What Can the Bacterial Community of Atta sexdens (Linnaeus, 1758) Tell Us about the Habitats in Which This Ant Species Evolves?

Studies of bacterial communities can reveal the evolutionary significance of symbiotic interactions between hosts and their associated bacteria, as well as identify environmental factors that may influence host biology. Atta sexdens is an ant species native to Brazil that can act as an agricultural pest due to its intense behavior of cutting plants. Despite being extensively studied, certain aspects of the general biology of this species remain unclear, such as the evolutionary implications of the symbiotic relationships it forms with bacteria. Using high-throughput amplicon sequencing of 16S rRNA genes, we compared for the first time the bacterial community of A. sexdens (whole ant workers) populations according to the habitat (natural versus agricultural) and geographical location. Our results revealed that the bacterial community associated with A. sexdens is mainly influenced by the geographical location, and secondarily by the differences in habitat. Also, the bacterial community associated with citrus differed significantly from the other communities due to the presence of Tsukamurella. In conclusion, our study suggests that environmental shifts may influence the bacterial diversity found in A. sexdens.


Introduction
Several studies have focused on understanding the factors that influence bacterial communities and their implications for the host, ranging from mammals [1] to ants [2][3][4][5][6][7][8]. Previous studies on vertebrate and invertebrate animals, including insects, have shown that bacterial communities are influenced by different factors such as host phylogeny [5,9,10], stages of development [4,11] and host environment [3]. In addition, the diet [2,12] and age [13] are other important host attributes that should be considered in studies on associated bacterial communities. However, in order to fully understand the ecological patterns and mechanisms behind these naturally established symbioses, more studies exploring new hosts are needed.
The composition and ecological drivers of host bacterial communities have been extensively studied in ants of the genera Camponotus and Cephalotes [2,4,9,10,14,15]. These symbioses can be so intimate and fundamental to the participating organisms that in many cases, hosts cannot survive without their symbionts, in this case called the primary. Examples include Blochmannia associated cane, citrus and eucalyptus) habitats, and 2) geographical location (West, Center, East, South and North, from São Paulo state in Brazil) can influence the A. sexdens microbiota. We expect all these factors to influence and explain the diversity of bacterial communities associated with this leafcutting ant.

Sample Collection
Samples were collected in natural habitats (n = 4 areas of Cerrado, n = 5 areas of Atlantic Forest) and agricultural habitats (n = 5 sugar cane; n = 5 citrus; n = 6 eucalyptus), located in São Paulo state -Brazil ( Figure S1). In each habitat, a linear transect was traversed along which foraging workers (with average size and weight between 10-20 mg and head width between 2.0 and 2.8 mm) were collected. The distance between colonies (n = 3 for each location) was at least 200 meters. We collected 12 workers in the Cerrado and 15 in the Atlantic Forest, 15 each in the sugar cane and citrus agricultural habitats, and 18 in the eucalyptus agricultural habitats (Figure 1; more details in Table S1). It is important to highlight that all agricultural habitats from this study are managed with agrochemicals: insecticides fungicides and herbicides for weed control. Synthetic fertilizers were also applied as part of cultural practices [42].
The taxonomic identifications were determined following morphological characters for species identification available in Mackay and Mackay (1986) [49] and vouchers were deposited in the collection de Mirmecologia do Alto Tietê. Samples used were collected in 95% ethanol in the field and kept at −20 °C until DNA extraction.

