Abstract Drosophila melanogaster has become an important model organism to study host-microbe interaction in the laboratory. However, the natural microbial communities that are associated with D. melanogaster have received less attention. Especially, information on inter-individual variation is still lacking, because most studies so far have used pooled material from several flies. Here, we collected bacterial 16S rRNA gene community profiles from a set of 32 individuals from a single population. We simulated pools from the individual data (i) to assess how well the microbiome of a host population is represented by pools, and (ii) to compare variation of Drosophila microbiomes within and between populations. Taxon richness was increased in simulated pools, suggesting that pools paint a more comprehensive picture of the taxa associated with a host population. Furthermore, microbiome composition varied less between pools than between individuals, indicating that differences even out in pools. Variation in microbiome composition was larger between populations than between simulated pools from a single population, adding to the notion that there are population-specific effects on the Drosophila microbiome. Surprisingly, samples from individuals clustered into two groups, suggesting that there are yet unknown factors that affect the composition of natural fly-associated microbial communities and need further research. Drosophila, bacteria, 16S INTRODUCTION Drosophila melanogaster has become an important model for the investigation of host-microbe interactions (Erkosar et al.2013). Interactions with bacteria can affect D. melanogaster phenotype in many aspects. Bacteria-mediated D. melanogaster phenotypic effects range from pathogenic effects (Galac and Lazzaro 2011), over effects on cold tolerance (Linderman et al.2012), and effects on D. melanogaster's nutritional status (Dobson et al.2015; Chaston et al.2016) to highly beneficial effects. For example, certain acetic acid bacteria (Acetobacteraceae) and Lactobacillaceae significantly promote D. melanogaster's larval growth on amino acid poor diet (Shin et al.2011; Storelli et al.2011) and can ensure longtime fertility and longevity under nutrient poor conditions (Téfit and Leulier 2017). Because bacteria from the same families that have these highly fitness relevant effects on D. melanogaster can also be found in association with wild-caught flies (Corby-Harris et al.2007; Cox and Gilmore 2007; Chandler et al.2011; Staubach et al.2013), it seems reasonable to assume that they could also play a role in fly evolution. However, in most studies bacterial strains and communities that were isolated from lab-reared flies are investigated, and we still know rather little about natural Drosophila-associated bacterial communities and the factors shaping them. Yet, only natural microbial communities can have played a role in Drosophila evolutionary history. Hence, deeper knowledge about natural Drosophila-associated microbial communities is important to gain insights into the role they might play for Drosophila in an evolutionary context. This is even more important since it was shown by Chandler et al. (2011) and Staubach et al. (2013) that bacterial communities associated with wild-caught flies are different from those associated with lab-reared flies. Under controlled laboratory conditions, host genetic make-up can influence D. melanogaster-associated bacterial communities (Unckless, Rottschaefer and Lazzaro 2015; Chaston et al.2016). Under natural conditions, the substrate that flies were collected from influences bacterial community composition (Chandler et al.2011; Staubach et al.2013), while species differences between D. melanogaster and D. simulans that also include host-genetic differences are detectable but have a smaller effect. These studies on natural microbial communities relied on pooling material from several flies for bacterial community profiling. The microbiomes of these pools are often treated as being representative for microbiomes of populations or even species. However, whether, and to what extent, the microbiomes of pools represent the microbiomes of populations is not known. Pools might still be affected by the microbiomes of individuals, in particular, when microbial loads vary between individuals. When individuals with high loads are included in a pool, they may contribute disproportionately to that pool's microbiome and, hence, becoming rather representative of that individual instead of the population. Therefore, it is necessary to collect information on intra-population variation of microbiomes for evaluating how representative pools are. This can be done by profiling the microbiomes of individual flies. Furthermore, microbiome profiles of individual flies can help to compare variation within and between populations. This is important to assess the significance of population differences reported before; if the variation between populations was not significantly larger than the variation within populations, doubt could be cast upon population differences in the microbiome structure that were based on pooled samples. This seems even more relevant in the light of recently reported stochasticity in the gut colonization process (Obadia et al.2017), as well as substantial intra-population variation in fungus and cactus feeding Drosophila (Martinson et al.2017; Martinson, Douglas and Jaenike 2017). Finally, controlling the natural diet by profiling the microbiomes of individuals from the same diet that were collected at the same time could help to evaluate whether there are factors that drive variation of microbiomes within natural D. melanogaster populations. Yet, contrary