Fast-decaying plant litter enhances soil carbon in temperate forests but not through microbial physiological traits

Microcosm preparation and incubation

Leaf litters were collected from Lilly-Dickey Woods, a mature eastern US temperate broadleaf forest located in South-Central Indiana (39°14′N, 86°13′W) using litter baskets and surveys for freshly senesced litter as described in Craig et al.⁵². Of the 19 species collected in Craig et al. (2018), we selected litter from 16 tree species with the goal of maximizing variation in litter chemical traits (Table S1). Litters were air-dried and then homogenized and fragmented such that all litter fragments passed a 4000 µm, but not a 250 µm mesh. Whereas leaf litters had a distinctly C3 δ¹³C signature of −30.1 ± 1.5 (mean, standard deviation), we used a ¹³C-rich (δ¹³C = −12.6 ± 0.4) soil obtained from the A horizon of a 35-yr continuous corn field at the Purdue University Agronomy Center for Research and Education near West Lafayette, Indiana (40°4′N, 86°56′W). The soil is classified as Chalmers silty clay loam (a fine-silty, mixed, superactive, mesic Typic Endoaquoll). Prior to use in the incubation, soils were sieved (2 mm) and remaining recognizable plant residues were thoroughly picked out. Soils were mixed with acid-washed sand—30% by mass—to facilitate litter mixing (see below) and to increase the soil volume for post-incubation processing. The resulting soil had a pH of 6.7 and a C:N ratio of 12.0.

We constructed the experimental microcosms by mixing the 16 litter species with the ¹³C-enriched soil. Each litter treatment was replicated four times in four batches (i.e., 16 microcosms per species, 272 total microcosms including 16 soil-only controls). Two batches (C budget microcosms) were used to monitor CO₂ efflux and to track litter-derived C into SOM pools, and two batches were used to quantify microbial biomass dynamics.

Incubations were carried out in 50 mL centrifuge tubes modified with an O-ring to prevent leakage and a rubber septum to facilitate headspace sampling. To each microcosm, we added 5 g dry soil, adjusted moisture to 65% water-holding capacity, and pre-incubated for 24 h in the dark at 24 °C. Using a dissecting needle, 300 mg of leaf litter were carefully mixed into treatment microcosms and controls were similarly agitated. This corresponds to an average C addition rate of 27.1 ± 1.1 g C kg⁻¹ dry soil among the 16 species. During incubation, microcosms were loosely capped to retain moisture while allowing gas exchange, and were maintained at 65% water-holding capacity by adding deionized water every week.

Carbon budget in microcosms

Respiration was quantified with an infrared gas analyzer (LiCOR 6262, Lincoln, NE, USA) coupled to a sample injection system. Our first measurement was taken about 12 h after litter addition (day 1) and subsequent measurements were taken on days 2, 4, 11, 19, and 30 for both batches and on days 46, 64, 79, 92, 109, 128, 149, and 185 for the second batch. Prior to each measurement, microcosms were capped, flushed with CO₂-free air, and incubated for 1–8 h depending on the expected efflux rate. Headspace was sampled with a gas-tight syringe and the CO₂-C concentration was converted to a respiration rate (µg CO₂-C day⁻¹). Total cumulative CO₂-C loss was derived from point measurements by numerical integration (i.e., the trapezoid method). To evaluate soil-derived CO₂-C efflux, we measured δ¹³C in two gas samples per litter type or control on a ThermoFinnigan DELTA Plus XP isotope ratio mass spectrometer (IRMS) with a GasBench interface (Thermo Fisher Scientific, San Jose, CA). Isotopes were measured on days 1, 4, 11, 30, 64, 109, and 185. On each of these days, a two-source mixing model⁷⁰ was applied to determine the fraction of total CO₂-C derived from soil organic matter vs. litter:

where (frac{{F}^{l}(t)}{F(t)}) is the fraction total CO₂-C efflux [(F(t))] derived from litter [({F}^{l}(t))] at time (left(tright)), (delta Fleft(tright)) is the δ¹³C of the CO₂ respired by each litter-soil combination, (delta {F}^{c}(t)) is the average δ¹³C of the CO₂ respired by the control soil, and (delta {C}_{l}) is the δ¹³C of each litter type. These data were used to calculate cumulative soil-derived C efflux via numerical integration and, for each litter type, average soil-derived C efflux was subtracted from total cumulative CO₂-C loss to determine cumulative litter-derived CO₂-C loss.

