Aedes albopictus microbiome derives from environmental sources and partitions across distinct host tissues

Abstract The mosquito microbiome consists of a consortium of interacting microorganisms that reside on and within culicid hosts. Mosquitoes acquire most of their microbial diversity from the environment over their life cycle. Once present within the mosquito host, the microbes colonize distinct tissues, and these symbiotic relationships are maintained by immune‐related mechanisms, environmental filtering, and trait selection. The processes that govern how environmental microbes assemble across the tissues within mosquitoes remain poorly resolved. We use ecological network analyses to examine how environmental bacteria assemble to form bacteriomes among Aedes albopictus host tissues. Mosquitoes, water, soil, and plant nectar were collected from 20 sites in the Mānoa Valley, Oahu. DNA was extracted and associated bacteriomes were inventoried using Earth Microbiome Project protocols. We find that the bacteriomes of A. albopictus tissues were compositional taxonomic subsets of environmental bacteriomes and suggest that the environmental microbiome serves as a source pool that supports mosquito microbiome diversity. Within the mosquito, the microbiomes of the crop, midgut, Malpighian tubules, and ovaries differed in composition. This microbial diversity partitioned among host tissues formed two specialized modules: one in the crop and midgut, and another in the Malpighian tubules and ovaries. The specialized modules may form based on microbe niche preferences and/or selection of mosquito tissues for specific microbes that aid unique biological functions of the tissue types. A strong niche‐driven assembly of tissue‐specific microbiotas from the environmental species pool suggests that each tissue has specialized associations with microbes, which derive from host‐mediated microbe selection.


| INTRODUCTION
Host-associated microbiomes or symbionts (i.e., organisms that rely on organisms as a habitat) are diverse ecological communities comprised of fungi, viruses, archaea, bacteria, and protozoans that often provide essential functions to their metazoan hosts (Douglas, 2015;Overstreet & Lotz, 2016). For instance, microbiome diversity has been attributed to many phenotypes that underlie overall host function, including nutrient provisioning (Gupta & Nair, 2020), metabolism (Wilke & Marrelli, 2015), and protection from natural enemies (Hegde et al., 2015;Hill et al., 2014;Tol & Dimopoulos, 2016). A primary feature of these host-associated microbiomes is substantial ecological diversity, including a large number of symbiont taxa in a microbiome assemblage and high turnover in the composition of these taxa among individual host microbiome assemblages (Alfano et al., 2019;Gupta & Nair, 2020).
The distribution and assemblage of microbes change in hosts through space and time, and understanding the drivers of community assembly remains one of the significant challenges in microbiome community ecology and has important implications for diseasemitigating strategies such as paratransgenesis (Gao et al., 2021;Jayakrishnan et al., 2018;Saab et al., 2020;Wilke & Marrelli, 2015). This is especially the case for complex organisms such as mosquitoes (Diptera: Culicidae) that spend time in both terrestrial and aquatic environments, metamorphose during their life cycle, and require discrete sources of food for maintenance and reproduction. Given this complexity, it is expected that the microbiota and their functional value in mosquitoes will change developmentally and even spatially, within the host.
The microbiome of mosquitoes consists of diverse populations of interacting symbiotic microorganisms that reside in mosquito hosts and may distribute differentially across mosquito tissues. The mosquito microbiome is a major modulator of host nutrition, metabolism, reproduction, development, immunocompetence, and behavior (Wang et al., 2011;Weiss & Aksoy, 2011). These effects can scale to impact the capacity of mosquito vectors to transmit infectious diseases (Alfano et al., 2019). Previous studies have demonstrated high diversity in the structure and composition of mosquito-associated microbial communities among mosquito species (Guégan et al., 2018;Wang et al., 2011), geographic sites (Muturi et al., 2017;Schrieke et al., 2022), developmental life stages (Wang et al., 2011), host sex (Wang et al., 2011), and host tissue . Except for endosymbiotic Wolbachia sp., which is acquired early on in embryogenesis, mosquitoes likely acquire the majority of this diversity from their environment (Duguma et al., 2013;Mancini et al., 2018). Indeed, recent studies have highlighted mosquito microbiota demonstrating habitat specificity (Schrieke et al., 2022). However, the environmental sources of these facultative symbionts and how they assemble within mosquito hosts remain inadequately resolved (Tawidian et al., 2021).
