Termites and subsocial roaches inherited many bacterial-borne carbohydrate-active enzymes (CAZymes) from their common ancestor

Abstract

Termites digest wood using Carbohydrate-Active Enzymes (CAZymes) produced by gut bacteria with whom they have cospeciated at geological timescales. Whether CAZymes were encoded in the genomes of their ancestor’s gut bacteria and transmitted to modern termites or acquired more recently from bacteria not associated with termites is unclear. We used gut metagenomes from 195 termites and one Cryptocercus, the sister group of termites, to investigate the evolution of termite gut bacterial CAZymes. We found 420 termite-specific clusters in 81 bacterial CAZyme gene trees, including 404 clusters showing strong cophylogenetic patterns with termites. Of the 420 clusters, 131 included at least one bacterial CAZyme sequence associated with Cryptocercus or Mastotermes, the sister group of all other termites. Our results suggest many bacterial CAZymes have been encoded in the genomes of termite gut bacteria since termite origin, indicating termites rely upon many bacterial CAZymes endemic to their guts to digest wood.

The phylogenetic trees of prokaryotic CAZymes found in termite gut mirror the termite phylogeny, suggesting termites rely on many CAZymes that were already encoded in the genomes of the ancestors of their gut prokaryotes.

Introduction

Termites, the oldest lineage of social insects with a fossil record dating back ~130 million years ago¹, are best known for their xylophagous habits. They descend from a wood-feeding cockroach ancestor^2,3, a diet many species have retained, except in the Termitidae, which include many species feeding on highly decomposed wood or soil^4,5. Although termite genomes encode a few cellulase genes^6,7, the ability of termites to digest and metabolize the wood lignocellulose largely depends on their symbiotic gut microbes, including bacteria, archaea, and lignocellulolytic protists present in all termite families but the Termitidae^8,9. In addition to these gut microbes, the termitid subfamily Macrotermitinae is associated with the lignocellulolytic fungus Termitomyces they cultivate inside their nest¹⁰.

Lignocellulose is a recalcitrant biopolymer composed of cellulose, hemicellulose, lignin, and a variety of minor components¹¹. Cellulose is a linear chain of glucose, hemicellulose is composed of various sugars linked by networks of bonds, and lignin is a biopolymer composed of cross-linked phenolic compounds¹². The degradation of the cellulose, hemicellulose, and lignin composing lignocellulose requires the action of distinct cocktails of Carbohydrate-Activate Enzymes (CAZymes) attacking various bonds of the polymer. CAZymes are enzymes that biosynthesize, break down, modify, or bind carbohydrates and glycoconjugates^13,14. They are divided into six classes based on their properties: Glycosyl Transferases (GTs) catalyze the formation of glycosidic bonds¹⁵; Glycoside Hydrolases (GHs), Polysaccharide Lyases (PLs), and Carbohydrate Esterases (CEs) cleave or rearrange glycosidic bonds^16,17; enzymes with Auxiliary Activities (AAs) act in conjunction with CAZymes, helping GH, PL, and CE gaining access to carbohydrates, for example by degrading lignin¹⁸; and enzymes of the Carbohydrate-Binding Modules class (CBMs) bind to carbohydrates and are associated with catalytic modules¹⁹. Therefore, GH, PL, CE, and AA are the classes of CAZymes involved in lignocellulose degradation.

Most CAZymes depolymerizing lignocellulose in the termite gut are produced by gut microbes^20–22. Termite gut microbes participate in the hydrolysis of cellulose through the production of various CAZymes, such as endoglucanases (EC 3.2.1.4) (e.g., GH9, GH45, and GH51) and β-glucosidases (EC 3.2.1.21) (e.g., GH1, GH3, and GH5)^22,23. They also participate in the degradation of hemicellulose, for example through the production of xylanases (e.g., GH10, GH11, and GH43) and endo-β-1,4-mannanase (e.g., GH8 and GH26), which respectively degrade xylans and glucomannans, two primary constituents of hemicellulose²⁴. Termite gut bacteria are also involved in other metabolic functions, such as nitrogen metabolism, including the fixation of atmospheric nitrogen and the recycling of nitrogen wastes^25,26.