Carbon budget microcosms were harvested on days 30 and 185 to track litter-derived C into mineral-associated SOC at an early and intermediate stage of decomposition. To do this, we used a size fractionation procedure^71,72 modified to minimize the recovery of soluble leaf litter compounds or dissolved organic matter in the mineral-associated SOC fraction. For each sample, we first added 30 mL deionized water, gently shook by hand to suspend all particles, and then centrifuged (2500 rpm) for 10 min. Floating leaf litter was carefully removed, dried for 48 h at 60 °C, and weighed; and the clear supernatant was discarded to remove the dissolved organic matter. The remaining sample was dispersed in 5% (w/v) sodium hexametaphosphate for 20 h on a reciprocal shaker and then washed through a 53 µm sieve. The fraction retained on the sieve was added to the floating leaf litter sample and collectively referred to as particulate SOC, while the fraction that passed through the sieve was considered the mineral-associated SOC. Both fractions were dried, ground, and weighed; and analyzed for C concentrations and δ¹³C values on an elemental combustion system (Costech ECS 4010, Costech Analytical Technologies, Valencia, CA, USA) as an inlet to an IRMS. As above, litter-derived C in the particulate and mineral-associated SOC was determined as follows:

where ({C}_{s}(t)) is the total particulate or mineral-associated SOC content in the sample at time ((t)), ({C}_{s}^{l}(t)) is the litter-derived C in the soil, (delta {C}_{s}left(tright)) is the measured δ¹³C value for each soil fraction, (delta {C}_{c}left(tright)) is the average δ¹³C for each fraction in control samples, and (delta {C}_{l}) is the δ¹³C of each litter type. In a few cases, mineral-associated δ¹³C was slightly less negative in the treatment than in the control soil. In these cases, litter-derived mineral-associated SOC was considered zero.

Total litter-derived SOC at each harvest date was calculated by subtracting the cumulative litter CO₂-C from initial added litter C. The difference between this value and the sum of litter-derived particulate and mineral-associated SOC was considered the residual pool which we assume mostly represents water-extractable dissolved organic matter.

Microbial biomass dynamics during incubation

Sample batches were harvested at days 15 and 100 to capture early- and intermediate-term microbial biomass responses to litter treatments. These times were selected to correspond with the middle of early and intermediate C budget microcosm incubations. We quantified microbial biomass as well as MGR, CUE, and MTR using ¹⁸O-labeled water^73,74 as in Geyer et al.⁷⁵.

Microbial biomass C (MBC) was determined on two ~2 g subsamples using a standard chloroform fumigation extraction⁷⁶. One subsample was immediately extracted in 0.5 M K₂SO₄ and one was fumigated for 72 h before extraction. After shaking for 1 h, extracts were gravity filtered through a Whatman No. 40 filter paper, and filtrates were analyzed for total organic C using the method of Bartlett and Ross⁷⁷ as adapted by Giasson et al.⁷⁸. The difference between total organic C in the fumigated and unfumigated subsamples was used to calculate MBC (extraction efficiency K_EC = 0.45).

To determine MGR, CUE, and MTR, we first pre-incubated two 0.5 g soil subsamples (one treatment and one control) for 2 d at 24 °C. Prior to this pre-incubation, samples were allowed to evaporate down to 53 ± 6% (mean, sd) water-holding capacity. After the pre-incubation, water was injected with a 25 µL syringe to bring each sample to 65% water-holding capacity. For one subsample, we used unlabeled deionized water. For the second subsample, enriched ¹⁸O-water (98.1 at%; ICON Isotopes) was mixed with unlabeled deionized H₂O to achieve approximately 20 at% of ¹⁸O in the final soil water. Each sample was placed in a centrifuge tube (modified for gas sampling), flushed with CO₂-free air, and incubated for 24 h. Headspace CO₂ concentration was then measured, and samples were flash frozen in liquid N₂ and stored at −80 °C until DNA extraction.

DNA was extracted from each sample using a DNA extraction kit (Qiagen DNeasy PowerSoil Kit, Venlo, Netherlands) following the protocol described in Geyer et al. (2019) which sought to maximize the recovery of DNA. The DNA concentration was determined fluorometrically using a Quant-iT PicoGreen dsDNA Assay Kit (Invitrogen). DNA extracts (80 µL) were dried at 60 °C in silver capsules spiked with 100 µL of salmon sperm DNA (42.5 ng µL⁻¹), to reach the oxygen detection limit, and sent to the UC Davis Stable Isotope Facility for quantification of δ¹⁸O and total O content.

