Environmental predictors of DOM

We measured 475 environmental variables in soils and their porewaters to partition variation in DOM composition. The variables were associated with microbial metabolic activity and biomass (n = 38), microbial diversity (n = 6), expression of CAZymes (n = 412), and soil-water chemical and physical properties (i.e., physico-chemical conditions: n = 19) (see Table ^S8 for full list).

We measured 38 variables related to microbial activity and biomass. Bacterial production (BP) was measured using ³H leucine incorporation after Bååth et al.^66,67 (Supplementary Methods). Decays per minute measured were converted to BP (mg L⁻¹ day⁻¹) using standard conversion factors⁶⁸. For microbial basal respiration, 1 g of field-moist soil was placed in a 10 mL glass vial and incubated for 24 h in the dark at room temperature (21 °C) without manipulating moisture levels. Incubations of live cells were initiated within 5 h of soil collection. Respired CO₂ (μL min⁻¹ g⁻¹ dry soil) collected in the headspace of the vial was measured with an infra-red gas analyser (QS102, Qubit Systems, Canada). We also measured the potential activity of four hydrolytic enzymes in cell suspensions: beta-xylosidase and beta-glucosidase that break down xylose and oligosaccharides, respectively, N-acetyl-B-D-glucosaminidase that degrades glycoside and amino sugars, and phosphatase that degrades proteins. All enzyme activities were assayed in 96-well plates under controlled conditions (pH 5, room temperature for 1-h) using 4-methylumbelliferone-fluorescence tagged substrates and measured with a Synergy H1 Hybrid spectrophotometer/fluorometer (BioTek Instruments, USA) as in existing protocols^69,70. To complement the enzyme assays, we measured microbial substrate use of 31 different carbon sources using Biolog® EcoPlates^TM (Biolog Inc., USA). Cell suspensions were prepared by adding 10 g dry soil equivalent to 95 mL of sterile NaCl solution (0.85%), 6 ceramic beads and 5 g glass microspheres. Samples were mixed on an orbital shaker for 30 min at 200 rpm, left to settle for 15 min, and a serial dilution on 1 mL supernatant performed to an end concentration of 1:1000 in saline. Inoculated plates were incubated in the dark at 25 °C, and absorbance read every 24 hours on the Synergy H1 microplate reader. Blank wells were always subtracted to reduce noise. Average well colour development (ACWD) was calculated as the rate of change over a 24 h period from day 1 to day 2, as day 2 produced the greatest ACWD indicating the least chance of substrate limitation. Finally, bacterial biomass was measured using flow cytometry. Bacterial cells were separated from soil matrices using buoyant density centrifugation adapted from ref. ⁷¹ (Supplementary Methods). Flow cytometry was performed on Accuri™ C6 Plus flow cytometer (BD Biosciences, UK) equipped with a 200 mW solid-state laser emitting light at 488 nm, measuring green fluorescence at 520 nm (FL1 channel). The FL1 and forward scatter detectors were used to reduce autofluorescence found in environmental samples. We ran samples in triplicate, passing 100 μL per technical replicate through the flow cytometer at a speed of 60 μL min⁻¹ to prevent overlap of scatter events. Gates were set by comparing scatter plots produced from stained and unstained samples. Flow cytometer counts were validated with Spherotech 8-peak and 6-peak beads. Bacterial biomass (mg g⁻¹ dry soil) was calculated from cell counts, assuming a conversion factor⁷² of 58 fg cell⁻¹.

Bacterial and fungal taxonomic diversities were assessed using exact sequence variants (ESVs) generated by amplicon sequencing of the 16S rRNA gene and ITS2 region, respectively. DNA was extracted from 250 mg of homogenised soil using the DNeasy PowerSoil Pro kit and the QIAcube® (Qiagen, Germany) automated platform. 16S rRNA and ITS2 libraries were prepared following ref. ⁷³ with well-established primers (Table ^S11). The one exception was that the first set of PCR reactions were set up by mixing 25 μL of HotStarTaq Plus Master Mix, 19 µL RNase-Free Water (Qiagen, homogenised), 0.5 μL of 10 μM primer and 5 μL of gDNA at 5 ng μL⁻¹. Indexed and purified amplicons were quantified using the Synergy™ Mx Microplate Reader (BioTek Instruments, USA) before pooling at equimolar concentration. Libraries were sequenced paired-end (2 × 250 bp) on the Illumina Miseq platform at the Aquatic and Crop Resource Development Research Centre, National Research Council Canada, Saskatoon at an average (±SE) read depth of 23594 (±863) and 34,017 (±2380) reads for 16S and ITS, respectively. Sequence data were processed using the MetaWorks pipeline version 1.4.0⁷⁴. To reduce potential bias introduced by both large differences in read depth (i.e. >10-times difference) and small, uneven libraries, we removed samples with <1000 reads and remaining samples were rarefied to the 15th percentile of reads (7150 and 11,103 for 16S and ITS, respectively) using the rrarefy function in vegan⁷⁵. Eight 16S and six ITS samples with slightly less than the 15th percentile were also kept (≥6097 and ≥7544 reads for 16S and ITS, respectively) based on rarefaction curves that showed saturation. Reads were then taxonomically annotated with the RDP classifier v2.13 and UNITE classifier v2.0 for 16S and ITS, respectively. We calculated bacterial and fungal diversity with the Shannon–Weiner index that accounts for relative abundances of ESVs in addition to their number using the diversity function in the R package vegan⁷⁵.