Termites acquire their gut microbes through vertical and horizontal transmission events, a process referred to as mixed-mode transmission²⁷. Vertical transmission is the primary mode of acquisition of lignocellulolytic protists and many bacterial lineages abundant in the gut of termites^28–32. It is mediated by nestmates exchanging gut fluid through trophallaxis from both ends of the digestive tract along with the microbes it contains, a behavior ensuring the transfer of gut microbes from parent colonies to the offspring colonies^33–35. Many bacterial lineages transferred vertically form clades endemic to the termite gut, and their phylogenetic trees generally present a strong cophylogenetic signal with that of their termite host^28,29,36. Termite gut bacteria acquired from the environment can readily be recognized in phylogenetic trees by their close evolutionary relationship to bacteria from non-termite environmental samples. These bacteria do not form termite-specific clades and do not present cophylogenetic patterns with termites^29,37,38. So far, these cophylogenetic analyses were performed using marker genes, and it remains unclear whether bacterial genes functionally relevant for the processing of glycans and lignocellulose digestion present similar cophylogenetic patterns with termites.

Some CAZyme gene families are ubiquitous in gut metagenomes of all termites and their sister group, the cockroach genus Cryptocercus²⁰. One potential explanation for the origin of bacterial-borne CAZyme families present in the gut of all modern termites is their presence in the genomes of the bacteria that initially colonized the gut of termites and Cryptocercus over the past 150 million years. This hypothesis entails that these CAZyme genes are encoded in the genomes of bacteria vertically transmitted across generations of termites. Therefore, the phylogenetic trees of these CAZyme genes are expected to present a cophylogenetic pattern with termites, similar to that found with bacterial marker genes²⁸. Alternatively, the CAZymes encoded in the genomes of bacteria present in the gut of modern termites may have been acquired more recently by horizontal transfers from bacteria living outside termite guts, in which case no cophylogenetic signals are expected. In this study, we analyzed the gut metagenomes of 195 termites and one Cryptocercus to reconstruct the evolutionary history of termite gut CAZymes. We built the phylogenetic trees of 180 CAZyme families using sequences derived from termites and Cryptocercus and sequences from the GTDB database not associated with termites. We carried out cophylogenetic analyses with the phylogenetic trees of CAZyme sequences forming termite-specific clusters (hereafter: TSCs) and one phylogenetic tree of termites reconstructed using 322 ultraconserved elements (UCEs) by Arora et al.²⁸. Our analyses revealed a strong cophylogenetic signal between termites and many clusters of CAZyme sequences only found in termite gut metagenomes, suggesting that the termite gut microbiota encodes unique CAZymes inherited across generations over the past 150 million years.

Results and discussion

The set of dominant CAZyme families is conserved across termite gut metagenomes

We found a total of 101,941 CAZyme sequences in the gut metagenome assemblies of 195 termites and one individual of Cryptocercus. Our sampling included species from all 13 termite families and 15 of the 18 subfamilies of Termitidae (as defined by ref. ³⁹ (Supplementary Data 1)). We found up to 135 CAZyme families per metagenome. In total, we detected 180 CAZyme families across all gut metagenomes, including 96 GHs, 42 GTs, 11 PLs, 14 CEs, 5 AAs, and 12 CBMs (Supplementary Data 2). 34 CAZymes were found in more than 70% of gut metagenomes, nine of which, including the lignocellulolytic GH3, GH5, GH13, GH43, and GH77, were present in more than 90% of gut metagenomes, confirming that the dominant CAZyme families are ubiquitous across the gut bacterial communities of all termite species, as described in more detail by ref. ²⁰.

Many bacterial CAZymes found in the gut of termites form clusters endemic to termite gut and present a cophylogenetic pattern with termites

We reconstructed the phylogenetic trees of each CAZyme family composed of more than 20 sequences derived from termite gut bacteria using sequences from the gut metagenomes of termites and Cryptocercus and sequences not associated with termites obtained from the GTDB database and identified with BLAST searches. For the 12 CAZyme families divided into subfamilies, we reconstructed one phylogenetic tree for each subfamily. Of the 201 reconstructed CAZyme trees, 116 contained up to 23 termite-specific clusters (TSCs) (Supplementary Data 3, Supplementary Data 4), which we defined as clusters including only sequences associated with termites and found in at least 20 termite and Cryptocercus samples. Some CAZyme trees included a dozen TSCs or more. This was the case of CAZymes ubiquitous amongst termites involved in cellulose degradation, such as GH3 (containing 16 TSCs) and GH5_2 (11 TSCs), and hemicellulose degradation, such as GH5_4 (12 TSCs), GH13 (23 TSC across 11 subfamily trees), and GH43 (17 TSCs across ten subfamily trees).

