A molecular survey of Australian and North American termite genera indicates that vertical inheritance is the primary force shaping termite gut microbiomes

Termite gut microbiota have been the subject of an increasing number of investigations over the past years using a suite of new molecular tools [7]; however, a large amount of termite diversity remains to be explored. Here, we present the first extensive culture-independent molecular survey of the gut microbiomes of Australian termites and expand our existing knowledge of North American termite gut microbial diversity. These data are then used to assess the relative effect of vertical inheritance and environmental factors (primarily diet). The 16 termite genera examined in this study have a set of core and accessory gut bacterial phyla that distinguish them from all other habitats (Table 2). This observation is consistent with previous culture-independent studies which show that the combination of these phyla is highly distinctive of the termite gut microbiome [7,23-25] particularly in comparison to other insect gut communities [26,27]. This distinctiveness is further underlined by the observation that the majority of operational taxonomic units (OTUs) identified in the present study cluster with sequences from previous termite surveys [28-38]. A recent extensive rRNA-based survey of gut bacteria in 34 termite species [24] allows direct comparison of the bacterial profiles of seven termite genera that overlap between the studies. The three higher termite profiles generally match well, but the four lower termite profiles have some conspicuous differences even at the relatively course phylogenetic resolution of phylum. In particular, the Dietrich et al. [24] profiles have higher proportions of Spirochaetes and lower proportions of Bacteroidetes and Firmicutes than the corresponding profiles in our study (Additional file 16: Figure S12). For Reticulitermes and Coptotermes, this may reflect real differences as different species were profiled; but for Mastotermes and Incisitermes for which the same species were examined, the more likely explanation is differences arising from methodology such as DNA extraction method [39] and/or PCR primers used [40]. A study by Sabree and Moran [41] using similar DNA extraction method and primers to ours produced a similar gut community profile for Mastotermes (Additional file 16: Figure S12).

With these methodological caveats in mind, key differences between higher and lower termite gut profiles are linked to the presence of protists in the latter group. The Bacteroidetes and Elusimicrobia are the most over-represented phyla in lower termites relative to higher termites because they harbor highly abundant members of the Candidatus genera Azobacteroides, Vestibaculum (Bacteroidetes), and Endomicrobium (Elusimicrobia; Figure 2), which are recognized protist symbionts [9,42,43]. Candidatus Azobacteroides pseudotrichonymphae, an endosymbiont of the parabasalid Pseudotrichonympha grasii, has previously been reported to comprise approximately 70% of the bacterial cells present in the gut of Coptotermes formosanus [44]. Here, we found phylogenetically distinct Candidatus Azobacteroides spp. comprise up to 66% of the bacterial gut profiles in Coptotermes, up to 63% in Schedorhinotermes, and up to 72% in Heterotermes (Additional file 10: Figure S8) and identified their putative Pseudotrichonympha hosts only in those termite genera (Figure 1 and Additional file 13: Figure S10), supporting the previously reported specific relationship between the two in multiple termite genera [9]. Candidatus Vestibaculum illigatum was first reported in Neotermes cubanus and was shown to be an epibiont of the flagellated protist Staurojoenina [45]. Here, we found abundant populations (8%–22%) of Candidatus Vestibaculum in Incisitermes and Marginitermes, both members of the family Kalotermitidae. Assuming that Candidatus Vestibaculum is a specific epibiont of Staurojoenina, this is consistent with the observation that Staurojoenina is only found in members of the family Kalotermitidae [46,47]. However, the other Kalotermitidae genus surveyed, Glyptotermes, lacked detectable populations of Candidatus Vestibaculum (Additional file 10: Figure S8) and Staurojoenina was not detected at all in our survey. The latter observation may be due to our primer set not targeting this parabasalid genus (two mismatches in the 926F primer to S. assimilis acc. AB183882).

Candidatus Endomicrobium was detected in all of the lower termite genera surveyed and was also found in low abundance in some of the higher termite genera (Figure 1) consistent with previous findings [48-50]. In Reticulitermes and Incisitermes, Candidatus Endomicrobium is a recognized cytoplasmic symbiont of the parabasalids Trichonympha and Metadevescovina, respectively [50]. Our data are consistent with these observations as high abundance populations of Candidatus Endomicrobium, and their respective host protists were detected in Reticulitermes and Incisitermes (Figure 1). The highest relative abundance of Candidatus Endomicrobium was found in Porotermes (65% of bacterial reads; Figure 2 and Additional file 10: Figure S8); however, no protist host sequences were detected, presumably due to primer mismatches as visual observation of Porotermes gut contents reveal a high diversity of protist morphotypes (unpublished observation).

