ABSTRACT
The diversity and structure of the intestinal microbial community has a strong influence on life history. To understand how hosts and microbes interact, model organisms with comparatively simple microbial communities, such as the fruit fly (Drosophila melanogaster), offer key advantages. However, studies of the Drosophila microbiome are limited to a single point in time, because flies are typically sacrificed for DNA extraction. In order to test whether noninvasive approaches, such as sampling of fly feces, could be a means to assess fly-associated communities over time on the same cohort of flies, we compared the microbial communities of fly feces, dissected fly intestines, and whole flies across three different Drosophila strains. Bacterial species identified in either whole flies or isolated intestines were reproducibly found in feces samples. Although the bacterial communities of feces and intestinal samples were not identical, they shared similarities and obviously the same origin. In contrast to material from whole flies and intestines, feces samples were not compromised by Wolbachia spp. infections, which are widespread in laboratory and wild strains. In a proof-of-principle experiment, we showed that simple nutritional interventions, such as a high-fat diet or short-term starvation, had drastic and long-lasting effects on the micobiome. Thus, the analysis of feces can supplement the toolbox for microbiome studies in Drosophila, unleashing the full potential of such studies in time course experiments where multiple samples from single populations are obtained during aging, development, or experimental manipulations.
INTRODUCTION
The microbial community of the metazoan intestine contributes substantially to the host's nutrition and energy balance (1). Dysregulation of the host-microbiota interaction is associated with various disease states, including chronic inflammation, obesity, or even cancer (2–4). As the complex interplay between host and microbiota is only fragmentarily understood, simple models, such as the fruit fly Drosophila melanogaster, are of particular interest. The fly combines a relatively simple microbial community with an unchallenged armamentarium of methods allowing manipulation of the host (5–7). Recent deep-sequencing approaches using 454 sequencing confirmed the general finding that the bacterial community of lab-reared flies is dominated by slightly more than a few operational taxonomic units (OTUs) (8). In Drosophila, the microbiota has been shown to modulate different aspects of intestinal homeostasis, including stem cell activity and epithelial immunity (9–11). Moreover, the intestinal microbiota influences life span and growth rate (12–14).
These aforementioned studies aiming to characterize the fly microbiome relied on isolation of bacterial material from whole flies or manually dissected intestines. Both approaches are invasive, and the flies must be sacrificed. Noninvasive approaches would offer the advantage that flies could be analyzed throughout their life span or before and after an experimental intervention.
The major aim of this study was to introduce a noninvasive sampling method that enables quantitative analyses of fly microbial communities and to compare this approach with those currently in use in terms of reproducibility and reliability. To this end, we evaluated the use of whole flies, isolated intestines, and fecal samples with both quantitative PCR (qPCR) and 454 sequencing of the bacterial 16S rRNA gene. Our results indicated that fecal sampling is comparable to the other two previously established means of sampling, with the additional benefits of being noninvasive and free from the influence of intracellular bacteria (e.g., Wolbachia spp.).
MATERIALS AND METHODS
Drosophila strains and sample collection.We performed our studies on three commonly used D. melanogaster strains: Oregon-R, Canton-S, and w1118 (Bloomington Stock Center, Bloomington, IN) that were kept at 25°C in a benchtop incubator on standard Drosophila medium (6.25% cornmeal, 6.25% yeast, 2% glucose, 3% sugar beet molasses, and 1% agar-agar, as well as 3% nipagin and 1% propionic acid as preservatives), as described earlier (15, 16). To compare sampling strategies for the characterization of microbial communities in the Drosophila intestine, we isolated DNA from (i) whole flies, (ii) manually dissected intestines (from about 20 animals per sample; only from the midgut, excluding the hindgut and the Malpighian tubules, in order to reproducibly prepare the same intestinal region), and (iii) fecal samples obtained from small cohorts of flies. Young mated females were used for whole-fly genomic DNA (gDNA) isolation, gut dissection, or feces collection. For isolation of feces, cohorts of 30 to 50 flies were transferred to a new medium vial for 24 h, and the fecal spots were removed from only the walls of the vial, to avoid contamination with bacteria growing on the fly medium, by using a sterile buffer-saturated swab (Greiner Bio One, Germany). Subsequently, the swab was cut, and genomic DNA isolation was done using the PowerSoil DNA isolation kit (MoBio Laboratories, Carlsbad, CA).
