Deep-rooted perennial crops differ in capacity to stabilize C inputs in deep soil layers

Experimental design and crop management

The study was conducted during 2019 in a field experiment on an arable soil (classified as Luvisols) in the deep root experimental facility at the University of Copenhagen, Denmark (Supplementary Table S4). The experiment was conducted with two diverse perennial deep-rooted species: the tap-rooted forage legume lucerne (Medicago sativa L. (cv. Creno); Family: Fabaceae) with the capacity to fix N₂ and the intermediate wheatgrass (Thinopyrum intermedium; Family: Poaceae) kernza developed by the Land Institute (Salina, Kansas, USA). Kernza was initially sown on April 11th, 2015 and lucerne on September 9th, 2016 with a seeding density of 20 kg seeds ha⁻¹. Every year, kernza was fertilized with NPK fertilizer (21:7:3; NH₄:NO₃ = 1.28) as a single dose in early spring (before the onset of plant growth). Kernza was harvested every year in August using a combine harvester and lucerne three times per year in June, August, and October. Plants were rainfed with a subsurface drain installed at both 1 and 2 m depth running between the plots.

For each species, fixed frames of 0.75 m² were inserted in the soil (ca. 5 cm) within each field plot. Specifically, three field plots of lucerne (with observable root nodulation) and kernza were used where each of the three kernza field plots contained two subplots of N fertilized kernza at 100 kg N ha⁻¹ (K₁₀₀) (i.e., the standard fertilization within this field) and N fertilized kernza at 200 kg N ha⁻¹ (K₂₀₀) (i.e., within the range of standard fertilization practices for kernza). Before the onset of plant growth, all plots received ¹⁵N (as ¹⁵NH₄Cl; 98 atom%) in trace amounts (corresponding to 1 kg N ha⁻¹) to trace N allocation from the surface to deeper layers.

¹³C/¹⁴C-CO₂-labeling

Within each fixed frame, the ¹³C/¹⁴C-CO₂-labeling was conducted using an atmospheric labeling chamber⁴¹. Labeling with C-tracers was done with multiple-pulse labeling (three times per week) over two months until first harvest (May 2nd to June 20th 2019). Glass beakers containing ¹³C labeled bicarbonate (0.1 g mL⁻¹ labeling solution; 99 atom%), and ¹⁴C labeled bicarbonate (11 kBq mL⁻¹ labeling solution) within a solution of NaOH (1 M) were added within each of the labeling chambers. Once chambers were sealed, hydrochloric acid (HCl; 2 M) was added to the labeling solution (in equivalent amounts) via a syringe promoting ¹⁴CO₂/¹³CO₂ evolution. Chambers remained sealed for one to three hours (between 9 am and 12 pm) depending on weather conditions (i.e., the duration and intensity of sunshine). The amount of added labeling solution sequentially increased with increasing plant growth (i.e., 5 mL per 20 cm increase in plant height) reaching a plant height of 100–120 cm at the termination of the labeling.

Shoot, root, and soil sampling

The labeling plots (0.75 m²) were harvested on June 20th, 2019 to obtain the aboveground biomass of lucerne and kernza (K₁₀₀ and K₂₀₀). The aboveground biomass in addition to samples obtained from unlabeled parts of the field was directly stored at − 20 °C until drying at 105 °C for two days. For each plot and unlabeled samples, the plant biomass was homogenized and ball-milled for subsequent isotopic analyses.

Soil cores to 1.5 m depth were taken inside all labeling plots, and cores were subdivided into four depth intervals: 0–25, 25–50, 50–100, and 100–150 cm. The soil coring was conducted in 25 cm intervals using a soil auger (6 cm inner diameter). Specifically, per depth three soil samples were taken and stored at 4–5 °C (ca. two days) and then immediately processed and stored at -20 °C until analyses. Roots, bulk soil and rhizosphere soil (adhering to the roots), were separated by sequential sieving of the soil with finer mesh sizes to 1 mm as described by Peixoto, et al.²⁶. A subsample of the bulk soil (ca. 150 g) from each depth in all labeling plots was washed on a 250 µm sieve to recover root fragments for subsequent isotopic determination in unrecovered root fragments. Soil samples (and associated roots) from unlabeled parts of the larger field plots were used to determine natural abundance of ¹³C/¹⁴C/¹⁵N with depth. The collection of plant material complied with relevant institutional guidelines and seeds were gifted by University of Copenhagen.

