Browsing by Subject "Microbiome"

Developing flexible tools that apply to datasets with large size and complex structure while providing interpretable outputs is a major goal of modern statistical modeling. A family of models that are especially suitable for this task is the P\'olya tree type models. Following a divide-and-conquer strategy, these tree-based methods transform the original task into a series of tasks that are smaller in size and easier to solve while their nonparametric nature guarantees the modeling flexibility to cope with datasets with a complex structure. In this work, we develop three novel tree-based methods that tackle different challenges in Bayesian hierarchical modeling. Our first two methods are designed specifically for the microbiome sequencing data, which consists of high dimensional counts with a complex, domain-specific covariate structure and exhibits large cross-sample variations. These features limit the performance of generic statistical tools and require special modeling considerations. Both methods inherit the flexibility and computation efficiency from the general tree-based methods and directly utilize the domain knowledge to help infer the complex dependency structure among different microbiome categories by bringing the phylogenetic tree into the modeling framework. An important task in microbiome research is to compare the composition of the microbial community of groups of subjects. We first propose a model for this classic two-sample problem in the microbiome context by transforming the original problem into a multiple testing problem, with a series of tests defined at the internal nodes of the phylogenetic tree. To improve the power of the test, we use a graphical model to allow information sharing among the tests. A regression-type adjustment is also considered to reduce the chance of false discovery. Next, we introduce a model-based clustering method for the microbiome count data with a Dirichlet process mixtures setup. The phylogenetic tree is used for constructing the mixture kernels to offer a flexible covariate structure. To improve the ability to detect clusters determined not only by the dominating microbiome categories, a subroutine is introduced in the clustering procedure that selects a subset of internal nodes of the tree which are relevant for clustering. This subroutine is also important in avoiding potential overfitting. Our third contribution proposes a framework for causal inference through Bayesian recursive partitioning that allows joint modeling of the covariate balancing and the potential outcome. With a retrospective perspective, we model the covariates and the outcome conditioning on the treatment assignment status. For the challenging multivariate covariate modeling, we adopt a flexible nonparametric prior that focuses on the relation of the covariate distributions under the two treatment groups, while integrating out other aspects of these distributions that are irrelevant for estimating the causal effect.

Advances in high-throughput DNA sequencing allow for rapid and affordable surveys of thousands of bacterial taxa across thousands of samples. The exploding availability of sequencing data has poised microbiota research to advance our understanding of fields as diverse as ecology, evolution, medicine, and agriculture. Yet, while microbiota data is now ubiquitous, methods for the analysis of such data remain underdeveloped. This gap reflects the challenge of analyzing sparse high-dimensional count data that contains compositional (relative abundance) information. To address these challenges this dissertation introduces a number of tools for Bayesian inference applied to microbiome data. A central theme throughout this work is the use of multinomial logistic-normal models which are found to concisely address these challenges. In particular, the connection between the logistic-normal distribution and the Aitchison geometry of the simplex is commonly used to develop interpretable tools for the analysis of microbiome data.

The structure of this dissertation is as follows. Chapter 1 introduces key challenges in the analysis of microbiome data. Chapter 2 introduces a novel log-ratio transform between the simplex and Real space to enable the development of statistical tools for compositional data with phylogenetic structure. Chapter 3 introduces a multinomial logistic-normal generalized dynamic linear modelling framework for analysis of microbiome time-series data. Chapter 4 explores the analysis of zero values in sequence count data from a stochastic process perspective and demonstrates that zero-inflated models often produce counter-intuitive results in this this regime. Finally, Chapter 5 introduces the theory of Marginally Latent Matrix-T Processes as a means of developing efficient accurate inference for a large class of both multinomial logistic-normal models including linear regression, non-linear regression, and dynamic linear models. Notably, the inference schemes developed in Chapter 5 are found to often be orders of magnitude faster than Hamiltonian Monte Carlo without sacrificing accuracy in point estimation or uncertainty quantification.

The human gut microbiome is increasingly recognized as having a causal or contributing role in a wide variety of diseases. While mechanisms by which the microbiome contributes to or triggers disease processes are myriad, short-chain fatty acid (SCFA) production has been found to be a powerful regulator of inflammation and gastrointestinal (GI) function, and may be central to the link between host and microbiota. Supplementing the diet with microbially accessible carbohydrates, termed prebiotics, is one mechanism by which SCFA production can be augmented or altered. While prebiotic therapies to increase SCFA in the gut have shown some promise in treating or preventing disease, treatment potential is limited by substantial inter-individual variation in responses to prebiotics. Determining the cause for this variation is necessary to develop treatment approaches that maximize patient responsiveness. Ultimately, tools to predict an individual’s response to a prebiotic and to guide treatment options must be developed. Here, I seek to understand the drivers of inter-individual and intra-individual variation in prebiotic response and to develop strategies to predict this response. In Chapter One, I introduce the human gut microbiome and its roles in maintaining host health and contributing to disease processes. I also present the existing evidence for substantial variation in SCFA productive response to prebiotic supplementation and highlight the need for a more nuanced understanding of the drivers of such variation. In Chapter Two, I explore the contributions of host factors and prebiotic choice to variation in SCFA production. This chapter introduces a novel in vitro fiber fermentation system, which is used throughout this thesis, and shows our methods validation of such. We find not only that host identity and prebiotic choice both impact SCFA production, but that the interaction of these terms is a significant contributor, introducing the possibility of the need for personalization. We then identify multiple host factors, including microbiota community composition and baseline SCFA metabolic state of stool, that explain some portion of inter-individual variation in prebiotic response. In Chapter Three, this relationship is further explored during the first in vivo triple-crossover prebiotic supplementation study. We supply 28 healthy adults with three different prebiotic supplements in a balanced and uniform crossover design, measuring SCFA as the primary outcome. This study makes the major contribution of quantifying the relative contribution of individual identity and prebiotic choice to butyrogenic response, and identifying individual as the vastly stronger predictor. We also identify habitual diet and baseline fecal SCFA concentrations as potential predictors of prebiotic efficacy. As a secondary analysis, we apply co-inertia analysis to draw associations between dietary choices and fecal SCFA metabolism. Together, these works highlight the need for personalization of prebiotic therapy and introduce potential biomarkers of responsiveness. In Chapter Four, we apply the concept of prebiotic therapy to graft-versus-host disease (GVHD) and show efficacy in a murine model. Importantly, we show that efficacy of prebiotics in this model of GVHD is dependent on the starting state of the microbiota, as observed through community composition analysis and functional in vitro fiber fermentation.