We identified 420 TSCs comprising an average of 120 sequences, the largest of which was composed of 1080 sequences of the amylomaltase GH77 primarily belonging to Breznakiellaceae (phylum Spirochaetota, previously family Treponemataceae) (Supplementary Data 5, Supplementary Data 6). We carried out cophylogenetic analyses between each TSC and one phylogenetic tree of termites reconstructed using 322 UCE loci. The topology of our termite phylogenetic tree was consistent with previous phylogenetic trees reconstructed with transcriptome and UCE data^40,41. We used three cophylogenetic methods: PACo⁴², the generalized Robison-Foulds metric⁴³, and the method of Nye et al.⁴⁴. 404 of the 420 TSCs showed a significant cophylogenetic signal with the three methods, 333 of which were highly significant (p < 0.001) for all three methods (Supplementary Data 6). TSCs highly significant (p < 0.001) for all three methods included the 20 TSCs composed of more than 500 sequences, all of which, besides four TSCs composed of GTs, were involved in lignocellulose degradation (two GH3, four GH5, one GH10, one GH13, one GH18, one GH20, two GH57, one GH77, one GH94, one GH130). On average, 42.3% of the CAZyme sequences derived from the contigs composing each metagenome assembly and 44.5% of CAZyme raw reads generated with the Illumina sequencing platform belonged to TSCs (Supplementary Data 2), indicating that they are an important component of the cocktail of bacterial CAZymes found in termite guts. Note that these values represent underestimations of the bacterial CAZymes forming clusters endemic to the termite gut environment given our conservative definition of TSCs, which included sequences of at least 20 samples, and our non-exhaustive sampling effort of the termite diversity. The high relative abundance of bacterial CAZymes composing TSCs is reminiscent of past studies performed on the 16S rRNA gene and bacterial marker genes that demonstrated many key members of the gut microbiota of termites, such as the Breznakiellaceae (phylum Spirochaetota) and the Ruminococcaceae (phylum Bacillota), belong to lineages endemic to termite guts and presenting cophylogenetic patterns with termites^28,29,36. In addition, most TSCs were composed of CAZyme families involved in the lignocellulose degradation, such as GH5, GH9, GH13, or GH43. Therefore, many CAZyme genes encoded by the gut microbiota of termites are only found in termite guts, are involved in lignocellulose degradation, and mirror cophylogenetic patterns found for bacterial marker genes.

Related termite species harbour termite-clade specific CAZyme clusters

Bacterial contigs from the gut metagenomes of termites and Cryptocercus comprising CAZymes were taxonomically annotated with DIAMOND BLASTx searches⁴⁵ against the GTDB database Release 207⁴⁶. Many bacterial CAZyme sequences composing TSCs were involved in lignocellulose digestion and found to belong to taxa dominating the gut microbiota of termites and known to present strong cophylogenetic signals with their termite hosts. This is well illustrated by many TSCs mostly comprised of sequences assigned to Spirochaetota (mostly Breznakiellaceae) and Fibrobacterota (mostly Candidatus Fibromonas) (Table 1, Fig. 1A–D), two bacterial lineages dominating the gut microbiota of many termite species and involved in the digestion and fermentation of wood fibers⁴⁷. Notably, CAZymes assigned to Breznakiellaceae and Candidatus Fibromonas form large termite clade-specific clusters associated exclusively with Microcerotermes, a genus of termites represented by 58 samples in this study (Fig. 1A–D). Termite clade-specific CAZyme clusters annotated as Breznakiellaceae and Candidatus Fibromonas were also found in the Nasutitermitinae (Fig. 1A–D). These results corroborate those of Arora et al.²⁸, who found termite clade-specific lineages of Breznakiellaceae and Candidatus Fibromonas exclusively associated with Microcerotermes and the Nasutitermitinae. The similar cophylogenetic patterns found across many genes of Breznakiellaceae and Candidatus Fibromonas and involving the same termite hosts highlight the stability of these genomes over tens of millions of years of association with specific termite lineages.