Determination of ¹³C/¹⁴C/¹⁵N enrichment, and C and N quantity

For each defined depth, samples of roots and soil were homogenized, freeze-dried (except PLFA samples that were stored at − 20 °C), and ground in a ball-mill for the determination of total C and N, ¹³C, ¹⁵N, and ¹⁴C activity. Total C, N, ¹³C, and ¹⁵N were measured with a FLASH 2000 CHNS/O Elemental Analyzer (Thermo Fisher Scientific, Cambridge, UK) combined to a Delta V Advantage isotope ratio mass spectrometer via a ConFlo III interface (Thermo Fisher Scientific, Bremen, Germany) at the Centre for Stable Isotope Research and Analysis (Georg August University Göttingen, Göttingen, Germany).

All δ¹³C values are standardized to the Vienna PeeDee Belemnite international isotope standard and δ¹⁵N values standardized to the δ¹⁵N values of atmospheric N₂. ¹³C and ¹⁵N enrichment is expressed as atom% excess as calculated by the atom% difference between the respective labeled and unlabeled samples. The ¹⁴C activity was determined by combustion in a Hidex 600 OX Oxidizer (Hidex, Turku, Finland) and counted on a liquid scintillation counter (Tri-Carb 3180TR/SL, PerkinElmer, Waltham, MA, USA). ¹⁴C enrichment is determined by the difference in the ¹⁴C activity (Bq g⁻¹) between the respective labeled and unlabeled samples.

Calculation of root C and net rhizodeposition

The amount of root C (mg C kg⁻¹ soil) was calculated based on the root dry matter and C concentration divided by the quantity of soil sampled³⁸. For the determination of net rhizodeposition, ¹⁴C was used due to lower detection limits in deeper soil layers⁴². A modified tracer mass balance approach described by Rasmussen, et al.⁴³ with adjusted unrecovered root fragments⁴¹ was used to determine the net rhizodeposition based on the following equations where the %ClvR is the relative proportion of rhizodeposition expressed as the percent C lost via rhizodeposition:

The ¹⁴C soil content was the sum of the adjusted bulk soil ¹⁴C and rhizosphere ¹⁴C content for each soil sample. The ¹⁴C rhizosphere and bulk soil content for each soil sample were determined by multiplying the total quantity of C by the ¹⁴C enrichment of the soil. The adjusted bulk soil ¹⁴C content was calculated as the difference between the bulk ¹⁴C soil content by the ¹⁴C root washed content as determined by the multiplication of ¹⁴C enrichment in root fragments recovered from a subsample of soil by the total C content within the entire soil volume sampled. The ¹⁴C root content was determined by multiplying the total quantify of C in roots by the ¹⁴C enrichment. Similar equations were used to calculate the net rhizodeposition of N based on ¹⁵N enrichment within the soil and roots.

Biomarker analyses

Phospholipid fatty acid (PLFA)

The analysis of PLFAs was done according to a modified protocol by Frostegård, et al.⁴⁴ with a detailed description of the modifications provided by Gunina, et al.⁴⁵. In brief, 25 μL of 1,2-Dinonadecanoyl-sn-Glycero-3-Phosphatidylcholine (C19:0) (1 mg mL^–1) were added to each of the samples and used in the quantification of recovery of the phospholipids. The lipid fraction from 5–6 g of rhizosphere soil was extracted twice using a one-phase Bligh-Dyer extractant⁴⁶ of chloroform, methanol (MeOH), and citrate buffer (pH 4) (1:2:0.8, v/v/v). To isolate the phospholipid fraction, a solid-phase extraction with activated silica gel and methanol elution was conducted. The derivatization into fatty acid methyl esters occurred via a sequential hydrolyzation with 0.5 mL sodium hydroxide (NaOH) (0.5 M) in MeOH for 10 min at 100 °C and methylation with 0.75 mL of boron trifluoride (BF₃) (1.3 M) in MeOH for 15 min at 80 °C. An external standard stock solution containing 28 individual fatty acids (ca. 1 mg mL^–1 per fatty acid) used in the quantification of PLFA content was simultaneously derivatized with the samples. The residues were dissolved in 185 μL of toluene, and 15 μL of the internal standard 2, tridecanoic acid methyl ester (C13:0) (1 mg mL^–1) were added to each sample prior to measurement using an Agilent 7820A GC coupled to an Agilent 5977 quadrupole mass spectrometer (Agilent, Waldbronn, Germany). The sum of all PLFAs was used as a proxy of the living microbial biomass based on the direct relation between PLFAs and microbial biomass.

