Microbial communities of the ant Formica exsecta and its nest material

In this study, we investigated the bacterial and fungal microbiomes of the ant Formica exsecta (Hymenoptera, Formicidae), and assessed whether the microbial communities inside the ants differ from those in their nest material. Furthermore, we investigated whether the microbial communities inside the ants are conserved across time. To achieve this, we sequenced the bacterial 16S rRNA, and the fungal ITS region in entire adult worker ants and their nest material by Illumina MiSeq. We found that both the bacterial, and the fungal microbiomes form communities discrete from those in the surrounding nest material. In addition to the differences in species composition, we also found that bacterial species diversity, species richness, ζ diversity, and evenness were lower in ants than in the nest material. For fungi, only species richness was lower in the ants than in the nest material. The rate of within‐colony species turnover across sampling events was not statistically significant for bacteria, but highly significant for fungi. This suggests that the fungal communities in the ants are less stable than the bacterial ones. Four bacterial taxa (Alphaproteobacteria, Proteobacteria, Staphylococcus, and Streptococcus), and two fungal taxa (Davidiella and Cryptococcus) formed a core microbiome, being consistently present and more abundant in the ants, but absent in the nest material. In all other cases differences in community composition and structure were due to taxa that were more consistently present and more abundant in the nest material, and frequently absent in the ants. Furthermore, we found 36 unique OTUs identified as Proteobacteria, and 82 unique OTUs identified as Alphaproteobacteria in the ants, representing 2.5% and 5.8% of all bacterial OTUs and 24.6% and 41% of the total number of bacterial sequences. This suggests that F. exsecta harbours a considerable bacterial diversity that so far remains unexplored.


| INTRODUCTION
A microbiome can be described as the microbial community markedly associated with an organism, its environment, or with specific parts of the organism (Berg et al., 2020;Hall et al., 2018), such as the gut or the integument (Brune & Dietrich, 2015;Kucuk, 2020;Mattoso et al., 2012). The microbiome can influence the ecology and evolution of its host, by exchanging genetic material or secondary metabolites with the host, contributing to improved nutrition, or potentially providing protection from pathogens (Brinker, Fontaine, et al., 2019;Brune & Dietrich, 2015;Koch & Schmid-Hempel, 2011;Kwong & Moran, 2016;Shade et al., 2013;Smith et al., 2015;Soucy et al., 2015). The host-associated microbiomes generally differ substantially from those in the immediate external environment (Adair & Douglas, 2017;Douglas, 2015), although microbiota may continuously be exchanged between the two. Organisms may also alter the microbiota in their immediate environment (Boots et al., 2012;Dean et al., 1997;Lucas et al., 2017;Travanty et al., 2022;Voulgari-Kokota et al., 2019), which may evolve into a two-way exchange of microbiota, between the gut of an organism, and its environmental matrix (Billiet et al., 2016;Brune & Dietrich, 2015). For example, burying beetles have been shown to seed the carcasses they use for rearing their larvae with bacterial groups which are part of their core microbiome, transfer carcass bacteria to their carrion nest that envelopes the carcass, and remove some groups of bacteria from the carcass (Duarte et al., 2018). This leads to a dynamic balance, whereby the host may gain fitness advantages by appropriating new microbial symbionts, which are available in the environment (Brune & Dietrich, 2015;Koch & Schmid-Hempel, 2011). The levels of associations between hosts and their microbiomes may vary from tight symbioses to loose associations, shaped by multiple dynamic processes (Shade et al., 2013;Smith et al., 2015), but often with identifiable core communities (Kwong & Moran, 2016;Shade & Handelsman, 2012).
Host-microbiome interactions are under intense study especially in animals and plants, and has expanded from mapping individual microbes, towards functional aspects and processes associated with host phenotype (Adair & Douglas, 2017;Brune & Dietrich, 2015;Douglas, 2015;Hall et al., 2018), and the temporal and spatial variation of microbiomes within host species (Anderson et al., 2012;Stegen et al., 2018;Sullam et al., 2015). In particular, the microbiomes of nest-building organisms may differ from the soil matrix or building material that lines the nest. For example, in the Eurasian reed warblers (Acrocephalus scirpaceus) the bacterial microbiome differed between the nest material and eggs or nestlings (Brandl et al., 2014). In the solitary alkali bees (Nomia melanderi), bacterial diversity was higher in the nest soil, than in the bees, although some transfer of microbial taxa occurred between the nest material and the bees (Kapheim et al., 2021). Sociality may play a key role in the transmission in microbiota across generations in corbiculate bees (Apidae), yet gut microbial communities are dominated by environmentally acquired bacteria in sweat bees (Halictidae) (McFrederick et al., 2014).