Mosquitoes interact with different microbial-rich environmental substrates at different stages of their amphibious life cycle. Research has shown that mosquito larvae have similar microbiome composition to their aquatic habitats (Dennison et al., 2014). During the terrestrial adult life stage, a portion of the gut bacteria of mosquitoes, such as Asaia sp., are acquired from the microbiomes that form on sugary plant nectars (Bassene et al., 2020). Indeed, removing mosquitoes from natural settings and allowing their microbiome to recruit in more sterile lab settings is associated with dramatic shifts in the mosquito microbiome (Akorli et al., 2019;Saab et al., 2020). Broadly, these examples suggest that the environment plays a major role in structuring mosquito microbiotas and may be the ultimate source of its diversity (Tawidian et al., 2021;Thongsripong et al., 2018).
As microorganisms from the environment colonize mosquitoes, they may sort differentially among organs and tissue types. For example, a recent study found crop and midgut-specific microbiotas within Aedes mosquitoes, which may be due to mosquito physiology (Villegas et al., 2023). The mechanisms and consequences of this assembly process remain inadequately described in mosquitoes.
However, research in other invertebrate systems indicates that such differentiation might be related to niche effects associated with the microbe in different environments within the host as well as hostmediated microbe selection during the formation of the symbiosis.
For example, Euprymna scolopes (Sepiida: Sepiolidae) is exposed to millions of bacteria in seawater, yet only a single strain of Vibrio fisheri (Vibrionales: Vibrionaceae) colonizes the light organ (Mandel & Dunn, 2016;Ruby & Lee, 1998). Previous studies have demonstrated that the host sets up a series of chemical and physical barriers to microbial colonization that selectively promote the assembly of V.
fisheri and inhibit the colonization of potential microbial competitors in the light-organ environment. Within the light organ, V. fisheri bioluminesces and reduces host mortality associated with predation through countershading in the water column. This system indicates that invertebrate hosts may actively develop and maintain symbiosis in distinct host tissues by selecting microbes that provide a beneficial function (Brooks et al., 2016).
Here, we use ecological network analyses (including nestedness, modularity, and specialization) to characterize attributes of the Aedes albopictus microbiome among individual host tissues and the mosquito's environment to better understand how mosquitoassociated symbionts source from environmental substrates and partition among different host tissues (Amend et al., 2019;Dormann & Strauss, 2014;Wright et al., 1997). Our study estimates nestedness, modularity, and specialization within the A. albopictus mosquito as a model. This mosquito is widespread globally, a medically relevant vector of several arboviruses, found across diverse habitat types, and previous studies have shown high diversity in the microbiome among individual hosts within and among populations (Medeiros et al., 2022;Seabourn et al., 2020). Overall, our analyses test a series of hypotheses that purport how microbes source and disseminate from the environment across distinct tissues of the mosquito host. We hypothesize that the environment is a source of microbes that enter a symbiosis with mosquitoes, and we predict that the lower diversity mosquito microbiomes will be a nested taxonomic subset of the higher diversity environmental microbiomes. We further hypothesize that these microbes will partition differentially within the mosquito among different organs and tissues and that a pattern of differential assortment among tissues may suggest different underlying ecological mechanisms of assembly. For instance, nested microbiomes along the mosquito alimentary canal may form in response to strong forces of dispersal from environmental sources and host tissue-specific filtering in the distinct tissues (Amend et al., 2019). Alternatively, very strong patterns of tissuespecific filtering would overwhelm the role of dispersal and lead to specialization and/or modularity among mosquito tissue microbiomes, perhaps associated with the biological functions of each tissue. Clarifying aspects of the environment that contribute to mosquito microbiomes will shed light on the origin of microbial symbionts and their process of assembly to form a host-associated microbiome and can be used to inform vector control strategies such as paratransgenesis (Gao et al., 2021;Wang et al., 2017;Wilke & Marrelli, 2015).  Table S1). Sites 1-9, 15, and 20 were primarily urban and were proximal to human habitation with natural and artificial mosquito oviposition sites detected (Supporting Information: Table S1). Sites  Table S1).