Amino sugars (AS)

Amino sugars were extracted according to a modified protocol by Zhang and Amelung⁴⁷ with a detailed description of the procedure by Peixoto, et al.²⁶. In brief, 0.8–1.5 g of freeze-dried rhizosphere soil were hydrolyzed with the addition of 11 mL of 6 M HCl for 8 h at 105 °C. Following hydrolysis, soil samples were filtered and HCl was removed via rotary evaporation at 45 °C to dry the filtrate. Prior to derivatization both iron precipitates and salts were removed from the filtrate and 25 μL of the internal standard 1, methylglucamine (MeGlcN) (1 mg mL^–1) was added and used for quantification of recovery. The derivatization into aldononitrile acetates was conducted as described by Zhang and Amelung⁴⁷. For the quantification of AS, an external standard stock solution containing the AS: N-acetylglucosamine (GlcN) (2 mg mL^–1), N-acetylgalactosamine (GalN) (2 mg mL^–1), N-acetylmuramic acid (MurN) (1 mg mL^–1), mannosamine (ManN) (2 mg mL^–1), and MeGlcN (1 mg mL^–1) was derivatized and analyzed with the samples. The residues were dissolved in 185 μL of ethyl acetate-hexane (1:1, v/v), and 15 μL of the internal standard 2, tridecanoic acid methyl ester (1 mg mL^–1), were added to the samples for measurement using an Agilent 7890A GC coupled to Agilent 7000A triple quadrupole mass spectrometer (Agilent, Waldbronn, Germany). Total amino sugars content was calculated as the summation of the four detected amino sugars: GlcN, MurN, GalN, and ManN.

Amino acids (AA)

Amino acids were extracted from both freeze-dried rhizosphere soil and root samples according to the protocol by Enggrob, et al.⁴⁸. In brief, 0.8–3 g of rhizosphere soil and 0.02 g of root were hydrolyzed with the addition of 2 mL of 6 M HCl for 20 h at 110 °C to break the peptide bonds. Samples were subsequently purified via the removal of lipophilic and solid compounds by the addition of 4 mL n-hexane/dichloromethane (6:5, v/v) to the soil and root samples. Following centrifugation, the aqueous phase was filtered through glass wool and rinsed with 2 × 0.5 mL 0.1 M HCl into new glass tubes with the addition of 300 μL of the internal standard, norleucine (2.5 mM). The samples were freeze-dried and the residues dissolved in 1 mL 0.01 M HCl prior to the separation of amino acids and amino sugars (i.e., N containing compounds) on a polypropylene column with a cation exchange resin. The amino acids were eluted with a 2.5 M ammonium hydroxide solution and freeze-dried prior to derivatization of the amino acids as described by Enggrob, et al.⁴⁸. For the quantification of AA, an external standard stock solution containing 14 AA was derivatized and analyzed with the samples. The amino acids were measured using a trace GC Ultra mounted with a TriPlus autosampler (Thermo Scientific, Hvidovre, Denmark) coupled via a combustion reactor (GC IsoLink, Thermo Scientific) to an isotope ratio mass spectrometer (Delta V Plus IRMS, Thermo Scientific). The total AA content of the rhizosphere soil and roots was based on the summation of the AA: alanine, Asx (asparagine and aspartate), Glx (glutamine and glutamate), glycine, isoleucine, lysine, phenylalanine, Pro/Thr (proline and threonine), serine, tyrosine, and valine.