PCR amplifications.These different DNAs were used as the template for qPCR analyses with bacterial species-specific primers and in-depth analyses via 454 sequencing of the 16S rRNA gene (17). Relative quantification of the amounts of bacterial DNA that was isolated from the three sources (whole flies, intestines, and feces) was performed with general bacterial primers (S-*-univ-0027-a-S-20 and S-*-univ-0338-a-A-19). DNA from three biological replicates per sample isolated from the flies that were cultivated in independent vials was used as the qPCR templates. Two qPCRs were carried out per biological replicate (technical duplicates).
To quantify relative abundances of bacterial species known to be part of the Drosophila microbiome, we performed qPCR analyses with species-specific oligonucleotide primer pairs for the following bacterial species: Acetobacter tropicalis (8), Commensalibacter intestini (S-S-C-intest-1018-a-S-19 and S-S-C-intest-1130-a-A-19), Lactobacillus brevis (18), Lactobacillus plantarum (8), and Guconobacter sp. (19). To obtain comparisons of the relative abundance of each species, we corrected for primer efficiencies and determined the threshold cycle (CT) values relative to those for total bacteria in each sample, i.e., we averaged the CT values between technical duplicates and then applied the following equation: ΔCT = CT(V2 primer set) − CT(sample bacteria-specific primer sets). The median ΔCT of the three biological replicates was then used to transform this relative logarithmic measurement of bacterial abundance to a linear scale.
454 procedures.Bacterial communities of w1118 and Canton-S were analyzed in depth based on a 16S rRNA gene pyrosequencing approach using a 311-nucleotide sequence flanking the hypervariable V1 and V2 regions, amplified with general bacterial primers (S-*-univ-0027-a-S-20 and S-*-univ-0338-a-A-19). The 454 procedures to produce libraries and to obtain sequence information as well as data analyses were performed as described earlier (17). For each isolation method, we used 5 independent biological replicates and generated a total of ∼100,000 sequence reads after quality filtering. The 454 reads were sorted into groups according to their MID tags, by using the MOTHUR program v1.23.1 (20). During the process, tags and primer sequences were removed. Only sequences perfectly matching the MIDs and the bacterial primers were kept. The resulting sequences were quality filtered under the following conditions: minimum average quality of 35 in each 50-bp window, minimum length of 260 bp, and homopolymers no longer than 8 bp. Filtered sequences were aligned to the SILVA reference database (21) by using the MOTHUR-implemented kmer algorithm with standard settings. Sequences not aligning in the expected region were removed. Passing sequences were filtered for sequencing errors by using the MOTHUR pre.cluster command. Sequences were classified into bacterial taxa with the classify.seqs command in MOTHUR, based on the SILVA reference database and taxonomy. Results were plotted using the R statistics package v.2.13.1 (22). Clustering of sequences into OTUs was performed by using MOTHUR with the average neighbor algorithm.