The kingdom Fungi is one of the major groups of the plant microbiome(Hardoim et al., 2015; Vandenkoornhuyse et al., 2015; Peay et al., 2016). Of the various plant-fungus interactions, mycorrhizal fungi that form mutualistic associations with host plants are the best studied symbiotic system(Bonfante & Genre, 2010; van der Heijden et al., 2015). Fungal endophytes represent another major type of plant-fungus symbioses(Rodriguez et al., 2009; Porras-Alfaro & Bayman, 2011). Defined as endosymbionts inhabiting a wide range of plant and lichen hosts without causing obvious symptoms, endophytes are now considered both ubiquitous and hyperdiverse (Stone, 2004; Rodriguez et al., 2009; U'Ren et al., 2012). Yet most of these fungi have to be identified using a phylogenetic approach (Arnold et al., 2009; Gazis et al., 2012; Chen et al., 2015) and remain unknown at lower taxonomic ranks (e.g., genus and species) and undefined in terms of their function in their symptomless hosts(Arnold et al., 2003; Busby et al., 2016). It is now understood that some endophytes are capable of switching to pathogenic(Wipornpan Photita et al.; Ávarez-Loayza et al., 2011) or saprotrophic(U'Ren et al., 2010; Zuccaro et al., 2011; Kuo et al., 2014) modes, but the genetic mechanisms of these switches remain unexplored. Bryophytes are a major component of the vegetation in boreal and arctic regions, where ecosystems are most vulnerable to global climate change(Turetsky et al., 2012; Jassey et al., 2013). It has been proposed that early land plants adopted a terrestrial lifestyle with the help of fungi(Heckman et al., 2001; Field et al., 2015). Mosses do not have mutualistic fungal symbionts such as mycorrhizal fungi(Davey & Currah, 2006; Field et al., 2015), but they are known to harbor diverse fungal endophytes of uncertain functions(U'Ren et al., 2010; Davey et al., 2012; Davey et al., 2013). The growth form of the moss Dicranum scoparium provided an ideal system for studying functional transitions between endophytism and saprotrophism across a senescent gradient. My PhD thesis focuses on the evolutionary history (Chapter 1) and functionality (Chapter 2, 3) of endophytic fungi.

In Chapter 1, I investigated the phylogenetic placements of fungal endophytes within the pharmaceutically and agriculturally important class Eurotiomycetes. The class Eurotiomycetes (Pezizomycotina, Ascomycota) includes various fungi with different ecological traits, including animal pathogens, saprotrophs, ectomycorrhizae, plant pathogens, rock-inhabiting fungi, lichens and endophytes(Geiser et al., 2006; Schoch et al., 2009; Gueidan et al., 2015). Phylogenetic affiliations of eurotiomycetous fungal endophytes with their ecologically diverse relatives had not been evaluated, leaving a gap in our understanding of the major evolutionary trends and ecological breadth of Eurotiomycetes as a whole. To fill this gap, we recently inferred the phylogenetic and taxonomic affinities of representatives of class 3 endophytes within Eurotiomycetes (Chen et al., 2015). Our results based on seven loci and 157 taxa revealed an undescribed new order (Phaeomoniellales) composed mainly of fungal endophytes and plant pathogens, and to a lesser extent, endolichenic and lichen-forming fungi. However, most of the deep nodes within this order were poorly supported. Interestingly, while described species of the order Phaeomoniellales are mostly plant pathogens on angiosperms (e.g., Genera Vitis, Nephelium and Prunus(Groenewald et al., 2001; Damm U. et al., 2010; Rossman et al., 2010; Thambugala et al., 2014)), endophytes within this order were mostly isolated from leaves of gymnosperms (Fig.1). These results, first-authored by the Co-PI, have been published in the journal Molecular Phylogenetic and Evolution(Chen et al., 2015).

In Chapter 2, I used metatranscriptomes of fungal ribosomal RNA to detect active fungal communities across a gradual gradient of senescence in wild-collected gametophytes of Dicranum scoparium (Bryophyta) to understand the distribution and the active component of fungal communities at a given time in adjacent living, senescing, and dead tissues. My results suggested that Ascomycota generally were more prevalent and active in living tissues, whereas Basidiomycota were more prevalent and active in senescing and dead tissues. Differences in community assembly detected by metatranscriptomics were echoed by amplicon sequencing of cDNA and compared to culture-based inferences and observation of fungal fruit bodies in the field. The combination of metatranscriptomics and amplicon sequencing of cDNA is promising for studying symbiotic systems with complex microbial diversity, allowing simultaneous detection of microbial presence, abundance and metabolic activity in symbiotic systems.