Compound-specific stable isotope probing

To determine the ¹³C enrichment of biomarkers, all raw δ¹³C were measured individually for AS and PLFA using a Delta V Advantage isotope ratio mass spectrometer via a ConFlo III interface (Thermo Fisher Scientific, Bremen, Germany). For AA, all raw δ¹³C were measured using a trace GC Ultra mounted with a TriPlus autosampler (Thermo Scientific, Hvidovre, Denmark) coupled via a combustion reactor (GC IsoLink, Thermo Scientific) to an isotope ratio mass spectrometer (Delta V Plus IRMS, Thermo Scientific). For each sample, chromatogram peaks identified based on retention times specific for the measured amino sugars, PLFA, and AA were integrated using Isodat v. 3.0 (Thermo Fisher Scientific). All raw δ¹³C values were corrected for dilution by additional C atoms added during the derivatization, amount dependence, offset, and drift (for PLFA samples)^49,50,51. To determine the ¹³C incorporation into each biomarker, the ¹³C excess for each biomarker as determined by the difference between the ¹³C of the labeled and unlabeled biomarker was multiplied by the C content of the specific biomarker.

Relative microbial stabilization (RMS)

The relative microbial stabilization is based on the relation of rhizodeposited ¹³C in the PLFA and amino sugar pools as described in detail by Peixoto, et al.²⁶. The underlying assumption is that ¹³C incorporation into the amino sugar pool indicates the transformation of rhizodeposited C into necromass^52,53, and the ¹³C incorporation into the PLFA pool (i.e., the living microbial biomass) represents a temporary C pool as PLFAs are immediately exposed to degradation following cell lysis⁵⁴. The relative microbial stabilization (RMS) is calculated as follows:

where the average weighted atom% ¹³C excess is determined by the total ¹³C incorporation divided by the total C content of the respective PLFA or amino sugar pools. Accordingly RMS < 0 reflects C allocated into living microbial biomass being a short-lived C pool, whereas RMS > 0 is indicative of higher stabilization of C based on the dominant entry of C into the microbial necromass. However, the RMS indicator does not imply the absolute stability of rhizodeposited C, but rather signifies the potential for microbial stabilization among contrasting experimental variables (i.e., depth and plant species).

Molecular analysis

DNA extraction

From each sample, 0.5 g of freeze-dried rhizosphere soil was used for DNA extraction using the Fast DNA Spin kit for Soil (MP Biomedicals, Solon, OH, USA) according to the manufacturer’s protocol with a single modification. Following, the addition of Binding Matrix, the suspension was washed with 5.5 M Guanidine Thiocyanate (protocol from MP Biomedicals) to remove humic acids that could inhibit preceding polymerase chain reaction (PCR) steps. The DNA was eluted in DNase free water and purified using the NucleoSpin gDNA Clean-up kit following the manufacturer’s protocol (Macherey–Nagel, Düren, Germany). The purity and concentration of DNA were checked on Nanodrop and Qubit, respectively.

Amplicon sequencing

Extracted DNA was sent to Novogene Europe (Cambridge, United Kingdom) for library preparation and amplicon sequencing. For 16S rRNA gene amplicon sequencing of the V3-V4 regions, the primer pair 341 F and 806 R were used (Supplementary Table S5). To identify the fungal communities, we targeted the Internal Transcribed Spacer (ITS) Region 1, using the primer pair ITS1 and ITS2 (Supplementary Table S5). The constructed libraries were sequenced using a Novaseq 6000 platform producing 2 × 250 bp paired-end reads. Raw sequences were deposited in the NCBI Sequence Read Archive (Bioproject number PRJNA736561).

Quantitative PCR

Copy numbers of the 16S rRNA gene were determined by quantitative PCR (qPCR) using the primers 341F and 805R (Supplementary Table S5) on an AriaMX Real-Time PCR System (Agilent Technologies, Santa Clara, CA, USA). An external plasmid standard curve was made based on the pCR 2.1 TOPO vector (Thermo Fisher Scientific, Waltham, MA, USA) with a 16S rRNA gene insert amplified from bulk soil. The PCR reaction was performed in 20 µl reactions containing: 1 × Brilliant III Ultra-Fast SYBR green low ROX qPCR Master Mix (Agilent Technologies, Santa Clara, CA, USA), 0.05 µg/µl BSA (New England Biolabs Inc., Ipswich, MA, USA), 0.4 µM of each primer and 2 μl of template DNA. The thermal cycling conditions were 3 min at 95 °C followed by 40 cycles of 20 s at 95 °C and 30 s at 58 °C, and a final extension for 1 min at 95 °C. A melting curve was included according to the default settings of the AriaMx qPCR software (Agilent Technologies). The reaction efficiencies were between 97 and 102%. Fungal quantification was done by qPCR amplification of the Internal Transcribed Spacer 1 (ITS1) using the primers ITS1-F and ITS2 (Supplementary Table S5). A plasmid standard curve was made using the pCR 2.1 TOPO vector containing an ITS1 region from Penicillium aculeatum. Reaction mixture and cycling conditions were as described above for the 16S rRNA gene (Supplementary Table S5). The reaction efficiency was 84%.