At each site, mosquitoes were collected using a battery-powered aspirator at one time point at each site (Supporting Information: Table S1). Samples were immediately placed in a plastic vial and placed in an insulated cooler with ice. The mosquitoes were identified as species using the Darsie and Ward taxonomic key (Darsie, 2005), and their sex and gravid status was determined using a standard ×10-40 dissecting scope. Female A. albopictus that were fed and/or gravid (determined by the presence of an enlarged abdomen) were excluded from downstream processing.
From the sample, a total of five host-seeking female mosquitoes were selected for further analysis. The samples were kept on ice and were processed within 4 h of collection. Individual mosquitoes were dissected to extract tissues for genetic analysis. Before dissection, whole mosquitoes were surface sterilized with a 75% ethanol wash, followed by two rinses in sterile 1× phosphate-buffered saline. Each mosquito was placed on ice and dissected with EMS High Precision and Ultra Fine Tweezers (Electron Microscopy Sciences) to expose the internal organs for dissection. The crop, midgut, Malpighian tubules, and ovaries were removed and placed separately in a sterile 1.5 microcentrifuge tube in the −80°C freezer for storage (i.e., a single tube for each tissue section). Samples of the same tissues from five same individual females captured at the same sampling event were pooled (i.e., crop, midgut, Malpighian tubule, and ovary tissue pools from the same five females collected from the same site), placed in a sterile 1.5 microcentrifuge tube, and stored at −80°C. For each of the 20 sites, there was a single sample for each of the four tissue types. However, the ovarian tissue sample from Site 1 failed to exit the molecular pipeline described, leaving a total of 79 mosquito tissue samples for analysis. Substrates that produce plant nectar, herein referred to as "plant nectar sources" (abbreviated as PNS), were collected by rubbing sterile flocked swabs (Puritan) on flower nectaries and the exudate of fallen fruits (Supporting Information: Table S2). Flocked swabs were rubbed against the surface of a single flower or fruit 10 times while rotating clockwise and placed in a sterile microcentrifuge tube. The majority of PNS samples were collected from flowers (n = 55) as compared to fruit (n = 5). Flowers and fruit were occasionally sampled from the same tree. Where possible, 250 µL of water samples were sampled from a single larval habitat (i.e., tree holes, pooled water generally <500 mL) using sterile pipettes. All environmental samples were immediately placed in an insulated cooler with ice and transported to the laboratory and stored in a −80°C freezer. While seasonal variation likely exists, our balanced design controlled for this intrinsically (i.e., we sampled all tissues and environments at each site/time).

| Bioinformatics
Bioinformatic processing of sequencing reads was performed using the bioinformatic pipeline to identify amplicon sequence variants (ASVs) (Arisdakessian et al., 2020). After quality control, sequencing resulted in 17.1 million reads (per sample mean: 59,749; per sample median: 60,128).
ASVs found in the reagent negative controls and no-template PCR controls were removed from the analysis as they represent contaminants along the molecular analysis pipeline. Exceedingly rare ASV taxa were removed if not seen more than one time (i.e., singletons), and samples found in less than 5% across all observations (i.e., rows in the data set) using the "filter_taxa" function implemented in package Phyloseq in R.
Following this filter step, 153 ASVs remained in the inventoried microbiomes of mosquito tissue samples and 1055 ASVs remained among the environmental samples were used to assess diversity measurements (Supporting Information: Tables S3 and S4).

