Certain microbes have the remarkable ability to form direct electronic connections with their extracellular environment, harnessing existing redox gradients to drive their own life processes. The recognition of extracellular electron transfer (EET) as a widespread microbial phenomenon that plays an important role in biogeochemistry is recent, and is accompanied by new understanding of how this process impacts physical and chemical characteristics of the environment. Our current knowledge of the mechanisms and the microbial physiology behind EET is based largely on studies of two well studied isolates, Geobacter sulfurreducens and Shewanella oneidensis. Nevertheless, nature is not well represented by isolated organisms. Identification of the larger set of “electroactive” microbes from across much of the phylogenetic spectrum is critical for complete understanding of microbial controls on redox chemistry in natural environments. Genomic information for these organisms has the potential to expand our understanding of the mechanisms of growth that rely upon EET. This information should be particularly useful if it is obtained in an experimental setting that enables manipulation of redox potential.

In this doctoral dissertation, microbial electrochemical cells (MXCs) are employed as a tool to select for and characterize the subset of electroactive microbial community members. The main advantage of MXCs is that they use a solid, inert electrode in place of natural minerals as the terminal electron acceptor (TEA) for microbial metabolism. This enables direct control over the redox potential and overcomes the chemical heterogeneity and variation in surface reactivity of natural minerals. Experiments are inoculated using microbial communities from subsurface sediments sampled from a major Department of Energy research site near the town of Rifle, Colorado, USA. This site has general relevance as a riparian zone aquifer; prior research has generated thousands of microbial genomes from samples collected across multiple subsurface environment types and impacted by a variety of experimental manipulations (including acetate amendment). Complex electroactive microbial communities, including anode-associated biofilms, are cultivated in the MXC experiments using the sediment inoculum and acetate as the electron donor. Experiments include two independent MXCs with duplicate compartments.

In Chapter 1, we subjected the biofilms and associated planktonic cells to large changes in anode redox potentials and characterized the responses of these organisms using electrochemical methods. Community composition was documented using 16S rRNA gene-based surveys. Our results strongly suggested that TEA redox potential is an important determinant of the makeup of mineral-respiring communities.

Chapter 2 tested the hypothesis that microbes populating anode-associated biofilms electrochemically adapt to imposed redox conditions on a short timescale (minutes). Through a series of cyclic voltammetry experiments, the overall biofilm responses to changing redox potential were observed. The voltammetry data suggested inherent redox flexibility in the microbial EET network, very likely conferred by multiheme c-type cytochromes known to be critical for extracellular respiration in most known electroactive organisms. We performed a series of experiments that allowed us to rule out within-biofilm pH changes as the cause for the observed shifts.

In Chapter 3, metagenomic methods were employed to genomically characterize the MXC microbial communities. Genome-resolved metagenomics enabled the recovery of comprehensive genetic information for the majority of community members without the requirement for isolation of individual organisms. Hundreds of near-complete genomes were reconstructed and used to compare biofilm and planktonic organisms across experimental systems. Specifically, biofilm vs. planktonic growth preferences for most organisms were defined. Importantly, genomes were used to predict genes that could enable growth on electrodes. Biofilm-associated bacteria are known to have genomes that are strongly enriched in multiheme cytochromes and to have associated outer-membrane porins implicated in electron transfer complexes. In addition to organisms related to known electroactive species, several novel organisms (from an expected undiversity of phyla) were implicated in electrode growth. The extensive metagenomic data assembled previously from the Rifle site was used to link specific organisms enriched in the MXCs to those from Rifle, and thereby yielding insights into their roles in the natural environment.

Coupling electrochemical methods with genomic and metagenomic DNA sequencing provided new information about microbial growth in MXCs and the response of the overall microbial community to imposed changes in redox potential. The hypothesis that TEA (and by proxy, mineral) redox potential shapes community composition was supported. The question of adaptation of microbes to changing environmental redox conditions was investigated in community context, without reliance on isolation methods. The findings suggested that large multiheme cytochromes evolved to cope with intrinsic mineral redox heterogeneity and fluctuating redox conditions. Finally, genome-resolved metagenomics enabled the identification of novel, putatively electroactive organisms functioning as part of complex microbial consortia, with widely different metabolisms from those of the well-studied type strains. This contributes to basic knowledge of EET in natural systems, and opens new avenues for the application of electroactive organisms in biologically-based technology, engineering and remediation.