to humans, where collecting individual microbial profiles has been in the focus (Consortium THMP 2012), such data, to our knowledge, does not exist for a D. melanogaster population sample. In order to assess bacterial community variability between individuals under otherwise constant conditions, we collected flies from the same substrate and location at the same time and profiled the bacterial microbiomes of individual male flies using 16S rRNA gene sequencing. We compared these microbiomes to microbiomes from pooled flies to evaluate the representativeness of pools for host population-associated microbiomes. Furthermore, we collected pools from different populations to place variation within a population into the context of variation between populations. We aimed at a comprehensive assessment of variation between populations. Therefore, we sampled populations from different substrates and locations at different time points. MATERIAL AND METHODS Fly samples Flies were collected as described previously (Staubach et al.2013). In short, live flies were collected and brought to the lab in empty vials within 5 h of collection. Male D. melanogaster were identified based on morphology and frozen at −80°C until DNA extraction. We focused on males because only D. melanogaster males can reliably be distinguished from D. simulans. This is crucial when working with wild-caught flies, because these two species often co-occur in the field. Samples from oranges, peaches and apple1 were the same as in Staubach et al. (2013). See Table S1 (Supporting Information) for a full list of sampling locations. All pools were based on five flies except ‘orange3΄, for which we were able to obtain only three flies. ‘Orange3΄ was no outlier in any of the analyses and was included with the pools of five flies. DNA extraction, PCR and sequencing DNA was extracted either from individual male D. melanogaster or pools of five males, with the exception D. melanogaster orange sample 3 (orange3), for which we were able to retrieve three males only. DNA extraction was performed using the Qiagen QIAamp DNA extraction kit (Qiagen, Carlsbad, CA) using bead beating as described in Staubach et al. (2013) and running negative extraction controls (without fly material) in parallel. Barcoded broad range primers, 515F (5΄GTGCCAGCM GCCGCGGTAA3΄) and 806R (5΄GGACTACHVGGGTWTCTAAT3΄), as described in Caporaso et al. (2010) and used for the Earth Microbiome Project were used to amplify the V4 region of the bacterial 16S rRNA gene. DNA was amplified using Phusion® Hot Start DNA Polymerase (Finnzymes, Espoo, Finland) and the following cycling conditions: 30 sec at 98°C; 30 cycles of 9 sec at 98°C, 60 sec at 50°C and 90 sec at 72°C; final extension for 10 min at 72°C. In order to reduce PCR bias, amplification reactions were performed in duplicate and pooled. PCR products were run on an agarose gel for quantification and pooled in equimolar amounts. Extraction control PCRs were negative and excluded. The resulting pool was gel extracted using the Qiaquick gel extraction kit (Qiagen, Carlsbad, CA) and sequenced on an illumina MiSeq sequencer reading 2 × 250bp. Data analysis Sequencing data were analyzed using mothur (Schloss et al.2009) (v 1.36.0) following the MiSeq SOP on mothur.org. Main data processing steps were quality filtering of raw sequence reads, denoising, removal of chimeric sequences, taxonomic classification and OTU clustering at 97% sequence similarity. Sequences were taxonomically classified using the SILVA reference database (Pruesse et al.2007) as implemented in mothur. Alpha-diversity measures were computed with mothur. The ecodist (Goslee and Urban 2007) R (R Core Team 2017) package was used to calculate Bray–Curtis distances. The vegan R package (Oksanen et al.2016) was used for Jaccard and UniFrac distances incorporating the GUniFrac package (Chen 2012). The pvclust package was used for cluster analysis (Suzuki and Shimodaira 2015). We applied the SIMPER method to identify signature taxa that differed between the two clusters found in the 2013 plum population. A detailed analysis script with all mothur and R commands can be found in File S5 (Supporting Information). During initial analysis, we found sequences classified as Halomonas in our samples. As this is a halophile neither expected on rotting fruit nor flies, we Blast searched a representative sequence from the largest Halomonas OTU. The best hits were plant mitochondria (see File S4 for blast search, Supporting Information). The occurrence of plant mitochondrial sequence in our samples appears much more likely than Halomonas. Because we had no interest in analyzing the organelles, Halomonas sequences were removed from the analysis. Simulations We simulated pools of flies from the individual community data. Because bacterial loads vary in wild-caught flies (Elya et al.2016), it is reasonable to assume that individuals contribute differently to the community of a pool. To take this into account, we used the distribution of bacterial loads from wild-caught flies from Elya et al. (2016) for our simulations. The data were kindly provided by Carolyn Elya. Implementation of these data was necessary because 16S rRNA gene sequencing provides only relative abundance data, but does not allow us to draw conclusions about bacterial loads. For the simulations, in short, we sampled bacterial loads from the empirical distribution in a first step. In a second step, we sampled sequences from randomly drawn individuals according to the bacterial loads sampled in the first step. This procedure was conditioned to add up to 967 sequences per simulated pool. This is the same number of sequences that was analyzed for