In Chpater3, I investigated the functions of D. scoparium across its naturally occurring senescence gradient and the associated fungal nutrient transporter (carbon, amino acid, phosphorus and nitrogen) activities. Higher fungal nutrient-related transporter activities were detected toward the bottom layer of the moss gametophytes. Among the four fungal nutrient types (Amino acid, carbon, nitrogen, phosphorus), the activities of nitrogen-related transporters had a drastic increase proportionally toward the bottom layer. In parallel, nitrogen breakdown was detected as the most enriched Gene Ontology term of D. scoparium for those transcripts having higher expression in the bottom layers. I analyzed the most abundant fungal nitrogen-related transporters in my dataset, the ammonium transporters, using a phylogenetic approach. I revealed that all ammonium transporters actively expressed in association with D. scoparium belong to the MEPg clade. Different sets of potential plant-microbe communication/defense/symbiosis-related genes are highly expressed in top vs. bottom layers, which suggest different mechanisms are involved in plant-fungus associations in photosynthetic vs. decomposing tissues.

Ballast water is a known vector for the global translocation of microorganisms. Research into the ballast microbiome recently accelerated following a ballast-associated outbreak of Vibrio cholerae in Peru during the 1990s that killed over 10,000 people. Over the last two decades there has been increasing regulation surrounding ballast water treatment with the aim of protecting human and environmental health, recently culminating in the approval of the International Maritime Organization Ballast Water Management Convention in September 2017. The Convention requires shipowners to install and use of ballast water treatment systems within an established timeline. However, many basic questions remain surrounding the composition of the ballast water microbiome. This dissertation strives to address several of these questions, which will allow shipowners, regulators, and enforcement agencies to make more informed decisions in an uncertain space.

The first aim of this dissertation is to characterize the bacterial microbiome of ballast water aboard vessels arriving at several ports, and identify characteristics to explain observed variations. Published research that utilizes high throughput sequencing (HTS) technology to explore microbial community dynamics is relatively rare. In this study, 16S rRNA gene sequencing and metabarcoding were used to perform the most comprehensive microbiological survey of ballast water arriving to hub ports to date. In total, 41 ballast, 20 harbor, and 6 open ocean water samples were characterized from four world ports (Shanghai, China; Singapore; Durban, South Africa; Los Angeles, California). In addition, total coliforms, Enterococcus, and E. coli were cultured to evaluate adherence to International Maritime Organization standards for ballast discharge. Five of the 41 vessels – all of which were loaded in China – did not comply with standards for at least one indicator organism. Dominant bacterial taxa of ballast water at the class level were Alphaproteobacteria, Gammaproteobacteria, and Bacteroidia. Ballast water samples were composed of significantly lower proportions of Oxyphotobacteria than either ocean or harbor samples. Linear discriminant analysis (LDA) effect size (LEfSe) and machine learning were used to identify and test potential biomarkers for classifying sample types (ocean, harbor, ballast). Eight candidate biomarkers were used to achieve 81% (k nearest neighbors) to 88% (random forest) classification accuracy. Further research of these biomarkers could aid the development of techniques to rapidly assess ballast water origin.

The first portion of the second aim of this dissertation evaluates the prevalence of indicator organisms and antibiotic resistance genes (ARGs) in ballast water compared to harbor and ocean water. The Ballast Water Management Convention, which sets forth guidelines regarding indicator organisms in ballast water, entered into force in September 2017. Notably, antibiotic resistance is absent from the Convention. We collected a total of 74 ballast and harbor samples from Singapore; Shanghai, China; Durban, South Africa; and Los Angeles, California. Eight ocean samples were collected for comparison. This research examines the concentration of indicator organisms and prevalence of three antibiotic resistance genes (ARGs). The ARGs examined in this study range from ubiquitous (sul1 – sulfonamide) to common (tetM – tetracycline) to rare (vanA – vancomycin). In ballast samples, there were significantly higher concentrations of E. coli in Singapore and China when compared to South Africa (Singapore, p = 0.040) and California (Singapore, p < 0.001; China, p = 0.038). Harbor samples from China had significantly higher concentrations of E. coli than Singapore (p = 0.049) and California (p = 0.001). When compared to ocean samples, there were significantly higher concentrations of normalized tetM in ballast samples from California (p = 0.011) and Singapore (p = 0.019) and in harbor samples from California (p = 0.018), Singapore (p = 0.010), and South Africa (p = 0.008). These findings indicate that there are differential microbial loads in different ports. Furthermore, there appears to be elevated levels of certain ARGs in ballast and harbor water when compared to ocean water, which may indicate that ballast is either translocating higher concentrations of certain ARGs or that conditions in the ballast tanks are placing selective pressure in favor of some ARGs.

