The study site was located in Xianlin campus of Nanjing Normal University, Nanjing, Jiangsu Province, China, having subtropical monsoon climate with abundant rainfall (avg. annual precipitation of 1200 mm) and the annual average temperature of 15.4°C. The formation of fruiting bodies of Termitomyces fungi mainly occurs during July to September each year. The position of termite nests was estimated according to fruiting bodies of Termitomyces fungi. Termite nests were collected under Broussonetia papyrifera (Linn.) L'Hér. ex Vent., Cedrus deodara (Roxb.) G. Don, Cinnamomum camphora (L.) Presl., Cunninghamia lanceolata (Lamb.) Hook, Pinus massoniana Lamb and Trachycarpus fortunei Wendl, respectively. Fungus combs were covered by hypha of Termitomyces fungi and distributed by lots of little white nodules (hyphal aggregates) in termite nests (Fig. 1a-i). The Termitomyces fungi cultured by the fungus-growing termite are Termitomyces clypeatus identified by amplifying and sequencing ITS region (ITS1-5.8S rRNA-ITS2) using the primer pair ITS1 (5'-TCCGTAGGTGAACCTGCGG-3') and ITS4 (5'-TCCTCCGCTTATTGATATGC-3') (Fig. 1e) and the FGTs found inside the termite nest are Microtermes pakistanicus, identified by amplifying and sequencing mitochondrial cytochrome oxidase subunit II (COII) gene using the primer pair COII1 (5'-CAGATAAGTGCATTGGATTTCGG-3') and COII2 (5'-GTTTAAGAGACCAGTACTTG-3') (Fig. 1j). The soil (the surface of the termite nest, at a depth from 0 to 10 cm) above the termite nest was collected as the surface soil (SS). The soil (with a thickness of ~ 0.5 cm, Fig. 1a) inside of the termite nest was defined as the termite nest wall (TNW). The texture of TNW is very solid, differing from the surface soil significantly. A total of nine termite nests were sampled in this study. Samples from three different structures of the termite nest including the fungus comb (FC), TNW and SS were collected, respectively. Then, all samples were transported to the laboratory on ice. Each sample was homogenously mixed and divided into two subsamples, one subsample passed through a 2-mm nylon sieve was used for the determination of physical and chemical properties, the other was stored at –80°C for extracting soil genomic DNA.

The moisture content of samples was calculated through loss of weight of 5 g sample oven-dried at 105°C for 12 h. Soil pH was determined at a 1:2.5 (w/v) soil-to-solution ratio using 1 M potassium chloride (KCl) solution. The oxalate content of samples was measured using a colourimetric method (Jin et al. 2007). Briefly, 25 ml of 1 M hydrochloric acid (HCl) were added to 10 g of dried sample for an overnight reaction on a reciprocating shaker and after shaking, pH was adjusted to 2 with 6 M HCl. The samples were centrifuged at 6000 rpm for 10 min and the supernatant was analysed using the method of Sun et al. (2019). The carbonate content was evaluated using a back-titration method according to the method of Sun et al. (2019). Mineral composition of samples was determined by X-ray powder diffraction (XRD) using an X-ray diffractometer BTX-526 (Olympus, Japan) using Co-Kα radiation (30 kV, 0.3 mA) from the 5° to 55° 2θ interval with 0.04° 2θ per step.

The total genomic DNA of samples was extracted using PowerSoil^® DNA Isolation Kit (QIAGEN, Germany). DNA samples were transported to Shanghai Sangon Biotech in dry ice for Illumina MiSeq sequencing. Primer pair frc171-F (5'-CTSTAYTTCACSATGCTSAAC-3') and frc627-R (5'-TGCTGRTCRCGYAGYTTSAC-3') were used to specifically amplify frc fragments of OxB (Sun et al. 2019). PCR amplification was performed in triplicate using a 30 μl reaction mixture containing 15 μl 2× KAPA HiFi Hot Start Ready Mix (Takara, Japan), 1 μl of each primer (10 μM),= and 20 ng of template DNA. PCR was performed using the following programme: 1 cycle of denaturing at 95°C for 3 min, first 5 cycles of denaturing at 95°C for 30 s, annealing at 45°C for 30 s, elongation at 72°C for 30 s, then 20 cycles of denaturing at 95°C for 30 s, annealing at 55°C for 30 s, elongation at 72°C for 30 s and a final extension at 72°C for 5 min. The PCR products were checked using electrophoresis in 1% agarose gels. Library construction used universal Illumina adaptors and index sequences. Before sequencing, the DNA concentration of each PCR product was determined using a Qubit^® 2.0 Fluorometer (Invitrogen, USA) and it was quality controlled using an Agilent Bioanalyzer 2100 system (Agilent Technologies, USA). The amplicons from each reaction mixture were pooled in equimolar ratios, based on their concentration and sequenced on an Illumina MiSeq PE300 platform adopting a paired-end sequencing strategy.