Quantification of functional genes involved in N cycling

The five bacterial genes amoA, nirK, nirS, nosZ, and nifH coding for enzymes involved in N-cycling were quantified by qPCR on an AriaMx Real-Time PCR System (Agilent Technologies). Reaction mixtures and cycling conditions were as described above for the 16S rRNA gene (Supplementary Table S5). The standard curves were prepared as described in Garcia-Lemos, et al.⁵⁵. The reaction efficiencies were in the range 87%-105%.

Sequence processing

Raw reads were treated using DADA2 version 1.14.1⁵⁶. In brief, reads were quality checked and primers were removed using Cutadapt v. 1.15⁵⁷. We followed the protocol DADA2 using default parameters, with a few modifications. For 16S rRNA sequences, the forward and reverse reads were trimmed to 222 and 219 bp, respectively, while the maxEE was set to 2 and 5 for forward and reverse reads, respectively. Detection of amplicon sequence variants (ASVs) was done using the pseudo-pool option and forward and reverse reads were merged with a minimum overlap of 10 bp. Merged reads in the range of 395–439 bp were kept, as reads outside this range are considered too long or too short for the sequenced region. Taxonomy was assigned using the Ribosomal Database Project (RDP) classifier⁵⁸ with the Silva database v.138⁵⁹. For ITS region 1, quality filtered reads shorter than 50 bp were removed prior to merging the forward and the reverse reads, with maxEE set to two for both forward and reverse reads. During merging, the minimum overlap was set to 20 (default). Taxonomy was assigned with the RDP classifier using the Unite v. 8.2 database⁶⁰ after removal of chimeras.

As ITS region 1 has a variable length, reads can be lost during merging. Hence, to validate our dataset we ran only the forward reads through the DADA2 pipeline and compared the overall community structure with the dataset from the merging using a Mantel test. No significant changes were observed in the community structures between the two datasets (r = 0.99; p = 0.0001). To obtain the highest taxonomic resolution, the dataset based on the merged reads was used. Further analysis was done using the phyloseq v. 1.30.0 R package⁶¹.

Statistical analysis

Analyses of variance (ANOVA) were conducted to examine the effects of N fertilized kernza at 100 kg N ha⁻¹ (K₁₀₀) and kernza at 200 kg N ha⁻¹ (K₂₀₀) as well as to test the effect of the deep-rooted plant species: kernza and lucerne, and soil depth on each of the dependent variables. An average across the two subplots within each of the three kernza field plots was used when measured variables did not significantly differ between K₁₀₀ and K₂₀₀. Subsequent pairwise comparisons of the means was conducted using the TukeyHSD post-hoc test. Homogeneity of variance and normality were confirmed (data log-transformed when required) for all comparisons using the Fligner-Killeen test of homogeneity of variances⁶² and the Shapiro–Wilk test of normality⁶³. A permutational multivariate analysis of variance (PERMANOVA) using the Bray–Curtis dissimilarity matrix with the adonis function in the vegan R package was used to test the effect of K₁₀₀ and K₂₀₀, lucerne across both K₁₀₀ and K₂₀₀, and depth on the bacterial and fungal communities. The multivariate homogeneity of group dispersions or variances were confirmed for all comparisons using the function betadisper in vegan. The bacterial and fungal communities in response to the ascribed variables were visually represented as ordination plots with a Principle Coordinates Analysis (PCoA). Unique ASVs were defined for each depth and between K₁₀₀, K₂₀₀, and lucerne as ASVs only present in those samples belonging to a specific depth and treatment. Significance testing was conducted at p < 0.05. All statistical analyses and graphics were conducted in R⁶⁴.

Source: Ecology - nature.com