all samples. The procedure was performed for 5 and 30 individuals/loads. We decided to simulate 30 pools, as this seemed a reasonable number that does not overstretch the empirical data basis given that 31 individuals were analyzed. A detailed R script for the procedure can be found in supplementary File S5 (Supporting Information). RESULTS We assessed bacterial community composition and diversity by sequencing the V4 region of the 16S rRNA gene for 32 individuals and 13 pools of five wild-caught D. melanogaster. (Fig. 1 and Table S1, Supporting Information). A total of ∼ 2240 000 sequences passed quality filtering. ∼1050 000 sequences matched the Wolbachia 16S rRNA gene sequence and were removed. Individual fly sample #6 was removed because only 11 sequences remained after Wolbachia removal. At least 967 sequences per sample were analyzed for the remaining samples. Figure 1. View large Download slide Overview of sampling locations and substrates. Fruit indicate sampling locations and substrates. (A) Most samples were collected in California, USA. (B) Two samples were collected in Rhode Island. Numbers indicate the number of samples collected from that area. See Table S1 (Supporting Information) for detailed sampling information, Map data from Google 2017. Figure 1. View large Download slide Overview of sampling locations and substrates. Fruit indicate sampling locations and substrates. (A) Most samples were collected in California, USA. (B) Two samples were collected in Rhode Island. Numbers indicate the number of samples collected from that area. See Table S1 (Supporting Information) for detailed sampling information, Map data from Google 2017. BACTERIAL RICHNESS WAS HIGHER IN SIMULATED POOLS THAN IN INDIVIDUAL FLIES AND DIFFERED BETWEEN POPULATIONS We were interested in finding out whether individual flies carry reduced bacterial diversity or a skew in bacterial abundance patterns, as might be expected from stronger stochastic effects in smaller samples, when compared to pools of flies. For comparing bacterial alpha-diversity between individual flies and pools, we grouped sequences into 97% identity operational taxonomic units (OTUs). In order to paint a more comprehensive picture of the representativeness of pools for a local population, we simulated additional pools by combining community data from individuals (see materials and methods for the simulation procedure). The individuals were sampled from a single population living on plums in 2013. In short, we used the distribution of bacterial loads in wild-caught flies from Elya et al. (2016) to draw sequences from five individuals matching the size of the empirical pools. The simulated pools revealed higher OTU-richness (Chao 1984) and diversity (Shannon's H, Shannon et al.1949) than the individual samples (Table 1A and Fig. 2A). This indicated that the microbiomes of individuals contained a smaller subset of the microbial taxa associated with the host population. We found no difference in the equitability of taxon abundance between pools and individual flies (Shannon's E and Simpson's E, Table 1A). This indicated no significant skew in taxon abundance resulting from pooling versus using individuals. Table 1. Comparison of alpha-diversity (A) between individuals and simulated pools and (B) within populations and between populations. Means +/− standard deviation are listed. P-values were calculated using the Mann–Whitney U test. Sample #OTU P Chao P Shannon H P Simpson D P Shannon E P Simpson E A Individual flies 42.1 ± 18.0 72.8 ± 32.7 1.87 ± 0.63 0.31 ± 0.21 0.50 ± 0.13 0.11 ± 0.04 Simulated pools 51.4 ± 13.4 0.012 140.9 ± 88.9 3.77E−06 2.16 ± 0.27 0.031 0.21 ± 0.07 0.18 0.55 ± 0.07 0.22 0.10 ± 0.03 B Pools of 5 across substrates 36.6 ± 20.8 65.0 ± 40.7 1.80 ± 0.77 0.33 ± 0.25 0.50 ± 0.17 0.13 ± 0.05 Simulated pools 51.4 ± 13.4 0.0067 140.9 ± 88.9 2.50E−04 2.16 ± 0.27 0.074 0.21 ± 0.07 0.21 0.55 ± 0.07 0.91 0.10 ± 0.03 Sample #OTU P Chao P Shannon H P Simpson D P Shannon E P Simpson E A Individual flies 42.1 ± 18.0 72.8 ± 32.7 1.87 ± 0.63 0.31 ± 0.21 0.50 ± 0.13 0.11 ± 0.04 Simulated pools 51.4 ± 13.4 0.012 140.9 ± 88.9 3.77E−06 2.16 ± 0.27 0.031 0.21 ± 0.07 0.18 0.55 ± 0.07 0.22 0.10 ± 0.03 B Pools of 5 across substrates 36.6 ± 20.8 65.0 ± 40.7 1.80 ± 0.77 0.33 ± 0.25 0.50 ± 0.17 0.13 ± 0.05 Simulated pools 51.4 ± 13.4 0.0067 140.9 ± 88.9 2.50E−04 2.16 ± 0.27 0.074 0.21 ± 0.07 0.21 0.55 ± 0.07 0.91 0.10 ± 0.03 View Large Figure 2. View largeDownload slide (A) Rarefaction curves of 97% identity OTUs for individual flies and pools of five flies simulated from the same population (plum 2013). The orange line shows the mean number of OTUs discovered for 31 individual fly samples with the shaded area indicating the standard deviation. The purple line represents simulated pools of five flies. (B) Rarefaction curves of simulated pools and 13 pools of five flies collected across substrates and locations. The green line represents the mean and the shaded area the standard deviation. Purple line as in A. Figure 2. View largeDownload slide (A) Rarefaction curves of 97% identity OTUs for individual flies and pools of five flies simulated from the same population (plum 2013). The orange line shows the mean number of OTUs discovered for 31 individual fly samples with the shaded area indicating the standard deviation. The purple line represents simulated pools of five flies. (B) Rarefaction curves of simulated pools and 13 pools of five flies collected across substrates and locations. The green line represents the mean and the shaded area the standard deviation. Purple line as in A. Next, we wanted to test whether alpha-diversity differed between populations, when taking into account