Table 1.

Results of the cophylogenetic analyses performed on the 420 termite-specific bacterial clusters (TSCs)

Cophylogeny	Fibrobacterota	Spirochaetota	Bacteroidota	Firmicutes A	Others
PACo p-value < 0.001	82	55	49	26	142
0.05 > PACo p-value ≥ 0.001	5	3	27	5	19
PACo non-significant p-value	1	1	2	0	3
Nye et al. p-value < 0.001	85	59	56	25	153
0.05 > Nye et al. p-value ≥ 0.001	2	0	14	6	8
Nye et al. non-significant p-value	1	0	8	0	3
Robinson–Foulds p-value < 0.001	87	59	58	27	156
0.05 > Robinson–Foulds p-value ≥ 0.001	1	0	16	4	6
Robinson–Foulds non-significant p-value	0	0	4	0	2
Total	88	59	78	31	164

P-values were estimated using three cophylogenetic analyses (PACo, generalized Robinson Foulds (RF) metric, and Nye et al.’s method). TSCs were assigned to a bacterial phylum when more than 95% of sequences were assigned to this phylum. The phylum Firmicutes is split into multiple categories in the GTDB database, including Firmicutes_A, one category abundant in termite guts.

Fig. 1. Four of the 420 maximum-likelihood phylogenetic trees of termite-specific bacterial clusters (TSCs).

All four trees showed strong cophylogenetic signals with termites. The trees included several termite clade-specific CAZyme clusters only found in Nasutitermitinae and Microcerotermes. Phylogenetic trees of (A) GH2 Cluster 10 composed of 97.4% of Spirochaetota, (B) GH77 Cluster 6 composed of 98.1% of Spirochaetota, (C) GH9 Cluster 7 composed of 100% of Fibrobacterota, and (D) GH8 Cluster 4 composed of 100% of Fibrobacterota. E Maximum-likelihood phylogenetic tree of termites inferred from UCEs. Black dots indicate CAZyme sequences assigned to a different bacterial phylum.

Termite clade-specific CAZyme clusters were also found to be associated with termite lineages sampled less intensively. For example, several TSCs annotated as Bacteroidota comprised subclades of termite clade-specific CAZyme clusters annotated as Candidatus Azobacteroides (Fig. 2A), a bacterial endosymbiont of the cellulolytic protist Pseudotrichonympha⁴⁸, confirming the exclusive association of these bacteria with all genera of Neoisoptera excluding Reticulitermes and the Termitidae, which do not harbor Pseudotrichonympha⁴⁹. Several TSCs also included subclades primarily associated with the Kalotermitidae (Fig. 2B, C), suggesting that termite-clade-specific bacterial CAZyme clusters are present across the termite tree of life. We expect future studies, relying on a comprehensive sampling of termite lineages not sampled intensively in this study, to reveal the existence of additional termite-clade-specific bacterial CAZyme clusters.

Fig. 2. Three of the 420 maximum-likelihood phylogenetic trees of termite-specific bacterial clusters (TSCs).

All three trees showed strong cophylogenetic signals with termites and included termite clade-specific CAZyme clusters associated with Kalotermitidae or non-Termitidae Neoisoptera. Phylogenetic trees of (A) GH57 Cluster 7 composed of Bacteroidota only and including the genus Candidatus Azobacteroides, (B) GH10 Cluster 13 composed of 98.5% of Bacteroidota, and (C) GH73 Cluster 3 composed of 97.6% of Bacteroidota. D Maximum-likelihood phylogenetic tree of termites inferred from UCEs. *CAZyme sequences annotated as Candidatus Azobacteroides; **CAZyme sequences assigned to Candidatus Azobacteroides with BLAST search against the GenBank database; X CAZyme sequences originally annotated as Candidatus Azobacteroides but with conflicting BLAST search against the GenBank database. Black dots indicate CAZyme sequences assigned to a different bacterial phylum.

