Soil was sampled from the Yangliu Agro-Ecological Experiment Station, Suixi County, Anhui Province, China (33°37′ N, 116°45′ E) in June 2020 (immediately following the harvest of winter wheat). The samples were collected from the top 20-cm soil layer using an auger, air-dried, and passed through a 2-mm sieve. The soil was classified as a Shajiang Calci-Aquic Vertosol according to the Chinese Soil Taxonomy. The soil contained 14.24 g kg⁻¹ of organic C, 0.99 g kg⁻¹ of total N, and 0.67 g kg⁻¹ of total P (Li et al., 2014).

We selected three types of OAs which comprise lignin-dominant OAs (LM), cellulose-dominant OAs (CM), and OAs with equal content of lignin and cellulose (MM). The molecular structure of OAs was determined using pyrolysis-GC/MS (MDGCMSMS TQ8050, SHIMADZU, Japan). All identifiable pyro-lysates were categorized into eight chemical groups: lignin compounds, phenol compounds, carbohydrates, saturated hydrocarbons, unsaturated hydrocarbons, nitrogen-containing compounds, aromatic hydrocarbons, and poly-aromatic hydrocarbons, according to the degree of chemical similarity; and four types of organic compounds were further classified based on chemical structure of organic C skeletons: alicyclic compounds (Alic_C), aliphatic compounds (Alip_C), aromatic compounds (Arom_C), and heterocyclic compounds (Hete_C; Supplementary Table S1).

Soil was added with LM, MM, or CM, and soil without addition of OAs was used as the control (Ctrl). The experiment included four treatments with each six replicates. OA was added to 80 g of soil (oven-dry weight) at a rate of 5 mg C g⁻¹ soil (Savy et al., 2020), and then pre-incubated at 25°C in dark for 30 days. N and P nutrients were provided according to C:N:P stoichiometric ratios (C:N:p = 10,000:833:200; Kirkby et al., 2013). Soil moisture was adjusted to 60% of water-holding capacity (WHC) using sterile deionized water during the incubation period. After the pre-incubation, the three replicates of each treatment received ¹³C-labeled straw (δ¹³C 99‰), and another three replicates without addition of ¹³C-straw were used to partition the mineralization and microbial assimilation of the ¹³C-straw. The addition rate was 1 mg straw-C g⁻¹ soil, equivalent to the amount of straw returned into soils per year. An open scintillation vial containing 10 ml of 3 M sodium hydroxide (NaOH) was placed to absorb CO₂ respired during the incubation. All soils were incubated at 25°C and 60% WHC for 28 days. Previous studies have demonstrated that soil microbes efficiently used easily-degradable component of straw in 1 week and labile straw-C was almost anabolized by soil microbes in a month (Chen et al., 2014; Zhong, 2020). Therefore, the soils were sampled in 7 and 28 days of the incubation to analyze CO₂, microbial biomass C (MBC), 16S rRNA and ITS genes, and the abundance of cellulose-degrading genes.

CO₂ was trapped in 10 ml 3 M NaOH trap solution to measure soil respiration at 7 and 28 days of incubation as previously described (Zibilske, 1994). BaCO₃ pellets were dried at 50°C and stored for ¹³C analysis using an isotope ratio mass spectrometer (Delta Plus, IRMS; Thermo Fisher Scientific, Bremen, Germany). MBC at 7 and 28 days of incubation was determined using the chloroform fumigation-extraction method (Makarov et al., 2015). For ¹³C measurement of microbial biomass, an aliquot (ca. 10 mL) of the K₂SO₄ extract was freeze-dried, and the solid material was then analyzed.

The microbial use efficiency of straw-C is the ratio of C incorporated in microbial biomass and used for respiration. The C used for respiration was based on the accumulation of CO₂ in 7 or 28 days. The proportion (f%) of CO₂-C or MBC derived from ¹³C-straw was calculated using a two-component isotopic mixing model based on Eq. (1) (Balejdent and Balabane, 1992):

where δ is the δ¹³C of CO₂-C or MBC in the ¹³C-straw amended soils (‰), δ_soil is the δ¹³C of the ¹³C-straw treatments (‰), and δ_straw is the δ¹³C of the ¹³C-straw (‰).

The δ ¹³C of MBC was calculated using Eq. (2) (De et al., 2011):

where δC_fum and δC_nfum are the δ¹³C levels in fumigated and non-fumigated soil extracts, respectively, and C_fum and C_nfum are the total MBC contents of the fumigated and non-fumigated soil extracts, respectively.

The CO₂ emission or MBC amount in the ¹³C-straw was calculated by multiplying f% with the total CO₂ emission or total MBC amount.