RESULTS
We performed our studies on three commonly used D. melanogaster strains: Oregon-R, Canton-S, and w1118. To compare sampling strategies for the characterization of microbial communities in the Drosophila intestine, we isolated total DNA from (i) whole flies, (ii) manually dissected intestines, and (iii) feces samples obtained from small cohorts of flies. DNA isolated from these three sources was used as the template for qPCR analyses with bacteria species-specific primers. Manually isolated intestines (midgut region) contained between 1 and 7% of the amount of bacterial DNA typically present in whole flies, but it has to be kept in mind that the hindgut was excluded. In feces samples, this percentage dropped further, to 0.01 to 0.5%, depending on the strain used (Fig. 1). All values for each strain (whole fly versus intestine, whole fly versus feces, and intestine versus feces) were significantly different from each other (P < 0.001). We aimed to determine whether the bacterial species known to be abundant in the intestinal community could be quantified by using DNA isolated from these different sources. To this end, we performed qPCR analyses with species-specific oligonucleotide primer pairs for Acetobacter tropicalis (Fig. 2A), Commensalibacter intestini (Fig. 2B), Lactobacillus brevis (Fig. 2C), Lactobacillus plantarum (Fig. 2D), and Guconobacter sp. (Fig. 2E), with the three different Drosophila strains and the three sources of bacterial DNA.
Relative quantification of the amount of bacterial DNA that can be isolated from different sources varies. Adults from three different strains (Canton-S, Oregon-R, and w1118) were used, and DNA isolated from whole flies (black bars), isolated intestines (gray bars), and feces samples (white bars) served as the template for the quantitative analyses. qPCR was performed with bacteria-specific primers targeting the 16S rRNA gene. The median amount of bacterial DNA detectable in material from a single whole fly was set as 100% for each strain, and the equivalents found in 1 isolated intestine (midgut only), as well as in the fly feces produced by one fly during a 24-h period, were calculated relative to the whole fly data. Bars represent the medians of three independent experiments ± standard deviations (within all Drosophila strains, values between flies and intestine, fly and feces, as well as between intestine and feces were significantly different [P < 0.001]).
(A to E) Relative abundances of selected bacterial species detected in material isolated from whole flies, isolated intestines, and feces samples of small cohorts of flies. Adults from three different strains (Canton-S, Oregon-R, and w1118) were used, and DNA isolated from whole flies (black bars), manually dissected intestines (gray bars), and feces samples (white bars) served as the templates for quantitative analyses. Results for quantitation of Acetobacter tropicalis (A), Commensalibacter intestini (B), Lactobacillus brevis (C), Lactobacillus plantarum (D), and Gluconobacter sp. (E) are shown. Bars represent the medians of three independent experiments ± the SEM. All values obtained from flies, intestines, and feces differed significantly among each other (P < 0.05) for each Drosophila strain, except between intestines and feces for L. brevis (Oregon R and w1118), L. plantarum (Canton S), C. intestini (w1118), and Gluconobacter sp. (Canton S). (F and G) Principal coordinate analysis (PCoA) of Bray-Curtis dissimilarities between samples, including Wolbachia-derived sequences (F) and after their removal (G).
The presence of individual bacterial species was reproducibly detected with material derived from all isolation methods, including isolation from feces (Fig. 1B to F). The DNA isolated from feces usually gave very similar relative abundances compared with material from dissected intestines. Most values obtained for flies, intestines, and feces differed significantly among each other (P < 0.05), except between intestines and feces for L. brevis (Oregon R and w1118), L. plantarum (Canton S), C. intestini (w1118), and Gluconobacter sp. (Canton S). Interestingly, for one Drosophila strain (w1118), we observed strongly reduced relative abundances in material derived from whole flies (this is especially clear in Fig. 2A and C and 1D and E). This surprisingly small relative concentration may be attributed to the presence of Wolbachia observed in the w1118 strain.
In order to analyze the bacterial communities in more detail, we choose the w1118 and Canton-S strains for an in-depth 16S rRNA gene pyrosequencing analysis. Whole-fly and intestinal samples of the w1118 strain revealed a dominance of Wolbachia sequences (up to 97%), with which infection is known to be present in more than 60% of wild-caught flies (23). In feces samples of the w1118 strain, we found only a negligible fraction of reads that could be assigned to Wolbachia. The Canton-S strain was not infected with this bacterium, as Wolbachia-derived sequences could not be identified. After removal of Wolbachia-derived sequences, the majority of remaining sequences could be assigned to the Acetobacter, Acinetobacter, Sphingomonadacea, Bacteroides, and the enteric bacterial clades. To analyze the relatedness of the communities obtained via the three different isolation methods, we performed a principal component analysis based on Bray-Curtis distances, including either all sequences (Fig. 2F) or after removal of all Wolbachia sequences (Fig. 2G). In Fig. 2F, all samples that contained Wolbachia clustered together on the left side. Pco1 differentiated the samples containing Wolbachia (w1118 whole fly and gut) from those that did not (P < 0.01, Welch's t test). Remarkably, the bacterial communities isolated from feces were very similar between the two strains, irrespective of Wolbachia infection status. In addition, we found a signature of all major bacterial genera known to be highly abundant in the fly's microbiota. Nevertheless, it has to be mentioned that feces samples differed from intestinal and whole-fly samples, but this differentiation was also observed between whole flies and dissected intestines (Fig. 2F and G).
To demonstrate the advantages of the noninvasive approach, we performed time series experiments with cohorts of flies (adult w1118) that were challenged with nutritional interventions (24 h of starvation or 3 days on a high-fat diet, i.e., normal medium supplemented with 15% coconut oil). The microbial community was analyzed before, directly after the intervention, and after an additional 2 days on normal medium, by collecting feces samples (Fig. 3). Total bacterial content (measured with general bacterial primers [Fig. 3A]), as well as the relative abundances of two highly relevant bacterial species (L. plantarum [Fig. 3B] and C. intestini [Fig. 3[C]) were tested by qPCR, and the results correlated with the number of fecal spots produced (expressed as the percent relative to the starting concentration on day zero). Values reported are means of 3 independent biological replicates (± the standard error of the means [SEM]). Under control conditions (normal medium [NM]), the total relative number of bacteria varied only marginally throughout the experiment (Fig. 3A). In contrast, the relative abundance of bacteria increased following a high-fat diet (HF), an effect that lasted up to the end of the experiment (significantly different from matching normal medium controls, P < 0.05). Starvation on the other hand induced only minor changes in the relative abundance of bacteria, with a drop in bacterial numbers at the end of the experiment (Fig. 3A). Regarding L. plantarum, the situation was completely changed (Fig. 3B). As for all bacteria, flies held on NM showed only marginal changes in L. plantarum abundance during the experimental time (Fig. 3B). Moreover, the high-fat diet only slightly increased the abundance of this bacterium, and starvation induced a boost in bacterial abundance (not statistically significant [Fig. 3B]). The abundance of C. intestini (Fig. 3C) followed, on the other hand, a kinetics that was very similar to that observed for all bacteria. In this case, the high-fat diet induced a massive and long-lasting increase in the abundance of this bacterial species (significantly different from matching normal medium controls, P < 0.05 [Fig. 3C]).
Time course experiments of microbial communities from the same cohorts of flies, following either a high-fat diet or starvation. Abundances of all bacteria (A), L. plantarum (B), or C. intestini (C) were measured by qPCR and are expressed as a percentage relative to the starting amount (time point 0 days). The first feces samples were taken on control medium (time point zero). One set of cohorts was held on normal medium throughout the experiment (NM; black dots). After treatment (3 days on high-fat [HF] medium or 24 h of starvation [St]), feces samples were taken from the same cohorts of flies (day 3). Subsequently, the same flies were transferred to normal medium for 2 days, and the next feces sample was taken on day 6. Each line represents the mean data obtained from 3 cohorts of flies throughout the experiment. Controls were set to 100%. Values are mean values from three independent experiments (± SEM). *, P < 0.05.