Humans spend approximately 90% of their time indoors, yet we know very little about the microbial ecosystem of the built environment and how it impacts occupants. Here we use infants hospitalized in a neonatal intensive care unit (NICU) as a model system to track the exchange of microbes between room and occupants. By leveraging high-throughput sequencing and other “omics” technologies, we conducted four major studies to broadly address the composition of microbes populating NICU surfaces, how these microbes migrate to the infant gut, and once in the gut, how these microbes compete for resources. Over the course of these campaigns we collected and processed over 5,000 samples from hospital room surfaces and over 300 infant fecal samples creating the largest collection of hospital samples to be interrogated with next-generation sequencing techniques. Using an approach that reassembles the entire 16S rRNA gene from room amplicons and gut metagenomics data, we discovered several organisms on room surfaces before their detection in the infant gut. Once in the gut, we used a metaproteomics technique to investigate the metabolisms of early infant gut colonizers. Unlike the anaerobic gut environment of older children and adults, we discovered a relatively high utilization of aerobic pathways in many of the facultative anaerobes colonizing the infant gut. We also observed niche partitioning amongst closely related Citrobacter strains in our strain- resolved proteomics data, providing insight into how early colonizers compete in the nascent infant gut. To better understand biomass trends inside and outside the gut, we developed an assay to quantify 16S rRNA gene copies using droplet digital PCR (ddPCR). We discovered a surprising amount of variation in bacterial densities across different NICU environments. These data also allowed us to adapt a novel in silico data cleaning method that leverages the quantification of negative controls to provide data less impacted by the inherent noise of low- biomass amplicon workflows. Cleaner data allowed us to apply a machine learning classifier that showed each infant’s room had a distinct microbial fingerprint. To validate this result, we conducted a metagenomics campaign on pooled room samples from six different infants. After assembly and binning, we were able to recover hundreds of high quality genomes. Utilizing genomes from this dataset and previously isolated genomes from our lab, we discovered the same strains in the room as in infants. Further, we found several taxa frequently isolated from infant gut samples in this NICU are the same strains in the NICU room metagenomes. Overall, the analysis from this work suggests that where a premature infant is born and the history of room occupancy can impact its gut microbiome development.

By sequencing environmental DNA and reconstructing microbial genomes, we can obtain insight into the previously hidden microbial world. This approach, known as genome-resolved metagenomics, has been utilized to study microorganisms in a variety of environments. Small sample sizes were common in genome-resolved metagenomics studies of the past, and thus few statistical methods of analysis were applied to the data resulting from these small-n studies. Instead, the analyses were focused on other aspects that did not require statistical methods, such as the identification of metabolic pathways possessed by the genomes and the phylogenetic relationships between organisms. However, in recent years, decreased sequencing costs and greater availability of computational resources have enabled scientists to sequence and process hundreds of samples for a single study. This dissertation demonstrates the application of several statistical and machine learning methods for the interpretation and strategic analysis of data from high-throughput genome-resolved metagenomic studies. Through the combination of new methods with previously existing methods, this work illustrates potential benefits that quantitative methods of analysis can offer to the field of genome-resolved metagenomics.

The first chapter of this dissertation serves as an example of a traditional genome-resolved metagenomics study, using primarily manual methods of analysis after the main steps of the data processing pipeline (including assembly, binning, and annotation) are complete. The manual methods of analysis applied in this small-scale study enable us to understand what microbes are present in a particular bioreactor community, and what metabolic functions these microbes are capable of. This contrasts with the much more data-intensive studies in the latter chapters, in which manual analyses would not be an efficient use of the data.

The second and third chapters, which are both focused on very large-scale data from the premature infant gut microbiome, illustrate the use of statistical methods for deciphering relationships in complex systems. This includes machine learning techniques applied to metagenome-associated genomes to make predictions that may potentially be useful in determining optimal care for a patient, as well as more basic statistical methods that allow us to better understand the gut microbiome and how it is influenced by external factors.