Microbial growth rate (MGR) was calculated following Geyer et al. (2019). Specifically, atom % of soil DNA O (at% O_DNA) was determined using the two-pool mixing model:

where at% is the atom % ¹⁸O and O_DNA+ss, O_DNA, and O_ss are the concentration of O in the whole sample, soil DNA, and salmon sperm, respectively. Atom percent excess of soil DNA oxygen (APE O_soil) was calculated as the difference between at% O_DNA in the treatment and control samples. Total microbial growth in terms of O (Total O; µg) was estimated as:

where at% soil water is the atom % ¹⁸O in the soil water. MGR in terms of C (µg C g⁻¹ soil d⁻¹) was calculated by applying conversion mass ratios of oxygen:DNA (0.31) and MBC:DNA for each sample, dividing by the soil mass, and scaling by the incubation time. Assuming uptake rate (Uptake) is equal to the sum of MGR and respiration, CUE and MTR were calculated by the following equations.

Data analysis for microcosm experiment

Litter decay constants were calculated for each species using litter-derived CO₂-C values to estimate litter mass loss over time. After it was determined that a single exponential decay model provided a poor fit, we fit litter decomposition data using the double exponential decay model:

where s represents the labile or early stage decomposition fraction that decomposes at rate k₁, and k₂ is the decay constant for the remaining late stage decomposition fraction.

To reduce the dimensionality of litter quality and microbial indicators, indices were derived by principal component analysis (PCA; Fig. S1A, B) using the ‘prcomp’ function in R. The first axis of a PCA of decomposition parameters (s, k₁, and k₂) and litter chemical properties (soluble and AUR contents; AUR-to-N and C-to-N ratios; and the lignocellulose index [LCI]) was taken as a litter quality index. Whereas this index highly correlated with indicators of C quality (AUR, soluble content, and LCI), the second axis of this PCA correlated with C:N and AUR:N and was therefore taken as a second litter quality index representing variation in N concentration. The first axis of a PCA of MGR, CUE, and MTR was taken as a microbial physiological trait index.

Bivariate relationships were examined using simple linear regressions on average species values at each harvest (n = 16). To examine relationships between microbial physiological traits and mineral-associated SOC, data from the early-term (day 15) and intermediate-term (day 100) microbial harvest were matched with early-term (day 30) and intermediate-term (day 185) C budget microcosms, respectively. In addition to examining total mineral-associated SOC formation, we also estimated the efficiency of litter C transfer into the mineral-associated SOC pool as the fraction of lost litter C (i.e., litter C lost as CO₂, recovered in the mineral-associated SOC fraction, or in the residual pool) retained in the mineral-associated SOC. Path analyses were used to evaluate the hypothesis that microbial physiological traits mediate the effect of litter quality on mineral-associated SOC formation and mineral-associated and particulate SOC decay. We hypothesized that the litter quality index would be positively associated with the microbial physiological trait index (representing faster and more efficient microbial growth) and microbial physiological traits would, in turn, be positively associated with the rate and efficiency of mineral-associated SOC formation. We expected that this mediating pathway would reduce the direct relationship between litter quality and SOC. This analysis was conducted using the LAVAAN package⁷⁹ to run path analyses for total litter-derived mineral-associated SOC, mineral-associated SOC formation efficiency, and soil-derived mineral-associated and particulate SOC for both early and intermediate stage harvests. All analyses were performed using R version 3.5.2.

Field study design and soil sampling

We worked in the Smithsonian’s Forest Global Earth Observatory (ForestGEO) network⁸⁰ in six mature U.S. temperate forests varying in climate, soil properties, and tree community composition (Fig. 1a): Harvard forest (HF; 42°32′N, 72°11′W) in North-Central Massachusetts, Lilly-Dickey Woods (LDW; 39°14’N, 86°13’W) in South-Central Indiana, the Smithsonian Conservation Biology Institute (SCBI; 38°54′N, 78°9′W) in Northern Virginia, the Smithsonian Environmental Research Center (SERC; 38°53′N, 76°34′W) on the Chesapeake Bay in Maryland, Tyson Research Center (TRC; 38°31′N, 90°33′W) in Eastern Missouri, and Wabikon Lake Forest (WLF; 45°33′N, 88°48′W) in Northern Wisconsin, USA. Land use history across the six sites consisted mostly of timber harvesting which ceased in the early 1900s. Soils are mostly Oxyaquic Dystrudepts at HF, Typic Dystrudepts and Typic Hapludults at LDW, Typic Hapludalfs at SCBI, Typic or Aquic Hapludults at SERC, Typic Hapludalfs and Typic Paleudalfs at TRC, and Typic and Alfic Haplorthods at WLF. Further site details are reported in Table S5.