DNA Extraction and Bacterial DNA Sequencing
In total, we used 75 whole ant workers for DNA extraction using the DNeasy Power Soil Kit (Qiagen, Germantown, MD, USA) following recommended methods from Moreau [50] to avoid contamination. Furthermore, three negative controls were added to the extraction methods and to subsequent 16S rRNA amplifications and library sequencing. These negative controls are blank samples (corresponding to water instead of DNA in the polymerase chain reaction (PCR)) that were used to remove contaminants that may have been added in the process of obtaining the data. Amplification of the V4 region of 16S rRNA (515F/806R primers) was performed in triplicate following Earth Microbiome Project (EMP) protocols (http://www.earthmicrobiome.org/emp-standard-protocols/), also described in Caporaso [51]. PCR reactions followed recommendations from the Taq Platinum Kit (Invitrogen, Insects 2020, 11, 332 4 of 20 Schwerte, Germany) with 2.5 µL Buffer 10×, 1.0 µL MgCl 2 50 mM, 1.0 µL dNTPs 10 Mm, 1.0 µL BSA (bovine serum albumin) 1 mg/mL, 0.5 µL 515F primer 10 uM, 0.5 µL 806R primer 10 uM, 0.5 µL Taq Platinum 5U/µL, 5 µL template DNA (10 ng/µL), 13.0 µL PCR Water (Certified DNA-free). The reaction was submitted to the thermal cycler under the following conditions: 95 • C for 3 min, 30 cycles at 95 • C for 30 s, 60 • C for 30 s and 72 • C for 30 s, with final of 72 • C for 10 min and 4 • C hold. Purification was performed with the Ampure X beads and adapter connections (NEXTERA XT) were also performed following the manufacturer's recommendations. The samples were quantified via Nanodrop (Thermo Fisher Scientific, Waltham, MA, USA). Pooling and normalization were performed so that the samples had the same concentration. Then, the DNA pool was diluted down to 2 nM, denatured, and diluted to a final concentration of 16 pM with a 20% PhiX for sequencing on the Illumina MiSeq with run 2 × 250 pb at the Centro de Genômica (ESALQ, Piracicaba, SP, Brazil). All raw sequence data are publicly available in NCBI SRA accession number PRJNA556881 and study SAMN12373152.

Bioinformatic Analysis
Data were imported into Qiime2 [52] following the Casava 1.8 paired-end demultiplexed fastq protocol. The Dada2 plugin was used to create a feature table and quality control [53,54]. Paired-end sequence reads were trimmed (forward reads trim 30 and trunc 245, reverse reads trim 30 and trunc 220) for better results and SILVA_132_QIIME database [55,56] with 99% similarity was used to access ASVs (amplicon sequence variants) for the taxonomic information from the data sequenced. We trained our specific classifier based on our sample preparation and sequencing, using the "feature-classifier classify-sklearn" command [57].
Negative controls were filtered by the prevalence method using the Decontam package in R software (R Development Core Team, Vienna, Austria, 2019) following default settings [58]. Once our feature table was filtered, an extra filtering step was applied in Qiime2 in order to remove mitochondrial, chloroplast, and low abundance ASVs. Reads were aligned with the "align-to-tree-mafft-fasttree" command [59] and the phylogeny was inferred before alpha and beta diversity analyses [60,61]. Samples were normalized at 2000 reads in subsequent analyses.
The categories used in order to test for potential influences in leaf-cutting ants were conducted in two classes: 1) natural habitats (Cerrado and Atlantic Forest) and main agricultural habitats (sugar cane, citrus and eucalyptus), and 2) geographical location where the colonies were collected. To test the influence of geographical location within São Paulo State (Brazil), we divided the data according to their cardinal points of collection: West, Center, East, South, and North of the state and samples were assigned to those locations independent of their status belonging to natural or agricultural habitats ( Figure S1). Permutational multivariate analysis of variance (permanova) tests with the "p-pairwise" parameter were conducted with Unifrac distance, a qualitative measure that includes phylogenetic relationships between features [62]. Analysis of non-metric multidimensional scaling (NMDS), principle coordinates analysis (PCoA), and related statistics were implemented to better visualize our results of relationships between ecological communities and in the PAST3 software package [63][64][65] and Qiime2 in Emperor software [61]. Similarity percentage (SIMPER) analysis [63] was used to verify the ASVs' contribution to structuring the bacterial community found in the study. Network Analysis and heatmap plots of bacterial communities were conducted in R software [66] with the packages phyloseq [67] and ggplot2 [68].

Data Accessibility
All raw sequence data are publicly available in NCBI SRA accession number PRJNA556881 and study SAMN12373152.

Results
Samples were normalized to a minimum of 2000 reads per sample with sequencing reaching the expected depth ( Figure S2). After this cutoff, 75 samples remained and represented a total of 6,383,699 Insects 2020, 11, 332 5 of 20 reads distributed in 273 different ASVs. Sampling ranged from a maximum of 177,158 reads to a minimum of 2244 reads, confirming the successful sequencing of the bacterial community associated with A. sexdens.
Subsequent analyses were performed at the family taxonomic level of the bacterial communities or at the high hierarchy taxonomic level when available in the database. The ASV table and the overview of relative abundance can be seen in Table S2 and Figure S3, respectively. The main bacterial community recovered revealed Mesoplasma as the predominant bacterial genus (7%), followed by Tsukamurella (6%), Aeromicrobium (Nocardioidaceae) (5%) and others in smaller quantities. Overall when we include all natural and agricultural habitats, the bacterial community of A. sexdens consists of several taxa per sample (Figure 2), which makes it difficult to detect any bacterial structure or community patterns. Samples were grouped according to the habitat (natural or agricultural) to which they belong. Bar graphs for each library (one column = community from a sample) show the abundance of sequence reads classified to selected 99% similarity. Each color represents a distinct bacterium. ** indicates samples in natural habitat.