The fourth chapter is focused on the development of a new method that takes the hierarchical structure of genome-resolved metagenomic data into account. With genes in pathways, pathways in genomes, and genomes in communities of microorganisms, traditional ways of comparing samples fail to fully elucidate the biological systems because not all levels of the hierarchy are accounted for. To address this problem, the new concept described here allows for the inclusion of both functional and phylogenetic data to best utilize the wide breadth of information available in genome-resolved metagenomic data. The combination of quantitative approaches with genome-resolved metagenomics may lead to a more robust understanding of microbial communities.

Culturing isolated microorganisms can be challenging, not only because usually the detailed environmental conditions where organisms grow optimally are not known, but also because many of them need to grow in the presence of other organisms. High-throughput sequencing and other `omics' technologies provide important approaches for the study of microorganisms in their natural environments. Specifically metagenomics methods enable culture-independent surveys of organisms and functions in microbial consortia, and can yield near-complete genomes of the most abundant community members, and partial genomes of lower abundance organisms. When coupled to community proteomic and/or transcriptomic analyses it is possible to predict what functions are being expressed within the community. Therefore, `omics' technologies provide a means for the study of community physiology and ecology in natural systems.

Acid mine drainage (AMD) is a mining-related problem caused by sulfide mineral dissolution coupled to microbial iron oxidation, which leads to acidification and metal contamination of the environment. The Richmond Mine AMD community is currently the best-studied AMD system. Bacteria of the genus Leptospirillum, of the Nitrospira phylum, generally dominate Richmond Mine AMD microbial communities. Current studies show that Leptospirillum rubarum (group II) tends to dominate early-formed biofilms, and Leptospirillum group II 5way CG (a genotype related to L. rubarum) or L. ferrodiazotrophum (group III) increase in abundance as environmental conditions change.

In chapter 1, community genomics was used to reconstruct the near-complete genome of Leptospirillum ferrodiazotrophum, and report the genome annotation and metabolic reconstruction of L. ferrodiazotrophum, Leptospirillum rubarum and an extrachromosomal plasmid associated to these bacteria. In addition, proteomic analyses were used to evaluate protein expression patterns in three AMD biofilms. Results indicate that, despite sharing only 92% identity at the 16S rRNA level, L. rubarum and L. ferrodiazotrophum share more than half of their genes. Both bacteria are motile, acidophilic iron-oxidizers, as evidenced by the presence of cytochrome Cyt572 and an electron transport chain. They are chemoautotrophs, using a reverse tricarboxilic acid (TCA) cycle for carbon fixation. Their metabolic potential indicates that L. rubarum and L. ferrodiazotrophum are capable of amino acid and vitamin biosynthesis, fatty acid biosynthesis, flagella biosynthesis, synthesis of polymers such as cellulose, and the synthesis of compatible solutes for osmotic tolerance. Only L. ferrodiazotrophum is capable of nitrogen fixation, although proteins were not detected by proteomics in the analyzed biofilms. Proteomic analyses indicate that core metabolic proteins are similarly expressed in both bacteria, however high expression of many hypothetical proteins unique to each Leptospirillum might contribute to their differentiation within the biofilms.

In chapter 2, the partial genome reconstruction of a new Leptospirillum bacterium, which is closely related to L. ferrodiazotrophum, is reported. The bacterium represents ~ 3% of the sequenced community, and comparison of its 16S rRNA gene sequence with those of other Leptospirilli identifies it as a new group within the Leptospirillum clade: Leptospirillum group IV UBA BS. The bacterium grows in unusually thick, Archaeal-dominated biofilms where other Leptospirillum spp. are found at very low abundance. It shares 98% 16S rRNA sequence identity and 70% amino acid identity between orthologs with L. ferrodiazotrophum. Its metabolic potential indicates that it too is a motile, iron oxidizing chemoautotroph capable of nitrogen fixation, although nitrogen fixation expression was not observed. Leptospirillum group IV UBA BS is distinguished from the other Leptospirilli in that it contains a unique multicopper oxidase likely involved in iron oxidation, and the presence of two clusters of hydrogenase genes. The cytoplasmic hydrogenase is likely used to take up H2 during nitrogen fixation, while the membrane-bound hydrogenase might be involved in anaerobic H2 oxidation for energy generation. Community transcriptomic and proteomic analyses confirm expression of the multicopper oxidase, as well as the expression of many hypothetical proteins and core metabolic genes. Transcription of hydrogenases in the only biofilm in which the nitrogen fixation operon in L. ferrodiazotrophum is transcribed points to potential cooperative interactions between the two bacteria.