The most prevalent (ubiquitous) genus in the gut survey was Treponema (Figure 2), which comprises most of the Spirochaetes phylum signal. Treponema has been reported in every termite gut investigation to date [7] and contributes substantially to the distinctiveness of the termite gut microbiome. Numerous Treponema OTUs were found in the present survey, many of which flourished in the higher termites (Additional file 10: Figure S8) likely following the evolutionary loss of protists from the hindgut [51]. It has been shown that Spirochaetes are essential for the survival of higher termites and that their removal results in a shorter life span [52]. Metagenomic, metatranscriptomic, and metaproteomic analyses of two higher termite genera, Nasutitermes and Amitermes, indicate that Treponemas are involved in all of the major functions in the hindgut, including fiber hydrolysis, fermentation, homoacetogenesis, and nitrogen fixation [23,53] which may explain their success (ubiquity) and long term co-habitation with their termite hosts. However, Treponema is a phylogenetically broad genus [54-56] and it seems likely that not all species will be capable of all key functions.

Two Deltaproteobacteria (Desulfovibrio and Desulfarculus-like OTUs) were present in low abundance in over half of the samples tested (Figure 2). Desulfovibrio has previously been reported as a widespread constituent of termite guts mainly based on cultivation studies, with proposed functions including oxygen removal and nitrogen fixation [57]. However, the Desulfarculus-like OTU was more prevalent (Additional file 10: Figure S8) and a member of this group has recently been inferred to be primarily responsible for the first step in CO₂-reductive acetogenesis [58]. If this key functionality in the Desulfarculus-like group is conserved across different termite genera, it may explain their widespread distribution among the surveyed termites. Less expected was the widespread occurrence of a Pirellula-like planctomycete OTU (Figure 2). Planctomycetes have been reported in alkaline gut segments of soil-feeding termite genera, where they are speculated to play a role in degradation of humus-associated biopolymers such as N-acetylglucosamine [32]. No soil-feeding genera were surveyed in the present study, although the planctomycete OTU may be associated with alkaline segments known to be present in several higher termite genera [59]. The planctomycete OTU was also detected in three lower termite genera (Additional file 10: Figure S8) which are not known to have alkaline gut segments, suggesting that planctomycetes are not strictly associated with higher pH in termites.

Archaea have been reported to constitute only a small fraction (up to 3%) of the termite gut ecosystem [60]; however, we found much higher percentages in the amplicon profiles of a number of lower termite genera and one higher termite genus (Table 1). We cross-checked our findings with an alternative forward primer broadly targeting bacteria and archaea (803F) and confirmed that the result was not an artifact of the universal primer pair (926F and 1392R). Also considering that many samples had archaeal proportions in the anticipated range (<3%; Table 1), we suggest that the higher values are not artifacts of the primers or of the DNA extraction method used. The observed variability in archaeal abundance between samples belonging to the same termite genus, e.g. Schedorhinotermes (1.4%, 24.5%, and 32.9%), suggests that archaeal abundance may be more variable between specimens than previously appreciated, possibly reflecting environmental factors or simply temporal dynamics (‘archaeal blooms’). Only hydrogenotrophic methanogens, dominated by Methanobrevibacter in most cases, were detected in the surveyed termite guts consistent with previous reports [61-63], suggesting that acetoclastic methanogenesis is likely unfavorable in this habitat. Phylotypes closely related to a recently described phylogenetically novel methanogenic genus related to the Thermoplasmatales, Methanomassiliicoccus, were detected in several termite genera raising the possibility that these methanogenic populations may have an obligate requirement for methanol [64].

Eukaryotes were not the primary focus of this study, and our data are likely an underestimate of protist diversity in the surveyed species due to primer mismatches [65,66]. Also, rRNA-based relative abundance estimates will likely not reflect cell counts (e.g. Reticulitermes [67]) due to the much higher number of rRNA gene copies in protists relative to bacteria [22], the former of which is not currently corrected by CopyRighter [17]. However, some interesting qualitative observations were made including putative protist host-bacterial symbiont pairings described above. It is commonly reported that higher termites lack flagellated protists, which are primary agents of lignocellulose digestion in lower termites [7,68,69]. Instead, bacteria and to a lesser extent, the termite itself, provide the enzymes necessary for lignocellulose hydrolysis in higher termites [7,70]. Unexpectedly then, a phylotype related to the ciliate Clevelandella, previously reported in wood-feeding cockroach intestinal tracts [71], was detected in the higher termite genus Gnathamitermes (Additional file 13: Figure S10). An older microscopic study of higher termite gut ecosystems supports our findings with the identification of small numbers of a closely related ciliate, Nyctotherus, in Amitermes [72], although no protists were detected in the Amitermes community profiles in the present study. The presence of low abundance protist populations in some higher termite genera suggests either reacquisition after the major evolutionary transition to bacteria-dominated gut communities in the higher termites or low-level persistence of some protist species across this transition. It will be interesting to determine if these ciliates are directly involved in lignocellulose digestion.