Raw data were treated referring to Sun et al. (2019). In brief, raw reads were merged into complete sequences using FLASH 1.2.11 software (Magoč and Salzberg 2011). Barcode sequences were subsequently removed from each sample and sequences < 200 bp were discarded using Cutadapt v.1.14 software (http://cutadapt.readthedocs.io). The processed sequences were clustered into operational taxonomic units (OTUs) at 97% similarity using USEARCH (Edgar 2013). A representative sequence with the highest abundance in each OTU was selected for taxonomic assignment. Taxonomic annotation of representative sequences was performed using the BLASTn method in the National Center for Biotechnology Information (NCBI) non-redundant database and sequences with similarity < 90% and coverage < 90% were defined as unclassified. The raw data were uploaded to the NCBI SRA database (http://www.ncbi.nlm.nih.gov/sra) with the retrieval number SRR8835650.

OxB on fungal combs were isolated by dilution coating on two-layer Schlegel solid media with calcium oxalate as the sole carbon source (Aragno 1992). Some larger colonies that produced transparent haloes were transferred to fresh Schlegel solid medium for further screened and purification. Finally, the isolate TA1 with the largest transparent halo was selected for subsequent studies. The micromorphology of the isolate TA1 was observed using a scanning electron microscope (SEM) S-3400 N (Hitachi, Japan). The isolate TA1 was also identified by 16S rRNA gene sequencing using the method of Sun et al. (2019). The 16S rRNA gene sequence of the isolate was deposited in GenBank with the accession number MK850375.

Since the isolate TA1 could not grow normally in liquid Schlegel medium, we selected modified King’s B medium to analyse the oxalate degrading ability of the isolate TA1 after trying different media. The components of modified King’s B medium (/l) were: CaC₂O₄ 50 g, tryptone 10 g, K₂HPO₄ 1.5 g, MgSO₄·7H₂O 1.5 g. The isolate TA1 was inoculated on solid Schlegel medium and incubated at 30°C for 5 days. Then, the bacteria were collected with sterile deionised water (SDW), washed thrice using SDW and diluted to 10⁸ cells per ml to serve as inoculum. The inoculum was inoculated into an Erlenmeyer flask (250 ml) with 100 ml modified King’s B medium (containing 25 mM CaC₂O₄) and incubated at 180 rpm at 30°C in an orbital shaker for 3 days. After incubation, the precipitates were collected and washed thrice with deionised water and anhydrous ethanol, respectively and dried at 60°C for mineralogical analysis. The SEM Zeiss-Supra55 (Zeiss, German) equipped with an energy dispersive X-ray energy spectrometer (EDS) was used to detect microscopic morphology and elemental composition of the precipitates and the X-ray diffractometer BTX-526 was used to detect mineral phases.

The primer pair frc 171-F and frc 306-R (5′-GDSAAGCCCATVCGRTC-3′) was used for quantifying frc copy number and primer pair 338F (5′-ACTCCTAGGGGGCGAGCAG-3′) and 518R (5′-ATTACCGCGGCGTGCTGG-3′) was used to detect the copy number of 16S rRNA gene in the total bacteria. The specific quantitative steps used were according to Sun et al. (2019).

In R version 3.3.2, the vegan package was used to calculate α diversity indices (Chao1, Shannon, Simpson) and β diversity analysis (including the principal coordinate analysis (PCoA), Adonis analysis, canonical correspondence analysis (CCA) and Mantel test. The heatmap was generated using the pheatmap package. T-test and one-way ANOVA were performed using SPSS 20 (IBM, USA). Data are expressed as mean ± standard deviation.

pH value, humidity, oxalate and carbonate contents in different structures of the termite nest were measured. The pH value of FC was 3.93 ± 0.01, significantly lower than that of TNW (5.74 ± 0.27) and SS (6.20 ± 0.22, Fig. 1k). The moisture of FC was 56.37% ± 0.80%, significantly higher than that of TNW (20.81% ± 1.55%) and SS (20.16% ± 2.59%, Fig. 1l). The oxalate content in the fungus comb was 0.12 ± 0.02 mg/g dry sample, which was significantly higher than that in the nest wall and surface soil (Fig. 1m). The highest calcium carbonate content was found in TNW (19.67 ± 3.22 mg/g soil), followed by FC (17.99 ± 3.14 mg/g soil) and SS (15.08 ± 1.65 mg/g soil, Fig. 1n).

The results of XRD showed that the main mineral in different structures of termite nests was quartz, but oxalate minerals (whewellite and weddellite) were also detected (Fig. 2). The characteristic peaks of calcite and aragonite appeared in FC by XRD analysis, Fig. 2a), but not in TNW or SS, indicating crystallised calcium carbonate exist in termite nests. In addition, clay minerals, such as montmorillonite, were detected in FC and illite was detected in TNW and SS.