AMD has long been considered a simple, low-diversity ecosystem. In chapter 3, a new view of the diversity of organisms in AMD was obtained by deep sequencing of the small subunit (SSU) rRNA from 13 biofilm communities. A total of 159 taxa, including Archaea, Bacteria, and Eukaryotes, were identified. Leptospirillum spp. dominate the samples, and members of diverse phyla, such as Actinobacteria, Acidobacteria, Firmicutes, Alpha-, Beta-, Gamma-, and Delta-Proteobacteria, Chloroflexi, and Deferribacter were present at low abundance. Interestingly, members related to Magnetobacterium spp. of the Nitrospira phylum were detected. These bacteria have not been reported present in AMD environments, and they have not been identified in community genomics datasets. However, the presence of magnetosome-like structures observed by cryo-TEM in some AMD biofilms supports the transcriptomics results. The findings indicate that it is dominance by a few taxa, and not lack of complexity of the system that has made AMD environments model systems for the study of microbial physiology and ecology.

In Chapter 4, non-ribosomal transcriptomic reads were mapped to several genomes of AMD organisms, including Archaea, Bacteria, and viruses, in order to evaluate the expression profiles of genes and non-coding regions (ncRNAs) in biofilms at increasing stages of development. More than 95% of the genes in the most abundant Leptospirillum group II and group III bacteria were detected by at least one transcriptomic read, indicating that the whole genome is transcribed at some level. More than half of the genes in the Archaea G-plasma and Ferroplasma Type II, and a virus associated to Leptospirillum were also detected. Transposases, cytochromes, and ncRNAs were among the most highly expressed genes in all samples. Gene expression profiles indicate that Leptospirillum group II 5way CG and L. ferrodiazotrophum prefer growth at higher pH and lower temperature, conditions generally present in bioreactors, while the opposite is true for L. rubarum and G-plasma, who prefer conditions found in early to mid-developmental stage environmental biofilms. High levels of expression were observed for a novel ectoine riboswitch predicted in the Leptospirillum group II genome, as well as for other non-coding RNAs. Results provide new insight into understanding functioning and adaptation of acidic ecosystems.

Viruses of Bacteria (bacteriophages) and Archaea have the ability to significantly alter the structure and function of microbial communities. Thus, it is critical to obtain a greater understanding of the dynamic interaction between microbial hosts and their associated viral populations. The CRISPR-Cas system in Bacteria and Archaea serves as a method to connect viruses to their hosts and provides insight into virus-host interactions. A clustered regularly interspaced short palindromic repeat (CRISPR) locus and CRISPR-associated (Cas) proteins function together in the CRISPR-Cas adaptive immune system. Transcripts of the spacers that separate the repeats in the CRISPR locus confer immunity through sequence identity with targeted viral, plasmid, or other foreign DNA. The CRISPR locus can be interpreted as a historical timeline of virus exposure as spacers are incorporated in a unidirectional manner at the leader end of the CRISPR locus.

Metagenomic approaches were employed to simultaneously analyze CRISPR loci in microbial hosts and sequences of their associated viruses to examine: 1) the viral response to CRISPR spacer diversification in a closed model system, 2) the retention of older spacers without co-existing targets through time, 3) the information about population histories revealed from the CRISPR locus, and 4) the factors impacting phage and viral diversity in a model natural system. The host-phage system of Streptococcus thermophilus DGCC7710 and phage 2972 was used as a closed model system whereas the low diversity microbial communities growing as biofilms atop acid mine drainage (AMD) within the Richmond Mine at Iron Mountain, CA, USA was used as a model natural system.

S. thermophilus DGCC7710 was challenged with phage 2972 and the resulting host and phage populations were examined after one week of co-culturing in order to explore how co-existing, co-evolving hosts and phage populations establish. Additions of new spacers converted the clonal CRISPR locus into a diversified locus, with multiple sub-dominant CRISPR strain lineages present in the final S. thermophilus population. All phage mutations that circumvented three early-acquired spacers were localized in the proto-spacer adjacent motif (PAM) or near the PAM end of the proto-spacer, suggesting a strong selective advantage for the phage that mutate in this region. The sequential fixation or near fixation of these single mutations indicates selection events so severe that single phage genotypes ultimately gave rise to all surviving lineages.