The second portion of the second aim of this dissertation evaluates the prevalence of fungal pathogens in ballast water compared to harbor and ocean water. Several recent studies have explored the ballast water microbiome, but few have examined the fungal mycobiome and, to our knowledge, no studies have examined fungal pathogens in ballast. The fungal mycobiome was characterized by collecting a total of 65 ballast, harbor, and ocean samples from four major ports and sequencing the fungal internal transcribed spacer (ITS) region. A literature review of the resulting taxa was performed to identify well-studied fungal pathogens. Hosts of the identified pathogens included corals, humans, animals, plants, and crops. Of ballast samples, 21.4% had at least one fungal taxon pathogenic to corals and 81.0% had at least one fungal taxon pathogenic to humans. The majority of the fungal community in 19% of ballast samples were pathogenic taxa. A significantly higher proportion of the fungal community was composed of pathogens in Shanghai compared to all other sample sites (p = 0.025). The identification of fungal pathogens in ballast, especially those affecting corals and humans, highlights the need to further research the ballast microbiome to protect human and environmental health from the threat of fungal pathogen introductions via ballast.

The first portion of the third aim of this dissertation examines correlations between the bacterial and fungal microbiome. This chapter utilizes high throughput sequencing (HTS) and machine learning to examine and integrate the 16S and 18S rRNA genes and fungal ITS region. These sequencing regions were examined using the SILVA v132 and UNITE reference databases. The highest correlation was found between the communities in Silva_16S and UNITE_ITS (0.74). There was a higher proportion of positive inter-kingdom correlations than positive intra-kingdom interactions (p = 0.032). Understanding the reasons for this difference will require additional research under more controlled conditions. Finally, a machine learning model was used to examine the accuracy of assignment when using each sequencing region and reference database. There was significantly higher accuracy when using SILVA v132 (0.814) when compared to UNITE (0.664) (p < 0.001). In the short term, future research with the goal of classifying ballast water samples based on location or ballast water residence time should be performed using the 16S rRNA gene and SILVA v132 reference database. Future research to curate other sequencing regions or the UNITE reference database in the aquatic ecosystem may improve the utility of these tools when attempting to classify ballast water.

The second portion of the third aim of this dissertation examines correlations between the bacterial microbiome and non-target chemical analysis. To our knowledge, no literature is available that examines the interaction between microbes and chemicals in ballast water. This study addresses this gap in the literature by examining correlations between bacterial taxa and non-target chemical compounds. All strong interdomain and intradomain Pearson correlations (i.e. |r| > 0.7) were positive (54 interactions); however, the majority of Pearson correlations at all levels were negative (25,497 of 33,920; 75.2%). The reasons for this pattern are unclear and further research to isolate specific bacterial taxa and non-target chemicals may provide useful insight. In addition, machine learning was performed using bacterial, chemical, and bacterial and chemical markers. The bacterial markers appeared to perform well at differentiating California, China, and South Africa; however, accuracy was poor when classifying Singapore samples. Chemical markers appeared to supplement this deficiency, and the lowest out-of-bag error was achieved using a combined bacterial-chemical marker set with 6 features (6.67%). Further research with a larger sample size is necessary to appropriately test the markers identified in this work; however, this research serves as a proof-of-concept for a combined bacterial-chemical machine learning classification approach to ballast and harbor water samples.

Over the last few decades, there has been an increasingly large body of research focused on the ecology and function of the gut bacteria, collectively known as the gut microbiome. This work has focused on the human gut microbiome as well as other animals including livestock and model organisms that can be genetically and experimentally manipulated. These organisms include laboratory rodents, fruit flies, worms, and pigs, to name a few. The background and future directions of this field are reviewed in Chapter 1 of this thesis. My work in the laboratory of Dr. John Rawls has focused on several aspects of the gut microbiome in different contexts, including how the gut microbiome is affected by nutritional challenges, host diseases, and lifestyle interventions, as well as the factors affecting microbiome composition, which might inform how we can develop strategies to manipulate the microbiome.In Chapter 2, I focus on the question of what genetic factors affect microbiome composition. To do so, I focus on a specific gut microbe known as Phocaeicola (Previously Bacteroides) vulgatus, or Pvu. Pvu is among the most abundant Bacteroidaceae species. Pvu also has myriad health associations in human studies, is an early life colonizer, and an efficient long-term colonizer in both humans and mice. However, the genes required for Pvu to establish itself in a complex microbiome are unknown. To address this gap in knowledge, I present experiments using transposon mutagenesis and insertion sequencing (INSeq) to understand Pvu colonization of the mammalian gut. This reverse genetics approach identifies several potential pathways that Pvu might use to colonize and persist in a complex microbiome. I further elucidate the functions of a hypothetical secreted protein, Pvu777, that is required for competition in vivo in a complex microbiome. In vivo competition experiments using genetically engineered Pvu strains recapitulate these findings in Pvu777 as well as the downstream putative fatty acid transporter Pvu776. Comparative genomics suggests that the operon containing Pvu777, which consists of the predicted DNA Binding/Histone-like protein Pvu778, Pvu777, and Pvu776 may be unique to Pvu and closely related gut Bacteroides and Phocaeicola species. RNA Seq approaches link Pvu777 to outer membrane and envelope functions. In conclusion, we identify a variety of pathways required for Pvu to colonize and persist in a complex microbiome using an INSeq screen, and elucidate the potential functions of one of the genes emerging from this screen using a range of experimental approaches. These findings could be used to inform further fitness-based studies of Pvu, but could also be used to inform methods to control its in vivo abundance, in addition to suggesting mechanisms that could be used to design efficiently colonizing engineered gut bacteria. In Chapter 3, I focus on the question of how nutritional challenges affect the gut microbiome using a zebrafish model of starvation. Starvation is a widespread nutritional challenge for which animals possess many physiological adaptations. However, current research into animal starvation has focused mainly on tissue histopathologies associated with starvation, excluding the physiological changes in the GI tract as well as the gut microbiome. In Chapter 3, we used RNA sequencing and 16S rRNA gene sequencing to uncover changes in the intestinal transcriptome and microbiome of zebrafish subjected to long-term starvation and refeeding compared to continuously fed controls. Starvation over 21 days led to increased diversity and altered composition in the intestinal microbiome compared to fed controls, including relative increases in Vibrio and reductions in Plesiomonas bacteria. Starvation also led to significant alterations in host gene expression in the intestine, with distinct pathways affected at early and late stages of starvation. This included increases in the expression of ribosome biogenesis genes early in starvation, followed by decreased expression of genes involved in antiviral immunity and lipid transport at later stages. These effects of starvation on the host transcriptome and microbiome were almost completely restored within 3 days after refeeding. Comparison with published datasets identified host genes responsive to starvation as well as high-fat feeding or microbiome colonization, and predicted host transcription factors that may be involved in starvation response. Overall, the results presented in Chapter 3 demonstrate that there may be distinct stages of starvation that lead to specific changes in gut microbial ecology and host GI tract transcriptome. These stages of starvation are largely reversible upon refeeding and the ensuing changes in host gene expression and microbiome composition may be an adaptive response to recover from starvation. This work could thus inform future research investigating the roles of specific bacterial taxa in host starvation, as well as mechanistic studies looking at the roles of specific host genes in starvation and refeeding using genetically modified hosts. In Chapter 4, I suggest studies that could extend from the work presented in Chapter 2, focusing on the role of individual genes in the Pvu 777 operon in vivo as well as within Pvu. I also suggest potential roles for the predicted DNA-Binding/Histone-like protein Pvu778 in the regulation of Pvu gene expression and Pvu fitness. I conclude by considering the evolutionary conservation of the 777 operon among Pvu and close relatives, and methods to investigate the fitness requirements of the 777 operon in these related bacterial species as well. Structural studies of Pvu777 are also proposed, which would help clarify the function and potential binding partners for the Pvu777 hypothetical protein. Thus, the studies proposed in Chapter 4 would help provide a clearer picture of the functions, regulation, and evolutionary history of the 777 operon, which would underscore its importance as a potentially conserved operon involved in Pvu fitness in vivo.