Social insects, such as termites, ants, and bees, provide an excellent test case for microbiome structure at both the individual and the colony level (Engel & Moran, 2013). In most species colonies are founded and headed by a single queen, which has mated for life in a single nuptial flight (or a queen and her male mate in the case of termites). As a result, the offspring workers are related (Crozier & Pamilo, 1996), which creates a genetically more homogeneous substrate for cohabiting microbes. Social insects also build and maintain a nest by excavating soil, collecting materials from the environment, and in the case of honey bees by producing the building material themselves (Gösswald, 1989;Hansell, 1984;Hölldobler & Wilson, 1990;von Frisch, 1974). Social insects encompass some of the most distinctive, diverse, and consistent microbial communities, with specialized beneficial functions in nutrition and protection (Mueller et al., 2005). The fungus-gardening ants of the tribe Attini furthermore exemplify multipartite interactions between beneficial and harmful microbes, actively acquiring and maintaining microbes that aid in defences against pathogens (Andersen et al., 2013;Andersen et al., 2015;Barke et al., 2010;Currie, 2001;Kellner et al., 2015;van Borm et al., 2002).

HIGHLIGHTS
• We analysed the bacterial and fungal microbiomes of the ant Formica exsecta • We found that the microbial community of Formica exsecta is consistent across colonies and time • The microbial community of ants was less diverse than that in the nest material • The fungal communities of the ants were temporally more variable than the bacterial ones • The high abundance of unidentified taxa indicated many unexplored taxa Nest-building ants live surrounded by the materials that form the nest. The nest materials of many ant species have their own specific microbiome, which is distinct from that of the surrounding soil environment (Boots et al., 2012;Boots & Clipson, 2013;Brinker, Weig, et al., 2019;Lindström et al., 2019). Wood ants of the genus Formica are a typical species of northern boreal forests, and build large, perennial, above-ground mounds constructed from twigs, conifer needles and other plant litter (Jurgensen et al., 2008). These mounds are regularly inhabited for 10-50 years (Pamilo, 1991;Sundstrom & Vitikainen, 2022). An earlier study on Formica exsecta found that the nest microbiomes differ from the surrounding top-soil environment (Lindström et al., 2019;Lindström et al., 2021), probably caused by manipulation of litter and soil performed by the ants during nest construction and maintenance. Soil perturbation and selective picking of litter particles changes the physicochemical conditions inside the nest mounds, compared to the surrounding top-soil (Dauber et al., 2001;Dignac et al., 2017;Frouz & Jilkov a, 2008;Jílkov a & Frouz, 2014;Lindström et al., 2021). The ants also shape the microbiome via the food they collect, for example, all wood ant species collect sugar-rich substances, thereby altering the pH, the rate of decomposition, and the respiration processes inside the nest (Jílkov a et al., 2012;Rosengren & Sundström, 1987). In subarctic regions, some wood ants also regulate the temperature inside the core part of their nests during the breeding period, keeping it above +20 C despite ambient sub-zero temperatures .
Contrary to the ants inhabiting the nest, the microbes in the sub-arctic bulk soil and the peripheral parts of the nest are strongly influenced by seasonal variations in temperature and humidity. During winter, the temperature in the top soil layers drops well below zero, and only taxa with high tolerance of cold can survive (Margesin & Miteva, 2011;Rankinen et al., 2004). Furthermore, the microbial taxa in boreal soils follow the seasonal cycles of plants. The processes associated with plants govern the function and abundance of many microbial communities in soil (Nacke et al., 2016;Prescott & Grayston, 2013;Zhou et al., 2017), such as mycorrhizal fungi (Sietiö et al., 2018;Timonen et al., 2017), microbial decomposers (Baldrian, 2017) or nitrogen fixers (Faucon et al., 2017). The processes within nests of wood ants apparently alleviate seasonal effects on the microbial communities (Lindström et al., 2021).
The nest environment in our study population of the wood ant Formica exsecta (Nyl.) shares a highly concordant and stable microbiome in nests across several islands spanning an area of ca. 2 km 2 ( Figure 1). Yet, the microbial communities of the nest environment are significantly distinct from the surrounding bulk soil, with representatives for especially Actinobacteria, Proteobacteria, and Ascomycota classified as core indicators for the nest material, in comparison to the surrounding bulk soil (Lindström et al., 2019). In a second study we found that the bacterial communities in the nest material are stable across years, whereas the fungal communities show higher rates of change over time (Lindström et al., 2021). This study makes a crucial contribution to the earlier studies by addressing the composition of the microbiome of the ants themselves, and compared this with their nest material. As we simultaneously analyse both the bacterial and fungal communities, we are able to compare the community structure of the two. The ants, the nests, and the surrounding soils, each with their distinct microbiomes, are potentially in close interaction with each other. This begs the question, whether the microbiomes of the nest material and the ants themselves are concordant, or whether the ants possess a microbiome distinct from that of the nest, manifest as a stable community composition of both bacteria and fungi.