Beta Diversity and the Influence of Natural and Agricultural Habitats
The results of the beta diversity analysis suggest that the bacterial community of leaf-cutting ants do not differ between natural habitats and agricultural habitats (permanova weighted unifrac Samples were grouped according to the habitat (natural or agricultural) to which they belong. Bar graphs for each library (one column = community from a sample) show the abundance of sequence reads classified to selected 99% similarity. Each color represents a distinct bacterium. ** indicates samples in natural habitat.

Beta Diversity and the Influence of Natural and Agricultural Habitats
The results of the beta diversity analysis suggest that the bacterial community of leaf-cutting ants do not differ between natural habitats and agricultural habitats (permanova weighted unifrac distance: p-value = 0.177, pseudo F = 1.329, Figure 3A; permanova unweighted unifrac distance: p-value = 0.423, pseudo F = 1.014, Figure 3B). When beta diversity analysis was conducted only with samples from natural habitats, we also did not find statistically significant differences (permanova weighted unifrac distance: p-value = 0.599, Pseudo F = 0.858, Figure 3C; permanova unweighted unifrac distance p-value = 0.086, Pseudo F = 1.186, Figure 3D). Overall, there is no difference in the bacterial community of A. sexdens associated with the Cerrado or Atlantic Forest.
Insects 2020, 11, x 7 of 21 (permanova weighted unifrac distance: p-value = 0.04, Pseudo F = 1,802) ( Figure 3E). Unweighted analysis differences were not significant (permanova unweighted unifrac distance: p-value = 0.151, Pseudo F = 1.095) ( Figure 3F), confirming that the abundance of bacteria recovered in weighted unifrac distance are important in helping explain associations with agricultural habitats, but not the bacterial composition.  When analyzing only leaf-cutting ant bacteria associated with agricultural habitats (cane, citrus and eucalyptus) the beta diversity results indicated an influence in the weighted bacterial community (permanova weighted unifrac distance: p-value = 0.04, Pseudo F = 1,802) ( Figure 3E). Unweighted analysis differences were not significant (permanova unweighted unifrac distance: p-value = 0.151, Pseudo F = 1.095) ( Figure 3F), confirming that the abundance of bacteria recovered in weighted unifrac distance are important in helping explain associations with agricultural habitats, but not the bacterial composition.
To explore this result more deeply, we performed the paired permanova test (weighted unifrac distance) that allows all samples associated with the agricultural habitat to be analyzed in pairs (see Figure 4). The only pair that obtained a significant result was of bacterial communities associated with citrus and eucalyptus agricultural habitats (p-value = 0.04 and Pseudo F = 3.045), contributing to the difference found in Figure 3. In addition, through the results of SIMPER analysis, the main bacteria (ASVs) associated with citrus and eucalyptus agriculture are identified in Table 1 (for all groupings see Table S3). Note that Tsukamurella is responsible for 12.5% of the difference found in this pair (citrus and eucalyptus), followed by Aeromicrobium with 7.07% and others in smaller abundance. Our heatmap analysis gives a better view of Tsukamurella's association with A. sexdens, especially citrus ( Figure 5).
Insects 2020, 11, x 8 of 21 To explore this result more deeply, we performed the paired permanova test (weighted unifrac distance) that allows all samples associated with the agricultural habitat to be analyzed in pairs (see Figure 4). The only pair that obtained a significant result was of bacterial communities associated with citrus and eucalyptus agricultural habitats (p-value = 0.04 and Pseudo F = 3.045), contributing to the difference found in Figure 3. In addition, through the results of SIMPER analysis, the main bacteria (ASVs) associated with citrus and eucalyptus agriculture are identified in Table 1 (for all groupings see Table S3). Note that Tsukamurella is responsible for 12.5% of the difference found in this pair (citrus and eucalyptus), followed by Aeromicrobium with 7.07% and others in smaller abundance. Our heatmap analysis gives a better view of Tsukamurella's association with A. sexdens, especially citrus ( Figure 5).