The composition and function of mammalian gut microbiomes are shaped by complex endogenous and exogenous factors that present on evolutionary and proximate timescales. In the Anthropocene era, host-associated microbiota are inevitably, yet differentially, influenced by natural and anthropogenic factors that vary across individuals and populations. In this dissertation, I used descriptive and experimental approaches, largely within a single species, the ring-tailed lemur (Lemur catta), to probe the roles of host physiology, environmental conditions, anthropogenic perturbation, and microbial environment in shaping primate microbiota across scales. First, I conducted a broad investigation of ring-tailed lemur gut microbiota and soil microbiota across 13 lemur populations (n = 209 individuals) spanning this species’ natural range in Madagascar, as well as multiple captivity settings in Madagascar and the U.S. By analyzing the lemur and soil microbiota, I showed that lemur gut microbiota vary widely within and between wild and captive populations, and that lemur and soil microbiota covary, suggesting a role for environmental acquisition in shaping interpopulation variation. Second, I analyzed vaginal, labial, and axillary microbiota of female ring-tailed lemurs and Coquerel’s sifakas (Propithecus coquereli) at the Duke Lemur Center (DLC) to demonstrate the influences of stable traits (e.g., species identity and mating system) and transient traits (e.g., ovarian hormones and forest access). We found that the effects of transient traits build on underlying differences mediated by stable traits. Third, and further focusing on DLC lemurs, but with a concentration on anthropogenic influence, I worked with a team of researchers to perform an experimental manipulation in ring-tailed lemurs to determine the influence of antibiotic treatment, with or without subsequent fecal transfaunation, on lemur gut microbiomes. I applied ecological frameworks to show that different facets of lemur microbial communities, such as bacterial diversity and composition, followed different recovery trajectories following antibiotic treatment. Fourth, I expanded my focus back to multiple ring-tailed lemur populations in natural and captivity settings to investigate the links between anthropogenic disturbance and antibiotic resistance genes (ARGs). I analyzed ARGs in wild and captive lemurs and soil from their habitats to show that lemur ARGs were correlated with anthropogenic disturbance and covaried with soil ARGs; lemur resistomes reflects multiple routes of ARG enrichment, including via antibiotic treatment or environmental acquisition. Integrating across these four data chapters, my results reveal that (a) the foundations of lemur-associated microbiomes are structured according to broad environmental conditions (e.g., wild vs. captive populations), but that between and within these broad categories, lemur microbiota are sensitive to more nuanced environmental variation, (b) lemur microbiota and resistomes co-vary with environmental microbiota, demonstrating the potential role of environmental acquisition in shaping host-associated communities across varying environments, and (c) integrating host microbial data across scales (e.g., at the individual and population level) with data on multiple facets of microbial communities (e.g., diversity, composition, membership, and resistomes), was key to providing a holistic perspective on host-associated and environmental microbe interactions across different microbial landscapes.