within population variation. Therefore, we compared the simulated pools from the 2013 plum population to pools collected across substrates and locations. The total number of OTUs and Chao's richness estimate were higher in pools from the 2013 plum population (Table 1B), indicating that (i) this population was relatively taxon rich and (ii) there were significant population-specific effects on taxonomic richness. BACTERIAL COMMUNITIES WERE DOMINATED BY ACETIC ACID BACTERIA AND VARIED IN COMPOSITION BETWEEN INDIVIDUALS FROM THE SAME POPULATION For a detailed view of the variability in bacterial community composition of individual and pooled fly samples, we classified the 16S rRNA gene sequences taxonomically (Fig. 3). Figure 3. View largeDownload slide Relative abundance of bacterial taxa as assessed by 16s rRNA gene sequences. Wolbachia sequences were excluded. Bacterial genera of abundance <3% have been removed for clarity. Figure 3. View largeDownload slide Relative abundance of bacterial taxa as assessed by 16s rRNA gene sequences. Wolbachia sequences were excluded. Bacterial genera of abundance <3% have been removed for clarity. The bacterial communities were dominated by acetic acid bacteria (Acetobacteraceae 63.0%) representing four of the five most common genera (Saccharibacter 24.3%, Gluconobacter 18.2%, Acetobacter 13.5% and Gluconacetobacter 6.7% average relative abundance). The relative abundance of the different taxa was highly variable between individuals from the same population (Saccharibacter 2.0%–94.3% relative abundance, Gluconobacter 1.7%–65.3%, Acetobacter 0.5%–37.3% and Gluconacetobacter 0.4%–44.3%). Please note that bootstrap support for the Saccharibacter classification was often relatively low (between 40% and 60%, Table S2, Supporting Information) indicating that there were several sequences in the SILVA database that matched sequences from this taxonomic group similarly well. Blast search for representative sequences from this taxonomic group produced perfect matches to bacteria classified as Commensalibacter in Chandler et al. (2011) and Acetobacter in Corby-Harris et al. (2007) (see File S1 for blast search results, Supporting Information). Acetic acid bacteria represent a major bacterial group associated with wild-caught flies (Corby-Harris et al.2007; Chandler et al.2011; Staubach et al.2013). Enterobacteriaceae were also common (17.0% average relative abundance) with Buttiauxella (8.3% average relative abundance) and Serratia (6.8% average relative abundance) being the most common. Because Buttiauxella was very common in the flies, we collected from grapes, but we could not find convincing evidence for the presence of this genus on grapes in previous studies, we performed a Blast search for a representative sequence. The best hit was a sequence classified as Enterobacter (File S2, Supporting Information) that was isolated from D. suzukii (Vacchini et al.2017). Members of the genus Enterobacter have been found in fermenting grapes (Barata, Malfeito-Ferreira and Loureiro 2012) before. Members of the genus Serratia can act as Drosophila pathogens and occur at high relative abundance in individual samples. This pattern of low abundance or absence in most samples and sharp increase in individual samples (sample orange2, 44.0% rel. abundance) has been associated with Drosophila pathogens (Staubach et al.2013). A similar pattern is visible for Enterococcus in sample orange1 (46.5% rel. abundance compared to 7.0% average relative abundance) and the tomato sample (53.7% rel. abundance). Enterococcus can reach high titers in Drosophila and cause mortality (Cox and Gilmore 2007). A representative sequence of the unclassified gammaproteobacterium (grey in Fig. 2) that can be found in several samples matched perfectly with sequences from uncultured enterobacteria isolated from social corbiculate bees and nectar feeding bats (File S3, Supporting Information). The high sugar content that nectar and rotting fruit have in common might favor the growth of similar bacteria. Orbales, as found to be associated with wild-caught Drosophila by Chandler et al. (2011), were not among the 100 best blast hits. A full list of the 25 most common OTUs and representative sequences can be found in Table S2 (Supporting Information). THE MICROBIOMES OF SIMULATED POOLS FROM THE SAME POPULATION WERE MORE SIMILAR TO EACH OTHER THAN THOSE OF INDIVIDUALS As described above, microbial community composition varied substantially even between individual flies from the same population. In order to elucidate how well this variation was represented by a pooling approach, we analyzed beta-diversity of individual flies and simulated pools using Bray–Curtis community distances (BCD). Bacterial communities from the simulated pools were more similar than the communities of individual flies (Fig. 4A, P < 2 × 10−16, Mann–Whitney U test), suggesting that compositional differences between communities from individuals might even out in pools. This result was confirmed by weighted UniFrac, was less pronounced for unweighted UniFrac (Figs S1 and S2, Supporting Information), but were not confirmed for Jaccard distances (Fig. S3, Supporting Information), suggesting that abundance information plays a role in generating this pattern. Figure 4. View largeDownload slide (A) Pairwise Bray–Curtis distances for individual and pooled samples from the 2013 plum population. Each dot represents the distance between two communities. Distances for all possible pairs in the respective groups are shown. P-values were computed using the Mann–Whitney U test. (B) Same as in A but this time distances within the 2013 plum population is compared to variation between populations