The origin of fungi-culture in termites was associated with a shift to a mycolytic gut bacteria community

Termites forage on a range of substrates, and it has been suggested that diet shapes the composition and function of termite gut bacterial communities. Through comparative analyses of gut metagenomes in nine termite species with distinct diets, we characterise bacterial community compositions and identify biomass-degrading enzymes and the bacterial taxa that encode them. We find that fungus-growing termite guts are enriched in fungal cell wall-degrading and proteolytic enzymes, while wood-feeding termite gut communities are enriched for plant cell wall-degrading enzymes. Interestingly, wood-feeding termite gut bacteria code for abundant chitinolytic enzymes, suggesting that fungal biomass within the decaying wood likely contributes to gut bacteria or termite host nutrition. Across diets, the dominant biomass-degrading enzymes are predominantly coded for by the most abundant bacterial taxa, suggesting tight links between diet and gut community composition, with the most marked shift being the communities coding for the mycolytic capacity of the fungus-growing termite gut.


35
Termites are widespread in tropical, subtropical and warm temperate regions (Buczkowski and 36 Bertelsmeir 2016) and form a diverse group of more than 3,000 described species in 281 genera and 37 seven families (Eggleton 2000;2001;Kambhampati and Eggleton 2000;Inward et al. 2007). They 38 have major impacts on their environments (Buczkowski and Bertelsmeir, 2016), and this success 39 has been attributed to their capacity to use nutritionally-imbalanced, recalcitrant food sources, ). This division of labour is consistent with gut bacteria being of importance mainly when 59 the comb material passes through the termite gut in a second passage (cf. Nobre et al. 2011;Poulsen 60 et al. 2014;Poulsen, 2015), but recent work has suggested that partial lignin breakdown may also be 61 accomplished during this first gut passage in Odontotermes formosanus (Li et al. 2017).

142
The bacteria encoding the most abundant glycoside hydrolase families 143 To gain further insight into the functional contribution of gut microbiota members to fungal 144 digestion, we grouped enzymes in GH families by fungal cell wall components targeted and 145 bacteria of origin (Table 2; Figure 3). The bacterial orders Clostridiales and Bacteroidales 146 contributed most fungal cell wall-degrading enzymes in fungus-growing termite guts (78%), while extent, Spirochaetales in wood feeders. In the remaining termite species, most of the fungal cell 155 wall-degrading enzymes were contributed by Clostridiales, Bacteroidales, and Spirochaetales (17%-156 92%), but the overall abundance of these enzymes was far lower than in fungus-growing termites 157 and wood feeders (Table 2; Figure 3).

159
Distinct protease profiles in termite gut metagenomes 160 We identified proteases in termite gut metagenomes and classified them by catalytic types to 161 investigate the proteolytic potential of gut communities. Proteases were most abundant in fungus-  Table S4). One of the most distinct differences we observed was for glutamic peptidases and mixed 165 peptidases, which were enriched only in the guts of fungus-growing termites. The wood feeder Cu. 166 ugandensis displayed the lowest protease abundance, followed by the litter-feeder Cornitermes sp. 167 As observed for fungal and plant cell wall-degrading enzymes, the most abundant taxa were also 168 predicted to encode for most of the proteases ( Figure 4A). Clostridiales and Bacteroidales, 169 Clostridiales and Spirochaetales, and Spirochaetales were the main contributors encoding proteases 170 within the metagenomes of fungus-feeding, dung-feeding, and wood-feeding termites, respectively.  Fungus-growing termite gut mycolytic enzymes were primarily coded for by Clostridiales and 189 Bacteroidales, which dominate the core gut microbiota of the Macrotermitinae (Hongoh et al. 2005, 190 Hongoh 2010; Dietrich et al. 2014;Otani et al. 2014;Mikaelyan et al. 2015). The ancestor of the 191 Macrotermitinae likely had a lower termite bacterial gut microbiota (but without protists) (Brune 192 2014; Brune and Dietrich 2015). Fungiculture exposed the gut microbiota to higher amounts of 193 fungal biomass including proteins than to the ancestral lignocellulolytic diet, resulting in a gut 194 microbiota that converged to become more similar to those observed in extant cockroaches 195 (Dietrich, Kohler, and Brune 2014;Otani et al. 2014). This is likely a product of several factors. outcompeted/selected against (e.g., the genus Treponema). Third, novel lineages adopted from other 203 termites or the environment were likely co-opted when fungiculture evolved. This is consistent with 204 recent work demonstrating rampant horizontal transmission of gut bacterial lineages associated with 205 termites (Bourgoignon et al. 2018). Collectively, this led to a gut microbiota enriched in mycolytic 206 (e.g., α-mannanases, β-glucanases, chitinases and proteases) and reduced in lignocellulolytic (e.g., 207 cellulase, cellobiohydrolase, hemicellulose and laccase) enzymes. feeding termites suggests that the decaying wood these species feed on, harbour fungal biomass that 216 gut bacteria or the termite host may utilise. However, it is also conceivable that fungal cell wall-217 degrading enzymes, such as β-1,3-glucanase, serve to cleave fungal cell wall to protect against dung and wood feeders could be explained by the presence of fungal biomass in the substrate 227 ingested by these termites. This is consistent with the low abundance of these enzymes in termites 228 feeding on soil and humus with much lower protein content. The only under-represented proteases 229 in fungus-growing termites was threonine peptidases, which has comparable activities as serine and 230 cysteine peptidases (Powers et al. 2002). The absence of these enzymes could thus be compensated 231 for by the presence of other protease families; however, more work will be needed to test what these 232 enzymes indeed target.