Beta Diversity and the Influence of Geographic Location
Another factor that may be responsible for influencing the diversity of bacterial communities is the locality where the host was collected. Our results show the influence of the geographic regions of the state of São Paulo (Brazil) (West, Center, East, South, and North) on the sampled bacterial communities in terms of abundance (permanova weighted unifrac distance: p-value = 0.026, Pseudo F = 1.534) and presence/absence (permanova unweighted unifrac distance: p-value = 0.001, Pseudo F = 1.438). That is, there is a difference not only in the composition but also in the abundance of bacterial communities associated with leaf-cutting ants when considering the geographic regions of the state of São Paulo (Brazil) (Figure 6A,B). The network analysis shows the complexity of the bacterial community associated with A. sexdens ( Figure 6C).

Beta Diversity and the Influence of Geographic Location
Another factor that may be responsible for influencing the diversity of bacterial communities is the locality where the host was collected. Our results show the influence of the geographic regions of the state of São Paulo (Brazil) (West, Center, East, South, and North) on the sampled bacterial communities in terms of abundance (permanova weighted unifrac distance: p-value = 0.026, Pseudo F = 1.534) and presence/absence (permanova unweighted unifrac distance: p-value = 0.001, Pseudo F = 1.438). That is, there is a difference not only in the composition but also in the abundance of bacterial communities associated with leaf-cutting ants when considering the geographic regions of the state of São Paulo (Brazil) (Figure 6A and 6B). The network analysis shows the complexity of the bacterial community associated with A. sexdens ( Figure 6C).  We also tested for any significant differences between the habitats (natural: Cerrado and Atlantic Forest; and agricultural: sugar cane, citrus and eucalyptus) present within each geographic region.
For the Center region we found significant results in both the abundance analysis and composition analysis (which tests for presence or absence) ( Table 2). For the other regions analyzed separately, our results show that there is a significant difference in bacterial composition (presence and absence) associated with different habitats, but not in abundance. Table 2. Abundance (permanova weighted unifrac distance) and composition (permanova unweighted unifrac distance) analysis of the microbiota associated with A. sexdens collected at different geographical locations. This table shows significant differences mainly in the composition between the habitat (natural: Cerrado and Atlantic Forest; and agricultural: sugarcane, citrus and eucalyptus) in each geographical location.

Regions
Permanova When we conducted analyses between natural and agricultural habitats, although we were able to reveal interesting results regarding the Tsukamurella influence mainly associated with citrus, some data were scrambled and diluted in the vast amount of information generated by this technique. When first investigating the influence of geographic location (West, Center, East, South and North), we determined that it does have a large impact on the weighted and unweighted composition of the bacterial community associated with A. sexdens. However, when the analyses were conducted by region (separately), the results showed significant differences in the composition (unweighted) of bacterial communities associated with different habitats, i.e., natural (Cerrado and Atlantic Forest) and agricultural (sugar cane, citrus and eucalyptus).
To test which pairs of geographic regions contributed to the difference found in Figure 6, paired permanova tests verified abundance (weighted unifrac distance) and composition (unweighted unifrac distance) and found that several groupings showed significant differences (see Figure 7). In general, all regions (West, Center, East, South, and North) influenced the abundance and composition of bacterial communities associated with this leaf-cutting ant. Therefore, the pairs that obtained significant results were subjected to SIMPER analysis, where we obtained the main bacteria associated with these groupings (peer regions) (see Table 3). Our results are consistent, showing that Tsukamurella is one of the major bacteria responsible for differences in all groupings of geographic regions. These results were also confirmed by the analysis in Table 1, further supporting that this bacterium has a significant impact on the bacterial community of A. sexdens. In addition, Mesoplasma, the main bacterium found in the present study (see Figure 2), was also frequent and one of the major bacterial genera responsible for the differences found in almost all clusters, except Center vs. North. Insects 2020, 11, x 12 of 21

Alpha Diversity: Different Habitats and Geographical Locations
Alpha diversity also presented distinct results when the analysis was conducted comparing only the different habitats and geographic regions with the Faith phylogeny index. According to a Kruskall-Wallis test, only geographic region influences the alpha diversity associated with A. sexdens (H = 16,795, p-value = 0.002). As this richness index is qualitative, it takes into account the composition of the bacterial community (presence and absence). However, the same analysis conducted with the Shannon index regardless of abundance did not yield any significant results (see Figure S4). Therefore, the different geographical regions may influence the composition of alpha diversity associated with A. sexdens, but do not appear to influence the abundance.