sampled across different substrates and locations. Figure 4. View largeDownload slide (A) Pairwise Bray–Curtis distances for individual and pooled samples from the 2013 plum population. Each dot represents the distance between two communities. Distances for all possible pairs in the respective groups are shown. P-values were computed using the Mann–Whitney U test. (B) Same as in A but this time distances within the 2013 plum population is compared to variation between populations sampled across different substrates and locations. POPULATION-SPECIFIC EFFECTS ON MICROBIOME COMPOSITION AS INFERRED FROM COMPARING SIMULATED POOLS FROM THE SAME POPULATION TO BIOLOGICAL POOLS FROM DIFFERENT POPULATIONS In order to assess, if there were population-specific effects on microbiome composition, we compared community distances within the plum 2013 population to distances between populations. Communities from the 2013 plum population, as represented by simulated pools, were more similar to each other than communities from different populations (Fig. 4B, P < 2 × 10−16). This suggested that there were indeed population-specific effects on community composition. This was also supported by PCoA analysis (Fig. 5B); the 2013 plum population pools grouped together along PCo1. Figure 5. View largeDownload slide Principal coordinates analysis of Bray–Curtis distances. (A) Of pools and individuals from the plum population and (B) for pools from the 2013 plum population and populations from different substrates and locations. Figure 5. View largeDownload slide Principal coordinates analysis of Bray–Curtis distances. (A) Of pools and individuals from the plum population and (B) for pools from the 2013 plum population and populations from different substrates and locations. SIMULATED POOLS REPRESENTED ONLY A PART OF THE HOST POPULATION MICROBIOME In order to explore potential factors shaping fly-associated bacterial communities, a Principal Coordinate Analysis (PCoA) was carried out using BCD (Fig. 5). The PCoA suggested that the individual fly population-specific samples formed two groups (Fig. 5A). This grouping was supported by hierarchical clustering based on BCD (Fig. 6A), Jaccard distances (Fig. S4, Supporting Information) and UnifFrac distances (Figs S5 and S6, Supporting Information). Of note, the pool collected at the same time represented only one part of the population (dark blue in Figs 5A and 6A), while the pool from the year before (light blue in Figs 5A and 6A) represented the other. The notion that pools were representative for only a part of the host population was confirmed by the clustering of simulated pools (Fig. 6B); the simulated pools also formed two clusters, and again one cluster contained the pool from 2013, while the other cluster contained the pool from 2012. Figure 6. View largeDownload slide (A) Hierarchical clustering based on Bray–Curtis dissimilarities for individuals and empirical pools from 2013 and 2012. Empirical pools are in light blue (2012) and dark blue (2013). (B) for pools (empirical and simulated). Values at branches are AU (Approximately Unbiased) bootstrap support (Suzuki and Shimodaira 2015). Figure 6. View largeDownload slide (A) Hierarchical clustering based on Bray–Curtis dissimilarities for individuals and empirical pools from 2013 and 2012. Empirical pools are in light blue (2012) and dark blue (2013). (B) for pools (empirical and simulated). Values at branches are AU (Approximately Unbiased) bootstrap support (Suzuki and Shimodaira 2015). In order to assess if the variation between individual microbiomes was driving the variation between pools, we identified the taxa driving the clustering with the SIMPER method. We found that 9 out of 10 OTUs with the largest effect on the clustering were identical between clusters of individuals and simulated pools (Table S3A and S3B, Supporting Information). The largest contributor to the clustering of both simulated pools and individuals was OTU1, classified as Saccharibacter, explaining 24% and 29% of the dissimilarity between clusters, respectively. This suggested that the microbiome composition of pools was strongly affected by inter-individual variation. Because we noted that microbiomes of simulated pools were on average more similar than those of individuals (Fig. 4A), we thought that further increasing pool size would make pools more representative of the whole population. In order to test this, we simulated 30 pools with a pool size increased to 30 individuals. Although the larger pools moved closer to the center of the PCoA (Fig. S7, Supporting Information), we still recovered two distinct clusters (Fig. S8, Supporting Information). DISCUSSION In this study, we described bacterial community variation between individual wild-caught D. melanogaster. By analyzing a sample of 32 individual flies as well as pools from the same population, substrate, and at the same time, we addressed the questions (i) how well pools represent the microbial communities of a host population, (ii) whether there is evidence for factors that shape within-population variation when substrate is controlled for and (iii) compare within-population variation of bacterial communities to between-population variation. As before (Staubach et al.2013), we used entire flies for our study. This is important because fly pathogens can reach high titers in the hemolymph (Cox and Gilmore 2007; Galac and Lazzaro 2011) and might be overlooked by focusing on the gut. Although bacteria on the fly surface contribute only ∼10% to the total bacterial load of flies (Ren et al.2007), and hence their effect on bacterial community composition should be minor, they could still play a role in inoculating fruit