The marginal effect of soil biota on litter decomposition under warming in a primary forest

1 Experimental design and soil sampling The gradient warming experiment was established in May 2019 using a randomized block design. The site was divided into four experimental blocks, each containing five 2 m × 2 m plots. Each plot was randomly assigned one of five warming treatments: control (CK), +0.8 °C (W1), +1.5 °C (W2), +3.0 °C (W3), and +4.2 °C (W4). In the warming plots, three infrared radiators were suspended in a triangular arrangement 1 m above the ground. Dummy infrared radiators were placed above the control plots to simulate the shading effect of the heaters. Heaters operated continuously throughout the experiment (Fig. S1). A detailed description of the experimental design is provided in the Supporting Information.Litter was collected from the gradient warming quadrats in May 2021, focusing on fresh, undecomposed fallen leaves. The collected litter was initially dried at 85 °C to a constant weight. After identification, five dominant leaf litter species were selected: Eriobotrya bengalensis, Vaccinium duclouxii, Spiraea thunbergii, Schima reinw, and Lithocarpus hancei. The litters were then dried at 55 °C, and the dry weight proportions of the species were determined as approximately 3:3:3:2:1. These proportions were used to prepare each litterbag, with a total dry weight of about 6 g per bag. Litterbags were constructed with three mesh sizes to control soil biota access: 0.1 mm (low biota participation), 2 mm (moderate participation), and 4 mm (high participation). The 0.1 mm, 2 mm, and 4 mm mesh sizes allowed access to microbes and microfauna, micro- and mesofauna, and the full decomposer community, respectively, including detritivorous invertebrates (Makkonen et al., 2012). In each plot, four litterbags of the same mesh size were placed, yielding a total of 12 litterbags per plot. Across the entire experiment, 240 litterbags were deployed. Litter decomposition was monitored at 3, 6, 12, and 24 months, with one bag recovered from each quadrat at each sampling time, resulting in 60 litter bags collected per sampling.2 Soil and litter physiochemical properties Soil cores (5 cm diameter, 10 cm height) were collected from the 0–10 cm and 10–20 cm layers in each plot in August 2021 and 2022. Soil from the same depth within each plot was combined to form a single composite sample. Samples were placed in sterile plastic bags and transported to the laboratory under cool conditions. The soil was passed through a 2 mm sieve to remove visible rocks and plant debris, and then divided into two portions. The first portion was stored at 4 °C for measurements of soil pH and nutrient content. Ammonium nitrogen (NH4+-N), nitrate nitrogen (NO3−-N), and dissolved organic carbon (DOC) were extracted from all samples using 0.5 mol L−1 K2SO4. Soil available phosphorus (AP) was extracted with 0.5 mol L−1 NaHCO3 solution (pH 8.5), while total nitrogen (TN) and total phosphorus (TP) were determined using the sodium salicylate and molybdenum–antimony methods, respectively (Hou et al., 2025). Soil organic carbon (SOC) and DOC were measured using a Vario TOC analyzer (Vario TOC, Langenselbold, Germany), with SOC determined by dry combustion at 980 °C in solid mode and DOC determined in liquid mode. Concentrations of TN, NH4+-N, NO3−-N, TP, and AP were measured using an automated discrete analyzer (De Chem-Tech GmbH, CleverChem380, Hamburg, Germany). The soil C:N ratio was calculated as the ratio of SOC to TN. Soil pH was measured in water using a 1:2.5 (v/v) soil-to-solution ratio. Total carbon (TC), TP, and TN in litter were determined using the same procedures as for soil (Table S1).3 Soil biodiversity measurement The remaining soil subsample was stored at −80 °C for high-throughput sequencing to characterize microbial communities. Microbial DNA was extracted from 0.5 g of soil using the FastDNA® Spin Kit for Soil (MP Biomedicals, Solon, OH, USA). Bacterial and fungal communities were analyzed by sequencing the V3–V4 region of the bacterial 16S rRNA gene and the fungal ITS region of the 18S rRNA gene on an Illumina MiSeq platform (San Diego, CA, USA). Demultiplexed paired-end Fastq files were processed using QIIME2 with the DADA2 denoising method for downstream analysis. After trimming sequences to remove adapters and truncating low-quality reads, amplicon sequencing variants (ASVs) with <0.01% relative abundance were filtered out. Taxonomy was assigned using a Naive Bayes classifier against the SILVA database for bacteria and the UNITE database (UNITE 8.2) for fungi, with rare ASVs and singletons removed. Bacterial and fungal diversity was assessed using the Shannon, Simpson, and Chao1 indices. Soil bacterial and fungal biomass was determined by PLFA analysis, with the following PLFAs used as bacterial indicators: i14:0, a15:0, i15:0, i16:0, a17:0, i17:0, 16:1ω7c, 16:1ω9c, cy17:0, 17:1ω8c, 18:1ω7c, 18:1ω9c, and cy19:0, and 18:2ω6c as the fungal indicator. Taken together, all of the PLFAs listed above were considered representative of total microbial biomass (Chen et al., 2023). Owing to the high SOC content and low bulk density at the study site, soil nematodes were extracted from 50 g of well-mixed fresh soil using the Baermann method (Barker, 1985). Nematodes were extracted over 48 h, fixed in 4% formaldehyde, and counted under a light microscope (Leica DM3000, Germany). For each sample, 150 individuals were randomly selected for genus-level identification; although not all could be classified, abundance was expressed as the number of individuals per 100 g of dry soil.