degradation (e.g. phaeophytins) versus a fraction of added lipid being protected (e.g. by clays) versus retention of lipid-C within the microbes and their necromass. To better understand degradation of leaf lipids requires knowledge of how microbes degrade and utilise lipids, whether there is direct uptake of intact lipid or if lipids are first degraded extracellularly then specific moieties taken up.