The microbial use efficiency of straw-C was calculated according to Eq. (3) (Geyer et al., 2016):

where ¹³C-MBC is the MBC derived from ¹³C-straw (mg kg⁻¹ soil), and ¹³C-CO₂ is the cumulative ¹³C-straw (mg kg⁻¹ soil) mineralized by day 7 or 28.

DNA was extracted from 0.25 g fresh soil samples taken at days 7 and 28. The extraction was done using the Power Soil DNA isolation kit (MoBio, Carlsbad, CA, United States), according to the manufacturer’s instructions. The DNA concentrations were determined by a nanodrop spectrophotometer (NanoDrop, Wilmington, DE). The Illumina sequencing was done by Novogene. The fragments of bacterial 16S rRNA gene and fungal internal transcribed spacer (ITS) gene were amplified using the primer pairs 341F (5′-CCTAYGGGRBGCASCAG-3′)/806R (5′-GGACTACNNGGGTATCTAAT-3′) and ITS1F (5′-CTTGGTCATTTAGAGGAAGTAA-3′)/ITS2R (5′-GCTGCGTTCTTCATCGATGC-3′), respectively. A sample-specific barcode was added to the forward primer for distinguishing the amplicons from different samples. The PCR conditions were 94°C for 5 min; 30 cycles of 94°C for 30 s, 52°C for 30 s, and an extension of 72°C for 30 s, followed by 72°C for 10 min, and held at 4°C. PCR products from all samples were pooled and purified in equimolar concentrations, and sequencing was performed on an Illumina MiSeq instrument (Caporaso et al., 2012). The bacterial 16S rRNA gene sequencing and ITS gene sequencing data were uploaded to the NCBI SRA database under accession number PRJNA898503.¹

The raw FASTQ files were demultiplexed and quality-filtered using QIIME2. Subsequently, the amplicon variant sequences (ASVs) with a 97% similarity were clustered together using DADA2 (version 1.8). Poor-quality sequences of less than 200 bp and quality scores of less than 20 were removed, and the chimeras were removed using the UCHIME algorithm (Edgar et al., 2011). The phylogenetic affiliation of each rRNA gene sequence was determined using the UNITE database for the ITS1 rRNA region and the Ribosomal Database Project (RDP) Classifier² against the SILVA (SSU117/119) database for the 16S rRNA gene at an 80% confidence threshold (Wang et al., 2007).

The fungal cbhI gene and bacterial GH48 gene were selected as functional biomarkers of cellulolytic fungi and bacteria, respectively. The primers for quantitative PCR (qPCR) of bacterial GH48 and fungal cbhI genes have been described in a previous study (Edwards et al., 2008). The primers GH48 F8/GH48 R5 (GH48_F8: 5 GCCADGHTBGGCG ACTACCT 3; GH48_R5: 5 CGCCCCABGMSWWGTACCA 3) and cbhI F/cbhI R (cbhI F: ACCAAYTGCTAYACIRGYAA; cbhI R: GCYTCCCAIATRTCCATC) were used for the qPCR assay. The abundance of bacterial GH48 and fungal cbhI genes was quantified according to modified procedures (Zhang et al., 2017). The thermal profiles of qPCR for the target genes of GH48 and cbhI were as follows: 95°C for 5 min, 40 × (94°C for 30 s, 60°C for 45 s, and 72°C for 90 s), and data collection at 84°C for 10 s; and 94°C for 4 min, 40 × (94°C for 45 s, 50°C for 30 s, and 72°C for 60 s), and data collection at 81°C for 10 s, respectively. The initial concentrations of the two plasmids as the standards for bacterial GH48 and fungal cbhI gene abundance analysis corresponded to 1.85 × 10¹¹ and 2.65 × 10¹⁰ copies g⁻¹ dry soil, respectively.

The effect of different OAs on microbial use efficiency of straw-C, fungal-to-bacterial (F:B) ratio and cbhI-to-GH48 ratios was analyzed using a one-way analysis of variance (ANOVA). Statistical significance was set at p < 0.05. Principal coordinate analysis (PCoA) was performed based on Bray-Curtis distances to visualize the change in the bacterial and fungal community under different OAs (Bray and Curtis, 1957). The difference in microbial use efficiency of straw-C among the OAs groups was analyzed using ANOVA. A correlation matrix was constructed to describe the complex co-occurrence patterns in bacteria-fungi networks. The maps were created using the geom_polygon function in the “ggplot2” package in R software.