The diversity of OxB in FC, TNW and SS were determined using Illumina sequencing. After splicing and quality control of the raw data, a total of 2,512,669 valid sequences were obtained.The average sequence number obtained for each sample was > 29,000 with an average sequence length of 370 bp. The α-diversity indices of OxB in the termite nest are shown in Table 1.

The number of OTUs in FC was significantly lower than that in TNW and SS, which indicated that the species richness of OxB in FC was significantly lower than that in TNW and SS (Table 1). The Chao1 index showed that the predicted number of OTUs was greater than that of detected OTUs, indicating more OxB were unrevealed in the termite nest. The result of Shannon index also showed that the diversity of OxB in FC was significantly lower than that in TNW and SS.

OTUs were assigned to six bacterial phyla, including Actinobacteria, Bacteroidetes, Cyanobacteria, Firmicutes, Planctomycetes and Proteobacteria. Amongst them, Cyanobacteria and Planctomycetes were found to contain frc for the first time. Cyanobacteria was only detected in SS. Firmicutes was only found in FC. Planctomycetes was found in SS and TNW. In SS and TNW, Proteobacteria and Actinobacteria were the dominant phyla, accounting for 41.80%–55.07% and 3.27%–19.97% of the total sequences, respectively. In FC, more than 99% of the sequences had no annotation information (Fig. 2d). A total of 61 species of OxB were annotated in the NCBI non-redundant database. Fig. 3e shows the composition and abundance of the top 10 dominant OxB in the termite nest, amongst which Azospirillum, Bradyrhizobium, Burkholderia, Methylobacterium, Mycobacterium, Paraburkholderia, Rhodoplanes, Streptomyces and Variovorax are the dominant genera.

We compared the differences of relative abundance of the first 100 dominant OTUs in different samples. The results showed that the composition of OxB in FC samples was significantly different from TNW and SS (Fig. 3a). The results of PCoA showed that the community structure of OxB amongst FC, TNW and SS was divided into three clusters with significant difference identified by ADONIS, based on Bray-Curtis distance (Fig. 3b). The results of CCA showed the environmental variables (pH, carbonate and oxalate) were associated with OxB communities (Fig. 3c), but only 18.63% of the total variance in OxB community structure was explained. The Mantel test, based on Pearson's product-motion correlation, further showed that pH value had a moderate positive correlation with OxB community (r = 0.390, p < 0.001), while oxalate and carbonate content had a weak positive correlation with OxB community (r = 0.105, p = 0.057; r = 0.124, p = 0.045).

In addition, phylogenetic relationships of the top 50 dominant OTUs and their relative abundance in different samples, based on frc sequence, were conducted (Fig. 4). The results showed that OTUs in FC differed from those in TNW and SS in frc sequences and relative abundance. OTU1, OTU3, OTU6, OTU7, OTU11, OTU12, OTU29, OTU37, OTU46 and OTU14016 also existed in FC and were also the dominant group, accounting for 50.72% of the total sequences.

The copy numbers of the total bacteria and OxB in FC, TNW and SS were determined by absolute quantitative PCR (Fig. 5). The number of OxB in FC was 9.87 × 10⁸ ± 2.03 × 10⁸ copies/g dry soil, higher than that in TNW (5.70 × 10⁸ ± 1.92 × 10⁸ copies/g dry soil), but significantly lower than that in SS (3.59 × 10⁹ ± 1.22 × 10⁹ copies/g dry soil). The number of bacteria in FC was (5.55 ± 1.41) ×10¹⁰ copies/g dry soil, significantly higher than that in TNW (1.01 × 10¹⁰ ± 0.50 × 10¹⁰ copies/g dry soil) and SS (2.24 × 10¹⁰ ± 0.44 × 10¹⁰ copies/g dry soil).

A short-rod-shaped oxalate degrading isolate TA1, identified as Methylobacterium sp., based on 16S rRNA gene sequencing, was isolated from the fungus comb (Fig. 6a-c). Liquid culture showed that the oxalate was completely degraded after incubation for 72 h (Suppl. material 1), a greater oxalate degradation capacity than Streptomyces sp. NJ10 from the report of Sun et al. (2019). The pH of the medium increased significantly during incubation. SEM imagery displayed that TA1 could form irregular rod-shaped or star-like secondary precipitates during the degradation of calcium oxalate (Fig. 6d). The EDS analysis showed that the precipitates were mainly composed of carbon, oxygen and calcium and its atomic number ratio was about 1:3:1 (Fig. 6e). Further analysis by XRD showed the precipitates were calcite (Fig. 6f).