Discussion
There are several factors that can predict and influence the diversity (composition and abundance) of microbial communities associated with a given host [8]. The microbiota of leaf-cutting ants has already been addressed in several studies [23,27,[34][35][36][69][70][71], but there remains much to understand about the drivers of these symbioses. This lack of knowledge motivates the present study, which is the first to compare the bacterial communities of ants in natural (Cerrado and Atlantic Forest) to agricultural (sugar cane, citrus and eucalyptus) habitats, as well as consider geographical location (West, Center, East, South, and North of the state of São Paulo-Brazil), to assess the factors that may influence microbial diversity. We found differences in microbiota of the different samples from the same species in different habitats and geographic locations. While each strain of bacteria can be acquired in a different way [72], our results may reflect the fact that these bacteria are being acquired from different environmental conditions (e.g., landscape simplification, different plant availabilities, agrochemicals, and soil), through horizontal more than vertical transmission. That is, even with sophisticated defense systems and behavior utilized by A. sexdens [49,73,74], the bacterial communities of the present study were altered mainly due to differing geographical locations, and secondarily due to natural and agricultural habitats.
There are several studies that show that the composition and abundance of the microbiota is impacted by horizontal acquisition [3,47,70,[75][76][77]. Our study corroborates those that have found that the geographic location of the host influences the diversity of the bacterial community, whether in the case of mammals [75], birds [78,79], other insects [47,77,80], or also of ants [3,70,76,81]. For example, the bacterial diversity recovered from the mesosoma and head of Camponotus workers is influenced by the habitat where these ants were collected, which suggests that these bacteria are being acquired via horizontal transfer [3]. Another study that reinforces the impact of geographical location on the diversity of bacterial communities was conducted by Martins and Moreau [76] with Pheidole species collected around the world. The same pattern of acquisition of microbial communities was also reported in fourth-instar larvae of Solenopsis invicta ants [81] and the lower attine species Mycocepurus smithii [70].
Atta cephalotes workers also showed a diverse and unstable microbiota with Mollicutes being the main bacteria recovered [36]. The bacterial community recovered from A. sexdens is also very complex, and this study was able to identify several important bacterial taxa associated with this species. ASVs identified as Mesoplasma sp. (Mollicutes, Entomoplasmatales) is one of the most prevalent bacteria in our sample with 7% relative frequency. Our results corroborate a previous study [23] that reports evidence of this bacterium mainly associated with the post-pharyngeal gland (70%) and also in smaller quantities in the midgut of the queen and the whole body of the worker of A. sexdens. Meirelles et al. [35] describes the association of Mesoplasma sp. with infrabucal pellets carried by A. texana queens. Also in this study, through phylogenetic analysis, the authors show specificity of this bacterium associated with eight different genera of Attine, since it formed a monophyletic clade. Mesoplasma sp. associated to the army ants [82] was also grouped in another clade, also showing specificity to army ants [83]. The function of this ant-associated bacterium was investigated with the genome of two strains, and it was found that each strain was acquired independently from a different origin. These bacteria are capable of decomposing arginine and providing nitrogen-rich amino-acid [84]. In that same study, the authors also concluded that the relative abundance of Mesoplasma sp. is related to the type of substrate that the ant worker offers to the fungus garden. Our study found a high relative abundance of these bacteria, which suggests that it must have been facilitated by the substrate of A. sexdens.
The next most abundant ASV recovered in the present study is Tsukamurella sp. (Actinobacteria), which despite occurring in other habitats was strongly related to the citrus agricultural habitat. Although its function is not fully understood, there are studies reporting that these bacteria may be an endophytic growth promoter [85] minimizing the growth of phytopathogenic fungi [86], and also may act as a nematicide [87] used in agriculture. The third most common ASV was identified as Aeromicrobium sp. (Actinobacteria), which has also been reported on species of ants as well [88] but is generally associated with the soil [89]. Other studies have reported the presence of ant-associated Actinobacteria, especially the genus Pseudonocardia sp., often associated with other Attines ants [71,90]. These bacteria are capable of producing compounds with antibiotic properties that defend the fungal garden and the ant colony from disease [25,71,90]. The samples of A. sexdens from our study were also associated with these bacteria in smaller quantities compared to the literature, which suggests that perhaps A. sexdens is also relying on other measures to maintain the health of the fungus garden and the colony. Further studies are needed to address this issue.
The phylum Proteobacteria has also been reported in A. cephalotes Linnaeus 1758 and A. colombica Guérin-Méneville 1844 [69,91]. Pseudomonas and Burkonderialles are bacteria belonging to this phylum, and were associated with the workers of leaf-cutting ants of the present study, although in smaller abundances. Some bacteria present in the insect microbiome may be able to metabolize agrochemicals, making the host pesticide-resistant [92][93][94], and it has been shown that certain strains of the genus Pseudomonas sp. are associated with insect resistance to insecticides [95]. The bacterial community is subject to several environmental stressors that may alter its composition [96]. As an example, certain strains of Burkholderia sp. are capable of degrading insecticide fenitrothion (organophosphate) [97]. Tago et al. [98] found an increase of this bacterium in the soil after the application of the pesticide in sugar cane agricultural habitats.
In another study of Acromyrmex spp., besides Entomoplasmatales, the bacteria Wolbachia and Rhizobiales were highlighted in association with these ants. These latter two bacteria are highly associated with several genera of ants [2,8,36,99]. While the function of Wolbachia associated with ants is still unknown, it is transmitted vertically, although its diversity is influenced by geography [17,36,100]. On the other hand, Rhizobiales is also often associated with nitrogen fixation and recycling [2,69], which is extremely important especially in the case of herbivorous ants. Our results also recovered both bacteria associated with A. sexdens.
The microbiota associated with a host can be composed of mutualists, commensals and pathogens, and many of the bacteria associated with A. sexdens are probably transient bacteria that have no associated role in ants. With the 16S rRNA amplicon technique, we cannot be sure of the microbe-host relationships or their functions. However, overall these bacteria may have been acquired horizontally and thus may reflect the environment (habitats). In addition, although the bacterial community of A. sexdens is unstable and highly variable, we do not know how this complexity varies in the different castes within the colony, and its implications for the host. In the case of Atta cephalotes it was found that the bacteria Mollicutes passes via social transmission of workers to the larvae [36]. Future studies that consider the different castes and location of this microbiota in the ant body may help to elucidate the mode of transmission and consequently the function of these bacteria associated with A. sexdens.