Co-occurrence networks of bacteria and fungi in different OAs groups were constructed (Faust and Raes, 2012). We focused on the most dominant phylotypes, both abundant (with relative abundance higher than 0.1%) and ubiquitous (bacteria and fungi were present above half of all the samples). The filtering identified dominant phylotypes in soils and minimized potential spurious correlations from the rare taxa. The focus was on the dominant soil phylotypes because they are expected to have disproportionate functional importance in their ecosystems and are widely distributed. A valid co-occurrence was considered a statistically robust correlation between species when the correlation coefficient (r) was >0.6 or < −0.6 at p < 0.01. The p-values < 0.01 were adjusted by the Benjamini-Hochberg procedure to reduce the chances of obtaining false positives. The network modules were identified using the “igraph” package. The number and relative abundance of ASVs in the three most important modules were calculated to assess the effects of different OAs on specific phylotypes with strong interconnections. In addition, subnetworks were extracted by preserving the phylotypes of individual soil samples using the subgraph function in “igraph” package. Networks were constructed and visualized using the “igraph” package. Random forest modelling was performed using the “random-forest” package, and the model significance and predictor importance were determined using the “rfPermute” packages. The rest of the analyses were performed with the R software.

Py-GC/MS analysis revealed that the three OAs had considerably different chemical structures (Figure 1A). Hete_C was predominant in the CM, Arom_C was predominant in the LM, and Hete_C and Arom C were in equal proportions in the MM. The microbial use efficiency of straw-C was significantly (p < 0.001) higher in Hete_C rich CM compared with other treatments in 7 days. There was no significant difference in microbial use efficiency of straw-C among the three OAs treatments in 28 days. Therefore, straw-C decomposition efficiency was higher in the CM than in the LM and MM (Figure 1B).

The chemical structures of OAs significantly influenced the microbial community structure (Figure 2). The F:B ratios and cbhI-GH48 were higher in the CM (p < 0.001), which all declined over time (Figures 2A,B). F:B ratio declined because the abundance of fungi decreased sharply over time, whereas the abundance of bacteria increased (Supplementary Table S2). The cbhI-GH48 ratios declined because even though the abundance of both cbhI and GH48 increased over time, the increase in expression of GH48 was higher than that of cbhI (Supplementary Table S2).

There was no significant difference in the bacterial richness among the three OAs (Supplementary Figure S1A, p = 0.09), but fungi richness was considerably lower in the CM (Supplementary Figure S1B, p < 0.001). Principal coordinates analysis (PCoA) revealed that both fungi and bacteria diversity differed considerably among the three types of OAs (p < 0.001; Figures 2C,D). Furthermore, the degree of fungal partitioning (R² = 0.71) was greater than bacterial partitioning (R² = 0.39). In addition, the type of OAs had little effect on the relative abundance of bacteria at the phylum level (Supplementary Figure S2A) but significantly impacted the abundance of fungi at the phylum level (Supplementary Figure S2B). Specifically, OAs with high Hete_C proportion favored a significant increase in the relative abundance of Ascomycota.

To evaluate the effects of different OAs on the microbial co-occurrence patterns, we built a correlation-based network incorporating the dominant bacterial and fungal phylotypes. We identified three highly connected modules within the network (Figure 3A). The number of bacteria was larger than that of fungi in the three most important modules (i.e., modules 1–3; Figure 3B). The relative abundance of fungi was higher than that of bacteria in modules 1 and 3, while an opposite trend existed in module 2 (Figure 3C). Compared with the Ctrl and LM, the CM drastically increased the relative abundance of fungi in module.

The type of OAs influenced the microbial community indices and microbial use efficiency of straw-C. Hete-C and Arom_C reduced the microbial community indices (Supplementary Figure S3). Specifically, F:B, cbhI-GH48, Ascomycota, Proteobacteria, and microbial use efficiency of straw-C were strongly (p < 0.05) and positively correlated with Hete_C content but were strongly (p < 0.05) and negatively correlated with Arom_C content. Furthermore, in investigating the correlation of microbial community structure with microbial use efficiency of straw-C, we found that microbial use efficiency of straw-C increased significantly (p < 0.05) with F:B, cbhI:GH48, the abundance of module #1, and the relative abundance of Ascomycota, whereas there was a strong negative correlation between the abundance of module #3 members (p < 0.05) and microbial use efficiency of straw-C (Figure 4). In addition, there was a weak negative correlation between microbial use efficiency of straw-C and the abundance of Proteobacteria, but there was a strong negative correlation between the bacterial diversity (p < 0.05) and the fungal diversity (Supplementary Figure S4). The microbial community with greater microbial use efficiency of straw-C were predicted using random forest modelling. The importance of a variable was determined by removing it from the model and determining the subsequent change in the model’s mean square error (MSE). Our results indicated that module #1 was potentially more important in the microbial use efficiency of straw-C, accounting for 11.92% of the MSE (Figure 5, p < 0.01). Moreover, Ascomycota and F:B were the next most important factors influencing microbial use efficiency of straw-C, accounting for 10.68% (p < 0.01) and 7.16% (p < 0.05) respectively, of the MSE.