Each site contains a rich assemblage of co-occurring arbuscular mycorrhizal (AM)- and ectomycorrhizal (ECM)-associated trees (Table S6), which we leveraged to generate environmental gradients in factors hypothesized to predict microbial physiological traits within each site. Specifically, the dominance of AM vs. ECM trees within a temperate forest plot has been shown to be a strong predictor of soil pH, C:N, inorganic N availability, and litter quality^52,53,54. We established nine 20 × 20 m plots in each of our six sites in Fall 2016 (n = 54) distributed along a gradient of AM- to ECM-associated tree dominance. Plots were selected to avoid obvious confounding environmental factors. Where possible, we established our nine-plot gradient in three blocks (<1 ha), each containing an AM, ECM, and mixed dominance plot. Plots were generally located on upland areas except for TRC where all plots were located on toeslopes due to a lack of AM trees in upland areas. At HF, we established six of the nine plots outside the boundaries of the ForestGEO plot due to a lack of appropriate AM-ECM mixtures. At LDW, we used a random subset of previously established 15 × 15 m plots⁸¹. Basal area of living trees was determined using the most recent ForestGEO census (within the last 5 years) or by conducting our own inventory. ECM-dominance was quantified as the percentage of ECM-associated basal area relative to the total plot basal area.

Four soil cores (5 cm-diameter) were collected in July 2017 from each plot to a depth of 5 cm. A thin O horizon (Oe and Oa) was sometimes present and was collected as part of the 5 cm soil core. However, there was often a thick O horizon (>5 cm) at HF, which was removed before coring. Samples were also collected at 5–15 cm depth for soil texture analysis. We sampled from an inner 10 × 10 m square in each plot to avoid edge effects. All samples from the same plot were composited, sieved (2 mm), picked free of roots, subsampled for gravimetric moisture (105 °C), and air-dried, or refrigerated (4 °C) until analysis for microbial physiological variables and N availability.

Soil properties

We determined several physicochemical properties known to predict mineral-associated SOC. We measured soil pH (8:1 ml 0.01 M CaCl₂:g soil) and soil texture using a benchtop pH meter and a standard hydrometer procedure⁸², respectively. Organic matter content was high in some upper surface soils, so plot-level soil texture was determined from 5 to 15 cm depth samples. We quantified oxalate-extractable Al and Fe pools (Al_ox and Fe_ox) in all soil samples as an index of poorly crystalline Al- and Fe-oxides⁸³, which is one of the strongest predictors of SOM content in temperate forests⁸⁴. Briefly, we extracted 0.40 g dried, ground soil in 40 mL 0.2 M NH₄-oxalate at pH 3.0 in the dark for 4 h before gravity filtering and refrigerating until analysis (within 2 w) on an atomic-adsorption spectrometer (Aanalyst 800, Perkin Elmer, Waltham, MA, USA), using an acetylene flame and a graphite furnace for the atomization of Fe and Al, respectively.

We quantified potential net N mineralization rates as an index of soil N availability. One 5 g subsample per plot was extracted immediately after processing by adding 10 mL 2 M KCl, shaking for 1 h, and filtering through a Whatman No. 1 filter paper after centrifugation at 3000 rpm. A second subsample from each plot was incubated under aerobic conditions at field moisture and 23 °C for 14 d before extraction. Extracts were frozen (−20 °C) until analysis for NH₄⁺-N using the salicylate method and for NO₃⁻-N plus NO₂⁻-N after a cadmium column reduction on a Lachat QuikChem 8000 flow Injection Analyzer (Lachat Instruments, Loveland, CO, USA). Potential net N mineralization rates (mg N g dry soil⁻¹ d⁻¹) were calculated as the difference between pre- and post-incubation inorganic N concentrations.