The CRISPR loci of Bacteria and Archaea and their associated viral and plasmid populations from AMD biofilm microbial communities were examined from samples collected over eight years. Notably, CRISPR loci were present in most populations. It was shown that CRISPR loci in some AMD microorganisms retain older CRISPR spacers over long time periods, despite not targeting any abundant coexisting viruses or plasmids. In order to investigate this phenomenon, the two CRISPR loci from a dominant archaeal G-plasma population were reconstructed and viral targets were identified via spacer matches. A polyclonal bloom of viruses was detected, and G-plasma population with highly diverse CRISPR loci emerged. In collaboration, a mathematical model was developed to link documented patterns of genomic conservation in CRISPR loci to an evolutionary advantage against persistent viruses. The subset of hosts that retain old spacers seemingly lacking any matches to current viruses may indicate tuning of CRISPR-mediated immunity against low abundance viruses that may re-emerge.

Through retention of older spacers as well as the acquisition of new spacer sequences, CRISPR-Cas systems can provide insight into recent population history. The link between spacers and their viral targets and the relative position and order of spacers in the loci are key characteristics that can be uncovered via population metagenomic analysis. New bioinformatics methods were developed and used to analyze the CRISPR loci of Leptospirillum group II bacterial population and its associated bacteriophage AMDV1 in biofilms sampled over five years in the AMD site. Spacers throughout the locus target the same phage population (AMDV1), but there are blocks of consecutive spacers without AMDV1 targets and only newer spacers target plasmid populations. This suggests the consistent co-existence of the bacteria with the AMDV1 phage population, with periods when this phage was prominent, and a fluctuating plasmid pool. The approach of examining the pattern of CRISPR spacers with targets may have direct application to tracking the potential sources of medically- and defense-relevant microbial strains. Spacer matches can identify phage or plasmids previously existing in the host's environment and indicate phage and plasmid diversity levels over time, thus constraining the past history of the population.

CRISPR spacers can be used to identify sequences derived from viruses or plasmids within a metagenomic dataset. In fact, with metagenomic datasets sequenced from AMD biofilms, CRISPR spacer targeting enabled detection of a very wide variety of previously unknown viruses and plasmids. Fine scale examination of the AMDV1 phage population diversity over time revealed meter-scale spatial variation, but somewhat reproducible seasonal genotypic abundance patterns. There is evidence in CRISPR loci of Leptospirillum group II that excision events removed large blocks of spacers that only match phage or plasmid sequences present at earlier times. These CRISPR locus and phage diversity patterns suggest that there is viral-imposed selection for host strains as well as host-related selection for virus types.

The function of the CRISPR-Cas system was only recognized a little over five years ago. In the intervening period, the majority of research has focused on unraveling the biochemical and mechanistic underpinnings of Cas and spacer function. In this thesis, the emphasis has been on understanding the importance of population diversification for long-term CRISPR-Cas system-based immunity, the significance of spacer retention, and the utility of locus analysis for ecological and evolutionary studies. Hosts and viruses were examined in two different systems. The challenge experiment between S. thermophilus and phage 2972 isolate cultures demonstrated the rapid evolution of host and phage populations as the direct result of the addition of CRISPR spacers and the fixation of phage mutations. Examination of CRISPR loci and associated viruses and plasmids populations in the AMD system suggests CRISPR loci reconstruction enables investigation of past viral and plasmid exposure. It was shown that older spacers may be retained in order to target reappearing or low abundance viruses. Spacers with co-existing matches provide host immunity and offer a method for detection and genomic analysis of new viruses. Evaluating results from both systems has provided a more complete understanding of the CRISPR-mediated host-virus dynamics in microbial communities.

Microbial commensals can stably colonize the adult gut for years and play key roles in host metabolism and pathogenic defense. However, much less is known about persisting early colonizing strains in infants. Given their early-arrival time, it is plausible that they play critical roles in shaping the trajectory of the assembly of the infant gut microbiome and the maturation of the immune system. The importance of persisting early colonizers in the early life gut microbiome and overall infant health, thus, motivated us to identify those strains and to investigate factors that contribute to their persistence. In my first two chapters, I leveraged longitudinal metagenomics data to explore the topic of persistence in two separate yet evolutionarily intertwined entities: bacteria (Chapter 1) and viruses that predate bacteria, bacteriophages (Chapter 2).