and shaping the microbial environment of D. melanogaster. Concordant with many other studies (Cox and Gilmore 2007; Chandler et al.2011; Staubach et al.2013; Wong, Chaston and Douglas 2013), acetic acid bacteria dominated the bacterial communities. Also concordant with earlier studies on wild-caught flies, enterobacteria occur at sometimes high relative abundance in some samples. This pattern has been connected to pathogens before and could indicate systemic infections (Staubach et al.2013). VARIATION OF DROSOPHILA-ASSOCIATED BACTERIAL MICROBIOMES WITHIN AND BETWEEN POPULATIONS Previous studies have found that Drosophila-associated microbiomes differ between populations. These differences tend to correlate with diet (Chandler et al.2011; Staubach et al.2013). However, the assessment of within-population variation is limited in these studies. In the light of high variation of Drosophila-associated microbiomes within host populations (Martinson et al.2017; Martinson, Douglas and Jaenike 2017), we assessed whether evidence for microbiome differences between populations of D. melanogaster can still be found. We evaluated within-population variation by analyzing community profiles of individuals. In order to make these individual data comparable to pools, we simulated pools of the same size combining the data from individuals. This allowed us to compare the variation of microbiomes within the plum 2013 population to those of pooled samples from other populations that were sampled across space, time and diets. We chose this sampling scheme to include a comprehensive set of factors that could generate differences between populations. This is reasonable because in order to assess if differences between populations can be detected, there need to be differences between populations in the first place. The comparison revealed that both, alpha- and beta-diversity showed population-specific effects; taxon richness was higher in the 2013 plum population than in the other populations. At the same time, the microbiomes of samples from the same population were more similar to each other than those of samples from different populations, corroborating earlier studies that found differences between populations. While we did not aim at disentangling the effects of diet, location and time with our study setup, Fig. 5B was concordant with previous results; communities from the same diet seemed to preferentially cluster together. However, the grape samples that clustered together were also collected in proximity, as were the sample apple1 and the peach sample, suggesting that location specific effects or dispersal could also play a role for community composition. POOLING AS A TOOL TO ASSESS HOST POPULATION MICROBIOMES The comparison of alpha-diversity between individual and pooled fly communities from the same population revealed that taxon richness and diversity were higher in pools. This indicates that pools offer a more comprehensive picture of the microbial taxa that can be found in a host population. Shannon's diversity in individuals (H = 1.87 +/− 0.63), simulated pools (H = 2.16 +/−0.27) and pools across locations and substrates (H = 1.80 +/− 0.77) was comparable to that found associated with wild-caught flies in Staubach et al. (2013) (H = 1.79 +/− 0.44) and hence to that from Chandler et al. (2011) and Cox and Gilmore (2007) who also investigated bacterial communities of wild-caught flies. Please see Staubach et al. (2013) for extensive diversity comparisons between studies. A comparison of community composition (beta-diversity) revealed that the microbiomes of simulated pools were on average more similar to each other than those of individuals. This effect was more pronounced for methods that take into account the abundance of microbes (BCD and weighted UniFrac). A reasonable explanation for this pattern is that differences, especially in abundance, between individuals even out in pools. As a consequence pools might provide more of an average community composition of a host population. Despite pooled samples falling more in line, the clustering into two groups that was observed between individuals from the 2013 plum population was recovered by simulated, as well as empirical pools from 2013 and 2012. This was even true when we increased the simulated pool size to 30 individuals. This suggests that pools tend to represent only a part of the host population microbiome and that variation between individuals still has an effect on the microbiome of pools. This effect is expected to be larger when bacterial loads vary substantially between individuals. For example, when an individual with a systemic infection carrying a high bacterial load ends up in a pool, this could drive the community of the whole pool. The bacterial loads of wild-caught flies in Elya et al. (2016) vary up to 1000-fold. As the distribution of loads from this publication was the basis for our simulations our results might be similarly affected. The fact that the 2012 pool from the same location and substrate falls into the distribution of simulated pools suggests that the communities at this location might be rather constant at a one-year time interval. EVIDENCE FOR FACTORS THAT SHAPE WITHIN-POPULATION VARIATION UNDER CONTROLLED SUBSTRATE CONDITIONS Surprisingly, samples collected from the same location and substrate at the same time clustered into two groups. The clustering correlated with the most abundant OTU (OTU1, supposedly Saccharibacter). We can only speculate here how the difference in abundance of OTU1 or the clustering was generated. Assuming that flies continuously exchange microbes with their environment and that they replenish their gut microbiota through uptake