Species-specific signatures of the microbiome from Camponotus and Colobopsis ants across developmental stages

Symbiotic relationships between hosts and bacteria are common in nature, and these may be responsible for the evolutionary success of various groups of animals. Among ants, these associations have been well studied in some genera of the Camponotini, but several questions remain regarding the generality of the previous findings across all the members of this ant tribe and if bacterial communities change across development in these hosts. This study is the first to characterize the bacterial community associated with a colony of the recently recognized genus Colobopsis and three colonies of Camponotus (two distinct species) and show how different the composition of the bacterial community is when compared across the different genera. Our data reveal that Colobopsis (species: Co. riehlii) and Camponotus (species: Ca. floridanus and Ca. planatus) have distinct microbiota, and we were able to verify that the identity of the species contributes more to the bacterial diversity. We also demonstrated that there were no significant differences between colonies of the same species (Camponotus planatus), and between stages of development from different colonies. We did find that some developmental stages have distinct bacteria, confirming that each stage of development could have a specific microbiota. Our results show species are one of the factors that shape the bacterial community in these Camponotini ants. Additional studies of the intra-colonial microbiome of other hosts and across development may reveal additional clues about the function and importance of bacteria in colony recognition, individual and colony health, and nutritional upgrading.


Introduction
Symbiotic interactions are thought to be one of the factors responsible for the ecological success of many groups of animals and plants [1][2][3][4]. Symbiotic microbes can influence the host through the manipulation of the host's reproduction or provide direct benefits to the host through nutrition, defense, or even environmental tolerance [5][6][7][8]. Social animals often interact intimately with other members of their group and offspring through grooming and microbiota? 4) Are there differences between the stages of development within the same colony? And finally 5) Does the same stage of development from different colonies/species have similar bacteria? Leveraging next generation amplicon sequencing, we address these questions and document the diversity of bacteria to help identify the factors that structure the bacterial communities found across a diverse and widely distributed group of animals.

Materials and methods
Sample collection and determination of the different stages of development Three Camponotus colonies, representing two species (Ca. floridanus and Ca. planatus), and one Colobopsis riehlii colony were obtained from hollow twigs of trees and brought to the lab. The samples were immediately preserved in 95% ethanol and stored at -20˚C before DNA extraction. In order to determine the castes / stages of development, we selected the eggs, larva with variation of size (L1 = small larvae, below 2 mm; and L2 = large larvae-last larval instar, approx. 2-4 mm), pupae classified according to the pigmentation of the eye and body (P1, P2 and P3-from the white eye to total pigmentation, respectively), small and large workers (W1 and W2 to represent adult worker size polymorphism, with W1 as minor and W2 as major workers), queens and males [43,44]. Within each entire colony, the quantity of each caste/subcaste/stage was determined (see Table 1). The taxonomic identification were determined using taxonomic keys for Camponotus and Colobopsis species in the southeastern US (available in: http://mississippientomologicalmuseum.org.msstate.edu//Researchtaxapages/ Formicidaepages/Identification.Keys.htm#.WE7qIH31-3H-from Creighton 1950, Snelling 1988; Mark Deyrup, pers. comm., William MacKay-Camponotus website) and vouchers were deposited in the collection of the USP Zoology Museum in São Paulo, Brazil.

DNA extraction and bacterial DNA sequencing
Total DNA was extracted from 85 samples (three specimens of each caste and colony, when available) of entire individuals with the Qiagen DNeasy Tissue kit following the manufacturer's recommendations with filtered pipette tips and sterile techniques were applied to avoid contamination following Moreau [45]. Additionally, the samples were sterilized on the surface as described in Moreau [45]. Although we did not use the modification of the Qiagen DNeasy kit for Gram-positive bacteria, we did follow the extraction method recommended by Rubin et al. [46]. This method is able to detect Gram positive bacteria in large quantities, but this could still influence the diversity of bacteria we are able to detect. We amplified the V4 region of 16S rRNA using primers described in Caporaso et al. [47], following the Earth Microbiome Project (EMP) protocol (515f primer and 806r; http://www.earthmicrobiome.org/emp-standard-protocols/16s/