In Chapter 1, I performed strain-resolved analyses of fecal samples of premature and full-term infants collected throughout their first year of life and identified that ~11% of the bacterial colonizers persisted in the infant gut for nearly one year. I then evaluated factors associated with strain persistence and found that bacterial persisters, particularly Bacteroides and Bifidobacterium, were more prevalent in full-term than in preterm infants, and were primarily maternally transmitted. Comparative genomic analyses revealed that strains encoding genes for surface adhesion, iron acquisition, and carbohydrate degradation are more likely to persist in the infant gut than strains lacking these genomic traits.

Bacteriophages (phages) prey on bacteria for survival. Despite their early detection in the infant gut, little is known about the roles they play in gut microbiome assembly. In Chapter 2, I applied strain-resolved metagenomics to examine phage persistence in pre- and full-term infants during their first three years of life. The collection of maternal fecal samples that were three years apart enabled me to not only identify persisting phage strains in mothers but also evaluate the influence of maternal virome on the development of infant gut viromes. I found that maternal gut viromes were more stable than those of infants, with ~20% of phage strain persisting for three years, whereas only ~9% of phages persisted in infants. Similar to bacterial persisters during infants’ first year of life, I also observed that maternally transmitted phages were more likely to colonize the infant gut long-term. Population genetic analyses suggested that phage persisters had a significantly higher nucleotide diversity than non-persisters during the initial colonization phase. This could be partly attributed to the presence of diversity-generating retroelements (DGRs), as these genetic elements were enriched in phage persisters. Further, we found that phages that used an alternative genetic code (recoding of either the TAG or TGA stop codon) were more likely to persist than phages that used standard code. Overall, insights from the first two chapters have implications for developing more effective microbiome interventions such as probiotics and/or phage-based microbiome engineering.

High-throughput metagenomics studies have significantly advanced our understanding of microbial communities from all natural ecosystems. However, this type of analysis is susceptible to contamination, especially those that are internally derived. We noticed signs of within-study, cross-sample contamination when analyzing data for my first chapter, which motivated an in-depth investigation of contamination in metagenomics datasets. In my Third Chapter, I developed a strain-based framework for the detection of internally- and externally-derived contaminants and applied it to two large-scale clinical studies. I demonstrated insidious contamination in metagenomics datasets and flagged the necessity of routinely assessing contamination beyond negative and positive controls. The research provided a devised strain-resolved workflow that is generalizable and can be applied to any microbiome dataset.

Gut microbial interactions directly influence human health. However, the complexity and individuality of the gut microbiomes, as well as the lack of tractable gut microbiome models, have impeded us from linking pivotal gut consortium interactions to resulting host-associated phenotypes of interest. To decipher microbial interactions in their native communities, in my Fourth and Final Chapter, I established stable, reproducible, and person-specific laboratory gut consortia using clinically relevant infant stool samples. I then used these laboratory microbiomes, along with metagenomics functional predictions, to investigate mechanisms explaining infant-specific responses to 2'-fucosyllactose (2’FL), a prevalent human milk oligosaccharide (HMO) and a common infant formula additive. I focused on the growth of Bifidobacterium breve, a prevalent infant gut commensal that plays key roles in early-life gut microbiome assembly and immune system development. B. breve typically cannot metabolize 2’FL on its own, yet in some infants, it demonstrated 2’FL-induced growth. Leveraging infant gut microbiome cultivation and genome-resolved functional analyses, I found that the presence of community members encoding extracellular fucosidases was critical for the enrichment of B. breve when 2’FL was supplemented. Together, this work uncovered key, yet infant-specific, microbial interactions in metabolizing a common dietary component. Given the significant health impacts of Bifidobacterium in infant development, this work also provided strategies for enhancing the colonization of autochthonous B. breve and possibly other Bifidobacterium species. Over time, insights gained from these individualized in vitro microbiomes can lead to the development of more precise and effective personalized microbiome treatments.