Based on the outcome from our earlier studies (Lindström et al., 2019;Lindström et al., 2021) we hypothesize that the ants inhabiting the nest may have a core microbiome that significantly differs in community composition and/or community structure from that of the nest material, yet may show variation among colonies. The core microbiome of the ants would in this case be characterized by signature microbes, sensu Shade and Handelsman (Shade & Handelsman, 2012) that are absent or rare in the nest material, but are consistently present and significantly more abundant in the ants. Such a core microbiome would also be characterized by an absence of significant variation in community composition across years. Furthermore, based on our earlier observations (Lindström et al., 2021), we hypothesize that the fungal communities may show more variation and higher rates of turnover among taxa than the bacterial ones. To address these questions, we compared the bacterial and fungal communities of ants and their associated nest material. We collected ants repeatedly from the same colonies and analysed the bacterial and fungal taxa that characterize the ants, as well as the variation across sampling events. We also simultaneously collected ants and nest material from the same colony to assess whether the microbial communities of the ants differ from the nest material. To achieve this, we sequenced the bacterial 16S rRNA and the fungal ITS area of both ants and nest material by Illumina MiSeq to identify key taxa present in ants, to compare community composition between ants and their nest material, and to assess among-colony variation in community composition.
The habitats in this study area are typical for the islands of the SW Gulf of Finland, where granite cliffs and dry meadows alternate with patches of coniferous forests dominated by Scots pine (Pinus sylvestris) and Norwegian spruce (Picea abies). The soil type is mainly leptosol of varying depth, depending on the topography. The lower vegetation consists mainly of ericaceous shrubs and grasses belonging to Deschampsia and Festuca, together with several species of moss (Adair & Douglas, 2017;Lindström et al., 2018). The F. exsecta nest mounds in our study system are nevertheless free of live plants, regardless of the surrounding vegetation.
In an earlier study, we reported the optimized methods for this system and streamlined the procedures for handling the material that was collected (Lindström et al., 2018), and these procedures were also used in two previous publications (Lindström et al., 2019; Lindström  , 2021). To analyse the microbial communities within the ants, we collected F. exsecta workers from six colonies between 2011 and 2015 ( Figure 1). Four of these colonies were sampled three times, one was sampled four times, and one twice, giving 18 samples in total, with each colony sampled in different years. Each of the 18 ant samples comprised $15 individual ants, which were pooled into one sample before DNA extraction to secure sufficient amounts of material. To test for differences in the microbial communities between the ants and their associated nest material, we used nest material collected at the same time from the same six colonies as above. The nest samples each comprised $0.2 litres of material taken from a depth of 10-15 cm of the nest dome, as described in Lindström et al. (2018). The nest data were also used in two earlier studies (Lindström et al., 2019;Lindström et al., 2021). All samples were taken with sterile gloves, placed in sterile plastic zip lock bags, transported on ice to Tvärminne zoological station, and stored within an hour of collection at À80 C until further processing. DNA was extracted from both the nest material and the ants within 2 years of collection, and stored at À80 C, until further processing.

| Laboratory procedures and bioinformatics
The ant samples were surface sterilized by rinsing each ant in 70% ethanol for 30 s and then rinsed in diH 2 O 3 Â 1 min, to remove microbes that may originate from the nest material, prior to extraction of DNA. The Power-Soil ® DNA isolation kit (MoBio Laboratories Inc., Carlsbad, CA, U.S) was used to extract DNA from the ant and the nest samples, as per the manufacturer's instructions. However, instead of the method recommended by the manufacturer, a TissueLyser II (Qiagen GmbH, Hilden, Germany) treatment for 3 min at 20000 rpm was used during the cell lysis phase, according to the optimized procedures described in (Lindström et al., 2018).
The extracted DNA was standardized to equimolar amounts, and submitted to Illumina MiSeq sequencing at the DNA Sequencing and Genomics Laboratory, Institute of Biotechnology, University of Helsinki (Helsinki, Finland). Preparation of libraries, sequencing and bioinformatics pipelines were done according to Lindström et al. (2018). In brief, we used the primer pair 27F (AGAGTTTGATC[A/C] TGGCTCAG) (Chung et al., 2004;Weisburg et al., 1991) and pD 0 (GTATTACCGCGGCTGCTG) (Edwards et al., 1989) for the amplification of the bacterial 16S rRNA region. The fungal ITS2 region was amplified with the primers fITS7 (GTGA[A/G]TCATCGAATCTTTG) (Ihrmark et al., 2012) and ITS4 (TCCTCCGCTTATTGATATGC) (White et al., 1990). Negative procedural controls were included at every sequencing instance (three controls each for bacteria and fungi). The controls indicated that 0.1% of the unrarefied bacterial and 0.08% of the fungal sequences may be attributable to contaminations, suggesting that the risk of inflated read abundancies due to contamination was low.