A cost-saving strategy of large-scale algae cultivation that can lead to more economical production of algal food, feed, and fuels involves reusing the cultivation water after algae harvesting. Few studies have focused on predictors of algae growth success in reused water, or explained these results in terms of algal ecology. Factors such as dissolved organic matter accumulation and interactions with bacteria are also understudied in the context of water reuse, yet could inform cultivation decisions to maximize water reuse without losses in algal productivity. This dissertation investigated trends in previous studies and also used an experimental approach to determine how reusing cultivation water affects algae growth, dissolved organic carbon (DOC) accumulation, DOC release rates, and relative abundances of bacteria. Based on over 80 previous studies testing algae growth in reused water, algae taxon was the only factor significantly associated with algae growth response. A possible explanation for this result could be differences in the amount and composition of DOC excreted by different algae. I therefore experimented with three taxonomically and physiologically distinct algae, two diatoms and a green alga, to test their growth responses, DOC excretion rates, and the extent of DOC accumulation in reused water. Algae growth response in reused water varied by algae, yet was not correlated with DOC concentrations. Additionally, DOC concentrations steadily increased with each water reuse, suggesting a build-up of recalcitrant DOC after bacteria, or possibly algae, degraded more labile DOC. To further explore the extent of strain-specific growth responses, I tested the effect of reused water from a self-inhibitory algae strain on other algae strains. This reused water did not inhibit two other algae strains, suggesting that the inhibitory mechanism was strain-specific and was likely from build-up of a certain DOC compound. Across the three algae cultures, different bacteria taxa became enriched or depleted in reused water, despite all cultures being exposed to the same reused water source. DOC composition and concentration in reused water were therefore likely not driving observed differences in final bacteria communities, and DOC produced from the growing algae may be more influential. Overall, results from this dissertation support strain-specific features of algae growth responses, and suggest that algae screening processes should include tests in reused water. Identifying algae strains with uninhibited growth responses will be important for implementing water reuse for algae cultivation, to ultimately improve the economic feasibility of algae bioproducts.

Marine microalgae are photosynthetic microbes that are a potential source of fuels, animal feed, and other specialized products. Large scale cultivation of microalgae occurs in open, outdoor raceway ponds, which are exposed to the natural environment and these cultures quickly become a complex milieu of microbes. Microalgae interact with attached and free-floating bacteria found in their medium, with both positive and negative outcomes. To investigate the diversity and dynamics of microbes associated with these systems, samples were collected during multiple growth cycles of two biofuel-relevant microalgae strains, Desmodesmus sp. and Oocystis sp. in ~4,500 L outdoor raceway ponds. Microbiome community composition and diversity was dramatically different between ponds from the two algae and from the natural microbiome of the treated seawater used in pond medium. In spite of variable environments, the pond microbiomes were most similar to their inoculum PBR (photobioreactor) communities suggesting the importance of priority effects or environmental conditioning by the host algae. Ponds when both algae strains were grown were dominated by Rhodobacteraceae and Saprospiraceae while unhealthy microbiomes were dominated by Cytophagaceae and Puniceicoccaceae. Microbiome change was variable over time and resulted in different community structures at the time of algae harvest. Variation in the microbiome community structure was driven by the strain of algae grown, time, pond temperature and percent oxygen saturation. These results provide insight into this industrial ecology and are a foundation for future microbiome research to improve microalgae production.

As climate change continues to alter our global ecosystem, ecological disturbances will increase in both frequency and intensity. These ecological disturbances have been shown to significantly impact microbial communities, often altering composition and/or function. However, while disturbance events are commonly studied in host-associated and engineered systems, there has been relatively little research into the causes, persistence and population-level impacts of disturbance in the dynamic ocean. Furthermore, the research that has been performed on marine microbiome responses to perturbation has been limited to the a priori identification of disturbances, limiting the scope of current research to events considered disturbances from the researchers’ perspective (e.g. hurricanes, contamination, etc.). In this dissertation, I investigate ecological disturbances using a microbiome-centered approach focusing on the role of perturbation in shaping marine microbial assemblies. In my first chapter, I provide an overview of current and historical theoretical frameworks in disturbance ecology and discuss the potential limitations of these approaches. In my second chapter, I apply a microbiome-centered approach to identify disturbances in the coastal ocean. Using three years of observations from a coastal time series, we identified ten disturbance events based on the largest week-over-week changes in microbiome composition. These microbiome disturbances were not clearly linked to specific environmental changes and did not exhibit common responsive taxa. Further, a category 1 hurricane, the only event that would likely be classified a priori as an environmental disturbance, was not an outlier in microbiome composition, but did enhance a bloom in seasonally-abundant phytoplankton. Although large storms can have significant impacts on marine microbiomes, we propose that unmeasured disturbances may have large impacts on the microbiome and may only be detected from a microbial perspective. In my third chapter, I investigate a cyclonic, Gulf Stream frontal eddy as a potential spatial disturbance that can trap and transport microbial communities along this current. By parsing our samples based on microbial community composition, we were able to identify an eddy-specific microbiome even in highly dynamic frontal zones. Furthermore, the eddy microbiome was defined by a higher abundance of a specific high-light Prochlorococcus ASV that was also enriched in slope waters, suggesting that the eddy may harbor signatures of the nearshore community entrapped during eddy formation. Although the eddy microbiome is distinct from the Gulf Stream’s, especially in cyanobacteria (e.g. lower Trichodesmium and higher Prochlorococcus), it is most similar to the Gulf Stream, suggesting eddy microbiome assembly favors environmental filtering over historical contingencies. In my final chapter, I investigate the proximal drivers of hurricane-driven responses in coastal microbiomes using experimental incubations intended to mimic specific hurricane attributes. Coastal microbiomes experienced either a dilution, intended to mimic the initial dilution of the bacterioplankton community from extensive precipitation, or to a daily addition of dissolved organic matter (DOM) intended to mimic a hurricane-induced phytoplankton bloom. The dilution treatment, which represents a more immediate and stochastic effect of hurricane landfall, led to changes in several taxonomic groups, while the DOM treatment, which represents the more latent and deterministic phase of hurricane effects that peaks days to weeks after landfall induced a consistent change among a few taxa. Thus, manipulating specific hurricane-associated drivers can help to differentiate responses to different hurricane components. Together, my dissertation approaches disturbances in time, space and complexity from a microbiome-centered perspective. By broadening our concept of disturbance to include a microbiome-centered perspective, we can better identify how these events shape the microbiome. As global climate change and the associated increase in ecological disturbances i continues to alter our ecosystems, we must understand the full scope of these disturbance responses and the potential ramifications for critical biogeochemical cycles as we strive to predict and mitigate ecosystem changes.