| Statistical analysis
Nestedness describes the degree to which ecological communities with low species richness are subsets of those with higher species richness (Amend et al., 2019;Atmar & Patterson, 1993;Patterson & Atmar, 2008). A nestedness temperature (Patterson & Atmar, 2008) was estimated to test the hypothesis that low-diversity mosquito tissue (crop, midgut, Malpighian tubule, and ovary) microbiomes source from higher diversity environmental (soil, water, and plant nectar) microbiomes. Three weighted bipartite matrices were constructed: (1) environment samples unmerged (water, soil, plant nectar) + partitioned mosquito samples (i.e., crop, midgut, Malpighian tubule, and ovary), (2) environmental samples merged + partitioned mosquito samples, (3) partitioned mosquito samples completely excluding all environmental samples. Endosymbionts of Wolbachia (wAlbA and wAlbB) were excluded from the nestedness analysis as the focus of the analysis was to assess if environmentally acquired symbionts are a source for mosquito microbiomes (Ding et al., 2020).
To account for disparities in ASV abundances between mosquito and environmental samples, the bipartite matrices were resampled without replacement using "rarefy_even_depth" function in the phyloseq package in program R using "sample.size" of 2176 for the environment + mosquito matrix and 2000 for the mosquito matrix.
The nestedness ranged from 0 (perfectly nested) to 100 (random) based on a weighted bipartite matrix, implemented in the "nestedtemp" function in the vegan package in program R (Dixon, 2003). The nested temperature was compared with a distribution of randomized null communities simulated using the "vaznull" method. The Simpson's index of each mosquito and environmental microbiota was estimated with the function "diversity" in the package vegan.
Richness was measured as the total number of ASVs in a sample and was assessed on the bipartite matrices that were resampled without replacement to control for uneven sampling depth.
The QuanBiMo algorithm, following the method and code described in Dormann et al. (2009), was used to calculate modularity (Q) for the weighted mosquito tissue network. Modularity describes patterns that consist of partitions or modules of community members that have either (i) few to no interactions between other modules present or (ii) more interactions within an individual module (Dormann & Strauss, 2014;Patterson & Atmar, 2008). The null model function (100 randomizations; method "vaznull") was used to convert Q to a z-score and estimate a p value assuming one degree of freedom. A generalized linear mixed model (GLMM) implemented in the package glmmTMB in program R was used to test for the overall differences in the composition of the microbiota from mosquitoes across the four tissue types. The model included a random interaction between the tissue type variable and symbiont taxa, which permit the different symbiont taxa to have different responses in relative abundance between the different tissues and a zero-inflation component, which was allowed to vary across the levels of the symbiont taxa.
GLMMs were also used to understand how variance in the composition of the tissue-specific mosquito microbiomes was partitioned among hypothetical groupings of tissues. We created 14 dummy variables that included all potential groupings of the four tissue types, as well as a null model that included no tissue grouping variable and only a y-intercept (Table 1).
All models except the null model include a random interaction between the tissue grouping variable and symbiont taxa, which permits each symbiont taxa to have different responses in relative abundance between the tissue groups. Every model, including the null model, incorporated a fixed intercept, a fixed effect for the tissue grouping variable, a random intercept for symbiont taxa, and a zeroinflation component, which was allowed to vary across the levels of the symbiont taxa. Candidate models with different tissue grouping schemes were compared based on model fit, using corrected Akaike information criterion (AICc), and implemented in the package bbmle.
The model with the lowest AIC value was the best-fitted model. The fits of all models within 2 ΔAIC units were considered indistinguishable. ΔAICc > 2 were considered to have a better fit.
Models with ΔAICc > 7 were considered a poor fit and rejected.
Nonmetric multidimensional scaling (NMDS) analysis using the Bray-Curtis dissimilarity matrix was applied to visualize the similarity of tissue-specific mosquito microbiomes using the package "vegan" in program R.
Specialization is a metric that evaluates the extent to which microbial taxa are exclusively associated with a particular category of the host (Dormann et al., 2009). The extent to which certain tissues select for microbes was calculated using the network-wide H2 index under the "H2fun" in package vegan (Blüthgen et al., 2006). The H2 index returns a value from 0 (generalized) to 1 (specialized).