Microbial biomass dynamics in field plots

Microbial biomass carbon and microbial physiological traits were quantified within 10 days of collection as described above, with four minor differences. First, 30 g soil subsamples were covered with parafilm and pre-incubated for 2 d near the field soil temperature measured at the time of sampling (16.5 °C for WLF and HF, and 21.5 °C for LDW, TRC, SCBI, and SERC). Second, for CO₂ analysis, samples were placed in a 61 mL serum vial crimped with a rubber septum. Third, DNA concentrations were determined using a Qubit dsDNA BR Assay Kit (Life Technologies) and a Qubit 3.0 fluorometer (Life Technologies). Fourth, 14.5 g subsamples were used for microbial biomass analysis.

Soil organic matter characterization in field plots

Mineral-associated SOC was quantified as in the microcosm experiment, but without a pre-fractionation leachate removal step. We additionally measured soil amino sugar concentrations to estimate microbial necromass contributions to SOM. Amino sugars are useful microbial biomarkers because they are found in abundance in microbial cell walls, but are not produced by higher plants and soil animals¹⁹. Moreover, amino sugars can provide information on the microbial source of necromass. For example, glucosamine (Glu) is produced mostly by fungi whereas muramic acid (MurA) is produced almost exclusively by bacteria^61,85. Amino sugars were extracted, purified, converted to aldononitrile acetates, and quantified with myo-inositol as in Liang et al.⁸⁶. We used the concentrations of Glu and MurA to estimate total, fungal, and bacterial necromass soil C using the empirical relationships reported in Liang et al.⁸.

Leaf litter and fine roots in field plots

In Fall 2017, we collected leaf litter on two sample dates from four baskets deployed in the inner 10 × 10 m of each plot. Litter was composited by plot, dried (60 °C), sorted by species, weighed, and ground. We performed leaf litter analyses on at least three samples of each species at each site —unless a species was only present in one or two plots— to get a site-specific mean for each species. Some non-dominant species were not included in these analyses because an insufficient amount of material was collected. Fine roots (<2 mm) were collected at the time of soil sampling (at LDW, TRC, and WLF) or the following summer (HF, SERC, SCBI) due to issues with sample processing.

Leaf litter and bulk root samples were analyzed for C and N content. For leaf litter, we conducted a sequential extraction as above (sensu Moorhead & Reynolds⁸⁷) to determine the soluble fraction using hot water and ethanol, the AUR—the insoluble ash-free fraction that resisted degradation by a strong acid—which contains lignin and is commonly found to inhibit leaf litter decay⁴⁰. The LCI was estimated by dividing the AUR fraction by the total non-soluble ash-free fraction. Leaf litter trait values were calculated for each plot as the average of site-specific mean values for each species in the plot weighted by proportion of total plot basal area.

Data analysis for field study

As above, a litter quality index and microbial physiological trait index was calculated via PCA for each plot (Fig. S1C, D). We note, however, that decomposition parameters were not available for the litter quality PCA in the field study. We evaluated the extent to which microbial physiological variables were predicted by litter quality, N availability, stoichiometery, C availability, pH, or root biomass. We first examined variability in indicators of these biogeochemical drivers (i.e., the litter quality index, net N mineralization, soil C:N, dissolved organic carbon (DOC), soil pH, and fine root biomass) and evaluated whether they correlated with our mycorrhizal gradients. Linear mixed models were fit to the microbial physiological trait index using site as a random factor and the above indicators as fixed factors.

We also used linear mixed models to evaluate whether microbial physiological variables predicted significant variation in mineral-associated or necromass-derived C in the context of other hypothesized or known controls on mineral-associated SOC. In addition to the litter quality and microbial physiological trait indices, we included clay content, pH, Fe_ox, Al_ox, and fine root biomass as fixed factors. We fit models to mineral-associated SOC—on a total basis (g soil⁻¹) and as a proportion of soil C (mineral-associated C/total soil C)—as well as total, fungal, and bacterial necromass. Models were selected to avoid highly correlated predictors (i.e., r > 0.5). Fe_ox and Al_ox were correlated above this threshold and final models were selected to contain only Fe_ox based on AIC. Residuals were screened for normality (Shapiro-Wilk), heteroscedasticity (visual assessment of residual plots), and influential observations (Cook’s D). Based on this, MGR, MTR, and mineral-associated SOC were natural log transformed. For all mixed models, we centered and standardized predictors (i.e., z-transformation) so that the slopes and significance levels of different predictors could be compared to one another on the same axis⁸⁸.

Source: Ecology - nature.com