Bacterial quantification
We measured the amount of bacterial DNA present in all samples with quantitative PCR (qPCR) of the bacterial 16S rRNA gene using 515f (5'-GTGCCAGCMGCCGCGGTAA)and 806r (5'-GGACTACHVGGGTWTCTAAT) universal bacterial primers of the EMP (http://www. earthmicrobiome.org/emp-standard-protocols/16s/) to check Illumina sequencing efficiency. All qPCRs were performed on a CFX Connect Real-Time System (Bio-Rad, Hercules, USA) using SsoAdvanced 2X SYBR green supermix (Bio-Rad) and 2 μL of DNA. Standard curves were created from serial dilutions of linearized plasmid containing inserts of the E. coli 16S rRNA gene and melt curves were used to confirm the absence of qPCR primer dimers. The protocol for standardization following the recommendations of Rubin et al. [46]. All samples were analyzed in triplicates including a blank. The results were averaged before calculating the number of bacterial 16S rRNA gene copies per microliter of DNA solution (S4 Table).

Bioinformatic analysis
The sequences were analyzed in QIIME 1.9.1 [48]. We merged the forward and reverse sequences through SeqPrep. Demultiplexing was completed and QIIME defaults were used for quality filtering of raw Illumina data. We implemented the pick_open_reference_otus.py command using the SILVA 128 reference database with 97% identity [49,50] to call OTUs, and UCLUST to create the OTU table. Singletons were discarded. Chimera checking was performed with UCLUST [51] and PyNAST (v1.2.2) was used for sequence alignment [52].
To test whether the composition of the bacterial community is more related to the colony/ species itself, or whether it is more related to the different stages of development, we separated our analyses into two categories: different colonies/species (MOR#59, MOR#62, MOR#69, and MOR#73) and developmental stage (available in colonies/species sampled, see Table 1).
The summarize_taxa_through_plots.py command was used to create a folder containing taxonomy summary files. In order to standardize, all samples that obtained less than 400 bacterial sequences after quality filtering were excluded from the subsequent analysis. We started our analysis with 73 samples (triplicate of each caste when available, which obtained good DNA quality), and after filtering to a sequencing depth of 400, 63 samples passed this cutoff and were included in downstream analyses. Ten samples were excluded because they did not pass the cutoff of 400 sequences (see S6 Table), and they are identified with a yellow star in Fig 1. All analyzes started from the bacterial OTU data. We implemented an analysis of multidimensional nonmetric scaling (NMDS) and related statistics in the PAST3 software package [53] to illustrate the relationship between ecological communities [54,55]. Sorensen (Dice coefficient) and Bray-Curtis, similarity and dissimilarity indices, respectively [54] were used to test the variation and the structure of the bacterial communities, respectively. The samples were grouped according to the host colony/species and developmental stage. Analyses of similarity (ANOSIM) with Bonferroni correction was used to determine statistical significance [54,55]. The SIMPER analysis was conducted to verify the contribution of each OTU for grouping between colonies/species and different developmental stage [55].
The G test of independence (P, FDR_P and Bonferroni_P) was carried out to determine whether OTU presence/absence is associated with different colonies/species or different developmental stages. We also used Analysis of Similarity (ANOSIM), Adonis [56] to determine sample grouping and a redundancy analysis (RDA) to test the relationships between samples.
Observed species richness, Shannon diversity, the Chao1 nonparametric richness estimator, whole-tree phylogenetic diversity, Simpson, and equitability metrics were calculated to compare alpha diversities based on a two-sample t-test using non-parametric (Monte Carlo) methods to test differences in OTU richness among categories. Unweighted and weighted UniFrac distance matrices [57], which use phylogenetic information to calculate community similarity, were used to calculated beta diversity. These beta diversity metrics were used to compare community level differences between two categories: colonies/species and developmental stage. Jaccard dissimilarity metrics were calculated and the average was compared. A matrix of community pairwise distances was generated by UniFrac and used to cluster samples by principal coordinates analysis (PCoA).
Cytoscape v3.2.1 [58] edge-weighted spring embedded algorithm was used to visualize networks of bacterial community [59]. Connections were drawn between samples representing the shared significant OTUs. A heatmap was constructed with all OTUs that had 100 reads using heatmap.2 and the vegan package [60] in R [61]. The dendrogram was created with Bray-Curtis dissimilarity hierarchical clustering of bacterial communities in hclust.

Phylogenetic tree reconstruction
To investigate the possible relatedness of some of our unassigned bacterial OTUs (representative sequences from the dominant OTUs), we downloaded from GenBank the closest Blast hits for our selected sequences and other strains of Blochmannia available from different Camponotini genera. We were able to include Blochmannia from all Camponotini genera except Dinomyrmex and Overbeckia. Sequences were assembled and edited using Bioedit Sequence Alignment Editor [62] and aligned with the Clustal W tool [63]. We implemented a maximum likelihood analysis using PhyML 3.0 [64] on the CIPRES web portal [65]. The GTR+I+G model of sequence evolution was implemented. Branch lengths and bootstrap support are reported. To facilitate visualization, the clade of Wolbachia (brown) and Blochmannia (green) were colored in FigTree [66].

Results
In this study we used the open reference option to assign OTUs. Open reference is an important tool for studies that want to take into account the diversity of bacteria that many not be present in the database [42]. Since these results may return OTUs with no close species level assignments contained in the database, the open reference command assigns these OTUs to the previous taxonomic level (i.e. Genus, Family, etc.) where there is confidence in the assignment. So in our results some OTUs are only confidentially assigned to higher taxonomic levels like "Burkholderiales; Other", while others are assigned to the species level like Candidatus Blochmannia. From the four colonies analyzed we obtained 1,322 observed OTUs from a total of 152,500 reads from 73 samples from one colony of Colobopsis riehlii and three colonies of Camponotus from two species (one colony of Ca. floridanus and two colonies of Ca. planatus), which permitted analyses comparing different colonies/species and developmental stages. To visualize the diversity of OTUs found per sample we used the summarize_taxa_through_plots. py command (Fig 1). For Colobopsis, 19 samples across the stages of development were analyzed, resulting in 134 OTUs from a total of 16,591 reads, ranging from 206 to 3008 reads per samples. For analysis of the colonies of Ca. floridanus and Ca. planatus were recovered 1,188 OTUs resulting from 135,909 reads ranging from 10 to 13,989 reads, with the latter value belonging to one from the queens analyzed.
By analyzing the bacteria found across different stages of development within a colony (developmental stage), we recovered bacteria associated with only a specific stage of development such as the Pasteurellales, which is present in the second larval stage of Colobopsis riehlii, and Wolbachia present only in the queens and males of Camponotus planatus. Our data also reveal that the larval stage exhibits much greater microbial diversity than the other stages of development (S3 Fig).

Patterns that influence the bacterial community
We performed statistical tests (weighted and unweighted, depth 400, 63 samples included) to examine potential patterns that influence the bacterial community of these Camponotini samples, and for this we analyzed the following two variables: differences between colonies/species and developmental stage. A list of the 10 samples that did not reach the depth of 400 reads and were excluded from the analysis are included in S6 Table. From these we found different colonies/species (Table 2), can influence the bacterial community of these Camponotini samples, although for the developmental stage we did not obtain significant results for the weighted distance.
Through analyses of beta diversity (matrices UniFrac weighted distance) we observed bacterial communities among all Camponotini samples. PCoA analysis showed that the bacterial community becomes more distinct when comparing the different species than when comparing the stages of development across all species (Fig 2). The average Jaccard dissimilarity metric was 0.90 for Camponotus colonies (one of Ca. floridanus and two of Ca. planatus), which suggests only few of the bacterial community members are shared among all individuals of different developmental stages of Camponotus, but for Colobopsis riehlii Jaccard dissimilarity of 0.65 was inferred, which suggests more of bacteria was shared among the colony.
No significant changes in the composition of the bacterial community (Sorensen index) were observed between the colonies/species and among developmental stage (S1 Table). However, when we analyzed the bacterial community structure (Bray-Curtis index, stress 0.051, for different colonies, and 0.051 for different developmental stage), we found significant results such as difference between samples MOR#73 (Ca. planatus) vs. MOR#62 (Co. riehlii) and MOR#69 (Ca. planatus) vs. MOR#62 (Co. riehlii) (Fig 3, S1 Table). For these analyses we did not recover significant differences between developmental stages.
The SIMPER between-groups analysis revealed that the OTUs recovered in the comparisons between the different colonies/species, are essentially the same OTUs responsible for structuring these bacterial communities within significance groups (S2 Table). But we also observe that there are multiple strains of Candidatus Blochmannia, Enterobacteriaceae and Wolbachia present across these samples.
To examine the complicated associations between samples with shared significant OTUs, we used Cytoscape to construct a network graph in which each node represented a host sample. Network analyzes were performed using the spring-embedded edge-weighted algorithm (Fig 4), which approaches the samples according to the number of OTUs shared, and we colored the edges according to the different colonies/species (Fig 4A), and in the different stages of development ( Fig 4B). OTUs with less than 100 reads were hidden for easy viewing. From this analysis we observe greater structuring between species than across different stages of development.

Bacterial communities of Camponotus (Ca. floridanus and Ca. planatus) and Colobopsis riehlii are different
Our statistical results confirm that the bacterial community of Camponotus (Ca. floridanus and Ca. planatus) and Colobopsis riehlii are different (Table 3). This can clearly be seen in Figs 2 and 3. This result shows that even in closely related genera, the microbial communities are different, at least for the colonies/species analyzed in this study.

Camponotus planatus from distinct colonies have similar bacterial communities
Of all the colonies analyzed in the present study, the two Camponotus planatus colonies (MOR#69 and MOR#73), have the highest similarity, as observed from the statistical tests that resulted significant differences (Table 4), but a small difference if we compare with the other colonies. This result corroborates S1 Table.   There are microbiota differences in the stage of development between host species Statistical analyzes show that there are significant differences in the development stage across two of the species, Camponotus planatus (MOR#69 and MOR#73) and Colobopsis riehlii (MOR#62). This pattern could also be true for Camponotus floridanus, but unfortunately after rarefaction only a few individuals from this colony (MOR#59) could be included and therefore we were not able to conduct beta diversity analyses on this species. As the main bacteria across all of these colonies are Blochmannia and Enterobacteriaceae (for Camponotus: Ca. floridanus and Ca. planatus, and Colobopsis riehlii respectively; S2 Table), the abundance of OTU (weighted) may not be appropriate to test for significant differences across the developmental stages within each colony. Therefore the results of unweighted distances were presented on Table 5, and there are significant differences in the development stage across two of the species.

The same stage of development in different Camponotus species have similar bacteria
To address this question we binned our Camponotus samples (Ca. floridanus and Ca. planatus) into the following groups: larva (L1 and L2), pupae (P1, P2 and P3), workers (W1 and W2) and finally a mixed group with queens, males and eggs (all directly derived from the queen).
The results show that there were no significant differences when we analyzed each of these groups, (Table 6), which reveals that there is similarity in each of these stages of development, even when they were grouped from different colonies (See S1 Fig).