Deviations from null expectations were quantified using a distribution of null community matrices calculated using the quantitative r2table method in package vegan (Amend et al., 2019). The significance of individual independent effects on α-diversity indices was assessed through a log-likelihood ratio test that compared a full model and a nested model that lacked an independent variable of interest. We assumed the log-likelihood approximates using a χ 2 distribution. The conditional modes from these models were used to estimate the responses of each symbiont to each tissue type.  (Tables 2 and 3). For instance, there was higher sharing between the crop and the environmental substrates compared to the midgut, Malpighian tubule, and ovary (Table 2). In contrast, the ovary had the least amount of ASV co-occurrence with environmental substrates (Table 2). Similar patterns of co-occurrence were noted between the midgut and Malpighian tubules and the environment.

| Mosquito microbiome
A GLMM used to test the variation in the composition of the microbiome among mosquito tissues (i.e., crop, midgut, etc.) indicated significant variation across tissue types (p < 0.0001; Figure A1 and Supporting Information: Table S5). The conditional T A B L E 1 Fourteen models with hypothetical groupings of tissues.  Figure 2b).
However, when we removed the environmental data, and only compared mosquito tissues, this nestedness pattern was no longer detected (nested temperature = 0.87, p = 0.9, null expected range = 34.7-35.1 (95% CI); Figure 2c). This ruled out a hypothesis T A B L E 3 Number of ASVs that co-occur among environmental sample types (plant nectar, water, and soil).  (Figure 3). Two modules were identified from this analysis, crop + midgut and Malpighian tubule + ovary. Further corroborating this finding, GLMMs that tested how tissue-specific mosquito microbiomes partitioned among hypothetical groupings of tissues indicated that the models that incorporated the groupings (cg-mo and c-g-mo) best fit the data (Table 4). Respectively, models that also included different combinations of hypothetical groupings fit the data poorly (Supporting Information: Table S6). The model that incorporated a cg-mo grouping variable (AICc weight = 0.575) was 1.7 times more likely than the model that assumed a c-g-mo grouping variable (AICc weight = 0.34), though both had a delta AICc < 2.0.

Plant nectar Water
One module was shared between the crop and midgut tissues and

| DISCUSSION
The assembly of metazoan-associated microbiomes is a dynamic process that leads to substantial diversity and variation between individuals both within and among host populations. One of the pressing challenges of microbiome science is understanding the drivers of this variation (Falony et al., 2016;McLoughlin et al., 2016;Miller et al., 2018). Mosquitoes represent a particularly suitable system to explore the drivers of this variation, due to substantial changes in microbiome composition among individual mosquitoes. In this study, we use various ecological network analyses to clarify how the microbiome diversity of the mosquito compares to that of environmental substrates that mosquitoes interact with over their complex life cycle. Additionally, we clarify how this diversity partitions within mosquito host tissues. Our observed patterns (c) Only mosquito tissues (c, g, m, and o) are compared for nestedness temperature. Bacterial taxa presence in each sample type is represented as a blue rectangle. Bacterial taxa are organized by residence (left to right) across the different sample types, and rows are ordered from the highest species richness to the lowest species richness (top to bottom). The curved line represents the "Fill line." If all bacterial taxa occurred above the "Fill line," then the system would be considered to be perfectly nested (PN).
suggest that most of the mosquito microbiome is sourced and selected from the greater diversity of microbes within the environment, and that microbial taxa differentially colonize different tissues within the mosquito (Figure 4).
The diversity of the A. albopictus microbiome represents a nested subset of the diversity of microbes in the mosquito environment. In biogeography, patterns of nested biodiversity of distinct ecological communities are indicative of a source-sink relationship in species diversity (Wright et al., 1997), and these dynamics manifest as low-diversity communities nested in high- implications for mosquito microbiome assembly and the consequences for mosquito health and disease vectoring capability (Seabourn et al., 2020).
While we find that the cumulative diversity of the mosquito microbiome is largely shared with their environment, we also demonstrate that this diversity partitions differentially among distinct tissues in the mosquito corpus. While the associated microbiomes of the crop, midgut, Malpighian tubules, and ovaries were subsets of the environmental microbiome, they did not nest significantly within themselves (i.e., low diversity tissue microbiomes were not subsets of higher diversity tissue microbiomes within a host mosquito). Instead, microbial symbionts were significantly specialized among mosquito tissues and organized into two modules of commonly co-occurring bacterial symbionts: those in the digestive tissues of the crop and midgut and those in the Wolbachia-harboring tissues of the Malpighian tubules and the ovaries.