Microbial eukaryotes are important pathogens, environmental quality indicators, integral components of natural microbial communities, and critical for understanding our own evolutionary history. Yet, microbial eukaryotes are an often neglected component of microbial ecology studies. Common metagenomic techniques, such as 16S rRNA gene sequencing, fully omit eukaryotes, and they are frequently ignored in shotgun-metagenomic sequencing projects. A methodology was developed for recovering eukaryotic genomes from metagenomes that relies upon a newly developed machine learning-based method, EukRep, to separate Eukaryotic scaffolds from prokaryotic scaffolds prior to binning. In this way, eukaryotic gene predictors can be applied to eukaryotic scaffolds, eliminating one of the largest challenges to properly binning eukaryotes in shotgun metagenomic samples. The effectiveness of EukRep was tested on both mock communities constructed from reference bacterial, archaeal, and eukaryotic genomes in silico as well as on natural microbial community samples and shown to enable the recovery of near-complete eukaryotic genomes including high-quality fungal, protist, and rotifer genomes from complex environmental samples. Thus, this approach enables consistent genome reconstruction and prediction of metabolic and behavioral potential for eukaryotes as well as their associated communities in a culture independent, natural microbial community context.

A EukRep-based approach was used to investigate the effect of addition of organic carbon to a geyser-associated microbial community. Crystal Geyser, a CO2-driven geyser in Utah (USA), provides large volumes of deeply sourced fluids, thus is well suited for studying microbial communities in high CO2 environments. Upon addition of organic carbon there was a substantial change of the community metabolism, with selection against almost all candidate phyla bacteria and archaea and for eukaryotes. Near complete genomes were reconstructed for three fungi placed within the Eurotiomycetes and an arthropod. While carbon fixation and sulfur oxidation were important functions in the geyser community prior to carbon addition, the organic carbon-impacted community showed enrichment for secreted proteases, secreted lipases, cellulose targeting CAZymes, and methanol oxidation. The results demonstrate the broader utility of EukRep for reconstruction and evaluation of relatively high-quality fungal, protist, and rotifer genomes from complex environmental samples. This approach opens the way for cultivation-independent analyses of whole microbial communities.

Fungi are common members of the human microbiome, but are often excluded from metagenomic studies due to the large size and complexity of Eukaryotic genomes. Here, targeted Eukaryotic genome recovery was performed on over a thousand metagenomes from premature infant fecal samples and twenty-eight metagenomes from the neonatal intensive care unit (NICU) housing the infants. Samples were screened for the presence of Eukaryotes using a machine learning classifier, and de novo genome assembly, curation, and annotation was performed on identified samples. Seventeen distinct Eukaryotic genomes were recovered (median completeness 91%; median size 15.6 Mbp), including genomes from four strains of Candida albicans, seven genera of fungi, and two organisms (Diptera (fly) and Rhabditid (nematode)) with no previously sequenced genomes of the same family. Seven percent of infants were colonized by a Eukaryote during the first months of life, and prevalence was significantly associated with administration of maternal antibiotics and particular bacterial taxa. All NICU samples had detectable fungal communities (median relative abundance 2%, full range 0.3-24.1%), and different locations in the NICU had distinct Eukaryotic microbiomes. Near-identical genomes of Purpureocillium lilacinum were recovered from both infant and NICU samples (99.999% average nucleotide identity), highlighting the potential for environmental NICU fungi to colonize premature infants. Zygosity and potential aneuploidy were determined for all assembled genomes, and regions with loss of heterozygosity (indicative of recent genome evolution) were detected in some C. albicans genomes. This study resolved Eukaryote dynamics in the NICU and premature infant gut samples, and reveals potential reservoirs of unexpected eukaryotic diversity within the hospital environment.

Candida parapsilosis is the third most common cause of invasive candidiasis. C. parapsilosis infections have been continually increasing in prevalence over the past two decades, and at significantly higher prevalence in neonates than other at risk populations, marking its importance as an emerging pathogen. Despite this, C. parapsilosis is understudied. The recovered C. parapsilosis genomes contain small genomic regions with highly elevated levels of Single Nucleotide Variants (SNVs), which we refer to as SNV hotspots. SNV hotspots are shared between strains, with some unique to C. parapsilosis strains from a single hospital. Four of the C. parapsilosis genomes have a high copy number (4-16) RTA3 gene, a lipid translocase previously implicated in antifungal resistance, potentially indicative of adaptation to antifungal treatment. Additionally, time course metatranscriptomics and metaproteomics were performed on a premature infant with a documented C. parapsilosis blood infection, offering a rare look at the in vivo expression and protein landscape of a Candida species. C. parapsilosis in situ expression is highly distinct from culture settings, but also highly variable, demonstrating the importance of studying Candida in situ in addition to culture settings.