Some termite-specific clusters are present across termites and Cryptocercus, suggesting their history of association is ~150 million years old

Many bacterial CAZyme sequences forming TSCs were found in the gut metagenomes of diverse termites and Cryptocercus. For example, 131 of 420 TSCs comprised at least one CAZyme sequence associated with the gut metagenomes of Cryptocercus or Mastotermes, which were represented by only one sample each in this study, and 229 TSCs comprised CAZyme sequences from the gut metagenomes of both Neoisoptera and non-Neoisoptera termites (Supplementary Data 6; Fig. 3A–C). The presence of bacterial CAZyme sequences associated with phylogenetically distant termite species in many TSCs suggests they have an ancient history of association with termites and Cryptocercus, some dating back to the origin of these insects ~150 million years ago. There is evidence that termites acquired some of the symbiotic bacterial lineages populating their guts some 150 million years ago²⁸, roughly around the time they acquired their gut lignocellulolytic protists⁹. While horizontal transfers of gut bacterial CAZymes among unrelated termite species could theoretically explain their distribution across termites in some cases, the strong cophylogenetic signals between most TSCs and termites and the existence of numerous termite clade-specific CAZyme clusters within TSC trees suggest coevolution with vertical transfers is the dominant factor. Termite colony members frequently exchange gut fluid and the microbes it contains through a process called trophallaxis, which provides a stable route of vertical transfer from parent to offspring colonies^33–35. This specific mode of inheritance, coupled with the oxygen sensitivity and the specialization of termite gut bacteria to the gut environment, possibly makes termite gut bacteria unable to migrate outside their host, explaining the strong cophylogenetic patterns with their hosts. Following this scenario, many CAZymes forming TSCs were acquired together with the bacteria encoding them in their genomes and have remained exclusively associated with termite guts since then.

Fig. 3. Maximum-likelihood phylogenetic trees of three of the 131 termite-specific bacterial clusters (TSCs) containing at least one sequence of Cryptocercus and/or Mastotermes.

The three trees showed strong cophylogenetic signals with termites. Phylogenetic trees of (A) GH3 Cluster 4 composed of 97.3% of Spirochaetota, (B) CE1 Cluster 2 composed of 86.2% of Spirochaetota, and (C) GH29 Cluster 5 composed of 97.7% of Bacteroidota. D Maximum-likelihood phylogenetic tree of termites inferred from UCEs. *Cryptocercus kyebangensis; **Mastotermes darwiniensis. Black dots indicate CAZyme sequences assigned to a different bacterial phylum.

Some termite-specific clusters are unique to Termitidae, suggesting their history of association is 30 to 60 million years old

While our results indicate many bacterial CAZymes forming TSCs have been inherited from early termite ancestors, some, such as the 175 exclusives to Termitidae, may have been acquired more recently by the ancestors of Termitidae, some 30–60 million years ago. Two mechanisms of acquisition of these CAZymes are the acquisition of new gut bacterial symbionts together with the CAZyme repertoire encoded in their genomes and horizontal transfers from bacteria not associated with termite guts. We found evidence of the former mechanism in 26 TSCs restricted to Termitidae and including upward of 90% of CAZymes assigned to Chitinispirillaceae (phylum Fibrobacterota) (Fig. 4A, B), a family of bacteria recorded in no other termites than Termitidae²⁰. Similarly, 23 TSCs restricted to Termitidae were comprised of upward of 90% of CAZymes annotated as Candidatus Fibromonas (phylum Fibrobacterota) (Fig. 4C, D), a bacterial genus abundant in the gut of many Termitidae and rarely found in other termites²⁸, suggesting these CAZymes were encoded in the genome of Candidatus Fibromonas as it transitioned to become a termite gut symbiont. In contrast, five TSCs restricted to Termitidae are suggestive of the latter mechanism, as they included more than 90% of CAZymes annotated as Spirochaetota, most of which from the Breznakiellaceae (Fig. 4E, F), a bacterial family present across the gut of most termites²⁰. Future studies are needed to determine whether the Breznakiellaceae populating the gut of the ancestor of Termitidae acquired these CAZymes by horizontal transfer from bacteria not associated with termite guts.

Fig. 4. Maximum-likelihood phylogenetic trees of six of the 175 termite-specific bacterial clusters (TSCs) strictly associated with Termitidae.

The six trees showed strong cophylogenetic signals with termites. Phylogenetic trees of (A) GH77 Cluster 10 composed of 99.4% of Fibrobacterota, primarily of the family Chitinispirillaceae, (B) GH57 Cluster 10 composed only of Fibrobacterota, primarily of the family Chitinispirillaceae, (C) GH5_2 Cluster 9 composed only of Fibrobacterota of the genus Candidatus Fibromonas, (D) GH26 Cluster 7 composed only of Fibrobacterota, primarily of the genus Candidatus Fibromonas, (E) GH4 Cluster 4 composed of 94.9% of Spirochaetota, and (F) GH57 Cluster 2 only composed of Spirochaetota. G Maximum-likelihood phylogenetic tree of Termitidae inferred from UCEs. Black dots indicate CAZyme sequences assigned to a different bacterial phylum.

Conclusion

Our results show that a large fraction of the CAZymes encoded by the termite gut microbiota are only found in termites and have been associated with this niche at geological timescales. Some were likely encoded in the genomes of the first gut symbiotic bacteria of the common ancestor of termites and Cryptocercus and passed down to modern termites by means of vertical transfers. This includes many CAZymes involved in lignocellulose degradation, indicating that the cocktail of CAZymes allowing termites to digest wood is largely encoded in the genomes of vertically inherited bacteria, with limited contribution from bacteria living outside their guts. The uniqueness of termite gut bacterial CAZymes raises the possibility that the exceptional efficiency of termites at digesting wood is partly linked to the intrinsic characteristics of their CAZymes.

Materials and methods

Data collection and metagenome analyses

We used the gut metagenome assemblies of 195 termites and one Cryptocercus previously published by refs. ^20,28 (Supplementary Data 1). All contigs composing these assemblies and longer than 1000 base pairs were taxonomically annotated using DIAMOND BLASTx searches⁴⁵ with e-value 1e-24 against the GTDB database Release 95⁵⁰. The open reading frames coding for CAZymes were identified among these metagenome contigs using Hidden Markov model searches against the dbCAN2 database⁵¹. Fragments of CAZyme sequences shorter than 50% of the expected CAZyme length were not considered. We only considered hits with e-value lower than e-30 and coverage upward of 0.35. For CAZymes composed of several modules, we separated the domains corresponding to specific CAZyme families. CAZyme sequences from termite gut metagenomes were also searched against the GTDB database Release 207⁴⁶ using Nucleotide-Nucleotide BLAST v2.10.0+⁵² with default settings to obtain sequences not associated with termites. We retained a single copy of every sequence obtained from the GTDB database and not associated with termites using seqkit tool v2.0.0⁵³. These CAZymes sequences not associated with termites were analyzed together with CAZyme sequences derived from termite gut metagenomes.

Reconstruction of CAZyme phylogenetic trees

We reconstructed one phylogenetic tree for each CAZyme family comprising more than 20 sequences derived from termite gut bacteria. For the large families divided into subfamilies, we reconstructed one phylogenetic tree for each CAZyme subfamily composed of more than 20 sequences derived from termite gut bacteria. 12 families were divided into subfamilies: GH43 into 25 subfamilies; GH13 into 21 subfamilies; GH5 into 20 subfamilies; GH30 into seven subfamilies; PL8 and PL12 into three subfamilies; and AA3, PL1, PL6, PL9, PL10, and PL11 into two subfamilies. In total, we reconstructed 201 phylogenetic trees. We used sequences derived from termite gut metagenomes and sequences from the GTDB database not associated with termites. Nucleotide sequences were translated into amino acid sequences using the codon Supplementary Table 11 (bacterial and archaeal code) with Geneious prime v2022.2.0. Protein sequences of each CAZyme gene family were aligned using MAFFT v7.490 with the parameters “--auto setting”, which are recommended for aligning many sequences^54,55. Protein alignments were converted into nucleotide alignments using pal2nal v14.1-3⁵⁶ with the codon table 11 (bacterial and archaeal code). Maximum-likelihood phylogenetic tree reconstructions were carried out on nucleotide alignments using Fasttree v2.1.11-2⁵⁷ with the settings “-gtr -gama”. The phylogenetic trees of every CAZyme family were rooted using 20 sequences of related CAZyme families included in the analyses as outgroups and chosen based on information available on www.cazy.org¹⁴. The outgroup sequences were non-termite sequences obtained from the CAZyme database v11 available at bcb.unl.edu/dbCAN2/download/Databases/v11/. We verified that outgroup sequences cluster together in the phylogenetic analyses, allowing us to identify the root of the tree unambiguously.

Termite phylogenetic trees

We used the termite phylogenetic trees reconstructed by ref. ²⁸ with UCEs. The phylogenetic tree was reconstructed with 322 of the 50,616 termite-specific UCE loci⁴¹. An average of 186.8 UCE loci were found per termite gut metagenome, thence the completeness of the matrix was ~57%. These UCE loci matched, at least partly, singly-annotated exons from the draft genome of Zootermopsis nevadensis⁵⁸. The maximum-likelihood phylogenetic tree was reconstructed using IQ-TREE v1.6.12 with a GTR+G+I model of nucleotide substitution and 1000 ultrafast bootstrap replicates (UFB) to assess branch supports^59,60, as described in ref. ²⁸. The phylogenetic tree, reconstructed with 43% of missing data, was congruent with previous phylogenetic trees built using UCE and transcriptome data, which contained smaller proportions of missing data^40,41.

Identification of termite-specific CAZyme clusters

We searched the phylogenetic trees of all CAZyme families for clusters including sequences exclusively derived from the gut metagenomes of termites and Cryptocercus. We only considered clusters containing sequences from more than 20 termite and Cryptocercus samples. We refer to these clusters as termite-specific clusters (hereafter: TSCs). Clusters containing sequences from fewer than 20 samples were not considered in downstream analyses. To estimate the relative contribution of TSCs to termite wood digestion, we calculated the relative abundance of each TSC by mapping the trimmed sequencing reads onto the CAZyme sequences. The procedure was performed separately on sequences from TSCs and sequences that did not belong to any TSC. Reads were aligned using BWA-MEM v0.7.10⁶¹ and the resulting alignments were sorted (“sort”) and fixed (“fixmate”) with SAMtools v1.9⁶². The number of reads mapping to each set of CAZymes was extracted using the SAMtools “flagstat” command. We used these values to estimate the proportion of CAZymes belonging to TSCs for each gut metagenome analyzed in this study.

Statistics and reproducibility

We carried out cophylogenetic analyses between termites and all TSCs using three different approaches. The first approach was the Procrustean Approach to Cophylogeny implemented in the R package PACo⁴². For this approach, termite and TSC trees were converted into distance matrices using the cophenetic() function of the vegan R package⁶³. We ran the software using the backtrack method of randomization to conserve the overall degree of interactions between termite and TSC trees⁶⁴. The second approach was the generalized Robinson Foulds (RF) metric⁴³ implemented in the ClusteringInfoDistance() function of the TreeDist R package⁴³. The third approach was the method of Nye et al.⁴⁴ implemented in the NyeSimilarity() function of the TreeDist R package⁴³. For this approach, the termite and TSC trees were matched to find an optimal 1-to-1 map between branches. For the last two methods, implemented in the TreeDist R package, each termite tip was split into x tips of zero branch length, where x is the number of CAZyme sequences associated with the metagenome corresponding to that termite tip^65,66. Congruence between the termite and TSC trees was determined using 1000 random permutations.

Author contributions

A.B. and T.Bou. conceptualized the experiments and approach. T.Ber., J.A., and S.H. performed the bioinformatics analyses. T.Bou. wrote the paper. T.Ber., J.A., J.R.A., A.B., G.T., J.S., and S.H. read, commented, and accepted the final version of this manuscript.

Peer review

Peer review information

Communications Biology thanks Joao Paulo L. Franco-Cairo and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Sabina La Rosa and David Favero.

Data availability

Raw sequence data used in this study were previously published and are available in two MGRAST projects (https://www.mg-rast.org/mgmain.html?mgpage=project&project=mgp101108 and https://www.mg-rast.org/mgmain.html?mgpage=metazen2&project=mgp84199) (see Table S1 for individual IDs). The UCE sequences were previously published and are available from the Dryad Digital Repository: 10.5061/dryad.tmpg4f53w.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-024-07146-w.