Source of termites, combs and their DNA extraction
Different castes, nymphs and fungus combs of O. obesus were collected from two different mounds of the IISER Mohali campus (Fig. S1-S3). The combs were collected in sterile plastic bags, termites were removed and then crushed within them. These were divided into 0.5 g portions in sterile glass containers. To minimize environmental contamination and degradation, DNA was extracted from some of these portions within 3 hours of the collection while others were used for microbial isolation. Major workers, minor workers and nymphs collected from these combs were also kept in separate sterile vials. Male and female alates were collected during swarming, which began after the first rains during June 2018. DNA was extracted from individual termite using the CTAB (Cetyl Trimethyl Ammonium Bromide) method  (supplementary methods S1).
DNA extractions from fungus combs often have extensive humic acid contamination which can hinder downstream PCR reactions. To remove this humic acid, the precipitation step with Phenol: Chloroform: Isoamyl alcohol were repeated until the final pellet obtained was white in color .
Isolation and identification of bacterial strains
Workers, nymphs, and alates were washed in the sterilized water to clean them of soil particles and then homogenized in 500 µl of sterile 1X PBS buffer with sterile plastic pestles. 0.5 gm portions of comb were also similarly homogenized. A dilution series (1x - 10-6x) of these homogenates were then plated on Luria Bertani Agar (LBA) and Tryptic Soya Agar (TSA) for any bacterial growth at 30˚C for 24 hrs. For strain identification, DNA from individual bacterial colonies was extracted with CTAB buffer and amplified with the primer set 341F/806R  for a portion of the 16S rRNA gene (supplementary methods S1). The sequences thus obtained were used to identify the bacterial strains by a BLAST search in NCBI. These were submitted to GenBank (accession number MN908295- MN908332). Since, multiple strains within each bacterial genus were obtained, the nomenclature of these strains were modified to read as ‘Genera-NCBI Accession Number’ (Table S1).
Bacterial- fungal interactions
Bacterial growth curves for all the 38 isolated strains were determined at 30˚C using Freedom EVO (Tecan Life Sciences) to ensure that interaction with the fungi was done when they were in their respective log phases. To identify whether these bacteria prevent the growth of the parasitic fungus, paper disc diffusion assay was used in 90 mm diameter Petri dishes containing Potato Dextrose Agar media (PDA) . Paper discs soaked with 10 µl of bacterial cultures were inoculated at the center of the Petri dish along with actively growing Pseudoxylaria in two plugs. Control growth plates were set up for both the bacteria and fungus and incubated in dark at 30˚C in a HerathermTM Compact Microbiological Incubator (Thermo Fisher Scientific). The magnitude of inhibition by bacteria was evaluated by using the following formula from Royse and Ries :
Magnitude of Inhibition = (Total Area of fungal growth in the control in mm2) – (Total Area of fungal growth in the interaction in mm2) / Total Area of fungal growth in the control in mm2
This assay was slightly modified for interaction against Termitomyces, as it grows much more slowly than Pseudoxylaria. While Pseudoxylaria plugs can fill up a control Petri plate within 5 days, Termitomyces plugs can take over 30 days to do so (Fig. S4). To account for this disparity, a liquid suspension of Termitomyces nodules and mycelia was prepared by homogenizing a 4 cm2 block of actively growing culture in sterile 1X PBS and spread on plates with Potato-Dextrose with Yeast Malt (PYME) Agar . These plates were first incubated for 36 hours at 30˚C, before adding bacterial discs . Another modification was the use of four bacterial discs against Termitomyces compared to a single bacterial disc against Pseudoxylaria. This enabled the establishment of contact between Termitomyces and the bacterial strains much faster than would have been possible with the use of single discs. These two modifications standardized the growing conditions of the two fungi so that final data could be taken on the 7th day after inoculation. All the interaction plates were evaluated for any type of inhibition every 24 hours for 7 days post-inoculation by taking photographs using the Panasonic Lumix G2 camera (Fig. S5). Photographs from the 7th day were counted as the final data and were analyzed in Adobe Photoshop CS6 by measuring the area of fungal growth in pixels and converting it into millimeters (100 pixels = 1 mm2). All interaction assays were repeated with sample sizes ranging from 3-5 plates. Representative figures for all the bacterial-fungal interaction assays are given in supplementary figures (Fig. S6-S8).
Sample preparation, running the Nanopore platform and obtaining sequences
To identify the microbiota present within termite castes and fungus combs, the V3-V4 region of the 16S rRNA gene was amplified using primers 341F/806R . Six different DNA samples (major worker, minor worker, nymph, male alate, female alate and comb) were prepared for pan-microbiome analysis on the Nanopore platform ( supplementary methods S1).
Samples were sequenced on a MinION platform (MinION Mk1B) on FLO-MIN106 flowcells for 48 hours using MinKNOW software with the protocol NC_48Hr_sequencing_FLO-MIN106_SQK-LSK108_plus_Basecaller. After the completion of the run, sequences were separated and trimmed according to their barcodes and quality (supplementary methods S1). Cleaned sequences were deposited to NCBI Sequence Read Archive (SRA) under the BioProject PRJNA608773 (BioSample accessions: SAMN14208512 - SAMN14208517).
Taxonomic identification of the pan-microbiome
A customized microbial repository was made by combining all the available 16S rRNA gene fragments from NCBI FTP site and the DictDb v 3.0 database . LAST v 973 , was used to identify the sequences with the following parameters: match score of 1, gap opening penalty of 1, and gap extension penalty of 1 . To estimate the sequencing depth rarefaction curves were generated with the Vegan package v 2.5-4 in R v 3.6.2. The estimated community richness (Chao1) and diversity indices (Evenness, Shannon, Simpson, and Inv Simpson) [37-39] were also calculated in R.
We compared the gut and comb microbiota from the other known Odontotermes sp. [31, 40, 41] with the results obtained from this study using weighted principle coordinate analysis (PCoA) in phyloseq package of R. Nineteen different microbiota were used for the similarity analysis as these also amplified the same 16S rRNA gene region and used the same reference database.