of environmental bacteria (Blum et al.2013), we can think of two scenarios that could generate this pattern. In the first scenario, the plum substrate falls into two different categories from a microbiome perspective. These categories could be states of decay or states that result from the dynamic interplay of microbial metabolites (Fischer et al.2017). In the second scenario, we might have a cohort of flies that entered the population only recently and brought microbes with them from the previous substrate. We cannot disentangle these options with the current data. Similarly, because D. melanogaster communities change with age (Brummel et al.2004; Ren et al.2007; Clark et al.2015), distinct age cohorts in the population could cause distinct microbial communities. Furthermore, because flies shape their associated microbial communities (Wong et al.2015) and fly host genetic make-up affects the community composition at least under laboratory conditions (Dobson et al.2015; Unckless, Rottschaefer and Lazzaro 2015; Chaston et al.2016), genetic differences in the host population could play a role in generating the observed pattern. Finally, stochastic effects may not only be important for the assembly of the Drosophila microbiome in the laboratory (Obadia et al.2017), but could also affect natural communities. Together with priority effects this could lead to historical contingency of the fly-associated communities and to a limited number of distinct alternative states of microbiomes (Fukami 2015). Our results are in agreement with a recent study on fungus feeding Drosophila (Martinson, Douglas and Jaenike 2017). Similar to our study the authors controlled the diet by collecting flies from the same substrate. They observed that most of the variation of microbial communities can be found between individuals with only a very small part explained by host species, sampling location or time of collection. Interestingly, as in our study, the samples clustered into groups. CONCLUSION Pooling of samples from host species to get a population representative sample of the microbiome is common practice. We showed that pooling can offer a more comprehensive picture of the microbial taxa associated with a host population. At the same time, microbiomes of pools should be interpreted with care because they can still be influenced by individual variation and might only represent a part of the population. The two clusters we found in our sample of individuals collected from the same substrate, at the same time and location, suggest that there are important factors that shape natural D. melanogaster microbiomes that we do not understand yet. Because understanding natural communities is what matters for understanding D. melanogaster evolution, more research is needed to better understand the factors that shape natural D. melanogaster-associated microbial communities. DATA AVAILABILITY Raw data are available at NCBI SRA under project number PRJNA391254. SUPPLEMENTARY DATA Supplementary data are available at FEMSLE online. ACKNOWLEDGEMENTS We thank Dmitri A Petrov (Stanford University) in whose lab the data were collected. We thank Alan Bergland and Heather Machado for support in fly sampling. We thank Kabir Peay for providing barcoded primers. We thank Carolyn Elya for sharing the bacterial load data from wild flies that our simulation was based on. We thank Till Bayer, two reviewers and Sybil Staubach for helpful comments on the manuscript. We thank Jiaona Zou for illustrating fruit and flies. FUNDING This work was supported by the German Research Foundation (DFG) with a Forschungsstipendium to FS (STA 1154/1-1) and the Research Innovation Fund of the Freiburg University. Conflict of interest. None declared. REFERENCES Barata A, Malfeito-Ferreira M, Loureiro V. The microbial ecology of wine grape berries. Int J Food Microbiol 2012; 153: 243– 59. Google Scholar CrossRef Search ADS PubMed Blum JE, Fischer CN, Miles Jet al. Frequent replenishment sustains the beneficial microbiome of Drosophila melanogaster. mBio 2013; 4: e00860– 13. Google Scholar CrossRef Search ADS PubMed Brummel T, Ching A, Seroude L et al. Drosophila lifespan enhancement by exogenous bacteria. Proc Natl Acad Sci 2004; 101: 12974– 9. Google Scholar CrossRef Search ADS PubMed Caporaso JG, Lauber CL, Walters WAet al. Colloquium paper: gobal patterns of 16S rRNA diversity at a depth of millions of sequences per sample. Proc Natl Acad Sci 2011; 108: 4516– 22. Google Scholar CrossRef Search ADS PubMed Chandler JA, Lang JM, Bhatnagar Set al. Bacterial communities of diverse drosophila species: ecological context of a host–microbe model system. PLoS Genet 2011; 7: e1002272. Google Scholar CrossRef Search ADS PubMed Chao A. Nonparametric estimation of the number of classes in a population. Scand J Stat 1984; 11: 265– 70. Chaston JM, Dobson AJ, Newell PDet al. Host genetic control of the microbiota mediates the Drosophila nutritional phenotype. Appl Environ Microbiol 2016; 82: 671– 9. Google Scholar CrossRef Search ADS Chen J. GUniFrac: Generalized UniFrac distances. 