The reads were filtered and OTUs clustered (at 97% identity) using UPARSE v.8.1 (Edgar, 2013;Edgar & Flyvbjerg, 2015). The SILVAv132 (Quast et al., 2013), and the UNITE v7.2 (Kõljalg et al., 2013) databases were used as reference for the alignment of the bacterial and fungal sequences, respectively. For taxonomic classification, we used the RDP16S training set 16 v2.12 (bacteria) (Wang et al., 2007), and RDP ITS Warcup training set v4 (fungi) (Deshpande et al., 2016). A higher taxonomic resolution could have been achieved by using ASV (with a 99%-100% identity) (Callahan et al., 2017;Edgar, 2018), instead of OTU. For this report we, however, decided to use the exact same methodology which we used previously (Lindström et al., 2018;Lindström et al., 2019;Lindström et al., 2021), to retain full congruence with earlier results on the same colonies. The taxa were identified to the level of genus, or higher if the genus level was unavailable, and are collectively referred to as GOH (genus or higher). We discarded all singletons, doubletons and sequences not identified to the level of kingdom from the data prior to further analysis. We used the number of sequences (reads) as proxy for the abundance of taxa.

| Statistical analyses
We generated two separate sets of data files: one with ant material only, 18 samples from 6 colonies, henceforth "Ant_Data", and one with ant samples and corresponding nest samples taken at the same time (6 samples of each), henceforth "Ant&Nest_Data. From the sequenced data, we counted the number of reads, and the number of OTUs separately for the two data sets. Prior to further analysis, all data were rarefied to the lowest number of reads in the samples of the data set in question. In addition, we run a rarefaction analysis on the number of species, separately for bacteria and fungi, and the two data sets. The outcome of rarefying of reads is henceforth referred to as total abundance.
To test for overall differences in the microbial community structure across colonies (Ant_Data), and between samples originating from ants or nest material (Ant&Nest data), we first generated Bray-Curtis matrices, which were visualized in a Principal Co-ordinates Analysis (PCoA) (Legendre & Anderson, 1999). We then used one-way nonparametric permutational multivariate analyses of variance, (Anderson, 2001), with 9999 permutations to test for significant differences in community structure among nests (Ant_Data, n = 6 colonies, in total 18 samples, each with 15 pooled ants), and between sample origins (Ant&Nest_Data, n = 6 colonies, each with one sample with nest material, and one sample comprising 15 pooled ants), respectively. In the first analysis we added sampling occasion and in the second case colony ID as stratum to account for repeated measures. The analyses reported here were done on untransformed data, neither ln-transformation nor square root-transformation resulted in any changes in the outcome. These analyses were carried out in R-4.1.3 (R Core Team, 2017) with the functions specaccum, vegdist, capscale and adonis2, in the package vegan v.2.5-7 (Oksanen et al., 2019).
To assess differences in community structure among colonies (Ant_Data), and between sample origins (Ant&Nest_Data), we calculated Shannon-Wiener diversities, species richness, and Pielou evenness indices. We then tested for differences among colonies (Ant_Data), and between sample origins (Ant&Nest_Data) in oneway repeated measures ANOVAs with colony ID, or sample source (ant or nest) as the main effect, and sampling occasion or nest, respectively, as the repeated random effect. With the Ant_Data, we furthermore estimated the absolute rate of within-colony species turnover per bacterial or fungal taxon following Lande, 1996, and tested for differences among colonies, and between bacterial and fungal reads using One-way repeated measures ANOVAs on ln (x + 1)-transformed counts, with colony ID as the main effect and taxon as the repeated random effect. For both data sets we furthermore calculated the Zeta (ζ) diversity (Hui & McGeoch, 2014) across all combinations of one, two, three, four, five, and six colonies (Ant_Data), and across colonies and sample origins (Ant&Nest_Data). Differences in ζ diversities between ants and nests were tested with nonparametric Wilcoxon signed ranks tests on the individual counts, based on which the mean ζ diversities were estimated, as the requirements for parametric tests were not fulfilled. Parameter calculations for the analyses described above in this paragraph were carried out in MS Excel on taxa with a minimum of 100 reads in total (Ant_Data: 83 bacterial, and 117 fungal taxa; Ant&Nest_Data: 139 bacterial, and 117 fungal taxa). The analyses described in this paragraph were carried out in Statistix 10 (Analytical Software).
Finally, to identify signature species that may define the core microbiome of the ants sensu Shade and Handelsman (Shade & Handelsman, 2012), or may have contributed to potential differences in community composition, we complemented the analyses with One-Way ANOVAs on ln(x + 1)-transformed read counts, with either colony as the main effect and sampling occasion as the stratum (Ant_Data), or sample origin as the main effect, and colony ID as the stratum (Ant&Nest_Data). The corresponding analyses on the Ant_Data were carried out on all taxa with a total read count >400 (36 bacterial and 82 fungal taxa). The analyses on the Ant&Nest_data were carried out on a subset of taxa with a total read count >300 in either ants or nest material (77 bacterial, and 83 fungal taxa, including unidentified Bacteria and Fungi). Where applicable, significance tests were corrected for multiple testing with the Benjamini-Hochberg correction for false discovery rate (FDR) (Benjamini & Hochberg, 1995). The statistical analyses described in this paragraph were carried out in Statistica 10 (TIBCO StatSoft).

