Microbial Community Composition Reveals Spatial Variation and Distinctive Core Microbiome of the Weaver Ant Oecophylla smaragdina in Malaysia

The weaver ant Oecophylla smaragdina is an aggressive predator of other arthropods and has been employed as a biological control agent against many insect pests in plantations. Despite playing important roles in pest management, information about the microbiota of O. smaragdina is limited. In this work, a number of O. smaragdina colonies (n = 12) from Malaysia had been studied on their microbiome profile using Illumina 16S rRNA gene amplicon sequencing. We characterized the core microbiota associated with these O. smaragdina and investigated variation between colonies from different environments. Across all 12 samples, 97.8% of the sequences were assigned to eight bacterial families and most communities were dominated by families Acetobacteraceae and Lactobacillaceae. Comparison among colonies revealed predominance of Acetobacteraceae in O. smaragdina from forest areas but reduced abundance was observed in colonies from urban areas. In addition, our findings also revealed distinctive community composition in O. smaragdina showing little taxonomic overlap with previously reported ant microbiota. In summary, our work provides information regarding microbiome of O. smaragdina which is essential for establishing healthy colonies. This study also forms the basis for further study on microbiome of O. smaragdina from other regions.


Results
Data summary and alpha diversity analyses. A total of 12 16S rRNA gene libraries from O. smaragdina colonies sampled from different locations were subjected to Illumina paired-end sequencing (Supplementary Table S1). Sequencing of the libraries produced 22,999,352 raw reads. After stringent quality-filtering and chimera removal, we obtained 2,664,077 high quality sequences in our dataset with an average of 222,006 sequences per sample (minimum and maximum of 138,607 and 325,050 sequences per sample) (Supplementary Table S2).
Clustering the data with UCLUST 48 at 97% sequence similarity produced 1,397 OTUs (Supplementary  Table S3). When OTUs accounting for lesser than 0.5% of the total sequences were excluded, 97.8% of the sequences retained with 16 OTUs remained (Supplementary Table S4). From rarefaction analysis, the number of observed OTUs of all samples reached plateau at a sampling depth of 10,000 ( Supplementary Fig. S1). The Good's coverage estimates (Table 1) also indicated that the sequencing coverage was adequate in capturing the microbial diversity associated with O. smaragdina.
The Illumina sequencing revealed a relatively simple microbiota in the ant species. Only an average of 11 unique 97% OTUs per 50,000 sequences was observed in all the samples. Chao1, Shannon's and Simpson's diversity indices also indicated an overall low microbial diversity associated with the ant species (Table 1). Interestingly, lowest diversity was observed in O. smaragdina sampled from forest (mean Shannon index = 1.63; mean Simpson's index = 0.56) than urban area (mean Shannon index = 1.73; mean Simpson's index = 0.6) and patchy green area in urban city (mean Shannon index = 1.8; mean Simpson's index = 0.6), despite being relatively close in term of richness as indicated by Chao1 analysis.
At the OTUs level, we compared the results of taxonomy assignment to green genes 13_8-release database and SILVA128 SSU Ref NR database and most of the identity of OTUs from both databases were in consensus (Supplementary Tables S3 and S4). Further BLASTn analysis revealed the nearest identity of the 16 major OTUs at genus and species levels with varying range of similarity (91.32% to 100%) (Supplementary Table S4). Network analysis on the major OTUs showed a high degree of shared OTUs between the O. smaragdina from different environments, especially OTUs of family Acetobacteraceae with 4 OTUs found in every sample (except OTU1012 that was not found in two colonies from urban area) ( Supplementary Fig. S3).
Most microbial communities of O. smaragdina were dominated by family Acetobacteraceae (order Rhodospirillales) that harboured two core OTUs (OTU424 and OTU636, both were identified as genus Neokomagataea) with several OTUs that had lower relative abundance (Fig. 1, Table 2, Supplementary  Table S4). Further BLASTn search to NCBI 16S microbial database revealed that the sequences had ≥ 92.17% similarity with bacteria from the genera in family Acetobacteraceae. We observed that the relative abundance of family Acetobacteraceae in O. smaragdina differed by environment the colonies thrived in (F 2, 9 = 6.560, P < 0.05, Supplementary Table S5). Highest abundance of Acetobacteraceae occurred in colonies from forest area (mean = 88.72% ± 6.61), with a gradual decrease in colonies from patchy green area in urban city (mean = 77.21% ± 11.91) to colonies from urban area with low abundance of green area (mean = 47.56% ± 25.28) (Fig. 2, Supplementary Fig. S2).
On the other hand, the abundance of family Lactobacillaceae (order Lactobacillales) was variable in microbial communities of O. smaragdina. Relative abundance of Lactobacillaceae was low in most O. smaragdina from forest area (range from 1.42% to 14.17%). A second BLASTn analysis revealed that all OTUs under family Lactobacillaceae shared ≥93.67% similarity with bacteria from genus Lactobacillus. In two out of four colonies from patchy green area of urban city, the family constituted 21.57% and 23.65% of the sequences but the remaining two colonies had less than 5% (Fig. 1, Table 2). Highest abundance of Lactobacillaceae occurred in three  colonies from urban areas, and became the most dominant family in colony Urban2, although the observation was not consistent in colony Urban1 from urban area ( Table 2, Fig. 2). Other than the two dominant families, the family Mycobacteriaceae (order Actinomycetales) represented by OTU1052 and family Moraxellaceae (order Pseudomonadales) represented by OTU1193 were detected at high prevalence. Family Mycobacteriaceae was consistently low in relative abundance in all the samples. Family Moraxellaceae was only observed at high abundance in colony Urban1 although it had been observed in 100% of the samples (Fig. 1, Table 2).
In addition to the highly prevalent OTUs, five families distributed sporadically but occasionally high prevalence in the community of O. smaragdina. Of these, family Enterobacteriaceae (represented by OTU1277 and OTU633) had relative abundance as high as 45.69% in colony Urban1 and 27.46% in colony PGA3. Besides, family Leuconostocaceae (OTU965) was absent in most of the colonies but had a relative abundance of 1.50% in colony Urban2, 2.26% in colony PGA2, and 19.53% in colony Urban3. The family Anaplasmataceae (represented by OTU1176) was almost unique to colony Forest3 with relative abundance of 7.05%, although its presence was detected in other samples with lesser than 0.01% (Fig. 1, Table 2).   Beta diversity analysis. Abundance-weighted and unweighted UniFrac are beta diversity metrics that incorporate phylogenetic relationship of taxa in measuring community similarity 49 . From PCoA plot for unweighted UniFrac that accounts only for presence and absence of taxa, no pattern of correlation was observed among the communities (Fig. 3a). We related the result to presence of taxa that were sporadically distributed among the communities. As for abundance-weighted UniFrac analysis, clear separation of forest group colonies from the urban area colonies was observed. The microbial communities of forest O. smaragdina colonies were closely related and clustered together (Fig. 3b). Samples from patchy green area of urban city and urban area were distantly related from each other, with samples from urban area group showing the largest distance within group. The first two coordinates explained in total 84.35% of the variation in the data. We related the result to the variation in abundance of the most dominant taxa. Colonies from forest areas had characteristically highest abundance of family Acetobacteraceae and least abundance of family Lactobacillaceae. In contrast, high abundance of family Lactobacillaceae only occurred in colonies from urban areas and was predominant in Urban2 (Fig. 2).

Discussion
Our findings revealed the common bacteria taxa associated with O. smaragdina workers and showed that its community composition is generally stable as 97.8% of over 2 million quality-filtered reads were clustered into only 16 major OTUs which were assigned into 8 bacterial families (Supplementary Tables S2 and S4). We found two dominant taxa of acidophilic bacteria, families Acetobacteraceae and Lactobacillaceae to make up 87.2% of the O. smaragdina microbiome and are likely gut inhabitants. Genera from both families are known to tolerate various sugar-rich and acidic environments and constitute gut community in many insects 50,51 . We also relate the predominance of Acetobacteraceae and Lactobacillaceae to the high formic acid content in the body of O. smaragdina worker ants 52 . When a prey is encountered, the aggressive worker ants attack by biting and spraying formic acid on the area of bite from the tips of their abdomen 53 . Both dominant families Acetobacteraceae and Lactobacillaceae were represented by a few OTUs of varying relative abundance and assigned to different species in the family ( Table 2, Supplementary Table S4). These findings suggest that the microbiota of O. smaragdina is dominated by a few bacterial strains.
OTUs of the most abundant family Acetobacteraceae were closely related to acetic acid bacteria from plant-associated bacteria such as genera Neokomagataea 54 , Asaia 55 , Acetobacter 56 , as well as genus Gluconobacter previously found in insect 57 (Supplementary Table S4). Bacteria from this family had also been reported in microbiome of Camponotini ants 6 and Argentine ants 58 . More interestingly, several species of family Acetobacteraceae that were identified as bacterial symbiont in other insects are actually cultivable bacteria. For instance, Acetobacter tropicalis that is the major symbiont of the fruit fly Bactrocera oleae 59 , Asaia bogorensis and A. siamensis that are constantly associated with mosquito Anopheles stephensi 60 had all been successfully cultivated in laboratory following microbiome studies on the insect hosts. Most of these symbiotic acetic acid bacteria have been shown able to colonize the gut of insects and survive under acidic pH with the availability of diet-derived carbohydrates and oxygen 50 .
On the other hand, BLASTn analysis revealed that all OTUs under family Lactobacillaceae shared ≥93.67% similarity with bacteria from genus Lactobacillus and most of them were closely related to Lactobacillus found in fermented food [61][62][63] and plant 64 (Supplementary Table S4). Besides being a well-known commensal microorganism in human and other animals, Lactobacillus has also been detected in microbiome study of insects including bee 37 , fruit fly 30 as well as ants [65][66][67] . Although the role of Lactobacillus in O. smaragdina is unknown, Lactobacillus isolated from the crop of honeybees shown ability to produce potent antimicrobial and defends the host from microbial infections 68 . Among the remaining OTUs, families Enterobacteriaceae and Entomoplasmataceae were detected in microbiome of many ants as symbionts 38,65,69 , although Entomoplasmataceae that forms a unique lineage in army ants has no essential contribution to the growth or development of the host 69 .
This study assessed the spatial variation in microbiome of O. smaragdina from different environments. High sequencing depth had been allocated for each sample to obtain a holistic microbiome profile from all samples. This has been indicated by rarefaction analysis where individual curves reached plateau and saturation ( Supplementary Fig. S1). Despite the overall similarity in community composition, differences in term of abundance of the dominant families exist between O. smaragdina colonies from different environments, especially for the family Acetobacteraceae ( Supplementary Fig. S2). As shown in PCoA plot for abundance-weighted UniFrac distance, the forest group samples formed a cluster that is separated from the samples from other environments, while the urban area samples were distantly related to each other (Fig. 3b). We related the observation to the possible differences in feeding habits between O. smaragdina from forest area and urban area.
In Southeast Asia where O. smaragdina is dominant, the arboreal ant distributes across wide range of habitats from forests 70 , plantations 71,72 , urban areas 47 to mangrove swamp 73 . Food resources play important roles in supporting O. smaragdina colonies, which often comprise up to hundreds of thousands of workers at mature stage 18 . In addition to preying on other arthropods, O. smaragdina also consumes plant-derived extrafloral nectar and honeydew and tends homopterans in return for the honeydew produced 74,75 . As compared to arthropod preys, plant sap is generally more and in sufficient supply to compensate the high energy requirement in huge colonies of O. smaragdina 16 . However, nectar and honeydew are food sources that are rich in carbohydrate but scarce in protein (nitrogen). To obtain optimal colony growth both carbohydrate from honeydew and nectar and protein from arthropod preys are required 76 .
Unfortunately, urbanization causes fragmentation of forests into patchy green areas and destroys natural habitats of living species, leading to decline of biodiversity 77,78 . Clearing of large area of vegetation alters species composition as certain species become locally extinct 79 . Even if a species survives the process of urbanization, its functional roles are often affected in the urban environment. Among the affected species, ants have been recognized with remarkable ability to adapt to urban habitats 80 . However, the changes of vegetation and plant growth may affect the food choice and feeding habits of ants living in urban areas 81 .
Host diet plays an important role in shaping the microbiota of insects 82 . In a microbiome study that involved the ant Azteca trigona (Emerly, 1893), it was postulated that differences in diet caused significant microbiota variation across colonies 40 . Besides, Cephalotes varians (Smith, 1876) worker ants experienced significant changes in their microbiota when they were fed with pollen for one to two months 44 . In a recent diet manipulation study on Argentine workers ants (Linepithema humile, (Mayr, 1868)), a carbohydrate-rich and protein-poor diet resulted in dominance of bacteria of family Acetobacteraceae (order Rhodospirillales) in the gut communities. By comparison, gut communities of the ants experienced a reduction in the same group of bacteria when treated with low carbohydrate and high protein diet 58 .
In this study, similar bacteria taxon (family Acetobacteraceae, order Rhodospirillales) was observed in highest abundance within community composition of forest group O. smaragdina (Fig. 2). Colonies from forest areas are believed to have more access to plant-derived extrafloral nectar and honeydew and encounter higher density of homopterans that produce honeydew in their living environment. High consumption of carbohydrate-rich resources may therefore result in predominance of family Acetobacteraceae in their community. In contrast, O. smaragdina colonies from urban areas have the least abundance of family Acetobacteraceae. They are affected as mature trees that serve as its nesting site are removed for urban development. When diversity of plant trees and insects decreases in urban city, food resources of O. smaragdina colony are affected as plant-derived resources and honeydew-producing homopterans are scarce. In times of carbohydrate food scarcity, an omnivorous ant colony increases foraging activity for insect preys and thus relies more on protein-rich resources 83 .
Even though O. smaragdina has a wide distribution, the species is confined to tropical regions of Asia and having similar ecological niches. As samples were not collected for biogeographic comparisons, this study focused primarily on O. smaragdina colonies in Malaysia. Nonetheless, our findings obtained are in concordance to other reported ant species, whose microbiome comprised of dominant members consistently present across all samples and a core microbiome had been identified. Notably, a substantial amount of studies had demonstrated that geographical location of ant species had no correlation with overall microbiome diversity 5,6,58,65 .
To date, herbivorous species have been the focus of study on microbiota of ants and only a few reported for omnivorous species 13,40,41,58 26,84 . The community associated with omnivorous O. smaragdina was mainly dominated by bacteria from families Acetobacteraceae and Lactobacillaceae (Fig. 1). Unlike the herbivorous ants that rely solely on plant exudates, omnivorous O. smaragdina obtains nitrogen and protein from preying on other arthropods and possesses a distinctive community composition 2 .
Distinctive from herbivorous ants, greater taxonomic overlap is observed among community compositions of O. smaragdina and some omnivorous ants. Among omnivorous ants, the community composition of carpenter ants (genus Camponotus) is predominated by obligate bacterial endosymbiont Blochmannia species that provide essential amino acids for growth of the hosts 7,11,13,85 . Additionally, in a recent study that investigated the microbial composition across ants of genus Polyrhachis (Smith, 1857) that were omnivorous, multiple novel Blochmannia strains were recovered and some of the strains displayed host fidelity with different subgenera of the ant genus. In contrast, we observed no Blochmannia or any closely related taxa in community of O. smaragdina.
However, in addition to Blochmannia, family Acetobacteraceae have been reported in high prevalence in Camponotus chromaiodes (Bolton, 1995) and other members of the genus 6 . Furthermore, phylogenetic analysis on Acetobacteraceae across the genus Camponotus ants revealed that the bacteria formed a monophyletic clade with other ant-associated Acetobacteraceae with accelerated substitution rates in their 16S rRNA gene 6  Camponotini ants, bacteria Acetobacteraceae were among the dominant and persistent taxa of the omnivorous O. smaragdina in this study. From the BLASTn analysis, it was revealed that Acetobacteraceae in O. smaragdina were showing high evolutionary rates as these OTUs shared only 92.17% to 96.77% identity in 16S rRNA gene with described taxa of Acetobacteraceae in database (Supplementary Table S4).
Among the ant species with reported microbiota, the bacteria community composition of O. smaragdina has higher similarity with Argentine ants (Linepithema humile) 58 . Although both the omnivorous species were dominated by families Acetobacteraceae and Lactobacillaceae, Lactobacillaceae is the predominant taxon in Argentine ants. Furthermore, dominance of Lactobacillaceae in community of Argentine ants remains stable in both native populations and colonies in introduced regions despite a shift from typical carnivorous diet towards sugar-rich, nitrogen-poor foods in its introduced range 58 . In contrast, the predominant family Acetobacteraceae in O. smaragdina colonies varied between environments and was most abundant in the forest areas but reduced in the patchy green areas in urban city and urban areas.
O. smaragdina is known to be effective in controlling over 50 species of insect pests that infest many tropical tree crops 19,86 . Our findings in this study had shown that colonies of O. smaragdina from similar environments in Malaysia possess highly similar microbiome, especially among the group of colonies from forest with high abundance of Acetobacteraceae (Fig. 1, Table 2). The maintenance of a stable microbiome is vital to the development of O. smaragdina colonies in plantations of different crops. Observations from a study on mango orchards revealed that only trees with abundant level of O. smaragdina had significantly lesser shoots damage by the red-banded thrips, as compared to trees with fewer O. smaragdina ants and trees without 19 . The consistent association O. smaragdina with its microbiota might be the driving factor for its success as biological control agent against insect pests in different crop plantations. In addition to pest control, insecticides were sprayed less frequently in citrus orchards and expenditures on insecticides reduced by half when O. smaragdina is abundant 87 . This avoids the use of highly hazardous insecticides which could be harmful to the health of consumers and environments 87 .

Conclusion
The data presented in this study provide insight into the microbial community composition associated with the weaver ant O. smaragdina in Malaysia across different environments. The bacteria families Acetobacteraceae and Lactobacillaceae occurred in high prevalence and abundance in all colonies and constituted the core microbiota of the ant species. Despite community similarity in terms of bacterial taxa, abundance of the dominant families among ant colonies varied between environments, as observed with family Acetobacteraceae that has highest abundance in forest group colonies but lowest in colonies from urban area. Such differences may indicate change of functional roles from forest areas to fragmented forests in urban city and urban city with low abundance of green area, enhanced by the change in food choice and feeding habits in different living environments. This study formed the basis for continued research efforts which will be beneficial for elucidating the microbiota of O. smaragdina from other regions.

Methods
Specimen collection. O. smaragdina major workers were obtained from 12 distinct colonies from locations in Malaysia (Supplementary Table S1). All colonies are distantly separated and no interaction was observed among the colonies, including Forest3 and Forest4 (located 1 km from each other). While most samples were obtained from Selangor and Kuala Lumpur, we also included sample Forest1 from Perak state and sample PGA1 from Borneo Malaysia. These locations represent three different environments including forest, patchy green areas in urban city and urban areas with low abundance of greenery. Samples were collected from different environments to explore the influence of environments on the microbiota composition.
Sampling was performed within the period August to October 2016. O. smaragdina lives in the leaf nest of mature tropical trees which are often difficult to reach 19 . As the species exhibits worker polymorphism and the smaller minor worker ants generally stay inside the nest and tend to queen and brood 2,88 , we thus sampled only major worker ants that were foraging outside the nest. Multiple (10 to 15) major workers were captured from colony of every site and the specimens were immediately preserved in 95% molecular grade ethanol upon collection. The taxonomic identification were determined morphologically 47 and specimens were deposited in the collection of Microbiome Lab in University of Malaya. The remaining specimens were stored at −20 °C until DNA extraction.
DNA extraction. Whole ants were removed from ethanol and rinsed several times with nuclease-free water.
Three worker ants randomly picked from collection of each site were pooled into one sample and homogenized with sterile micropestle in microcentrifuge tube. Prior to DNA extraction, a two hours lysozyme (20 mg/ mL, Sigma-Aldrich) pre-treatment at 37 °C was performed on homogenized samples. Total genomic DNA was extracted with DNeasy blood and tissue kit (Qiagen, Valencia, CA). Purified DNA was eluted in elution buffer (EB) (Qiagen) and stored in −20 °C.
The quality-filtered sequences were clustered into Operational taxonomic units (OTUs) using UCLUST 48 in QIIME based on sequence similarity at 97%. A de novo approach was employed in OTUs picking to ensure all reads are clustered. A representative sequence was selected for each OTU and assigned for taxonomy with reference to the green genes 13_8-release database 93 and SILVA128 SSU Ref NR database 94 . The representative sequences were aligned using PyNAST 95 and a phylogenetic tree was built using FastTree 96 . An OTU table in biom format was created. OTUs accounting for less than 0.5% of the total number of sequences were excluded from further analysis as many of them consisted of singletons and were actually sequencing errors 90 . The major OTUs were blasted against the NCBI 16S microbial database for taxonomic assignment into genus and species level. A heatmap was constructed with OTUs that accounted for more than 0.5% of the total sequences using pheatmap R package 97 . To examine the interaction between OTUs and samples, an edge table was generated in QIIME and visualized using Cytoscape 3.6.1 98 .
Alpha diversity indices (including Chao1, Shannon, Simpson and observed OTUs) and Good's coverage 99 were calculated in QIIME on sample sizes normalized to 50,000 sequences per sample. To compare community from different colonies, samples were rarefied to the smallest dataset and beta diversity was analysed by abundance-weighted and unweighted UniFrac 49 . Weighted and unweighted UniFrac distances were calculated to examine the variation between microbial communities of O. smaragdina colonies. The resulting distance matrices were used for principal coordinates analyses (PCoA) and visualized in PCoA plots.
Data availability. The raw datasets for 16S rRNA gene amplicon sequencing generated for this paper have been deposited in the GenBank Sequence Read Archive (accession number SUB2165087).