Bacteria responsible for differences between colonies/species and development stages
Through the results of the G test (P, FDR_P and Bonferroni_P), we found bacterial community presence/absence is significantly different across developmental stage and colonies/species (S3 Table). Between colonies/species more OTUs were significantly different across samples than the other developmental stage category (different stages of development within a colony). However, the bacteria Enterobacteriaceae (multiple strains, including Candidatus Blochmannia), Wolbachia (multiple strains) and Pasteurellales were present across all categories (S3 Table). Separate G-test analyses of the different developmental stages within each Colobopsis  Table). According to our results of measures of Alpha diversity (Chao1, PD whole tree, observed OTUs, Simpson and Shannon), we found that the samples of Camponotus floridanus, Ca. planatus and Colobopsis riehlii are not very diverse, showing few different OTUs. Likely due to this low diversity, we did not obtain significant results when comparing alpha diversities based on a two-sample t-test using non-parametric (Monte Carlo) methods. Through the rarefaction curve analysis of OTUs observed, our sequencing coverage of the bacterial communities appears satisfactory for most samples. However for some samples, it was not possible to reach a plateau (S2 Fig).
For easy viewing on our HeatMap, we are presenting only OTUs with more than 100 reads. We grouped the samples according to the quantity and type of associated bacteria. Our results revealed that there are some OTUs specific to a particular colony, such as OTU AJ245591.  5). In less quantity the strain New.ReferenceOTU71-Wolbachia (difficult visualization in Fig 5) was found in both colony MOR#69 and colony MOR#62, at different stages of development. Phylogenetic Tree: Blochmannia and Enterobacteriaceae OTUs are related The inferred maximum likelihood phylogeny received high boostrap support across the major nodes placing our samples with their closest relatives. All the sequences of Blochmannia are grouped in a single clade with high bootstrap support (99%). In addition, the OTUs identified as Enterobacteriaceae in the present study are closely related to Blochmannia, corroborating the hypothesis that all Enterobacteriaceae are actually Blochmannia (Fig 6).