To identify CAZyme genes and quantify their transcripts, we used metagenome and metatranscriptome shotgun sequencing, respectively. Shotgun metagenomic libraries were prepared with the Nextera XT DNA library preparation kit and the Nextera XT Index kit v2 (Illumina, USA) following the manufacturer’s instructions using the same input DNA that was used for amplicon sequencing. DNA libraries were purified with Agencourt AMPure XP beads (Beckman Coulter, USA) and fragment size (250–1000 bp) verified on a 2100 Bioanalyzer with a high-sensitivity DNA kit (Agilent, USA). Libraries were quantified with the Qubit BR dsDNA assay kit and pooled at equimolar concentrations prior to pair-end sequencing (2 × 150 bp) at the Centre d’Expertise et de Services Génome Québec on an Illumina Novaseq platform. Metatranscriptomes were obtained by extracting RNA from 2 g of soil using the RNeasy® PowerSoil Total RNA Kit (Qiagen, Germany), except that the phenol/chloroform step was repeated twice. The pellet was suspended in 50 µL RNase/DNase-free water, treated with the RNA Clean & Concentrator-5 with DNase I treatment kit (Zymo Research, USA), and eluted in 15 µL of DNase/RNase-free water. RNA quality was verified with the 2100 Bioanalyzer using the RNA 6000 Nano or Pico assay (Agilent, USA), while RNA concentration was determined with the Qubit RNA high-sensitivity assay kit (Life Technologies, USA). Absence of residual DNA in RNA extracts was further confirmed by PCR amplification of the 16S gene. rRNA was depleted from RNA extracts using the Pan-Prokaryote riboPOOL-kit (siTOOLs Biotech, Germany) with hydrophilic streptavidin magnetic beads (New England Biolabs, USA). rRNA-depleted RNA was then purified with the RNA Clean & Concentrator-5 kit and eluted into 10 µL of DNase/RNase-free water. Libraries were prepared using the NEBNext Ultra™ II RNA Library Prep Kit for Illumina (New England Biolabs, USA) and the NEBNext Multiplex Oligos for Illumina kit (New England Biolabs, USA) following the manufacturer’s protocol for rRNA-depleted RNA. A quality check of the libraries was performed on the 2100 Bioanalyzer with the high-sensitivity DNA kit (Life Technologies, USA) prior to pooling and pair-end sequencing (2 × 125 bp) on an Illumina HiSeq platform at the Aquatic and Crop Resource Development Research Centre, National Research Council Canada, Saskatoon. Metagenomes were screened with Fastp⁷⁶ for read adaptor removal and co-assembled per sampling site with metaSpades⁷⁷ (v0.6.1) using Kbase⁷⁸ according to default parameters and including the BayesHammer option for read error correction⁷⁹. Gene sequences were identified on the assembled contigs using Prodigal⁸⁰ and then annotated as CAZymes using Hidden Markov Models from dbCAN (v9⁸¹, e-value < 1e−15; coverage > 0.35) in local searches with HMMER v3.1b1⁸². Metatranscriptomes were quality-filtered with Fastp according to default parameters⁷⁶ and mapped against gene sequences confirmed as CAZymes to obtain their expression profiles using CoverM (v0.6.1 using the ‘tpm’ option, https://github.com/wwood/CoverM). Transcript counts were normalised using the R package DESeq2 to correct for library size and composition and allow for comparison between samples⁸³.

Finally, major ions, nutrients, and metal concentrations were measured from a subset of the soil pore water at the Great Lakes Forestry Centre, Sault Ste. Marie, Ontario, according to methods outlined in Table ^S12. Soil moisture content was directly measured by weighing the change in the mass of ca. 10 g of soil before and after drying in an oven for 24 h at 105 °C degrees relative to the original mass.