Specimens of C. gestroi (Wasmanm) were collected from field colonies with traps of corrugated cardboard and maintained in the Termite Laboratory of the Biology Department, UNESP, Rio Claro, São Paulo, Brazil (22° 23′S, 47°31′W). Termites were kept at 25 ± 2°C and fed with pine wood chips with 10% of humidity.

Prior RNA extraction, termites were washed in saline solution (1% NaCl) to remove possible microorganism occurring in its exoskeleton. After, total RNA (10 μg) was extracted from the whole bodies of 50 workers and 50 soldiers using TRIZol reagent (Invitrogen) and purified using RNeasy Plant Mini Kit (Qiagen) under manufacture instructions. The quality of RNA was verified using RNAnano chip Bioanalyzer 2100 (Agilent). The cDNA from workers and soldiers were synthesized using Oligo-dT for mRNA enrichment (kit Superscript III RT™–Invitrogen) and sequenced using high throughput sequencing platform, under manufacture instructions, (GS FLX Titanium/Roche) generating single-end reads.

The pyrograms from workers and soldiers were processed using the sff_extract program (MIRA Package) to convert the sff file into a fasta file and remove low-quality or adaptor sequence ends. The cdhit-454 program was used to eliminate identical and nearly identical (>98% identical) duplicates reads (Niu et al., 2010). Finally, the BDtrimmer program (Baudet and Dias, 2006) was used to perform the sequence trimming (poly A/T, adaptor and low-quality regions). The trimmed non-ribosomal reads larger than 100 bp were assembled into contigs and singlets using the MIRA EST sequence assembler version 3.0.3 (Chevreux et al., 1999) with the default parameters for 454 data.

Sequence data of 454 reads from soldier and worker libraries was submitted to SRA/NCBI under the accession number SRR1774237 and SRR1774239, respectively. This Transcriptome Shotgun Assembly project was deposited at DDBJ/EMBL/GenBank under the accession GCET00000000. The version described in this paper is the first version, GCET01000000.

The EST unigenes were BLASTed against the protein sequence database (NCBI/NR) using an e-value threshold of 1e⁻⁵ and classified into four taxonomic groups (insect, fungi, bacteria, and protist). The taxonomic classification was performed by comparing homologous organisms (identified in the first hit of the BLASTx/NR output for each unisequence) with a list of all organisms described in the insect, fungi, and bacteria groups extracted from the NCBI/taxonomy database. The protist unigenes were defined as containing one homologous organism in the taxonomic group of Parabasalia (taxid 5719) that was represented by more than 121,414 entries.

For the discovery of genes related to CAZy and PAD enzymes, HMM-based methodology was applied to identify these genes among the EST unigenes from both castes of C. gestroi. The sequences related to the carbohydrate-active enzymes were downloaded from the CAZy database. Since this database has been organized into proteins families, these CAZy protein families were aligned using clustalW (Larkin et al., 2007) with the default settings, and the HMM models were calculated and calibrated using hmmbuild and hmmcalibrate from the HMMER package (Finn et al., 2011). For PAD enzymes identification, the HMM models related to a specific Pfam of interest were downloaded from Pfam database and used to searches. The list of Pfams used to categorize the classes of PAD genes was previously described in the literature (Tartar et al., 2009; Sethi et al., 2013b). Thus, the identifications of unigenes, correlated with CAZy and PAD enzymes, were performed by comparing the total EST unigenes and HMM models using hmm-search with an e-value cut-off of 1e⁻⁵, and a configuration appropriate for use with parameters of local searches. Additionally, the EST unigenes identified as CAZy and PAD enzymes were compared to the Pfam database using the RPS-BLAST program with an e-value threshold of 1e⁻⁵ to confirm our methodology. The dbCAN or CAT databases (Park et al., 2010; Yin et al., 2012) could be used for the identification of CAZymes, however both lack the PAD sequences. After the identification and annotation of CAZy and PAD unigenes, a read counting for each Pfam domain was performed and also classified based on their taxonomy origins in both castes.