2012. R package version 1.0. http://CRAN.R-project.org/package=GUniFrac. Clark RI, Salazar A, Yamada Ret al. Distinct shifts in microbiota composition during Drosophila aging impair intestinal function and drive mortality. Cell Rep 2015; 12: 1656– 67. Google Scholar CrossRef Search ADS PubMed Consortium THMP. Structure, function and diversity of the healthy human microbiome. Nature 2012; 486: 207– 14. CrossRef Search ADS PubMed Corby-Harris V, Pontaroli AC, Shimkets LJet al. Geographical distribution and diversity of bacteria associated with natural populations of Drosophila melanogaster. Appl Environ Microbiol 2007; 73: 3470– 9. Google Scholar CrossRef Search ADS PubMed Cox CR, Gilmore MS. Native microbial colonization of Drosophila melanogaster and its use as a model of Enterococcus faecalis pathogenesis. Infect Immun 2007; 75: 1565– 76. Google Scholar CrossRef Search ADS PubMed Dobson AJ, Chaston JM, Newell PDet al. Host genetic determinants of microbiota-dependent nutrition revealed by genome-wide analysis of Drosophila melanogaster. Nat Commun 2015; 6: 6312. Google Scholar CrossRef Search ADS PubMed Elya C, Zhang V, Ludington WBet al. Stable host gene expression in the gut of adult drosophila melanogaster with different bacterial mono-associations. PLoS One 2016; 11: e0167357. Google Scholar CrossRef Search ADS PubMed Erkosar B, Storelli G, Defaye Aet al. Host-intestinal microbiota mutualism: “learning on the fly”. Cell Host Microbe 2013; 13: 8– 14. Google Scholar CrossRef Search ADS PubMed Fischer CN, Trautman EP, Crawford JMet al. Metabolite exchange between microbiome members produces compounds that influence Drosophila behavior. eLife 2017; 6: e18855. Google Scholar PubMed Fukami T. Historical contingency in community assembly: integrating niches, species pools, and priority effects. Annu Rev Ecol Evol Syst 2015; 46: 1– 23. Google Scholar CrossRef Search ADS Galac MR, Lazzaro BP. Comparative pathology of bacteria in the genus Providencia to a natural host, Drosophila melanogaster, Microbes Infect 2011; 13: 673– 83. Google Scholar CrossRef Search ADS PubMed Goslee S, Urban D. The ecodist package for dissimilarity-based analysis of ecological data. J Stat Soft 2007; 22: 1– 19. Google Scholar CrossRef Search ADS Linderman JA, Chambers MC, Gupta ASet al. Infection-related declines in chill coma recovery and negative geotaxis in Drosophila melanogaster. PLoS One 2012; 7: e41907. Google Scholar CrossRef Search ADS PubMed Martinson VG, Carpinteyro-Ponce J, Moran NAet al. A distinctive and host-restricted gut microbiota in populations of a cactophilic Drosophila species. Appl Environ Microbiol 2017; 83: e01551– 17. Google Scholar CrossRef Search ADS PubMed Martinson VG, Douglas AE, Jaenike J. Community structure of the gut microbiota in sympatric species of wild Drosophila. Ecol Lett 2017; 20: 629– 39. Google Scholar CrossRef Search ADS PubMed Obadia B, Güvener ZT, Zhang Vet al. Probabilistic invasion underlies natural gut microbiome stability. Curr Biol 2017; 27: 1999– 2006.e8. Google Scholar CrossRef Search ADS PubMed Oksanen J, Blanchet FG, Friendly Met al. Vegan: Community Ecology Package. 2016. R package version 2.4-0. http://CRAN.R-project.org/package=vegan. Pruesse E, Quast C, Knittel Ket al. SILVA: a comprehensive online resource for quality checked and aligned ribosomal RNA sequence data compatible with ARB. Nucleic Acids Res 2007; 35: 7188– 96. Google Scholar CrossRef Search ADS PubMed R Core Team. R: A language and environment for statistical computing. 2017. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/. Ren C, Webster P, Finkel SEet al. Increased internal and external bacterial load during Drosophila aging without life-span trade-off. Cell Metab 2007; 6: 144– 52. Google Scholar CrossRef Search ADS PubMed Schloss PD, Westcott SL, Ryabin Tet al. Introducing mothur: open-source, platform-independent, community-supported software for describing and comparing microbial communities. Appl Environ Microbiol 2009; 75: 7537– 41. Google Scholar CrossRef Search ADS PubMed Shannon CE, Weaver W, Blahut RE et al. The Mathematical Theory of Communication , Vol. 117. Urbana: University of Illinois press, 1949. Shin SC, Kim S-H, You H et al. Drosophila microbiome modulates host developmental and metabolic homeostasis via insulin signaling. Science 2011; 334: 670– 4. Google Scholar CrossRef Search ADS PubMed Staubach F, Baines JF, Künzel Set al. Host species and environmental effects on bacterial communities associated with Drosophila in the laboratory and in the natural environment. PLoS One 2013; 8: e70749. Google Scholar CrossRef Search ADS PubMed Storelli G, Defaye A, Erkosar B et al. Lactobacillus plantarum promotes Drosophila systemic growth by modulating hormonal signals through TOR-dependent nutrient sensing. Cell Metab 2011; 14: 403– 14. Google Scholar CrossRef Search ADS PubMed Suzuki R, Shimodaira H. Pvclust: hierarchical clustering with P-values via multiscale bootstrap resampling. 2015. R package version 2.0-0. http://CRAN.R-project.org/package=pvclust. Téfit MA, Leulier F. Lactobacillus plantarum favors the early emergence of fit and fertile adult Drosophila upon chronic undernutrition. J Exp Biol 2017; 220: 900– 7. Google Scholar PubMed Unckless RL, Rottschaefer SM, Lazzaro BP. The complex contributions of genetics and nutrition to immunity in Drosophila melanogaster. PLoS Genet 2015; 11: e1005030. Google Scholar CrossRef Search ADS PubMed Vacchini V, Gonella E, Crotti Eet al. Bacterial diversity shift determined by different diets in the gut of the spotted wing fly Drosophila suzukii is primarily reflected on acetic acid bacteria. Environ Microbiol Rep 2017; 9: 91– 103. Google Scholar CrossRef Search ADS PubMed Wong AC-N, Chaston JM, Douglas AE. The inconstant gut microbiota of Drosophila species revealed by 16S rRNA gene analysis. ISME J 2013; 7: 1922– 32. Google Scholar CrossRef Search ADS PubMed Wong AC-N, Luo Y, Jing Xet al. The host as the driver of the microbiota in the gut and external environment of Drosophila melanogaster. Appl Environ Microbiol 2015; 81: 6232– 40. Google Scholar CrossRef Search ADS PubMed © FEMS 2018. All rights reserved. For permissions, please e-mail: firstname.lastname@example.org
FEMS Microbiology Letters – Oxford University Press
Published: Mar 1, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera