Microbiome structure of endemic Hawaiian Drosophila is shaped more by habitat than host identity

Abstract

The Hawaiian Drosophila radiation exemplifies rapid adaptation and species diversification. Many factors have been attributed to these phenomena, including allopatry, sexual selection, and ecological specialization. In recent years, the microbiome has come to the forefront as an important driver of adaptation that is capable of facilitating host survivorship, enhancing resilience to local environmental challenges, and enabling the use of different dietary resources. To determine the factors that contribute to microbiome community variation in natural populations, we conducted a survey of bacterial and fungal communities from over 500 wild flies collected from across six islands of the Hawaiian archipelago. These samples represent a breadth of host plant specializations, habitats, lifestyles, and endemicity. Our findings reveal that microbiome variation is largely driven by abiotic factors including elevation, temperature, rainfall, and evapotranspiration, but is not strongly constrained by phylogenetic relatedness. Identical species inhabiting three separate locations exhibited different microbiomes. By contrast, distantly related species inhabiting the same site had more similar microbiomes. The microbiomes of endemic species also differ from recently introduced invasive Drosophila in terms of diversity, composition, and predicted function. Given the myriad roles of the microbiome in nutrition, reproduction, and mate choice, these results provide a foundation for determining the roles of the microbiome in the ecological divergence of Hawaiian Drosophila.

INTRODUCTION

Animal evolution takes place in a natural world teeming with pathogenic and beneficial microorganisms, both of which shape a breadth of physiological functions including digestion, detoxification, immune response, circadian rhythm regulation, thermal tolerance, and reproduction (Baldassarre et al 2022, Kohl et al 2022, Kucuk 2020, Lee and Hase 2014, Li et al 2021, Suyama et al 2023, Tefit et al 2023, Walters et al 2020, Zhang et al 2024). The complex partnership between host and microbes has been hypothesized to facilitate adaptation by improving resilience to climate stress, offering protection from pathogens and parasites, and enabling a move to novel dietary niches (Gao et al 2023, Kohl et al 2014, Zhang et al 2024). How host-associated microbiomes assemble and contribute to adaptation remains an open question. In humans and other animals, environmental factors, especially diet, significantly influence the composition and diversity of the microbiome (Carmody et al 2015, Turnbaugh et al 2009). Additionally, organisms may co-diversify with microbial taxa over evolutionarily long periods of time, referred to as phylosymbiosis (Brooks et al 2016, Moran and Sloan 2015). Neotropical butterflies are one example of this phenomenon whereby host phylogeny plays a prominent role in predicting gut microbial communities (Ravenscraft et al 2019). Other genera exhibit different relative contributions of habitat versus phylosymbiosis in host microbiome assembly (e.g., (Lim and Bordenstein 2020, Martoni et al 2023, Perez-Lamarque et al 2022).

The rapid adaptive radiation of the Hawaiian Drosophila provides an excellent opportunity to investigate the relative contributions of endemic status, genetics, evolutionary history and environmental factors to the structure of microbial communities. The bulk of Drosophila diversification occurred in the Hawaiian Islands, where a recent radiation (~10 Ma; (Church and Extavour 2022, Magnacca and Price 2015) spanning the eight main islands of the Hawaiian archipelago has resulted in hundreds of species across a number of major clades, almost all of which are single-island endemics. Invasive drosophilids including cosmopolitan species such as D. suzukii and D. immigrans are also commonly found throughout Hawai‘i (Koch et al 2020, Mainland 1949, Medeiros et al 2025). The extreme geographic isolation of Hawai‘i as well as its diverse climate conditions contribute to the high endemism and biodiversity of plants and animals found in the archipelago (Barton et al 2021). Notably, Hawaiian Drosophila exhibit a variety of host plant specializations: the larvae of some species are generalists that subsist on multiple plants, while others are strict specialists and associate with a single plant clade (Magnacca et al 2008, Magnacca and Price 2012, Magnacca and Price 2015). Additionally, the Hawaiian Islands provide a variety of abiotic environmental conditions along a chain of islands of known age (Clague 1996). This “natural laboratory” allows island biogeographic hypotheses to be tested with respect to host-associated microbiomes, such as the prediction that taxonomic diversity increases with island age (Hembry et al 2021, Matthews et al 2021, Shaw and Gillespie 2016). Other geographical aspects including latitude, elevation, and area have also been shown to shape macroorganism diversity (Davison et al 2018, Grayson and Lennstron 2022, Losos and Ricklefs 2009) but there is limited evidence as to whether these patterns are recapitulated by host-associated microbiomes (Brown et al 2023, Dickey et al 2021, Henry and Ayroles 2022, Meyer et al 2018).

To determine the factors that influence the microbial communities of endemic Hawaiian flies, we used high throughput amplicon sequencing to characterize the microbiomes profiles of endemic and invasive species sampled from six islands of the Hawaiian archipelago and tested the role of the following factors in influencing microbial diversity and composition: endemic status, abiotic features, phylosymbiosis, and host plant specialization. Considering that Drosophila microbiomes are largely diet-derived (Douglas 2019, Wong et al 2013), we hypothesized that the microbiome composition of endemic flies will be i) distinct from invasive species in terms of composition and predicted function, ii) strongly influenced by environmental factors, and iii) characteristic of host plant specialization.

METHODS

Wild fly collection and processing

We sampled over 500 flies in the field, representing 60 native species (including 29 from the picture wing subgroup) and 6 introduced species from 20 locations across six islands (Supp. Table 1). Wild flies were captured on sponges baited with crushed Cavendish bananas (Musa acuminata) mixed with baker’s yeast at a ratio of approximately 5 g of yeast per banana as well as fermented mushroom (Agaricus bisporus) “tea.” The mushroom tea was made by combining 200 g mushrooms in 1 L of water and allowing the mixture to ferment at room temperature for 14 d. Live flies that landed on the bait were captured with a polypropylene tube and transferred to 95% EtOH within 10 min and stored in coolers with ice packs. Samples were transferred to −10 °C storage within 1-2 hrs of collection.

A subset of 488 flies (437 native, 51 invasive species) were used for 16S rRNA analysis and 528 flies (471 native, 57 invasive species) for ITS analysis. The elevations of collection sites ranged from 599 – 1536 m, average annual rainfall ranged from 750 – 6392 mm, average annual temperature ranged from 13 - 19.8 °C, and average annual evapotranspiration ranged from 599 −978 mm (Fig. 1). Data for Hawai‘i climate variables and topographic data were sourced from the Hawai‘i Climate Data Portal (McLean et al., 2020), TessaDEM database (https://tessadem.com/) and OpenTopography (USGS 2018).

Figure 1.

Drosophila collection sites, corresponding volcano ages, and environmental conditions. Environmental conditions are based on Giambelluca et al. (2013) and represent average annual values. Volcanic age estimates are based on Carson and Clague (1995); Mya: million years ago; HETF: Hawai‘i Experimental Tropical Forest; ET: evapotranspiration.

Genomic DNA extraction and library preparation

DNA was extracted using published procedures (Medeiros et al 2024). Briefly, flies were surface sterilized with 2 washes in 95% EtOH followed by 2 washes in sterile water. Individual flies were homogenized in ATL buffer from PowerMag Bead Solution (Qiagen; MD, USA) with 1.4 mm ceramic beads (Qiagen), using a bead mill homogenizer (Bead Ruptor Elite, Omni, Inc; GA, USA) and extended vortexing for 45 min at 4 °C. The homogenate was treated overnight with proteinase K (2 mg/ mL) at 56 °C and DNA was extracted using the MagAttract PowerSoil DNA EP Kit (Qiagen) according to manufacturer’s instructions. Bacterial diversity was characterized by PCR amplification of the 16S rRNA gene with primers to the V3-V4 region (515F: GTGYCAGCMGCCGCGGTAA; 806R: GGACTACNVGGGTWTCTAAT) (Parada et al 2016). Fungal diversity was characterized using primers to the internal transcribed spacer (ITS1f: CTTGGTCATTTAGAGGAAGTAA; ITS2: GCTGCGTTCTTCATCGATGC) (White et al 1990). The primers contain a 12-base pair Golay-indexed code for demultiplexing.

The PCRs were performed with the KAPA3G Plant kit (Sigma Aldrich, MO, USA) using the following conditions: 95 °C for 3 min, followed by 35 cycles of 95 °C for 20 seconds, 50 °C for 15 seconds, 72 °C for 30 seconds, and a final extension for 72 °C for 3 min. The PCR products were cleaned and normalized with the Just-a-plate kit (Charm Biotech, MO, USA). High throughput sequencing (HTS) was performed with Illumina MiSeq and 250 bp paired-end kits (Illumina, Inc., CA, USA). Each individual Drosophila collected was identified using both morphological features and PCR amplification of the approx. 800 bp cytochrome-c oxidase subunit 1 gene followed by a Genbank BLAST search.

Microbiome data analysis

Post-processing of HTS data (read filtering, denoising, and merging) was performed using the “MetaFlow|mics” microbial 16S rRNA pipeline for bacteria and the fungal ITS pipeline for fungi (Arisdakessian et al 2020). For bacterial samples, reads shorter than 20 bp and samples with fewer than 5,000 reads were discarded to standardize the sampling depth (see rarefaction curve in Supp. Fig. 1). Paired reads are merged if the overlap is at least 20 bp with a maximum 1 bp mismatch. The contigs generated by DADA2 (Callahan et al 2016) were processed using MOTHUR (Schloss et al 2009), and clustered at a 97% sequence similarity. We chose this similarity cutoff for determination of OTUs as we were attempting to investigate broad patterns rather than investigate the effects of specific strains of microbes. For fungal samples, the ITS1 target region is extracted using ITSxpress and only the forward reads are used. Contigs are generated and denoised with DADA2. Reads shorter than 20 bp are discarded and filtered to remove chimeric sequences. Fungal sequences were clustered at 97% sequence similarity and initially aligned and annotated using the SILVA v138 database (Quast et al 2013), NCBI BLAST, UNITE (Nilsson et al 2019), and MycoBank (Robert et al 2013) using a >95% sequence similarity cutoff. The ITS data were normalized by transforming sample counts to proportions (McKnight et al 2019). This method was selected to normalize variation in sampling depth without having to discard data. Analyses were performed after clustering at the genus level using the phyloseq package (McMurdie and Holmes 2013) in R version 4.3.1 (RCoreTeam 2022).

We generated several datasets to test focal questions, including a subset that includes only endemic Hawaiian Drosophila, and a subset of invasive flies introduced to Hawai‘i. For the endemic and invasive datasets, we first examined whether the dispersion (multivariate variance) in microbial community composition was unequal among fly species. We employed the function betadisper in the vegan package (Oksanen et al 2015) to test for heterogeneity of multivariate dispersions, using the anova function in the R stats package to test for significance. We calculated dispersion among samples based on Bray-Curtis dissimilarity values (Bray and Curtis 1957), which is expected to perform robustly as the metric exhibits high sensitivity to group-level differences and linearity in abundance transforms across samples (Kers and Saccenti 2021, Ricotta and Podani 2017). We selected alternative downstream statistical tests depending on whether the among-group dispersion was homogeneous (i.e. permutational multivariate analysis of variance, PERMANOVA), or heterogeneous (i.e. Mantel correlation and multiple regression on distance matrices, MRM). Although Mantel tests are sensitive to heterogeneity in dispersions, we employ them as an initial omnibus test in conjunction with logistic regression using MRM, which is considered to be robust to among-group variance and an unbalanced sampling design (Anderson and Walsh 2013, Lichstein 2007, McArtor et al 2017, Warton et al 2012).

Alpha diversity was evaluated at the 97% OTU level based on Chao1 and Shannon indices and were compared using a Mann-Whitney test. Relative abundance charts were constructed after grouping flies from the same geographical area or taxon. Univariate multiple testing with an F-test was used to test for significant differences in microbial taxa using the phyloseq package. Because we observed no differences in the microbiomes of female and male flies, these samples were treated as independent replicates unless otherwise noted. Venn diagrams were generated based on the ten most abundant genera (based on total counts) for each category of endemic or invasive flies using the “Venn Diagrams” software from Ghent University/ VIB (https://bioinformatics.psb.ugent.be/webtools/Venn/). There was large variation between individual flies in the same category in terms of the number of reads per genus; as such, individual flies may not necessarily contain all ten of the most abundant genera. Bacterial taxa were classified as pathogens or arthropod symbionts based on previous publications (Supp. Table 2).

Phylosymbiosis:

To test for phylosymbiosis, a pattern where the microbial community would have greater similarity among closely related species, we employed Mantel tests to examine the association between microbial composition dissimilarity and phylogenetic dissimilarity. Analyses were performed using a larger microbiome data set comprised of all detected microbes for each fly sample and a smaller data set comprised of the 25 most abundant bacterial taxa or three most abundant fungal taxa for each fly based on relative abundance. The relative abundance of an OTU is the number of reads for an OTU divided by the total number of reads in the sample. We included only endemic Hawaiian Drosophila samples to align with a recent Hawaiian drosophilid phylogeny based on phylogenomic data (Church and Extavour 2022). As three Hawaiian Drosophila are missing from the phylogeny, we placed them with known sister taxa as unresolved terminal branches: D. grimshawi was placed with D. pullipes and D. craddockae; D. oreas was placed as sister to D. conspicua. We calculated the pairwise phylogenetic distance among Drosophila taxa using branch lengths through the cophenetic.phylo function in the package ape (Paradis and Schliep 2019), where the phyogenetic tree and a phenetic tree based on microbiome similarity (Bray-Curtis distance) are compared. We used the mantel function in the vegan package to test for a correlation between the microbial community dissimilarity and phylogenetic dissimilarity. A non-parametric Spearman’s rank correlation was used in the mantel function and the data were permuted 9,999 times to calculate significance of the correlation coefficient.

Abiotic factors:

To test for the role of abiotic factors in microbial community variation, we first assessed a set of climatic predictors for independence using Pearson’s correlation, including temperature, elevation, rainfall, and evapotranspiration (Giambelluca et al 2013). As temperature and elevation were highly correlated, we focus on elevation as a predictor. We also assessed island age by assigning a value based on dates in Carson & Clague (Carson and Clague 1995), with Kauaʻi being oldest, followed by Oʻahu, Molokaʻi, Lānaʻi, Maui, and Hawai‘i Island being the youngest. Island size typically correlates with island age, with younger islands being larger than older ones, but this pattern does not apply to Molokaʻi and Lānaʻi, which used to be joined with Maui as one larger island. Lānaʻi was dropped from our analyses of island age and size due to collecting only two endemic Drosophila from that island.

Island age and community diversity:

To minimize potential bias from uneven sampling between each island, we randomly subsampled 45 individuals (for 16S rRNA comparison) and 56 individuals (for ITS comparison) from each island. Samples from Lānaʻi were omitted since only two endemic flies were collected.

Environment, geography, and phylogeny:

We used multiple regression to test the relative importance of phylogenetic distance, sex, elevation, evapotranspiration, rainfall, and island age as explanatory factors in microbial community composition. Each variable was analyzed using the MRM function in the ecodist package (Goslee and Urban 2007). We used the logistic method, as the among-group variance was heteroscedastic. A desirable feature of logistic regression is that it does not assume linear dependence between the dependent and independent variables, although it does assume that the independent variables are linearly related to the log odds. The effect size and statistical significance of each variable was calculated using 1,000 permutations of the data. The overall deviance of the logistic regression model can be used as a measure of the goodness of fit, with values closer to zero showing improved model performance. To help visualize the effect of each variable graphically, we employed distance-based redundancy analysis using the capscale function in the vegan package. We regressed a distance matrix of each predictor variable against the Bray-Curtis dissimilarity matrix of the microbes in a linear model, and used ordination to display the first two principal coordinates. Samples were visually grouped in the ordination according to their value of the predictor variable.

To further explore the predictive power of environmental variables on microbiome variation, we examined the effect of among-site variation within endemic species using PERMANOVA. PERMANOVA is a non-parametric multivariate statistical permutation test that employs distance matrices for response and predictor variables to test group-level differences. The predictor values are site, sex, elevation, rainfall, evapotranspiration. We restricted these analyses to two species with large sample sizes (>30 samples from at least 6 sites) to ensure sufficient degrees of freedom in the statistical tests: D. tanythrix (55 samples, 8 sites) and D. sproati (36 samples, 7 sites). We used the adonis2 function in the vegan package to conduct the PERMANOVA (McArdle and Anderson 2001). We also used D. sproati and D. tanythrix from three sites that exhibited the largest differences in abiotic variables to investigate whether beta-diversity of microbial communities differed within or between sites.

Host plant specialization:

Hawaiian drosophilids comprise of 6 major species groups, amongst which the picture wing clade is the best described in terms of phenotype and ecology. To test the impact of host plant specialization on microbiome composition, we examined a subset of Hawaiian drosophilids from the picture wing clade that are designated as generalists (16S rRNA: n = 30; ITS: n = 42) and specialists (16s rRNA: n = 184; ITS: n = 190) based on field observations of larval association with host plants (Magnacca et al 2008). Generalists in our data set include D. villospedis, D. crucigera, and D. grimshawi. For specialists, we include D. ochracea, D. sproati, D. prolaticilia, D. punalua, D. ambochila, D. inedita, D. hirtitibia, D. oreas, D. murphyi, D. cilifera, D. melanocephala, D. orphnoopeza, D. anomalies, D. quasianomalipes, D. picticornis, D. planitibia, D. neoperkinsi, D. bostrycha, D. limitata, D. setosimentum, and D. cilaticrus.

Functional analysis:

In order to determine if bacterial microbiome functional differences varied among fly species, we used PICRUSt2 (Douglas et al 2020). PICRUSt2 uses the 16S rRNA marker gene to match microbial OTUs and their reference genomes, where the functional annotation of each microbial genome is retrieved. Functional annotations of gene families present in each genome are based on orthologs within the Kyoto Encyclopedia of Genes and Genomes (KEGG). We then used STAMP v2.13 (Parks et al 2014) to analyze differences among fly groups based on functional abundance and to generate visualizations of the functional data. The functional profiles were filtered to retain features based on statistical significance (p = 0.01) and remove features with few supporting reads (maximum per group < 1).

RESULTS

Differences in microbiome composition between endemic and invasive drosophilid species in Hawai‘i

High-throughput amplicon sequencing of the bacterial marker 16S rRNA and fungal marker ITS revealed that the microbiomes of both endemic and invasive species contain similar numbers of bacterial genera (26 ± 6 (average ± range) in endemic species vs. 22 ± 7 in invasive species) and fungal genera (10 ± 4 in endemic species vs. 9 ± 3 in invasive species). Members of the Enterobacteriaceae class and Pseudomonas and Dysgonomonas genera were among the top 10 taxa in all flies regardless of endemic status (Fig. 2). With respect to the fungal microbiome (mycobiome), eight genera were shared between endemic and invasive flies (Fig. 2). The dominant class in both populations was the yeast Saccharomycetes, consistent with previous reports of native Hawaiian drosophilids (O’Connor et al 2014) (Supp. Table 2). Despite the overlap in common bacterial and fungal taxa, endemic flies exhibit bacterial communities with greater taxonomic richness and evenness compared to invasive flies (Supp. Fig. 2) and were compositionally distinct from invasive species (Fig. 3A). In particular, several previously identified arthropod pathogens and symbionts were more abundant in endemic flies (Fig. 3B; Supp. Table 2). To test whether a disparity in the sample sizes of invasive and endemic groups may have contributed to the difference in the beta-diversity of bacterial communities, we randomly subsampled the endemic group data set to the same sample number as the invasive group (n = 51). The compositions of each group were still significantly different (ANOSIM, p = 0.001, R = 0.0883; Supp. Fig. 2C).

Figure 2.

Venn diagrams of microbial taxa found in endemic and invasive drosophilids, based on the ten most abundant (A) bacterial and (B) fungal genera; *likely Enterobacteriaceae and Yeriniaceae. Of the Wolbachia-positive flies, most individuals contained fewer than 10 counts. However, several individual invasive flies were heavily infected by Wolbachia, resulting in Wolbachia being amongst the ten most abundant genera for invasive flies. Photo credit: D. grimshawi (left; J. Sartore and Hawai‘i Invertebrate Program) and D. suzukii (right; V. Stacconi (Stacconi 2022)).

Figure 3.

Visualization of differences in the composition and predicted function of bacterial communities in endemic Hawaiian Drosophila and invasive flies. (A) Ordination employing a distance-based redundancy analysis (dbRDA) of Bray-Curtis distance in relation to endemic species status. (B) Differential abundance analysis showing bacterial groups that differ significantly among the endemic and invasive fly species. Significant differences at the lowest taxonomic level are shown (g: genus; c: class; o: order; f: family). Genera previously characterized as arthropod symbionts (†) or animal pathogens (*) are indicated. (C) Predicted functional differences between endemic Hawaiian Drosophila and invasive species. Functional profiles are based on KEGG annotations of microbial reference genomes and are weighted by bacterial abundance in each fly sample.

To evaluate the effect of environmental variables and endemic status on microbiome composition, we employed a multiple regression on distance matrices (MRM). Our analyses reveal that bacterial communities varied significantly with endemicity, elevation, evapotranspiration, and island age, with elevation showing the strongest effect on microbiome dissimilarity among samples (Supp. Table 3; see Supp. Fig. 3 for a visualization of each variable’s effect in a distance-based ordination). As with bacterial communities, fungal communities were affected by elevation, evapotranspiration, and rainfall (listed in order of explanatory strength) but not endemic status (Supp. Table 4; Supp. Fig. 3).

Functional potential analysis of endemic and invasive bacterial microbiomes using the metagenome prediction pipeline PICRUSt2 suggests slight shifts in functional space, rather than the gain or loss of functional pathways (Fig. 3C). The shift in bacterial microbial diversity was associated with changes in its predicted functions related to several notable biosynthetic pathways. Invasive species host microbes that are enriched for predicted functions of peptidoglycan maturation, which is important for maintenance of cellular integrity and may enable survival in challenging environments. In contrast, endemic species host microbes that have predicted biosynthesis functions for thiazole and thiamin (vitamin B), essential micronutrients, as well as L-glutamate and L-glutamine biosynthesis (ammonia assimilation), L-arginine biosynthesis (production of metabolites and amino acids), and beta-D-glucuronide and D-glucuronate degradation, both of which are related to detoxification processes.

Variation in endemic fly microbiome composition across abiotic conditions

Given the contribution of endemic status as a major feature impacting variability in microbiome composition, we employed an MRM to separately test the contribution of genetic relatedness along with island age and other abiotic environmental conditions only using endemic populations. Elevation (which is highly correlated with temperature: r = −0.96, p = 2.2e-16), evapotranspiration, and island age had significant effects on structuring bacterial communities (Table 1; Supp. Fig. 3). The effect of phylogenetic relatedness was non-significant, although it explained a similar amount of variation as the significant variables in the model. For fungal profiles, elevation had the strongest effect, followed by evapotranspiration and rainfall (Table 2). Interestingly, two predictor variables differed between bacteria and fungi: first, island age, which had a strong effect on bacterial composition, was negligible for fungi. Second, phylogenetic relatedness was one of the strongest predictors of mycobiome composition. To test the possibility that bacterial communities that are specific to each site may have distinct functional properties, we compared the putative functions across significant environmental predictors. However, PICRUSt2 analysis found only slight shifts in the predicted functional space (Supp. Fig. 5). Notably, when the bacterial and fungal profiles of invasive species are examined separately for environment-induced variation using an MRM, no significant effects were found (Supp. Table 5; Supp. Fig. 3).

Table 1.

Multiple regression on distance matrices (MRM) analysis of endemic Hawaiian Drosophila bacterial profiles.

Variable†	β-value‡	p-value
Intercept	1.407	1.000
Sex	−0.013	0.230
Elevation	0.056	0.009
Rainfall	0.018	0.333
Evapotranspiration	0.044	0.001
Island Age	0.046	0.002
Phylogenetic relatedness	0.040	0.193

^†Predictor variables are measured as Euclidean distance values and significant predictors are italicized (residual deviance = 8865.5).

^‡Microbiome distance values among samples are based on Bray-Curtis dissimilarity values.

Table 2.

Multiple regression on distance matrices (MRM) analysis of endemic Hawaiian Drosophila fungal profiles and environmental and genetic predictors.

Variable†	β-value‡	p-value
Intercept	1.535	1.000
Sex	0.002	0.860
Elevation	0.299	0.001
Rainfall	0.084	0.004
Evapotranspiration	0.138	0.001
Island Age	−0.025	0.133
Phylogenetic relatedness	0.112	0.007

^†Predictor variables are measured as Euclidean distance values and significant predictors are italicized (residual deviance = 19919.54).

^‡Mycobiome distance values among samples are based on Bray-Curtis dissimilarity values.

Testing for phylosymbiosis in endemic Hawaiian Drosophila

Based on our finding that endemic status is a strong predictor of microbiome composition, we next asked whether the microbiome composition of endemic Hawaiian flies is linked to host phylogeny and evolutionary history. We looked for evidence of phylosymbiosis by testing whether host species that are more closely related will exhibit more similar microbial profiles based on Bray-Curtis distances. We found statistically significant, but weak relationships between bacterial and fungal diversity and phylogenetic relatedness (16S rRNA: non-parametric Mantel statistic r = 0.0727, p = 1e-04; ITS: r = 0.05454, p = 1e-04). In a comparison of the Hawaiian drosophilid phylogeny and a dendrogram of Bray-Curtis distance among microbiomes, the similarity between the topology of each tree is not random (16S rRNA Nye similarity = 0.3012 and ITS Nye similarity = 0.3056, when two random trees should equal 1) (Fig. 4A and Supp. Fig. 4A). However, the degree of phylosymbiosis appears limited among fly species and microbiota based on the topological relationships of terminal taxa. The lack of phylogenetic clustering is also evident when only the most abundant taxa are considered (Fig. 4B and Supp. Fig. 4B).

Figure 4.

Cophylogeny analysis of Hawaiian Drosophila and bacterial taxa. (A) The cophylogeny of Hawaiian Drosophila (left) and a dendrogram based on the similarity of bacterial community composition according to Bray-Curtis distance (right) shows weak evidence of phylosymbiosis due to the lack of topological congruence in the two trees. (B) A heatmap depicting the 25 most abundant bacterial taxa in Hawaiian Drosophila, which include a number of arthropod endosymbionts (†) and animal pathogens (*). The fly species are listed in the same order as the fly phylogeny. (C) The reproductive endosymbiont Wolbachia is not closely associated with speciation events among sister taxa of Hawaiian Drosophila and appears clumped to certain species groups. The presence of Wolbachia was inferred for a species if Wolbachia was detected in any of the samples in a species.

The endosymbiont Wolbachia has been implicated as a potential promoter of reproductive isolation and driver of genetic divergence (Kaur et al 2021). To explore a putative role for Wolbachia in Hawaiian drosophilid diversification, we looked explicitly for its occurrence in endemic microbiomes. Notably, the widely-distributed arthropod endosymbiont Wolbachia is present in a number of Hawaiian drosophilids. Yet, Wolbachia does not appear to be associated with speciation events among sister taxa, and instead, appears clumped to certain species groups (Fig. 4C). Overall, Wolbachia is a relatively minor component of our total data set, with its total abundance representing 1.13% of all 16S rRNA reads, resulting primarily from ten individual flies (seven endemic D. prolaticilia from Hawai‘i Island and three invasive flies). Many Wolbachia-positive flies had fewer than five counts, suggesting that while some individual Drosophila in Hawai‘i are heavily infected with Wolbachia, the majority are not.

Contribution of host identity vs. environmental factors

To directly assess the contribution of host identity vs. environmental factors in predicting microbiome variation, we compared the microbiomes of D. tanythrix and D. sproati, two endemic species that had large sample sizes (>30 samples) and were found in eight or seven sites, respectively (Fig. 5A). We asked first whether microbiomes of the same species differ between different sites, using PERMANOVA. We observed a strong effect of site on overall microbiome variation, but not of sex (Table 3; Supp. Table 6). Further analysis of abiotic factors showed that only rainfall explained a significant amount of variation in the bacterial composition of D. tanythrix and D. sproati (Table 3) whereas both rainfall and evapotranspiration contributed to fungal variation (Supp. Table 6). Next, given the impact of site on microbiome composition, we compared samples from three sites with significantly different abiotic features: Tom’s Trail, Ola‘a Pole 44, and Ka‘u Forest Reserve (Fig. 5B). At least seven individuals each of D. sproati or D. tanythrix were collected from each of these sites. D. sproati from each site exhibited different bacterial and fungal profiles (Table 4). D. tanythrix showed a similar trend for 2 of 3 sites, with one exception: microbiome composition of flies from Ola‘a Pole 44 and Ka‘u Forest Reserve approached but did not reach significance (PERMANOVA with Bray-Curtis: 16S rRNA p = 0.08, ITS p = 0.059).

Figure 5.

Bacterial and fungal community compositions are associated with rainfall levels for within-species comparisons of D. tanythrix and D. sproati microbiomes. (A) Images of D. tanythrix and D. sproati. (B) Map of Hawai‘i island showing average rainfall for January 2021 for collection sites used in analysis. Rainfall data sourced from Hawai‘i Climate Data Portal (McLean et al., 2020). (C, D) The top 20 bacterial taxa found in D. tanythrix and D. sproati whose differential relative abundance is associated with rainfall variation. (E, F) The top 20 fungal taxa whose differential relative abundance is associated with rainfall variation. Positive coefficients indicate an increase with rainfall, and negative coefficients indicate a decrease with rainfall. Photo credit: Karl Magnacca.

Table 3.

Outcomes from PERMANOVA testing for drivers of intraspecific and intra-site variation in the bacterial microbiome of two endemic Hawaiian Drosophila species.

			distBC ~ site + sex		distBC ~ elevation + rainfall +ET1
Species	Samples	Sites	Site R2	Sex R2	Elevation R2	Rainfall R2	ET R2
			(p-value)	(p-value)	(p-value)	(p-value)	(p-value)
D. tanythrix	55	8	0.185 (0.004)	0.017 (0.419)	0.017 (0.452)	0.034 (0.035)	0.031 (0.059)
D. sproati	36	7	0.213 (0.047)	0.054 (0.459)	0.041 (0.092)	0.046 (0.048)	0.031 (0.274)

¹ET: Evapotranspiration.

Table 4.

Pairwise comparisons of bacterial and fungal beta-diversity of microbiome for D. sproati and D. tanythrix from three locations on Hawai‘i Island.

Population A†	Population B†	p-value (16S r RNA)	p-value (ITS)
(T) sproati	(T) tanythrix	0.452	0.114
(T) sproati	(K) sproati	0.020	0.002
(T) sproati	(K) tanythrix	0.011	0.002
(T) sproati	(O) sproati	0.011	0.002
(T) sproati	(O) tanythrix	0.012	0.002
(T) tanythrix	(K) sproati	0.166	0.002
(T) tanythrix	(K) tanythrix	0.050	0.002
(T) tanythrix	(O) sproati	0.112	0.002
(T) tanythrix	(O) tanythrix	0.024	0.002
(K) sproati	(K) tanythrix	0.340	0.774
(K) sproati	(O) sproati	0.030	0.005
(K) sproati	(O) tanythrix	0.154	0.034
(K) tanythrix	(O) sproati	0.024	0.002
(K) tanythrix	(O) tanythrix	0.080	0.059
(O) sproati	(O) tanythrix	0.011	0.118

^†T: Tom’s Trail; K: Ka‘u Forest Reserve; O: Ola‘a Pole 44.

Among the top bacterial taxa associated with shifts in rainfall, three OTUs common to both species increased in abundance with rainfall (Otu0008: Pseudomonas nr. P. alcaligenes, Otu0014: Gluconobacter nr. G. oxydans, and Otu003: an unclassified Enterobacteriaceae) while the abundances of four species decreased with rainfall (Otu0033: Lactococcus nr. L. lactis, Otu0001: an unclassified Enterobacteriaceae, Otu0010: Pseudomonas nr. P. fluorescens, and Otu0016: Chryseobacterium sp.) (Fig. 5C, D). For fungal profiles, evapotranspiration and rainfall were significant factors governing composition in both D. tanythrix and D. sproati, while elevation was only a significant factor for D. sproati (Supp. Table 6). Ten of the top fungal taxa were influenced by each of the three ecological variables in both fly species (OTU97_56: Fusarium solani, OTU97_4: Wickerhamomyces anomalus, OTU97_25: Candida railenensis, OTU97_5: Candida asparagi, OTU97_13: Candida tropicalis, OTU97_190: Dipodascus geotrichum, OTU97_43: Zygoascus meyerae, OTU97_15: Papiliotrema flavescens, OTU97_70: Rhodotorula mucilaginosa, and OTU97_74: Cyberlindnera lachancei) (Fig. 5; Supp. Fig. 6).

Next, we examined whether the microbiomes of D. sproati and D. tanythrix found within the same site are similar to each other. With one exception, species from the same site tended to have similar bacterial and fungal communities (Table 4). No significant differences in beta-diversity were found when comparing D. sproati and D. tanythrix found within Tom’s Trail or within Ka‘u Forest Reserve (16S and ITS: Bray-Curtis with PERMANOVA; p > 0.05). The only differences between D. sproati and D. tanythrix found within-site was in the bacterial profile at Ola‘a Pole 44 (p = 0.01).

Role of host plant specialization and evidence for a core microbiome in Hawaiian Drosophila

Many Hawaiian Drosophila selectively oviposit and breed as larvae on a single endemic plant family (Magnacca et al 2008). To examine whether host plant specialization is associated with distinct microbial components, we compared known larval specialists with generalists. The microbiomes of adult flies generally did not differ in terms of composition or bacterial or fungal taxonomic alpha-diversity (Supp. Fig. 7) although host plant specialists exhibited higher evenness in bacterial profiles (Shannon p < 0.02).

Host plant specialization had a significant effect on both the bacterial and fungal microbiomes but only explained a small amount of among-species variation (16S data: R² = 0.009, p = 0.024; ITS data: R² = 0.007, p = 0.046; PERMANOVA). The assumptions of the PERMANOVA were robust since a test for homogeneity of multivariate dispersion was not significant (16S data: p = 0.498; ITS data: p = 0.79). A plot of the top 20 taxa contributing to the difference between generalists and specialists in each dataset shows a number of known symbiotic taxa and fewer putative pathogenic bacterial taxa in the specialist species (Fig. 6A, B). Specialist species were enriched for taxa with predicted glycolysis abilities and the breakdown of sugars, including bacteria such as Orbus, Gilliamella, Dysgonomonas, and Snodgrassella. Chryseobacterium were found in another specialized clade of Drosophila that subsist on rotting cacti (Chandler et al 2011, Martinson et al 2017a), as well as diverse insect species spanning Coleoptera, Hymenoptera, and Lepidoptera (Hammer et al 2020, Madden et al 2022, Shelomi et al 2023). Generalist species exhibited a stronger association with genera previously identified as arthropod pathogens, such as Wolbachia (Kaur et al 2021) and Pseudomonas (Teoh et al 2021).

Figure 6.

Bacterial and fungal communities differ between host plant specialists and generalists. The 20 most abundant bacterial (A) and fungal (B) taxa whose differential relative abundance diverges between Hawaiian Drosophila host plant generalists and specialists in the endemic picture wing clade. Positive coefficients indicate stronger association with host plant specialists, and negative coefficients indicate a stronger association with host plant generalists; *: animal pathogen. (C) Statistically significant functional differences in specialist versus generalist Hawaiian Drosophila in the endemic picture wing clade. Functional profiles are based on KEGG annotation of microbial reference genomes and are weighted by bacterial abundance in each fly sample.

Predictions of functional differences between generalists and specialists using PICRUSt2 suggest slight shifts in microbiome functional space (Fig. 6C). Specialist species host microbes that are predicted to be enriched for cellular energy production through central metabolic pathways, including glycolysis, homolactic fermentation, and formaldehyde assimilation II, as well as processing of nucleotides through adenine and adenosine salvage. In contrast, generalist species host microbes that have the potential for biosynthesis of heme, pyrimidine deoxyribonucleotides, and cysteine, as well as assimilation of sulfate to recover sulfur, an essential nutrient. Changes in the predicted functional space are not due to the gain or loss of functional pathways, however, only the relative expression of pathways.

DISCUSSION

Using a dataset comprised of over 500 Drosophila from six different Hawaiian islands, we report that of the phylogenetic, dietary and environmental features included in our analyses, abiotic factors are the main predictors of both bacterial and fungal microbial communities in endemic flies. Our findings are consistent with previous studies of wild-caught Drosophila (Chandler et al 2011, Chandler et al 2012, Henry and Ayroles 2022, Staubach et al 2013). Considering that Drosophila acquire microbes primarily from the environment (Douglas 2019, Wong et al 2013), it is not surprising that local habitat conditions are a major source of variation in the microbiome. Invasive status also contributed to bacterial composition, an outcome that contrasts with a study of Hawaiian spiders that found environmental context rather than endemic status to be a stronger influence on microbial communities (Pfau et al 2025). Distinct from endemic species, the microbiomes of invasive flies show no detectable response to changes in elevation, evapotranspiration or rainfall. Other studies of invasive species have suggested that adaptation to new niches is accompanied by the acquisition of microbes in the invaded range (Escalas et al 2022, Martignoni and Kolodny 2024, Medeiros et al 2025, Zhang et al 2024). The invasive flies in our study may have a greater resilience to variation induced by abiotic factors or depend less on the incorporation of local microbes for adaptation.

Shared features of the Hawaiian Drosophila microbiome

Endemic flies exhibit different and more diverse bacterial communities compared to invasive flies but exhibit some overlap with other wild drosophilids, despite differences in host plant, evolutionary lineage, and geography. In particular, Enterobacteriacea, Commensalibacter, Gluconobacter, and Dysgonomonas, are amongst the most abundant taxa found in natural populations of endemic flies and previous surveys of wild populations (Chandler et al 2011, Chandler et al 2012, Gale et al 2024, Henry and Ayroles 2022, Staubach et al 2013). By contrast, Pseudomonas was among the ten most abundant taxa found in both endemic and invasive populations in Hawai‘i. Members of the genus are known to be arthropod pathogens although, depending on the strain, some pseudomonads may play neutral or beneficial role for insect hosts (Teoh et al 2021). Interestingly, Lactobacillus, a major component of the microbiomes of lab-raised D. melanogaster (Shin et al 2011, Storelli et al 2011, Wong et al 2013) was not detected in Hawaiian drosophilids. Compared to two other endemic arthropod radiations, Ariamnes spiders and true bugs, Hawaiian Drosophila have almost no overlap in bacterial genera, with some exceptions such as Pseudomonas (Armstrong et al 2022, Poff et al 2017).

The taxonomic diversity of fungi did not differ between endemic and invasive hosts (Supp. Fig. 1, Supp. Table 2). Given that fungi, including yeasts, are used as a food source by Drosophila, it may be expected that similar fungal taxa are found in endemic and invasive flies (Stefanini 2018). Saccharomycetes yeast is frequently associated with feeders of rotting fruit and were previously identified in a survey of Drosophila populations from disparate locations (Chandler et al 2012) and a study of two endemic Hawaiian Drosophila species (D. neutralis also included in our study, and D. imparisetae, not in our study). Several of these yeast have antimicrobial properties (Chai et al 2024), enhance Drosophila development (Dmitrieva et al 2019, Murgier et al 2019), or contribute to the fecundity of lab-raised Hawaiian D. grimshawi (Medeiros et al 2024). However, whether similar benefits are provided in the natural environment for wild populations of flies remains to be determined. (O’Connor et al 2014). The overlap in fungal features between distantly-related species with varied diets and habitats supports the possibility that drosophilids actively select for members of the Saccharomycetes class, similar to desert-adapted drosophlids that specialize on rotting cacti (Fogleman et al 1981, Fogleman et al 1982). To evaluate this possibility, feeding preference assays together with microbiome profiling of host plants will be required.

Contribution of the microbiome to rapid adaptation

The Hawaiian Drosophila clade is estimated to contain as many as 1000 species at one point, originating from a single founder species (O’Grady and DeSalle 2018). To what extent could the microbiome have facilitated this explosive adaptation? For many animals, dietary specialization is enabled by microbial symbionts that provide critical functions for nutrition acquisition or detoxification (Brune and Dietrich 2015, Haanstad and Norris 1985, Ingala et al 2021, Kwong et al 2014). Experiments with lab-raised D. grimshawi reveal that manipulating microbiome composition with antimicrobials alters fecundity and mating behavior (Medeiros et al 2024). However, phenotypic effects resulting from natural variation in microbiome composition are likely to be subtler (Price et al 2025, Tefit et al 2023, Walters et al 2020). Considering the varied habitats and dietary habits exhibited by Hawaiian flies, microbes acquired from the local environment may facilitate niche expansion and local adaptation by enhancing nutrient acquisition and enabling the detoxification of novel food sources (Medeiros et al 2025, O’Connor et al 2014, Tefit et al 2023, Walters et al 2020). In support of this possibility, we identified slight but significant shifts in the predicted functions of bacteria from generalist and specialist flies. In particular, specialists show a shift towards pathways with greater cellular energy production, whereas generalists show a shift towards chemical synthesis. Several of the highly prevalent microbes in host plant specialist flies (e.g., Gilliamella) are also components of a core microbiome in insect species that have sugar-rich diets and require the degradation of complex polysaccharides or neutralization of toxic host plant allelochemicals. We caution, however, that functional inference based on amplicon sequences provides only broad generalizations and is biased towards taxa with reference genomes, overlooking rare and uncharacterized taxa. Moreover, metabarcoding with short amplicon sequences does not allow host-associated strain-level differences to be discerned, capture the diversity of metabolic functions, or necessarily reflect gene expression or activity in a microbial community (Comeault Aaron et al 2025). Microbial patterns of association and predicted function also can differ depending on the level of taxon resolution (Khatib et al 2025, Yang et al 2025). Notably, different taxa from the yeast genera Wickerhamomyces and Candida were enriched in both generalists and specialists, indicating potential species- or strain-specific functions. Whole genome sequencing and ultimately, biochemical characterization will be necessary to elucidate the contributions of microbiome members in local adaptation, host plant specialization, and pathogenesis.

The endosymbiont Wolbachia has been implicated as a mechanism that can enhance behavioral isolation and possibly increase the rate of speciation (Cruz et al 2021, Shoemaker et al 1999). From our survey, Wolbachia does not appear to be associated with speciation events among sister taxa and as such, does not lend support to an important role in reproductive isolation. We found Wolbachia in at least some individuals of the same species identified in a previous survey of endemic Hawaiian Drosophila, e.g. D. murphyi, D. ochracea, and particularly in D. prolaticilia, and not in others, e.g. D. sproati and D. tanythrix (Corpuz et al 2023). As in our study, Corpuz et al. (2023) did not report finding Wolbochia in every specimen of a given species, supporting the idea that Wolbochia is not necessarily present in all individuals of a species and that modes of transmission may be both horizontal as well as vertical.

Testing for phylosymbiosis in endemic Hawaiian Drosophila

Phylosymbiosis, which predicts that microbial communities will be more similar within a species than between host species (Bordenstein and Theis 2015), has generally been supported in other animal taxa, especially mammals and birds (Harrison et al 2021) but also in the bacterial gut microbiomes of Hawaiian spiders (Perez-Lamarque et al 2022) and neotropical butterflies (Ravenscraft et al 2019). By comparing correspondence between host lineage and compositional differences of microbial communities, we found weak, but significant evidence of both bacterial and fungal phylosymbiosis in endemic Hawaiian Drosophila. Phylosymbiosis can arise through the vertical transmission of microbes through generations (Guilhot et al 2021, Guilhot et al 2023). However, a pattern of co-diversification can also occur from other processes that do not require the co-evolution of host and microbes, such as selective filtering by the host of microbes from the local environment (Moran and Sloan 2015, Perez-Lamarque et al 2022, Pfau et al 2025). In this scenario, closely related hosts may share similar mechanisms of microbe selection that permit colonization by the same set of microbes. Additionally, both host and microbes in the environment can change in response to environmental factors, resulting in parallel changes that are independent of host-microbe interactions. To differentiate between these possibilities, it will be necessary to have a better understanding of the sources of environmental microbes used by adult flies, host filtering mechanisms, and characterization of microbes at the strain level, rather than genus or species. Considering that the microbiomes of D. sproati and D. tanythrix varied by site, it seems unlikely that at least for these two species, host mechanisms do not play a strong role in specifying microbiome members. Parsing by species subgroup or host plant specialization may reveal other mechanisms of microbiome assembly within the Hawaiian Drosophila radiation.

Biogeography of Hawaiian Drosophila microbiomes

The theory of island biogeography posits that biodiversity will scale with various environmental characteristics including extent of isolation, habitat heterogeneity, and island area and age (Carey et al 2023, Gray and Cavers 2014, MacArthur and Wilson 2001). To what extent do the communities of host-associated micro-organisms follow the rules of biodiversity for macroscropic organisms? Host-associated microbial diversity has been shown to be proportional with the size of islands, both virtual, such as host plants (Chlebicki and Pawel 2007, Dinnage et al 2019, Peay et al 2007, Peay et al 2010), and actual, as shown in a study of soil from islands of the Thousand-Island Lake in China (Li et al 2020). Research on Galapagos finches revealed that island size appears to play a role in host bacterial community composition, though a number of other factors, in particular, phylosymbiosis, may explain more variation (Loo et al 2019).

We found a small but significant effect of island age on gut bacterial diversity and community makeup (Table 1 & 3), as well as significant differences in fungal alpha-diversity between most islands. The beta-diversity of bacterial and fungal communities differed across all five islands in our analyses. However, no clear patterns emerged associating diversity with island age or size in contrast to observations for the diversity of animals and plants (Davison et al 2018, Green et al 2004). Perhaps this outcome is not surprising considering that microbial communities are subject to host-filtering effects and have dispersal limitations (Dickey et al 2021). Differences in habitat quality and urbanization (Zhou et al 2022) also have not been accounted for in our models and could impact the diversity and composition of environmental microbiome reservoirs and, by extension, Drosophila microbe communities.

Abiotic conditions are strong predictors of Hawaiian Drosophila microbiomes

A number of non-covarying abiotic factors were significant predictors of bacterial communities, particularly evapotranspiration, elevation, and island age. Similarly, evapotranspiration, elevation (highly correlated with temperature), and rainfall were predictors of fungal communities. Although little is known about the role of evapotranspiration rates and animal microbiomes, elevation was found to be a predictor of gut bacterial communities in animals ranging from other Drosophila (Brown et al 2023) and bees (Mayr et al 2021) to fish (Bereded et al 2022). Much less is known about the role of abiotic conditions on the gut mycobiome, although one study of mesquite spiny lizards revealed different fungal assemblages at different elevations (Montoya-Ciriaco et al 2020). When comparing only D. sproati and D. tanythrix, two species we collected from multiple locations on Hawaiʻi Island, rainfall was a strong predictor of bacterial community diversity, despite rainfall not being a predictor of bacterial composition for our larger dataset with all species included. D. tanythrix and D. sproati diverged between 7.5 - 11 Mya (Church and Extavour 2022, Magnacca and Price 2015). Additionally, each species has access to distinct microbial reservoirs: D. tanythrix utilize leaves of the host plant Cheirodendron trigynum (‘Ōlapa) on the ground for breeding, whereas D. sproati lay eggs in decaying C. trigynum bark and females likely do not approach leaf litter (Magnacca et al 2008). Despite these differences in evolutionary history and lifestyle, the microbiomes of D. sproati and D. tanythrix found within a site were not significantly different for two of three sites tested. By contrast, microbial communities were distinct for populations of the same species found in different sites. These observations lend additional support for the hypothesis that abiotic conditions, rather than host species identity, are the main drivers of microbiome community composition in endemic Hawaiian Drosophila. It is possible that differences at the species or strain level occur in sympatric species and may be missed by amplicon sequencing (Comeault Aaron et al 2025). In addition, individual variation within a host species may be obscured by combining individual profiles (Martinson et al 2017b).

CONCLUSION

The Hawaiian Drosophila clade is an iconic model for understanding how unique features of island ecology such as diverse habitats, steep environmental gradients, and geographic isolation contribute to rapid adaptation and speciation. Along with these features, the microbial environment is likely to play a prominent role in shaping evolution. Here, our microbiome survey of endemic and invasive flies from across the Hawaiian archipelago reveal that the microbiomes of endemic Hawaiian Drosophila are distinct from those of invasive species in terms of composition, greater diversity, and response to abiotic conditions. Microbiomes of the same species differ between sites on the same island, illustrating the robust influence of environmental features. The concurrent diversification of microbial communities, along with their hosts, has been hypothesized to play a role in local adaptation and even ecological speciation (Rennison et al 2019). The presence of local, environmentally-specific microbes may facilitate adaptation by enhancing host ability to exploit local food sources, enhancing fecundity, altering metabolism, or influencing mating behavior (Medeiros et al 2024). Whole genome sequencing, metabolomic analysis, and behavioral analysis of identical species from different sites or islands would allow this hypothesis to be tested. Additionally, careful characterization of host plant microbes will help to clarify the extent to which host plants serve as microbial reservoirs for Hawaiian flies and contribute to their diversification (Ort et al 2012).

Supplementary Material

Figures S1-S7

Supp. Fig. 1: Rarefaction curve for 16S rRNA (n=488) and ITS (n=528) data for native and non-native Drosophila samples.

Supp. Fig. 2: Alpha diversity analyses of bacteria and fungi found in endemic Hawaiian Drosophila (HD) and invasive Drosophila (nonHD).

Supp. Fig. 3: Ordinations employing a distance-based redundancy analysis (dbRDA) of Bray-Curtis distance of endemic and invasive Drosophila microbiomes in relation to ecological predictors.

Supp. Fig. 4: Cophylogeny analysis of Hawaiian Drosophila and fungal taxa based on ITS data.

Supp. Fig. 5: Principal component analyses examining the relationship between Hawaiian Drosophila bacterial microbiome function and ecological predictors.

Supp. Fig. 6: Fungal composition is associated with abiotic factors for within-species comparisons of D. tanythrix and D. sproati microbiomes.

Supp. Fig. 7: Comparison of microbiome communities from generalist and specialist Hawaiian Drosophila.

Tables S1-S4

Supp. Table 1: Summary of collection site locations and species collected at each site.

Supp. Table 2: Prevalent microbial taxa found in endemic Hawaiian Drosophila species.

Supp. Table 3: Multiple regression on distance matrices (MRM) analysis of endemic and invasive Drosophila bacterial profiles and environmental and genetic predictors.

Supp. Table 4: Multiple regression on distance matrices (MRM) analysis of endemic and invasive Drosophila fungal profiles and environmental and genetic predictors.

Supp. Table 5: Multiple regression on distance matrices (MRM) analysis of only invasive Drosophila bacterial profiles

Supp. Table 6: Outcomes from PERMANOVA testing for drivers of intraspecific and intra-site variation in the fungal microbiome of endemic Hawaiian Drosophila species.

DATA AVAILABILITY STATEMENT

Sequence data and metadata will be available in the NCBI Sequence Read Archive (Project PRJNA1268594) . Data for Hawai‘i climate variables and topographic data were sourced from the Hawai‘i Climate Data Portal (McLean et al., 2020), TessaDEM database (https://tessadem.com/) and OpenTopography (USGS 2018). Supplemental Tables (including metadata) are available at Zenodo (doi: 10.5281/zenodo.15486747).