Microbial communities associated with animals and plants (i.e., microbiomes) are implicated in the day-to-day functioning of their hosts, and there has been an explosion of research on them. Much of this research has focused on surveys of microbial diversities across a variety of host species, including humans, with a view to understanding how these microbiomes are distributed across space and time, and how they correlate with host health, disease, phenotype, physiology and ecology. Fewer studies have focused on how these host-microbiome associations evolve. In this dissertation, we develop a computational agent-based frameworks for modelling the long-term evolution and short-term dynamics of microbiomes within a population of hosts. Our frameworks allow different ecological processes and evolutionary forces to directly or indirectly affect microbiome composition. By incorporating a Wright-Fisher or Moran genealogical population model, we combine host phylogeny with microbiome assembly to consider the shared evolutionary history between hosts and their microbes. We also incorporate how hosts acquire their microbiomes, and how the environmental microbial community available to the hosts is assembled under both neutrality and selection. Under the selective models, we allow selection to operate on both microbes and hosts and observe how microbial diversities are gradually shaped by this evolutionary feedback between hosts and microbes. Furthermore, host population division and dispersal limitation are taken into account for our short-term neutral models. Our results indicate that the extent of parental contribution to microbial availability from one generation to the next significantly impacts the diversity of microbiomes over both long-term and short-term periods: with greater parental contribution, microbiome diversity within hosts tends to decline while microbiome diversity between hosts tends to increase. We also show that the implementation of selection further depresses microbial diversities and the comparison between host level and microbe level selection suggest that the evolutionary pressures directly acting on microbes is more dominant in shaping microbial diversity patterns. Finally, we show that host population division and dispersal limitation under high host contribution further shape the diversity patterns by elevating microbiome differences between hosts and depressing microbial diversity within hosts.

Background: Acute lower respiratory infection (ALRI) is the leading infectious cause of death among children worldwide, and respiratory syncytial virus is the most common cause of ALRI. The majority of deaths occur in low- and middle-income countries, but there is a dearth of data on risk factors for poor outcomes in these settings. We endeavored to identify clinical and microbiome-related factors associated with poor outcomes among children hospitalized with RSV-ALRI in Gaborone, Botswana. Methods: Children 1-23 months of age presenting for care at Princess Marina Hospital who met the World Health Organization (WHO) definition of clinical pneumonia were enrolled in this prospective study and followed until hospital discharge or death. Nasopharyngeal swabs were collected for respiratory viral testing and microbiome analysis. Results: Young age (<6 months), household use of wood as a cooking fuel, moderate or severe malnutrition, and oxygen saturation <90% on room air were independent predictors of clinical nonresponse at 48 hours. Among HIV-uninfected infants 6 months of age or younger, HIV exposure was associated with a higher risk of in-hospital mortality. Nasopharyngeal microbiome composition and diversity were not associated with clinical nonresponse. Conclusions: We identified several risk factors for poor outcomes from RSV-ALRI among children in Botswana. These data could inform future use of RSV vaccines and therapeutics in these populations.

Cells can sense and respond to various environmental cues. In the past 25 years, this ability has been exploited in engineering many innovative applications, ranging from bioproduction and metabolic engineering, to living therapeutics and biosensing. Despite tremendous advancements in complex genetic circuit development, the field still suffers from several limitations. For instance, evolutionary pressure can hamper the long-term genetic stability and functionality of circuits. The long incubation times required for cell growth serve as a fundamental rate limiting step for routine microbiology experiments and circuit engineering. Available biological parts, such as promoters and ribosome binding sites, often confer a limited dynamic range of gene expression levels and are incompatible, exacerbating the construction of higher order circuits.

Plasmids are extrachromosomal DNAs, usually circular, that replicate independently of the host genome. Because they are easy to manipulate and engineer, plasmids have served as a popular workhorse for programming desired functions in microbial populations. Plasmids can maintain steady average copy numbers in hosts, and a specific plasmid type can be chosen to express genes at a desired level. However, recent studies have focused on the dynamic modulation of plasmid copy number as a new engineering strategy, which is still underexplored. Understanding plasmid dynamics can provide insights to harness powerful tools for engineering microbial communities and offer a new avenue to overcome the current challenges in synthetic biology.