Multiple hypotheses might explain the apparent tissue specialization and modularity identified in this study. There is growing evidence that hosts actively maintain symbiosis that will support host organismal performance (Bascuñán et al., 2018;Grond et al., 2020).
Symbionts provide essential functions to their host (Martin et al., 2019), and selection may favor hosts that differentially associate with symbionts and harbor them in the right tissues to provide these functions (Mandel & Dunn, 2016). Interestingly, these data suggest a module between the microbiomes of the crop and midgut tissues, two tissues that have important functions associated with food digestion and nutrition (Calkins et al., 2017;Guégan et al., 2020). Given the shared physiological processes of the crop and midgut (i.e., nutritional functions) and their close proximity (i.e., connection via the alimentary canal), it is plausible that the host would select for symbionts capable of supporting this broad physiological function and that connectivity via the alimentary canal permits for frequent dispersal between these tissue types; however, the mechanism for this selection and indeed the function of many of the symbionts such as Burkholderia sp. would need to be studied further. The symbiont Burkholderia sp., Acetobacteraceae found within the crop and midgut tissues may assist with bloodmeal digestion by detoxifying ammonia in the gut (Feldhaar et al., 2007;Villegas et al., 2023).
Module formation may also recapitulate developmental pro-  (Chavshin et al., 2015). In this study, the Malpighian tubules have the second greatest species diversity, higher than the microbiomes of the midgut and ovaries (  (Faria & Sucena, 2013) While the environment within a host plays an important role in microbiome assembly and maintenance, microbial biology also influences microbial growth rates, which must contribute to shaping their distribution among mosquito tissues. For instance, NITRO1 (an unclassified Nitrososphaeraceae lineage), which is known to contain ammonia oxidizing strain, was entirely restricted to the crop, an area exposed to ammonia-rich substrates due to its proximity to the environment. Other studies have also found tight associations between Aedes crop tissues and Acetobacteraceae (Feldhaar et al., 2007). Furthermore, the ovaries were the only site for bacterial symbionts Pseudomonas, Acinetobacter, Methyloligellaceae (METHYL1), Sphingomonas, and Methylobacterium. Overall, these factors determine not only microbial survival but also microbe colonization potential that may be dependent on tissue structure, physiological environment, neighboring microbes, and the intramicrobial dynamics of mosquito microbiotas.
Broadly, these data demonstrate that mosquito tissue microbiomes are nested within environmental sources (soil, plant nectar, and water), indicating that mosquito microbiotas may be sourced from specific environmental sources like plant nectar, soil, and aquatic habitat (i.e., water) (Boet et al., 2020). Specifically, 98% of A.
albopictus microbiome diversity is shared with environmental sources.
Additionally, while mosquito microbial symbionts derive from environmental sources, they also assemble into distinct microbiomes across tissue types (crop, midgut, Malpighian tubule, and ovaries) and form specialized modules based on commonly co-occurring symbiont species across tissue groups (specifically, the digestive tissues of the crop and midgut and the Wolbachia-harboring Malpighian tubules and ovaries). While this study cannot differentiate the mechanisms that shape this selective assembly of mosquito symbionts from the environment, a plausible scenario would involve varied dispersal and filtering of symbionts from the environment, niche effects within the mosquito that sort species based on growth optima, and physiological mechanisms that allow hosts to select microbes and alter growth rates. Future studies should seek to clarify the specific mechanisms that govern microbiome assembly within mosquito hosts. Further understanding of the mechanisms driving module formation, nichedriven assembly, and host-mediated microbe selection demonstrated in this study will shed light on the complexity of ecological networks and can enrich our understanding of microbiome assembly within hosts. Additionally, clarifying drivers of microbiome variation in specific mosquito tissues has substantial implications for improving disease-mitigating strategies (Gao et al., 2021;Wilke & Marrelli, 2015).