Discussion
In Camponotini ants the presence of bacteria such as Blochmannia, considered a primary endosymbiont, and Wolbachia, as secondary, is already well known [21,22,37,67], but the diversity of the entire bacterial community has not been fully documented and differences across developmental stage have not been adequately explored. Although our study included a modest number of colonies (85 individual samples from four colonies) our results are the first to characterize the bacterial community associated with a colony of the recently recognized genus Colobopsis (species: Colobopsis riehlii) and three colonies of Camponotus (two distinct species: Ca. floridanus and Ca. planatus) and show how different the composition of the bacterial community is when compared across the different colonies/species (different genera and different species-collected in the same location), and how they are conserved when comparing across the different stages of development within a colony. In general, our data reveal that Colobopsis riehlii and Camponotus (Ca. floridanus and Ca. planatus) have distinct microbiota, although they are closely related ant genera. The OTUs from these two host genera are distinct. We were able to verify that the identity of the species contributes more to the bacterial diversity than the stage of development. A significant difference between species is likely due to the different bacterial communities between Camponotus and Colobopsis ant species. We also demonstrated that there were no significant differences between colonies of the same species (Camponotus planatus), and between stages of development from different colonies, confirming that each stage of development may have a specific  Note that there were no significant differences when we binned the same stage of development.  [32] in elevating Colobopsis as a separate genus from Camponotus.
In this study, bacterial community structure and composition in ants of the same colony were most similar to each other, a pattern recovered in other ant species [13,40,46,68]. This is likely because social insects live in densely populated colonies with highly related individuals [69], which may facilitate the sharing of the microbiota. In addition, it is often observed that Camponotini ants exhibits mouth-to-mouth (stomodeal) trophallaxis, i.e. the sharing of liquid nutrients through mutual feeding [70,71]. Nutrients stored in the crop or 'social stomach' are shared throughout the colony during trophallaxis [72], which is thought to be a primary means for horizontal bacterial transfer within a colony [26,[73][74][75]. This intense interaction and exchange of microbiota may reinforce colony-specific signatures [40,76], and also appears to occur with Camponotini ants.
Blochmannia, a member of the Enterobacteriaceae, is known to provide important functions in Camponotini ants, which includes, Camponotus, Colobopsis, Polyrhachis and others, whose phylogenetic trees of symbionts are congruent with those of their hosts across long periods of evolutionary time, indicating the coevolution of host and symbiont [77,78]. In addition to its nutritional role [23], especially in early life [79], it has also maintained genes needed to contribute to the metabolism of nitrogen, sulfur and lipids [80][81][82].
The high mutational rate of Blochmannia [83] may influence and disrupt the identification of OTUs at the bacterial genus level for the short sequences generated by most amplicon methods, therefore Enterobacteriaceae or Other Enterobacteriaceae-may in fact be Blochmannia. For Camponotus (Ca. floridanus and Ca. planatus) we detected high Candidatus Blochmannia abundance. We also expected this bacterium in high abundance for Colobopsis riehlii but our results did not reveal this at first. Our phylogenetic analysis of the main OTUs found in our study grouped in the same clade as Blochmannia and Enterobacteriaceae with high statistical support. All the individuals of Camponotus (Ca. floridanus and Ca. planatus) and Colobopsis riehlii analyzed in the present study have some type of Enterobacteriaceae as the main bacterium in their microbiota and based on our phylogenetic analysis is likely Blochmannia. Our study also found 44 samples of Camponotus planatus, from two colonies (MOR#69 and MOR#73), with two strains of Blochmannia (double infection). This result corroborates Ramalho et al. [67] finding of double infections of Blochmannia in Camponotus textor Forel, an exclusively Neotropical species.
In fact the Blochmannia (strain / OTU) of Colobopsis is different from the (strain / OTU) of Camponotus, at least for the fragment of 16S rRNA that we sequenced. All these strains belong to the genus Blochmannia, corroborating what has already been found in other studies that this bacterium is established in the tribe [21,22], but our data show that there is still a great difference / diversity of these strains of Blochmannia found in the samples of the present study (Co. riehlii, Ca. floridanus and Ca. planatus). This great diversity found within Blochmannia genus (16S rRNA gene) can be explained by its high evolutionary rate, which made possible the differences found in the present study. Functionally these distinct strains of Blochmannia are similar, supported by the genome [24] and the phylogenetics presented in this study. But the Blochmannia strain found in the samples of Colobopis and Camponotus of the present study are distinct, which suggests their microbiome to be distinct.
Wolbachia, a major invertebrate endosymbiont [14,[84][85][86] famous for manipulating the reproduction of the host [87], was the second most common endosymbiont found in all Colobopsis riehlii samples, occurring in all individuals of the colony, across all stages of development. There are many studies associating this bacterium with Formicidae, but its function remains unclear. Interestingly OTU GAUE02014372.1.1238 of Wolbachia was found only in the reproductive castes (queen and males) in the colony MOR#69 of Camponotus planatus but was not found in the other stages of development across the colony. Another OTU of Wolbachia (New.ReferenceOTU71) although in a lower concentration was found in Colobopsis riehlii (MOR#62) and in Camponotus planatus (MOR#69). This low infection rate by Wolbachia (1%) has also been found in another North American Camponotus [14], although is not true across the genus as Ca. textor was found to be highly associated with Wolbachia [67].
The next most common bacterium associated was Pasteurellales-EU348326.1.1455 found specifically in the larval stage (L2) in Colobopsis riehlii (2.20%). Pasteurellales are one of the major orders within the class Gammaproteobacteria [88,89]. This bacterium is often present in the respiratory, alimentary and reproductive tract of various birds and mammals, including humans [89,90]. This group of bacteria has been identified from another arboreal ant, Pseudomyrmex ferrugineus [91], but their function in ants is not clear.
Previous studies have reported the presence of other symbionts in Camponotus, including Spiroplasma which has been reported in other species of Camponotus [14], but was not found in our results. Acetobacteraceae was also recently found in Camponotus, and is believed to be strongly host specific [37]. We also recovered this bacterium (multiple strains) in 11 individuals of Camponotus (20.37%), but with few copies per individual, ranging from 1-4 reads.
Studies that try to understand the patterns that explain the association of microbiota and host inform us about the potential impacts and roles of these symbioses. In the present study we show that the Colobopsis riehlii microbiota is distinct from Ca. floridanus and Ca. planatus, a closely related genus. In general, the microbiota presented here appears as a species-specific signature, whereas most developmental stages do not have distinct microbiota. Although we present some differences across development, especially in the larval stage, the intense social interaction between individuals of a colony likely homogenizes the microbiota among colony members. Additional studies of the intra-colonial microbiome of other hosts and across development may reveal additional clues about the function and importance of bacteria in colony recognition, individual and colony health, and nutritional upgrading.