DISCUSSION
The major aim of this study was to establish a noninvasive method to analyze the dynamics of the fruit fly's microbiome. Use of feces samples from small cohorts of flies enabled us to reprobe the same cohort several times. Our study demonstrates that use of feces is well-suited to analysis of microbiome dynamics in response to different types of experimental interventions. All major bacterial species known to be relevant members of the fly's microbiome (8, 24) could also be identified in feces samples, and the communities appeared to remain constant over time. Until now, Drosophila microbiome studies have been based on whole flies or isolated intestines (8, 11, 24, 25) as the starting material. Use of whole flies is obviously the most convenient means of probe generation, but it has several disadvantages. Adult flies offer, in addition to the intestine, other niches for bacteria, including the fly's surface. We found only a small portion of fly-associated bacteria were in the midgut. Together with the bacteria of the hindgut, which were excluded from our analysis, the total proportion of the intestinal bacteria is always below 50%. On the other hand, surface bacteria account for less than 50% of the entire bacterial population (26). Use of whole flies as a starting material has the highest risk for the samples to be compromised by Wolbachia infection, which are widespread among laboratory strains and wild-caught flies (27). These intracellular bacteria are found in highest abundances in reproductive organs. Moreover, in infected animals, bacteria are also found in somatic tissues, including the intestine (28), whereas feces samples appear not to be compromised.
Feces samples have served as representatives for the intestinal microbiome in a great number of studies, including the majority of those dealing with the human microbiome. Although functioning as a proxy for the intestinal microbiome, the bacterial composition found in feces is never identical to that of the intestine. In addition, it has to be kept in mind that microbial communities vary along the intestinal tract (29); thus, the composition found in feces represents a simplified readout of the various region-specific communities of the intestine, with highest congruence to that found in the hindgut. In our study, the dissected intestines were part of the midgut and excluded the hindgut, so as to reproducibly remove the same part of the intestine in all experiments. Thus, our method could not be expected to identify identical bacterial communities if dissected midguts and feces samples were compared. Although starting with different relative abundances, changes in the intestinal microbiome should also be found in the feces microbiome. Although fecal samples of humans or animal models may be affected by differential bacterial growth prior to DNA extraction, a recent analysis revealed only minor effects of different storage conditions (30). In Drosophila, fecal spots are extremely small, which is seemingly a drawback but which may turn out to be advantageous. As a consequence of the small samples, fly fecal spots dry out quickly. This dryness should permit bacterial growth within fecal spots, which thus conserves the bacterial community present at the time of spot deposition.
The greatest advantage of our noninvasive approach is the possibility to follow the same cohort of flies over time. Recent studies revealed a striking heterogeneity of microbiomes between Drosophila strains, which negated the hypothesis that a core microbiome exists in the fly (31). Thus, the focus of studies using feces as the starting material cannot be on a description of the microbiome per se, but on the dynamics observed following different types of intervention. We observed a dramatic increase in bacterial abundances following feeding a high-fat diet, an effect that lasted longer. At the bacterial species level, we found different types of responses to this intervention. Effects of different diets on the microbiome have often been reported in humans, but the underlying mechanisms and the pathological outcomes remain unclear (32, 33).
The noninvasive approach will allow time course studies on the same cohort of flies, including studies observing age-dependent changes, responses to nutritional regimens, or even well-defined local interventions, based on, e.g., the GeneSwitch or the Gal4/Gal80 systems (34, 35). Thus, in summary, use of feces to characterize the fly microbiota will complement the current toolbox for fly microbiome studies, because this noninvasive sampling method is (i) easily applicable, (ii) allows for time series experiments, and (iii) is insensitive to Wolbachia and other intracellular bacteria.
ACKNOWLEDGMENTS
This work was supported by the DFG (SFB/TR 22, Teilprojekt A7), the Cluster of Excellence Inflammation@Interfaces, and DFG Forschungsstipendium STA1154/1-1 (F.S.).
We thank Britta Laubenstein and Katja Cloppenborg-Schmidt for excellent technical assistance.
FOOTNOTES
- Received 11 June 2013.
- Accepted 30 August 2013.
- Accepted manuscript posted online 6 September 2013.
- Copyright © 2013, American Society for Microbiology. All Rights Reserved.