The proteins from GH5, GH7, GH10, and GH11 families that contain sequence similarities with non-insect organisms were submitted for phylogenetic analysis. Firstly, these protein sequences were aligned against non-protein redundant (NR) database from NCBI using BLASTp program and top 5 protein hits for every query were downloaded. For each protein family, a FASTA file containing non-insect proteins and all blast hits were submitted for global alignments among amino acid sequences using MUSCLE (Edgar, 2004). The selection of amino acid substitution models was done using Akaike criteria implemented in maximum likelihood analysis on MEGA version 6 (Tamura et al., 2013). The phylogeny was reconstructed using Maximum Likelihood analysis implemented on RAxML (Stamatakis, 2014) with branch support estimated by 1000 bootstraps.

The protein was extracted from the whole bodies of 50 workers and 50 soldiers as previously described (Franco Cairo et al., 2011). The protein extract (75 μg) from each caste was loaded into a 12% SDS-PAGE gel and bands at 19, 26, 34, 50, and 90 kDa and above 90 kDa were excised, reduced (5 mM dithiothreitol, 25 min at 56°C), alkylated (14 mM iodoacetamide, 30 min at room temperature in the dark), and digested with trypsin (Promega). The samples were dried in a vacuum concentrator and reconstituted in 50 μL of 0.1% formic acid to extract the peptides from the gel. The supernatant was transferred to new tubes and 4.5 μL of the resulting peptide mixture was analyzed on an ETD-enabled LTQ Velos Orbitrap mass spectrometer (Thermo Fisher Scientific) coupled with LC-MS/MS by an EASY-nLC system (Proxeon Biosystems) through a Proxeon nanoelectrospray ion source. The peptides were separated by a 2–90% acetonitrile gradient in 0.1% formic acid using a PicoFrit Column analytical column (20 cm × ID75 μm, 5 μm particle size, New objective) at a flow rate of 300 nL/min over 27 min. The nanoelectrospray voltage was set to 2.5 kV, and the source temperature was 200°C. All instrument methods for the LTQ Velos Orbitrap were set up in the data-dependent acquisition mode. The full scan MS spectra (m/z 300–1600) were acquired in the Orbitrap analyzer after accumulation to a target value of 1e⁶. The resolution in the Orbitrap was set to r = 60,000, and the 20 most intense peptide ions with charge states ≥2 were sequentially isolated to a target value of 5000 and fragmented in the linear ion trap by low-energy Collision-Induced Dissociation–CID (normalized collision energy of 35%). The signal threshold for triggering an MS/MS event was set to 1000 counts. Dynamic exclusion was enabled with an exclusion size list of 500, exclusion duration of 60 s, and repeat count of 1. An activation q of 0.25 and an activation time of 10 ms were used.

The spectra were acquired using the software MassLynx v.4.1 (Waters - Milford, MA, USA), and the raw data files were converted to a peak list format (mgf) without summing the scans using the Mascot Distiller v.2.3.2.0 software (Matrix Science Ltd.). These spectra were searched against the C. gestroi database (181,554 unigenes; 42,520,001 residues—generated by the unigenes identified in the metatranscriptomic analysis described above) using the Mascot v.2.3.01 engine (Matrix Science Ltd.) with carbamidomethylation as the fixed modification, oxidation of methionine as a variable modification, one trypsin missed cleavage and a tolerance of 10 ppm for precursor ions and 1 Da for fragment ions.

All datasets processed using the workflow feature in the Mascot software were further analyzed in the software ScaffoldQ+ to validate the MS/MS-based peptide and protein identifications. Peptide identifications were accepted if they could be established at greater than 60.0% probability as specified by the Peptide Prophet algorithm (Keller et al., 2002). Peptide identifications were also required to exceed specific database search engine thresholds. Mascot identifications required at least both the associated identity scores and ion scores to be greater than 31. Protein identifications were accepted if they could be established at greater than 80.0% probability for peptide identification. Protein probabilities were assigned using the Protein Prophet algorithm (Nesvizhskii and Aebersold, 2004). Proteins that contained similar peptides and could not be differentiated based on the MS/MS analysis alone were grouped to satisfy the principles of parsimony. The scoring parameter (Peptide Probability) in the ScaffoldQ+ software was set to obtain a false discovery rate (FDR) of less than 2%. Using the number of total spectra output from the ScaffoldQ+ software, we identified the differentially expressed proteins using spectral counting. A normalization criterion, the “quantitative value,” was applied to normalize the spectral counts. All mass spectrometric raw file associated with this study is available for download via FTP from the PeptideAtlas data repository by accessing the following link: http://www.peptideatlas.org/PASS/00574.

Biochemical assays using the whole worker and soldier crude extracts were performed to evaluate the ability of this extract to hydrolyze natural polysaccharides and synthetics oligosaccharides. Total protein extractions for biochemical characterization was prepared from 100 whole bodies of worker and soldier. They were homogenized using Harvest-Potter with 2 ml of 50 mM sodium acetate buffer, pH 5.5. After extraction, the mixture was centrifuged at 20.100 × g for 20 min at 4°C, followed by addition of 1 μl of Cocktail Protease Inhibitor (Amresco) per ml of crude extract produced. The supernatant was collected and hereafter referred such as crude enzyme extract. For all assays, the protein concentration used was of 1 μg/μl and the concentration was determined for Bradford method (Bradford, 1976). All procedures were performed on ice. The assays were performed as previously described by Franco Cairo et al. (2011) with slight modifications. Enzymatic reactions consisted of 10 μl of crude enzyme extract incubated with 40 μl of 50 mM sodium acetate buffer pH 5.5 and 50 μl of 0.5% specific substrate (in water), in triplicate, at 37°C, for 60 min. Enzymatic assays were stopped after the addition of 100 μl of dinitrosalicylic acid–DNS (Miller, 1959) and heated at 99°C for 5 min. The measurement of color change was performed at 540 nm using a micro plate reader. The enzymatic activity assays results were expressed in mM of glucose equivalents produced. The enzymatic assays with p-nitrophenyl-carbohydrates (pNP) were performed as follows: 10 μl of crude extracts were incubated with 50 μl of 5 mM pNP substrates and 40 μl of 50 mM sodium acetate buffer pH 5.5. Assays were stopped after addition of 100 μl of 1 M Sodium Carbonate_.(Na₂Co₃). The measurement of color change was performed at 400 nm using a micro plate reader. The enzymatic assays were done in triplicates and results were expressed in terms mM of p-nitrophenyl released. Glucose and p-nitrophenyl were used for standard curve construction. Substrates were purchased from Megazyme and Sigma Aldrich: CMC (carboxymethyl cellulose); β-glucan from barley, lichenan from moss, laminarin from Laminaria digitata, xyloglucan from tamarind, xylan from oat spelt, rye arabinoxylan, mannan, pectin from Citrus; 4-nitrophenyl β-D-glucopyranoside (pNP-G); 4-nitrophenyl β-D-cellobioside (pNP-C); 4-nitrophenyl β-D-xylopyranoside (pNP-X); 4-nitrophenyl β-D-galactopyranoside (pNP-Gal); 4-nitrophenyl α-L-arabinofuranoside (pNP-A); 4-nitrophenyl β-D-mannopyranoside (pNP-M); 4-nitrophenyl-α-L-fucopyranoside (pNP-F).

Two cDNA libraries using oligo-dT were constructed from whole bodies of specimens of worker and soldier castes of C. gestroi without replicates aiming to perform only gene discovery of CAZy and PAD enzymes. The libraries were sequenced using the GS FLX Titanium system from Roche, producing approximately 800,000 single-ends reads for each library. All sequence data is summarized in Figure 1A. After trimming, high-quality expressed sequence tags (ESTs) were obtained from worker and soldier libraries. The MIRA EST sequence assembler configured with the default parameters for 454 data was used with success to perform the transcriptome assembly. The majority of contigs were shared by both castes (Figure 1A and Table S1). Moreover, the workflow applied in sequencing data analysis was also described (Figure S1).

The unigene annotation using BLASTx against NCBI/NR was performed and could assign protein hits for only 28.8% of the unigenes (Table S1). Based on homology searches using BLASTx against NR, the unigenes could be classified into insect, protist, bacteria, and fungi taxonomic groups (Figure 2A). As expected, most of the proteins were of insect group, but there was a contribution of protists and fungi organisms, as it was expected due to symbiotic organisms that inhabit the termite gut. Although, poly-A enrichment step was used, unigenes classified into bacteria group were also identified (around 3%). In addition, other taxonomic groups were identified, including Branchiostoma, Hydra, Strongylocentrotus, Ixodes, and some other genera (data no shown).

Although, a high number of unigenes were identified based on comparisons with other insect transcriptome projects available at the Gene Index Project website (Quackenbush et al., 2001; approximately 25,000 and 55,000 unique sequences from Apis mellifera and Drosophila melanogaster were identified, respectively), more detailed analyses of the sequence data were carried out. We further evaluated the number of contigs with BLASTx/NR hits as a function of the contig coverage and analyzed the ratios of the insect and symbiont hits found in these contigs. The distribution of reads/contigs revealed a high number of contigs that were assembled based only on a few reads, e.g., 48% of contigs were assembled based only on two reads (Figure 2B). In the same figure, the red line shows that the number of unigenes of symbiotic origin that showed similarity to known proteins (BLASTx/NR) increased as a function of the contig coverage, and only 15% of the unigenes assembled based on two reads resulted in hits to known proteins.

Another interesting result was observed when the percentage of insect and symbiotic protein hits was plotted as a function of the contig coverage (Figure 2C). For instance, almost 27% of the contigs assembled and based on two reads were from symbiotic organisms, and this number decreased as a function of the contig coverage. These results suggested that contigs assembled based on a few reads (2, 3, or 4 reads) mostly represent unknown proteins or non-coding RNAs that were probably highly expressed in symbiotic organisms, which were likely represented at low cell densities.

The metatranscriptomic analysis of workers and soldiers from the lower termite C. gestroi has revealed a total of 778 unigenes containing Pfam domains assigned to CAZy and PAD enzymes. The results revealed unigenes from 32 different glycoside hydrolases families (GH), 8 Auxiliary Activities enzymes subfamilies (AA), 8 carbohydrate esterase families (CE), 2 polysaccharide lyase families (PL), and 5 classes of PAD.

The glycoside hydrolases and PADs were the most represented unigenes in both castes (Figures 3A,B). The majority of these unigenes were homologs to insect sequences, however, there were a counterpart of sequences with homology to symbionts (Figure 3C), mainly from protists. Figures 3D,E show the distribution of reads throughout the CAZy families and classes of PAD for worker and soldier, respectively. Table 1 describes in detail the distribution of CAZy and PADs enzymes identified by our analysis, along with their respective Pfams and taxonomic origin, such as insect, protist and bacteria. Although, the analysis was based on sequences derived from poly-A (for eukaryotic organisms), the bacterial sequences discovered in this analysis were maintained in the results due to their representative information for the main objective of this work, which was the discovery of CAZy and PAD genes in C. gestroi castes. No fungi sequences related to CAZy and PAD enzymes were found.

Among the GHs, the predominant family in the worker caste was GH9, followed by GH13, GH1, GH18, GH39, GH25, GH16, and GH7. In soldier caste, the most abundant family was GH18, followed by GH9, GH1, GH13, GH7, and GH25. From these abundant families, GH9, GH39, GH16, and GH7 harbor typical members displaying enzymatic activity related to carbohydrates from plant cell wall. Regarding the unigenes related to cellulose degradation, endo-glucanases (GH5, GH9, and GH45), and beta-glucosidases (GH1) sequences were reported to insect, protist and bacteria taxonomic groups. Exo-glucanases (GH7) were reported only of protist origin.

The GH families found in both worker and soldier libraries involved with hemicellulose degradation were GH3, GH8, GH10, GH11, GH16, GH26, GH29, and GH31. Classical xylanases (GH10 and GH11) were classified only with protist origin. Beta-glucosidases/xylosidases (GH3), lichenases (GH8), and endo-mannanases (GH26) were classified with bacteria origin. Laminarases (GH16) and alpha-fucosidases (GH29) were only of insect origin and alpha-glucosidase/xylosidase (GH31) was assigned to insect and protist taxonomic origin. However, for instance, a termite enzyme classified as GH16 was previous described as Gram-negative bacteria-bind protein (DGNBP) in C. formosanus (Hussain et al., 2013).

Regarding the auxiliary activities enzymes, members of the families/subfamilies AA1_3, AA3_2, AA3_3, AA4, AA5_1, AA5_2, AA6, and AA8 were found in worker and soldier libraries. AA3_2 was the most abundant subfamily in both workers and soldiers, followed by AA3_3 and AA1_3. The sequences of AA1_3, AA3_3, and AA4 were only of insect origin, while AA3_2 sequences were both of insect and protist origin. The AA3 family contains GMC_ oxidoreductase domain (PF00732), which catalyzes the oxidation/reduction of glucose/alcohol/pyranose and generation of hydrogen peroxide as product. The literature has suggested this reactive oxygen specie as a co-factor for lignin peroxidases and Fenton reaction applied for lignocellulose breakdown (Levasseur et al., 2013). On the other hand, the ecdysone oxidases from insect Bombyx mori play a role in cuticle formation rather than lignocellulose degradation and this enzyme contains GMC domain (Sun et al., 2012).

The AA1_3 subfamily contains a Cu2_oxidase pfam domain (PF00394), which is also found in multi-copper laccases. AA5_1 and AA5_2 family members were also found from both castes, but sequences of protist origin were only identified for AA5_1 subfamily and for AA5_2 both insect and protist sequences were detected. The sequences from AA3_2, AA3_3, AA4, AA5_1, and AA5_2 are part of a family that contains glucose, alcohol, vanillyl-alcohol, glyoxal, and aldehyde oxidases also involved in hydrogen peroxide generation and lignin oxidation (Levasseur et al., 2013). Interestingly, AA6 and AA8 reads were identified of worker and soldier only with insect origin. These AAs feature protein domain are known to be involved in the generation of Fenton components and iron reduction, respectively (Arantes et al., 2012; Levasseur et al., 2013). Sequences related to the AA2 and AA3_1 families, which contains lignin peroxidases and cellobiose dehydrogenases respectively as well as for the families AA9, AA10, AA11, and AA13 classified as Lytic Polysaccharide Monoxygenases-LPMOs (Levasseur et al., 2013) were not identified in this study.

The predominant CE family found in our analysis was CE10, followed by CE4 and CE1. The identified CE10 and CE4 reads were only of insect origin, whilst CE1 reads were of both insect and protist origins. The unigenes from CE3 and CE6 families were only of protist origin. CE10 family is reported to contain only genes/proteins that are not involved with carbohydrate modifications. Wheeler et al. (2010) suggested that proteins with COesterase domain (PF00135), for instance members of CE10, could act as feruloyl esterases in termites. Otherwise, the families CE1, CE3, CE4, and CE6 are described as acetyl xylan esterases, directly involved in hemicelluloses modifications. Another class of CAZymes found in our analysis was PL (polysaccharide lyases), but only two families of this class were identified in this study, the PL11, a rhamnogalacturonan lyase, and PL1, a pectate lyase (Cantarel et al., 2009). The most abundant family was PL11, which had only sequences of insect origin.

Among the PAD components, the predominant class identified was p450, followed by aldo-keto reductase (AKR), glutathione transferase (GST), superoxide dismutase (SOD), and catalase (CAT). The reads classified as p450 members were predicted to be of insect and protist origin. The AKR reads in both castes also assigned with unigenes from insect and protist. In the case of the GST class, the results were similar to those for AKR. For the SOD and CAT classes, the results showed that both classes had sequences of insect and protist origins.

Considering that whole termite bodies were used for RNA extraction, it is important to mention that not all enzymes should be considered as components of the termite digestome. In order to investigate potential involvement in the termite digestome, further analysis to identify secretion signals was performed on all the unigenes classified as CAZy and PAD. As a result, our bioinformatic analysis suggested that the majority of unigenes from CAZy families and all PAD classes identified by metatranscriptomics display unigenes predicted with secretion signals (Table 1, Tables S2, S3).

Secretion signals were identified in glycoside hydrolase unigenes related to cellulose degradation (such as GH1, GH9, GH16), mannan degradation (such as GH2) and AA members (families AA1_3, AA3_2, AA3_3, AA5_2, and AA8) related to cellulose and lignin oxidation. Unigenes from CE families, which are directly involved in biomass degradation, such as CE1, CE4, and CE13, as well as PL11, were predicted with secretion signals (Table 1). As expected, all unigenes from protists indicated as cellulolytic enzymes, such as those from families GH5, GH7, and GH45 and the hemicellulolytic enzymes from GH10 and GH11, did not present secretion signals (Table 1, Table S3), since the degradation of lignocelluloses by protists has been reported to occur in phagocytic vesicles in the cytoplasm of these microorganisms.

To get further insights into the diversity and taxonomic origin of symbiotic CAZymes, phylogenetic analyses were performed with enzymes classified as members of GH5, GH7, GH10, and GH11 families (Figures S2, S3). The sequences coding endo-glucanases from family GH5 were grouped into three different clades (Figure S2A). The first clade is represented by protist organisms, grouping uncultured protists from different termite species and the protist Spirotrichonympha leidyi, suggesting the occurrence of this protist specie in C. gestroi guts. Bacterial organisms from Firmicutes and Bacteriodetes phylum represent the second and third clades, respectivelly. Collectively, these results revealed the substantial diversity of GH5 endo-cellulases in C. gestroi.

For the exo-glucanases from GH7 family, the phylogenetic tree for GH7 enzymes generated two distinct clades highly supported by bootstrapping values bigger than 90% (Figure S2B). Protist organisms represent the first clade that grouped a subset of GH7 sequences identified in this work with GH7 enzymes from Holomastigotoides mirabile and Pseudotrichonympha grassi, two major protist species described for Coptotemes genus. The second clade grouped set of GH7 sequences with fungal organisms.

The xylanases from GH10 (Figure S3A) and GH11 (Figure S3B) families identified in this work were grouped with xylanases from protist origin, separating them from GH10 and GH11 sequences of bacterial origin. Interestingly, protist's proteins from GH10 family clustered in a specific clade only containing sequences from a uncultured protist from C. gestroi (Figure S3A), while the other GH10 sequences clustered in individual clades, belonging to other termite species such as C. formosanus and Neotermes koshunensis. In the case of protist's proteins from family GH11, they grouped together with sequences of xylanases GH11 from the protist Holomastigotoides mirabilis, suggesting the origin of these sequences, as well, as the occurrence of this specie in C. gestroi guts.

Metaproteomic analysis was also performed to elucidate the profile of carbohydrate-active and PAD enzymes in the lower termite C. gestroi. Mass spectrometry-based proteomics was used, which is a consolidated method applied for metaproteomics (Burnum et al., 2011; Franco Cairo et al., 2011). The summary of the metaproteomic results of worker and soldier castes is provided in Figure 1B. In total, 1,420 proteins were identified in the metaproteome of C. gestroi (identified from 3,744 total peptides in both castes), applying the criterion of one unique peptide with a 2.0% False Discovery Rate (FDR) (Table S4). The use of one unique peptide in the metaproteomic analysis is acceptable in literature only if the FDR is below 5.0% (Tanca et al., 2013; Tang et al., 2014). Moreover, the number of identified proteins represented 1.3% of the predicted unigenes in the metatranscriptomic analysis.

Performing the analysis based on the criterion of one unique peptide, 433 peptide matches for CAZy and PAD enzymes were identified. These peptide matches were distributed across 73 different proteins: 28 classified as GH families, 27 as PAD, 6 as AA families, and 12 as CE (Figures 4A,B), moreover, the majority of these peptides were of insect origin (Figure 4C). The distribution of the peptides throughout taxonomic origin as well as the CAZymes families and PAD enzymes are also described in Table 2. These results are in agreement with the metatranscriptomic data, confirming that the GHs and PAD enzymes were highly abundant (based on spectrum counts) in both castes of C. gestroi, as well as, that the majority of the protein matches are of insect origin (Figures 3C, 4C).

The distribution of the peptides throughout the CAZy families and domains of PAD enzymes are described in Figures 4D,E, Table 2. The most abundant glycoside hydrolase family in both castes was GH9 (Figures 4D,E). The second most abundant family was GH7, followed by GH1. Several previous studies have reported that these GH families are directly involved in cellulose breakdown (Cantarel et al., 2009; Franco Cairo et al., 2013; Scharf, 2015). All GH9 and GH1 peptide matches, for both workers and soldiers, were of insect origin, and conversely, GH7 peptide matches were of protist origin. Our results also identified peptide matches for proteins involved in hemicellulose degradation in both castes, such as peptides from proteins of GH2 and GH29 families that were of insect origin, the GH3 and GH26 families that were of bacterial origin and the GH5 and GH10 families of protist origin.

The AAs identified in our metaproteomics data were from families AA3 and AA5 for both castes (Table 2), in which the most abundant in soldier was AA3 and AA5 in workers. All auxiliary activity enzymes identified were of insect origin. No AA1 (laccases) 1and AA4 family members were found in our metaproteomic analysis. Concerning the CE members, our analysis resulted in the identification of 4 families in the workers and soldiers, all of insect origin. CE1 family members were only identified in the soldiers. Likewise, CE4 protein was identified only in the workers. CE10 was the most abundant family identified in both castes.

The PAD enzymes were also assigned in the metaproteomic analysis of C. gestroi. Five different Pfam domains were identified in the workers and soldiers. Based on spectrum counts, GST was the most abundant PAD enzyme in both castes, followed by CAT. SOD and AKR was the third most abundant enzyme in workers and soldier, respectively. Regarding the taxonomy of the PAD enzymes, our results indicated that insect was the major contributor of these components. However, two SODs were of Crustacea and Actinopterygii origin, and one CAT was of Platyhelminthes origin.

Several proteins identified by proteomics contained secretion signals (Table 1), thus confirming our metatranscriptomics findings. Among them, proteins correlated to lignocellulose breakdown were found from insect and bacteria origin, such as GH1, GH2, GH9, GH16, GH26, and GH29. All proteins from protists, assigned as cellulolytic enzymes, did not present secretion signals, as expected. In addition, putative secretion signals were identified in proteins from families AA3 and AA5, as well as from PAD enzymes, such as SOD and AKR. Regarding carbohydrate esterases, one CE10 was identified as containing a secretion signal (Table 2).

The crude protein extracts of C. gestroi workers and soldiers were tested for their ability to breakdown natural polysaccharides and synthetic oligosaccharides. All these assays were performed with biological and technical triplicates (represented in the error bars). The biochemical assays using polysaccharides were performed using the same amount of protein source from both castes. In general, the results showed that the worker crude extract could breakdown all the natural substrates tested, as well as, the worker extract was more efficient than the soldier crude extract. According to Figure 5A, the crude extracts from both castes exhibited the highest activity toward β-glucan, followed by lichenin, rye arabinoxylan, xylan, mannan, xyloglucan, CMC, laminarin, and pectin. For the p-nitrophenyl-derivatives (pNP), the results showed pNP-F as the most hydrolyzed substrate by crude extracts from both castes, followed by pNP-G, pNP-M, pNP-C, and pNP-Gal. The activities against pNP-X and pNP-A were not conclusive (Figure 5B).

In Figure 5C, a scheme showing the deconstructive hydrolytic enzymes interactions in C. gestroi was generated based on the assessments for GHs found in our metatranscriptomic and metaproteomic data. In this scheme only GHs correlated with lignocellulose degradation were considered. The main enzymes from GH families identified in this work, which could be involved in the hydrolysis of glucose-polymers such as CMC, β-glucan, Laminarin, Xyloglucan, and Lichenin, were GH1, GH5, GH7, GH9, GH16, and GH45 families. For xylose polymers degradation such as xylan and arabinoxylan, the main families were GH3, GH8, GH10, and GH11. In the case of the mannose-based polymer, such as mannan, the putative GH families found in our data correlated with the hydrolysis of this substrate were GH2, GH5, and GH26. The activities against pNP-G (substrate for β-glucosidases) corroborated our findings for enzymes from families GH1 and GH3, pNP-C (substrate for cellobiohydrolases) to GH7 family, and pNP-M (β-mannosidases substrate) for GH2 (not exhibited in figure). Collectively, our biochemical assays agreed with our metatranscriptomic and metaproteomic data, thus, validating our overall results.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.