| Sequencing results
The sequencing of the 18 ant samples (Ant_Data) with Illumina MiSeq resulted in a total of 864,337 bacterial reads and 1418 bacterial OTUs, and 1,152,740 fungal reads and 704 fungal OTUs. After rarefaction, the read counts were 258,056 for bacteria, and 541,998 for fungi, based on which we identified 255 bacterial, and 187 fungal taxa to the level of genus, or a higher taxonomic level, that is <family, <order, <class, <phylum. The corresponding numbers of reads and taxa for the subset of six ant samples and the corresponding six nest material samples (Ant&Nest_Data) were 615,582 and 811,026 unrarefied bacterial and fungal reads, respectively. After rarefaction the read counts for the ants were 105,633 and 125,430, and for the nest material 105,622 and 123,275, for bacteria and fungi, respectively. Based on these we identified 364 bacterial and 250 fungal taxa. Of these, 208 bacterial and 119 fungal taxa were present in ants, whereas 299 bacterial, and 219 fungal taxa were present in the nest material. Following rarefaction on the reads and the species all samples had reached, or were close to reaching, the asymptote ( Figure A1, A2, A3). Thus, the most prominent taxa are all likely to be adequately represented in the sample.
In the 18 ant samples (Ant_Data) 478 (0.02%) of the bacterial reads, and 11,645 (2.1%) of the fungal reads remained unidentified. In the six samples from ants and nests (Ant&Nest_Data) 143 (0.1%) bacterial, and 2197 (1.8%) fungal reads remained unidentified in the ants, whereas 3348 (3.2%) bacterial and 4935 (4.0%) fungal reads remained unidentified in the nest material. In both bacteria and fungi, the number was higher in the nest material than in the ants (23-fold for bacteria and 2-fold for fungi). However, 49% of the total bacterial read count was assigned as Alphaproteobacteria, and 24% was assigned as Proteobacteria in ants, whereas the corresponding numbers for nests were 3.2% and 2.1%, respectively. These could not be assigned to a lower taxonomic level, but encompassed 36 unique OTUs identified as Proteobacteria, and 82 unique OTUs identified as Alphaproteobacteria (representing 2.5% and 5.8% of all bacterial OTUs, and 24.6% and 41.2% of the total number of bacterial sequences). For comparison, for the fungal phylum Ascomycota 10.6% of the reads in ants, and 23.7% of the reads in nests remained unidentified.

| Comparison between ants and nest material (Ant&Nest_Data)
The bacterial and fungal communities differed significantly between the ants and the nest material (Ant&Nest_Data) (permutational ANOVA: R 2 = 0.57, F 1,10 = 13.5, p = 0.0013, and R 2 = 0.23, F 1,10 = 3.06, p = 0.0026, for bacteria and fungi, respectively; Figure 2). The bacterial communities in the nest material were diverse, with 3 taxa encompassing ca 10% of the total read count, and an additional 29 taxa encompassing more than 1% of the total rarefied read count in the nest material. The bacterial communities of the ants were strongly dominated by Alphaproteobactera, Proteobacteria, and Pseudomonas, whereas the remaining taxa encompassed 1%-6% of the total rarefied read count (Figure 3). The fungal communities showed a similar, but less distinct pattern, with unidentified Ascomycota encompassing 30% of the total rarefied read count, and the remaining 17 taxa encompassed 1%-12% of the total rarefied read count in the nest material. In the ants Oidiodendron, Davidiella, and unidentified Ascomycota dominated with a 12%-20% shares of the total rarefied read count (Figure 4). Species diversity, species richness, and evenness were all significantly lower for bacteria in ants than in nest material, whereas for fungi only species richness was significantly lower in ants, than in the nest material (Table 1). The Zeta diversity (ζ), that is, the number of species present in 1, 2, 3, 4, and 5-6 samples (the last two combined), was significantly higher in the nest material than in the ants, both for bacteria and fungi (Table 2).
Delving deeper into the Ant&Nest_Data, we identified 40 bacterial, and 9 fungal taxa that differed significantly in the number of reads between the ants and the nest material ( Figure 5, Tables A1, A2). Of these, the bacterial taxa encompassed 39.3% and 33.0% of the total rarefied read count in ants and nest material, respectively. The corresponding figures for fungal taxa were 30.3% and 36.5%. The significant differences between the ants and their associated nest material notwithstanding (Figure 5a), only four bacterial taxa (the genera Staphylococcus and Streptococcus, the class Alphaproteobacteria, and the phylum Proteobacteria) showed significantly higher read counts in ants, compared to the nest material. Indeed, the genera Staphylococcus and Streptococcus were not detected in the nest samples. All remaining bacterial taxa showed higher read counts in the nest material than in the ants, and 24 of these taxa were not detected in the ants (Table A1). Of the 12 fungal taxa that showed significant differences in the number of reads between ants and the nest material, two (the genera Cryptococcus and Davidiella) showed higher read counts in ants, and were not detected in the nest material (Figure 5b). Ten taxa showed significantly higher read counts in the nest material, four of which were not detected in the ant samples (Table A2).