A primary motivation of our study was to determine the relative importance of vertical inheritance (host signal) versus diet on termite gut microbiota composition given the unusual status of termites among insects in terms of gut microbiome complexity [6] and the importance of termites as ecosystem engineers [23]. This question is not immediately addressable using field observations of lower termites as they are primarily wood feeders with the exception of Mastotermes [73]. However, we obtained sufficient specimens of polyphagous higher termite genera to evaluate the relative effect of diet and host signal. The strongest signal was clearly due to vertical inheritance, with termite genus and even family level associations being resolved based on gut community profiles alone, particularly in unweighted analyses (Figure 4). This is consistent with previous studies indicating that vertical transmission plays an important role in structuring termite gut communities, for example co-speciation of gut symbionts within the genera Reticulitermes and Microcerotermes [28] and a general host signal in whole gut community analyses of 34 termite and cockroach species [24]. Maintenance of host-specific microbial communities must be achieved via vertical transmission during trophallaxis or coprophagy, as there is no germline transfer in termites [74]. It is important to note that a dominant host signal in gut community composition does not imply that all component species are the product of vertical inheritance, ultimately resulting in co-speciation. The termite gut is an open system that would allow ingress of foreign microorganisms, which may be able to persist under favorable conditions. For example, it was speculated that some Firmicute populations in Amitermes have been laterally acquired from herbivore gut communities as a result of dietary specialization, i.e. dung feeding [23] (see below). These bacterial populations were then subsequently vertically transmitted in the Amitermes lineage. While fine-scale reconstruction of population co-evolution is not feasible with partial rRNA sequences, the clusters of identical reads identified in the most ubiquitous 97% OTU, Treponema OTU1, reflects the dominant overall host signal but also suggests that a minority of strains in the cluster may have been laterally transferred between termite genera (Additional file 11: Table S3).

The effect of diet on gut community structure has been addressed to a lesser extent in termites. No clear dietary signal was observed in unweighted analyses (Figure 4), but when the evenness (relative abundance) of gut phylotypes was taken into account, a secondary effect of diet on community structure became apparent in the well-sampled polyphagous termite genera. Specifically, Nasutitermes samples are partitioned into wood- and grass-feeding clades and Gnathamitermes into grass- and dung-feeding clades (Figure 5). Changes in phylum-level abundances could be correlated with the dietary differences such as an increased abundance of Spirochaetes and Fibrobacteres and decreased abundance of Firmicutes in wood-feeding relative to grass-feeding Nasutitermes (Additional file 9: Figure S7). This is consistent with previous reports of the importance of Spirochaetes and Fibrobacteres in the digestion of wood fibers [53,75]. He et al. [23] identified phylum-level shifts between dung-feeding Amitermes and wood-feeding Nasutitermes. Based on metagenomic and metatranscriptomic analyses, they explained these differences by inferring that Firmicutes play a greater role in hemicellulose hydrolysis and utilization of fixed-nitrogen compounds required for dung digestion and Spirochaetes play a greater role in cellulose hydrolysis and nitrogen fixation required for wood digestion. However, our data suggest that phylum-level differences attributed to diet were overestimated in the He et al. study [23] because of marked differences between the Amitermes and Nasutitermes gut communities due to vertical inheritance. We estimate that changes in the relative abundance of these phyla between wood- and grass-feeding Nasutitermes samples is only 4%–8%, as opposed to the 15%–34% differences seen between dung-feeding Amitermes and wood-feeding Nasutitermes (Additional file 9: Figure S7). Presumably in some instances, changes in relative abundances of gut populations occurred over evolutionary time scales in response to dedicated dietary specialization [76]. However, recent feeding trial studies of Reticulitermes flavipes indicate that such changes in population evenness can occur on short time scales allowing polyphagous termite species to adjust rapidly to changes in their diet due to seasonal variation or availability of foraged plant species [77,78].