Conclusions
Genomic approaches have become increasingly accessible and can be used as excellent tools to explain several biological phenomena such as intra-and interspecific relationships, species ecology such as diet and symbioses, as well as molecular, physiological, behavioral and evolutionary features of the host. These approaches have the potential to generate data that can contribute to the understanding of symbiotic interactions within the field of myrmecology [2,[101][102][103][104][105]. Recent next-generation sequencing techniques have made it possible to analyze the communities of bacteria present in several hosts. Specifically for ants, many studies have demonstrated the importance of these symbionts and how they can influence host attributes [2,6,8,10,20,106]. However, studies of microbial interactions are challenging because they involve dynamic systems that can be shaped by the sum of abiotic factors, as presented here. This is the first study that compared the bacterial communities of A. sexdens in natural habitats, such as Cerrado and Atlantic Forest, to those in agricultural habitats (sugar cane, citrus and eucalyptus) and that considered geographic location (West, Center, East, South, and North of the state of São Paulo-Brazil), showing that all these factors may influence microbial diversity. Knowledge about this insect-associated microbiota may help to develop strategies to increase insect pest resistance in agricultural habitats, as well identify bacteria that can act as bioremediation of pesticides and help recover degraded areas [107].
Supplementary Materials: The following are available online at http://www.mdpi.com/2075-4450/11/6/332/s1, Figure S1: Location of Atta sexdens collection areas in São Paulo State, Brazil, Figure S2: Rarefaction curves were used to estimate richness in the Shannon index for 16S rRNA. The vertical axis shows the Shannon index observed and the number of sequences per sample is shown on the horizontal axis, Figure S3: The overview of bacteria relative abundance recovered from this study of Atta sexdens, Figure S4: Alpha diversity associated with Atta sexdens, Table S1: Samples of Atta sexdens collected from different natural and agriculture habitats, Table S2: Amplicon Sequence variants (ASVs) table highlighting each bacterium recovered from each sample of Atta sexdens, Table S3: SIMPER analysis of main bacteria in all differents habitats. This test indicates the contribution of specific ASVs to the observed differences in community structure among different groups. Note that Tsukamurella remains the most influential bacteria associated with citrus crop.