| Variation within colonies (Ant_Data)
When considering the entire data set on ants (Ant_Data), neither the bacterial, nor the fungal communities differed significantly among colonies (permutational ANOVA: R 2 = 0.26, F 5,12 = 0.83, p = 0.67, and R 2 = 0.36, F 5,12 = 1.37, p = 0.06, respectively; Figure 6). Both the bacterial and fungal communities in this data set echoed the pattern found in the Ant&Nest data, with Alphaproteobacteria, Proteobacteria, and Pseudomonas dominating the bacterial profile, and Oidiodendron, Pseudogymnascus, and unidentified Ascomycota dominating the fungal profile (Figures 3, 4). Similarly, neither species diversity, evenness, nor richness differed significantly among colonies for either bacteria or fungi. Species turnover across colonies was not statistically significant for bacteria, but highly significant for fungi (Table 3). The overall rates of turnover among colonies did, however, not differ significantly between bacteria and fungi (mean ± SD: 0.44 ± 0.54, and 0.54 ± 0.68 for bacteria and fungi, respectively, F 1,10 = 0.89, p = 0.37). Twenty-seven bacterial and 22 fungal taxa were found in all colonies, whereas seven bacterial and 28 fungal taxa were found in only one colony (Table A3). None of the ζ diversities differed significantly between bacteria and fungi ( Pseudomonas, accounting for 72.9% of the total rarefied read count, and the remaining taxa represented <0.01%-8.2% of the reads (Table A3). In addition to the three most abundant taxa, six taxa (Actinomycetales, Mycobacterium, Proprionibacterium, Methylobacterium, Nitrobacteraceae, and Hyphomicrobiales), had a prevalence of 80% or higher (Figure 7a). None of the bacterial taxa showed significant variation among colonies after a correction for false discovery rate (FDR) ( (Davidiella, Oidiodendron, Malassezia, Mortierella, Pseudogymnoascus, and the phylum Ascomycota) together accounted for 51.4% of the total abundance (Table A4). The remaining taxa each represented 0.08%-2.8% of the total read count. In addition to the six most abundant taxa, five taxa, Venturiaceae, Penicillium, Eupenicillium, Talaromyces, Trichocomaceae, and the class Dothideomycetes had a prevalence of 60% or higher, and read counts above 1800 in the ants (Figure 7b). None of the fungal taxa showed significant variation among colonies after an FDR correction (Tables A3 and A4).

| Community structure and dynamics
Here we show, based on sequence data, that the microbial communities within the ants are distinct from those in their surrounding nest material, and that the bacterial and fungal communities differ in their dynamics. Thus, the bacterial diversity, evenness, and species richness were significantly lower in the ants, compared to the nest material. In fungi, only species richness was significantly lower in the ants than in the nest material, whereas neither species diversity, nor evenness differed between the ants and the nest material. Furthermore, both bacterial and fungal species ζ diversity was significantly lower in the ants than in the nests. Thus, the bacterial community inside the ants encompasses fewer taxa and is less even (i.e. dominated by a few taxa) than the one in the surrounding nest material, whereas the fungal communities are less distinct, subject to more variation, and more even than the bacterial ones. Previous studies have found differences in the microbiomes between ants and their nest material in Solenopsis invicta and S. geminata (Ishak et al., 2011), Azteca trigona (Lucas et al., 2017), and Lasius fuliginosus (Brinker, Weig, et al., 2019). Our study adds new insights in demonstrating differences in species richness for both bacteria and fungi between ants and their nest environment, as well as differences between bacteria and fungi in community patterns measured by species diversity (as measured by the Shannon-Wiener index and the Zeta diversity (ζ), and evenness.
In the ants, we found no significant differences among communities across colonies for either bacteria or fungi, as indicated by both the PCoA, and the indices for species diversity, evenness, richness, and ζ diversity. However, the rate of within-colony species turnover was not statistically significant for bacteria, but highly significant for fungi. This suggests that the relative abundances of at least some fungal species changes across time within colonies, therefore rendering the fungal communities less stable. This is in agreement with our earlier studies on nest material in this population (Lindström et al., 2021), which suggest that the fungal communities are temporally more variable. In contrast to these results, Lucas et al. (Lucas et al., 2017;Lucas et al., 2019) found that both the bacterial and fungal communities vary extensively across years in the ant Azteca trigona, presumably due to diet. In Formica ants the diet is exceptionally stable across years (Rosengren & Sundström, 1987), with honeydew as the main source of carbohydrates, and insect prey as a source of protein. This might stabilize the microbial communities, but instead we found that sequence read counts varied between zero and 100% of all reads, both within and among colonies (Table A4).
T A B L E 2 Mean number of bacterial and fungal taxa found in ants and their nest material (ζ1), and across all samples and nests, present in one (ζ2), two (ζ3), three (ζ4), four (ζ5), and five or six (ζ6) samples of ants, and nest material, respectively.

| Species composition
Four bacterial taxa stood out as almost exclusively present and significantly more abundant in ants, but nearly absent in the nest material ( Figure 3a, and Lindström et al., 2021). Two were identified only at the level of Class (Alphaproteobacteria) or Phylum (Proteobacteria). Together these encompassed 75.3% of the total read count and represented 36, and 82 unique OTUs, respectively, which indicates that the F. exsecta microbiome encompasses several bacterial taxa, which are absent from the data bases we used in this study. The remaining two taxa were identified at the level of genus, Staphylococcus, and Streptococcus. Of these, Staphylococcus has previously been found in association with Pharoah ants (Monomorium pharaonis) (Boursaux-Eude & Gross, 2000), whereas Streptococcus has been found in the transcriptome of F. exsecta (Johansson et al., 2013). Both taxa contain human and animal pathogens, but many are opportunistic, rather than obligatory pathogens (Du Toit et al., 2014; Seipke et al., 2012). Several other bacterial taxa were present in 16 or more samples (out of a total of 18) in the larger data set on ants. These include Actinomycetales, Hyphomicrobiales, Methylobacterium, Mycobacterium, Nitrobacteraceae, Pseudomonas, and Proprionibacterium. Of these, Methylobacterium, and Mycobacterium, have documented associations with ants (Baldrian et al., 2012;Barke et al., 2010;Ishak et al., 2011;Lindström et al., 2019;Lucas et al., 2017;Mattoso et al., 2012;Nacke et al., 2016). Pseudomonas is a known pathogen of insects (Flury et al., 2016), whereas Mycobacterium is a common saprophyte in soil, but the genus also contains pathogens (Demangel et al., 2009). Of the bacterial genera detected almost exclusively in the nest material three, Marmoricola, Gemmatimonas and Conexibacter (Marupakula et al., 2016) have been found in the ant genus Solenopsis, but their function remains unclear. The genus Opitutus has been suggested to participate in recycling and reincorporating nitrogen in ants of the genus Cephalotini ants (Anderson et al., 2012), but was not found in the ant-associated bacterial community in our study. Most of the genera, which were detected only in the nest material represent bacteria that have previously been recorded in soil and environmental samples, such as Rhizomicrobium (Marupakula et al., 2016), and TM7 (incertae sedis) (Lauber et al., 2009).
The fungal community of the ants (Figure 3b, Lindström et al., 2021) included only two taxa (Davidiella and Cryptococcus), which were present in a significantly higher fraction of reads in ants, compared to the nest material. Davidiella was present in all but one ant sample, whereas Cryptococcus was present in ten (out of 18 ant samples) (Table A4). Davidiella encompasses several species involved in decomposition of soil and litter, (Monard et al., 2013;Voříškov a & Baldrian, 2013). The abundant presence of Davidiella in our samples of Formica exsecta could signal a specific, as yet unknown, role in the ant microbiome. The genus Oidiodendron represented the highest individual fraction of the total abundance (16.4%), with 76% of the reads attributed to ants (and 24% in the nest material). It was found in high counts in all ant samples (Table A4), and has, together with Cryptococcus been classified as a core indicator of the nests of this F. exsecta population (Lindström et al., 2019, Lindström et al., 2021. Oidiodendron is considered to belong to the rhizosphere of ericoid plants (Sietiö et al., 2018), but given the high and consistent abundance of this taxon in both the nest material, and in the ants it may also have an unknown role in these ants. Several other taxa were also occasionally present at high abundances in ants (Table A4), but in only one or a few samples each. Such sporadic high abundances may represent transient infections, or accidental ingestion of spores. Of the fungal genera absent from the ant microbiome, but abundantly present in the nest material, nearly all, Tothia (Wu & Jaklitsch, 2011), Trichoderma (Meincke et al., 2010), Mycena (Santalahti et al., 2016), and Zalerion (Korkama-Rajala et al., 2008), have mainly been recorded in environmental samples. Cladosporium is an exception, as it has also been detected in the nest material of F. exsecta (Lindström et al., 2021), and in ants more generally (Duarte et al., 2014;Yamoah et al., 2008). These largely correspond to the core microbiome defined in Lindström et al., 2021, although the genera Malassezia, Pseudogymnascus and the class Dothideomycetes were present in lower abundancies in the present samples.
In the PCoA on the Ant_Data two samples (F12 and F103, sampled in 2011) stood out in both their bacterial and fungal species communities (Figure 2a,b). These deviations were mainly explained by the abundant presence of Pseudomonas, combined with low abundancies of Proteobacteria and Alphaproteobacteria, and high abundances of the fungi Mortierella and Hypocrea, combined with low abundancies of Oidiodendron and Davidiella in these samples. Possibly, the bacterium Pseudomonas, some species of which are known pathogens of insects (Flury et al., 2016), and the fungi Mortierella and Hypocrea, which are known as saprotrophs (Deacon, 2005;Webster & Weber, 2007), may have capitalized on the absence of some of the regular dominant taxa.