234
Similarities in host diet have been shown to drive convergence in the functional potential of gut 235 microbes in other organisms (Muegge et al. 2011;Delsuc et al. 2014 (Madden and Coutts, 1979;Hajek et al. 2013), and ambrosia beetles (Batra 1966; Hulcr and are also abundant and mycolytic in fungus-growing termites. The convergent prevalence of specific 251 protease-producing Firmicutes and Bacteroidetes taxa suggests that they were selected for high-252 protein host diets, consistent with findings in humans (Eckburg et al. 2005) and pigs (Leser et al. 253 2002), and likely contributing to the observed convergence in fungus-growing termite and 254 cockroach gut metagenomes (Dietrich et al. 2014;Otani et al. 2014;Schauer et al. 2012).

256
The prevalence of microbial communities with ample mycolytic capacities in the guts of fungus-257 growing termite species supports that the shift to a fungal diet was associated with both a functional 258 and compositional shift in gut microbial communities at the onset of fungiculture in termites. The 259 enrichment of GH families encoding fungal cell wall-degrading enzymes and proteases indicates 260 adaptations to the decomposition a fungal diet, consistent with this capacity being absent or less in 261 termites with predominantly plant-based diets. An exception is wood-feeding termites, for which 262 wood-degrading fungi also may comprise an appreciable component of the termite diet. Further 263 work will be needed to elucidate whether the functional capacities of the gut microbiota reflect the 264 amount of fungal biomass in the diet and potential differences in dietary properties of the fungal 265 species fed on. Furthermore, taxonomy assignments were reference-dependent and limited by the  Guts were ground in liquid nitrogen, after which DNA was extracted using the Qiagen Animal 283 Tissue Mini-Kit (Qiagen, Hilden, Germany) according to the manufacturer's description, with the 284 exception that a chloroform extraction step was followed by incubation with protease K. After 285 proteinase K digestion, one volume chloroform/isoamyl alcohol (24/1) was added; tubes were 286 incubated for 15min on a slowly rotating wheel, and centrifuged at 3,000 g for 10 min. The 287 supernatant was transferred to spin columns and the remainder of the manufacturer's protocol was 288 followed. The quality and purity of samples were determined using NanoDrop® (Thermo   HiSeq2500. The quality of raw sequencing reads was assessed before assembly. Reads containing 296 the adaptor, more than 10% N or more than 50% low quality bases (Q-score 5), were removed. To  Clean reads were assembled by IDBA-UD v1. 1.2 (Peng et al. 2010;2011) with an iterative set up 302 from k-mer size of 19 to 99 at step of 10 (--pre_correction --mink 19 --maxk 99 --step 10).