In this dissertation, I used mathematical modeling and synthetic biology approaches to develop methods for engineering microbial communities by exploiting and manipulating plasmid dynamics. First, I examined the sources of circuit failure and studied design strategies for enhancing synthetic gene circuits’ stability in microbial hosts for robust long-term performance. I summarized the engineering strategies into two categories: (1) to suppress the chance of mutant emergence by reducing the evolutionary pressure; and (2) to suppress the relative fitness of mutants by selecting against genetic variants.

Applying some of the identified engineering strategies, I developed synthetic gene circuits, named Red Queen circuits, that can modulate the host cell viability according to its growth rate. Using the circuit, I achieved a 250% increase in host cell growth rates at the end of a 100-day long-term adversarial laboratory evolutionary experiment, during which the circuit continuously suppressed slow-growing cells. The results suggest that the circuit can serve as an effective strain engineering strategy to accelerate biotechnology and molecular biology research.

Next, I constructed another gene circuits, named ADEPT system, to regulate the collective gene expression of an engineered microbial community by modulating plasmid dynamics. By dynamically tuning the plasmid loss rate, horizontal gene transfer rate, and plasmid-mediated fitness effects, I demonstrated that the ADEPT system can tune the total gene expression with a significantly amplified dynamic range.

Finally, in the Appendix, I engineered gene circuits for targeted conjugative plasmid elimination from microbial communities. The results illustrate the potential of plasmid dynamics modulations in engineering complex microbial communities.

All organisms live in complex habitats that shape the course of their evolution by altering the phenotype expressed by a given genotype (a phenomenon known as phenotypic plasticity) and simultaneously by determining the evolutionary fitness of that phenotype. In some cases, phenotypic evolution may alter the environment experienced by future generations. This dissertation describes how genetic and environmental variation act synergistically to affect the evolution of glucosinolate defensive chemistry and flowering time in Boechera stricta, a wild perennial herb. I focus particularly on plant-associated microbes as a part of the plant’s environment that may alter trait evolution and in turn be affected by the evolution of those traits. In the first chapter I measure glucosinolate production and reproductive fitness of over 1,500 plants grown in common gardens in four diverse natural habitats, to describe how patterns of plasticity and natural selection intersect and may influence glucosinolate evolution. I detected extensive genetic variation for glucosinolate plasticity and determined that plasticity may aid colonization of new habitats by moving phenotypes in the same direction as natural selection. In the second chapter I conduct a greenhouse experiment to test whether naturally-occurring soil microbial communities contributed to the differences in phenotype and selection that I observed in the field experiment. I found that soil microbes cause plasticity of flowering time but not glucosinolate production, and that they may contribute to natural selection on both traits; thus, non-pathogenic plant-associated microbes are an environmental feature that could shape plant evolution. In the third chapter, I combine a multi-year, multi-habitat field experiment with high-throughput amplicon sequencing to determine whether B. stricta-associated microbial communities are shaped by plant genetic variation. I found that plant genotype predicts the diversity and composition of leaf-dwelling bacterial communities, but not root-associated bacterial communities. Furthermore, patterns of host genetic control over associated bacteria were largely site-dependent, indicating an important role for genotype-by-environment interactions in microbiome assembly. Together, my results suggest that soil microbes influence the evolution of plant functional traits and, because they are sensitive to plant genetic variation, this trait evolution may alter the microbial neighborhood of future B. stricta generations. Complex patterns of plasticity, selection, and symbiosis in natural habitats may impact the evolution of glucosinolate profiles in Boechera stricta.

One of the fundamental goals of statistics is to develop methods which provide improved inference in applied problems. This dissertation will introduce novel methodology and review state-of-the-art existing methods in three different areas of applied statistics. Chapter 2 focuses on modelling subcommunity dynamics in gut microbiome data. Existing methods ignore cross-sample heterogeneity in subcommunity composition; we propose a novel mixed-membership model which models cross-sample heterogeneity using the phylogenetic tree and as a result is robust to mispecifying the number of subcommunities. Chapter 3 reviews state-of-the-art methods in recommender systems, including collaborative filtering, content-based filtering, hybrid recommenders, and active recommender systems. Existing literature has focused primarily on bespoke applications; statisticians have an opportunity to build recommender system theory. Chapter 4 proposes a novel method of accounting for time-based design inconsistencies in Bayesian network meta-analysis models and discovers non-linear time trends in the effectiveness of vancomycin as a MRSA treatment. Chapter 5 provides some concluding remarks.

The gut microbiome is a complex ecosystem of hundreds of species that is constantly subject to perturbations as a result of day-to-day dietary variation, among other factors. In some cases, disturbances to microbial communities have been associated with lasting impacts on microbiome structure. While much research has been done to uncover sources of inter-individual variation in the gut microbiome, less focus has been given to understanding the ecological mechanisms governing intra¬-individual variation. To address this, we carried out dietary intervention studies in human cohorts and analyzed microbiome composition, metabolism, and physical particulate structure. We also employed in vitro models of the gut microbiome to manipulate variables difficult to modulate in vivo, and to collect samples with a greater temporal resolution. As discussed in more detail in Chapter 2, we found that bacterial communities retain an “ecological memory” of past prebiotic exposures, which is encoded within one day by changes to the abundance and transcriptional state of primary degraders. Chapter 3 details findings from an investigation into fecal particle size, in which we found this metric to correlate with microbiome diversity and to be stable within individuals. Together, the results presented in this dissertation present fundamental new insights into the ecology of the gut microbiome.