| Characteristics and processes that may influence the microbiota
These results show that Formica exsecta exhibits a distinct microbial community both within the body (this study), and in the surrounding environmental matrix that constitutes their nest mound (this study and Lindström et al., 2021). These taxonomically distinct communities are constantly in contact with each other, given that the ants continuously handle nest material, transfer cuticular microbiota via social grooming and hygiene behaviours, and exchange crop contents with microbiota therein (Cremer et al., 2018). Yet the ants maintain a microbiome distinct from their surrounding nest material. The similarity of the F. exsecta microbiome across colonies, stands in contrast to Azteca trigona, in which variation among colonies is caused by variation in the relative abundance of Lactobacillus, apparently mediated by the diet of the ants (Lucas et al., 2017). In the fungus-growing ant, Trachymyrmex septentrionalis, the microbiome also varies extensively across locations and conditions (Green & Klassen, 2022). Several studies have recently screened the microbiomes of ants, and the T A B L E 3 Among-colony metrics for species diversity, richness, evenness, and within-colony turnover.  results point to extensive differences in the composition of these communities. For example, the bacterial taxa Actinobacteria, Lactobacillales, Enterobacterales, Clostridiales, Burkholderiales, Actinomycetales, Blochmannia, and Wolbachia, are prominent in the microbiomes of the ant taxa Camponotini Ramalho et al., 2019), Attini (Kellner et al., 2015;Vieira et al., 2017), Oecophylla (Hosmath & S.C., 2019), and Lasius (Brinker, Weig, et al., 2019). Representatives from all these taxonomic groups, except Blochmannia and Wolbachia, were among the 37 most abundant bacterial taxa in F. exsecta, but none constituted a major component of the microbiome. Instead, a very high fraction of bacterial reads representing several OTUs, representing 63% of the total read count and 74% of all Proteobacteria in the ants remained undetermined as Alphaproteobacteria or Proteobacteria, and were not found in any of the data bases we used. This suggests that F. exsecta harbours a considerable bacterial diversity that so far remains unexplored.
The question then arises, what maintains the microbial communities within workers of Formica exsecta, and what drives the microbial similarity across colonies, yet allows the maintenance of a microbiome distinct from the surrounding nest material. Cohesion in microbial communities within host animals may be driven by several factors, such as filtering by the host (Mazel et al., 2018), coevolution between host and microbes (McFall-Ngai et al., 2013), and host diet (Anderson et al., 2012;Hammer et al., 2017). Social insects distribute food via mouth-to-mouth feeding, and engage in cleaning behaviour (Hölldobler & Wilson, 1990), both of which facilitate the spread of microbes, as well as homogenizing the microbiota within colonies. The diet of Formica exsecta is dominated by sugar rich honey dew, supplemented with protein from insect prey (Hammer et al., 2017;Rosengren & Sundström, 1987). This may create a nutritional environment that may facilitate the transmission of microbes, some of which may be pathogenic. Indeed, formicine ants have evolved defence mechanisms against pathogens present in their food sources. For example, a related species, Formica fusca, has been shown to self-medicate upon the infection by pathogenic bacteria (Bos et al., 2015). Furthermore, Tragust et al., 2020(Tragust et al., 2020 showed that several formicine ant species, including five Formica species, use their highly acidic poison gland secretions to acidify crop contents, which decreased bacterial viability. Similarly, ants of the genus Formica use resin to disinfect their nest (Chapuisat et al., 2007), and enhance the impact applying formic acid on tree-collected resin (Brütsch et al., 2017). All of these behaviours may act as a strong filtering mechanism.
Our results add to our understanding of the ecological situation that ants face, with a strong incentive to maintain cohesion in their internal microbiota, and to evade the intrusion of undesirable microbes that may usurp social homeostasis, and the well-being of the entire colony. Several studies on ants, and bees have revealed an array of mutualistic, commensal, and parasitic relationships, in which mutualistic microbiota aid digestion of food, or help defend against parasites (Currie, 2001;Koch & Schmid-Hempel, 2011;Mueller et al., 2005). The interactions between social insects and their symbiotic microbes are exceptionally well-studied in termites (Aanen & Boomsma, 2006;Nombre et al., 2010), and in fungus-gardening ants (Currie, 2001;Kellner et al., 2015;Mueller et al., 2005), and have revealed complex tripartite or even quadripartite interactions between the insects and their mutualistic fungi, and their parasitic bacteria and fungi. Thus, being able to appropriate potential useful microbes that may become available in the environment can also open new avenues for adaptation, as has occurred in the fungus-growing ants (Brinker, Fontaine, et al., 2019). One may also ask: to what extent is there a two-way interaction between the ants and the soil, such that ants seed the soil in their nests with favourable microbes, at the expense of harmful ones? This sets the stage for a broader ecological and evolutionary perspective on the interaction among and between microbes and their hosts (Koskella et al., 2017;Koskella & Bergelson, 2020;Liu et al., 2019;Parfrey et al., 2018).

FUNDING INFORMATION
This work was supported by The Academy of Finland (grant numbers 251337, 252411, 284666) and the University of Helsinki.