Mono Lake, CA, is a high alkalinity, hypersaline lake with an unusually productive ecosystem largely supported by benthic and planktonic algae. A species of choanoflagellate from Mono Lake that forms a multicellular, hollow rosette filled with bacteria, but little was known about this choanoflagellate and its associated microbial community. This association is of interest given choanoflagellates are the closest living relatives to animals and the analogy between rosette-enclosed consortia and animal gut microbiomes. Metagenomic shotgun sequencing was performed in order to reconstruct genomes for the choanoflagellete and its associated community. EukRep was used for eukaryotic sequence identification and enabled genome recovery, genome completeness evaluation and prediction of metabolic potential of both the choanoflagellate nuclear and mitochondrial genomes. The nuclear draft genome measures 49 Mbp in length, contains 11052 predicted genes and appears to be near complete. Interestingly, its extracellular proteins have a higher isoelectric point compared to marine choanoflagellates, likely an adaptation to their saline, high pH environment. Characterization of bacterial communities leveraged samples taken from choanoflagellate rosette enriched and choanoflagellate rosette depleted samples in order to distinguish bacteria within and outside rosettes. Across all samples, 23 near-complete bacterial genomes were recovered, primarily belonging to Gammaproteobacteria, Bacteroidetes, and Spirochaeta. Of these, seven were found only in the choanoflagellete enriched samples, suggesting that these bacteria are partitioned into the rosette interior. Overall, the research provided insights into the composition and metabolic interactions between an ordered assemblage of single celled eukaryotes and its enclosed microbiome.

In this work, genome-resolved and culture-independent methods are employed to study microbial eukaryotes in a variety of natural community contexts, ranging from animal microbiomes, the hospital room, and environmental communities. The development of EukRep and subsequent incorporation into metagenomic pipelines represents an important methodological advance for the comprehensive study of the structure and ecology of natural microbial communities and provides new insights into community functioning.

Anthropogenic changes in climatic conditions, such as the timing and amount of rainfall, can have profound biotic and abiotic consequences on grassland ecosystems. Grassland's plant and animal phenology are adapted to the ecosystem's wet and cold winters and hot and dry summers and changes to this pattern will have profound consequences in aboveground community structure. Changes in climatic conditions and aboveground communities will also affect the soil biogeochemistry and microbial communities. Soil microbes are an essential component in ecosystem functioning, as they are the key players in nutrient cycling. This thesis investigated the direct and indirect effects of climate change on the structure, composition and abundance of grassland soil microbial communities. The research used the high-throughput technique of 16S rRNA microarrays (Phylochip) to detect changes in the abundances and activities of soil bacterial and archaeal taxa in response to changes in precipitation patterns, aboveground plant communities, and soil environmental conditions. The research took advantage of alongterm climate change experiment that simulated both an increase and an extension of the current winter season in northern California. Five years into the experiment, soil samples and aboveground plant diversity were collected before and after each treatment for two consecutive years. The variability in soil microbial communities after natural wet-dry rainfall events was also investigated. Results showed that, at the community level, soil microbial communities are very robust and resilient to intensified or extended rainfalls during the winter but under extreme and unusual weather events their community structure can be altered. On the other hand, an increased in moss biomass in the plots that received additional water during the spring and fluctuations in soil moisture content (precipitation models and wet-dry patterns) caused changes in soil environmental conditions which in turn affected the activity and abundance of some microbial taxa/guilds. Soil organic carbon and inorganic nitrogen were among the environmental variables that correlated the most with these changes in microbial groups. Considering the great importance soil microbes have in ecosystem functioning, the approach developed here will find application for monitor responses of keystone microbial species/guilds to future changes in climatic conditions. These responses should be taken in consideration for future soil management and conservation practices, and the impacts included in future climate change models.