The gut microbiome is an integral component of the human body that plays a role in many physiological processes. Dysbiosis, an imbalance of the microbiome, has been associated with disease states including inflammatory bowel disease, type II diabetes, and obesity, and moreover, contributes to the pathogenesis of these states. Understanding the functional mechanisms governing microbial ecology and microbe-host interactions is essential to understanding the microbiome’s role in health and disease. However, at present, functional genetic studies of diverse natural mammalian gut microbiomes remain challenging, due to a lack of genetic tools for bacteria outside of a handful of well-studied model organisms. Altering the metagenome of a complex microbial community requires novel platform technologies for genetic engineering which can operate in a generalized fashion across many different host organisms. In this thesis, I present two novel genetic tools designed for genetic modification of bacterial communities. The first, the Cas-Transposon platform, is a host-independent targeted genome editing tool that utilizes programmable, targeted transposases to mediate site-specific gene insertions into user-defined loci. The Himar1 transposase naturally inserts transposases into random TA dinucleotides in a genome, but when fused to the dCas9 RNA-guided, DNA-binding protein, the fusion protein Himar1-dCas9 targets transposon insertions to a single TA site. The activity of Himar1-dCas9 was characterized using in vitro experiments, demonstrating that site-specific transposition is dependent on guide RNA (gRNA) orientation relative to the target site and the sequence surrounding the target site, but robust to variations in DNA and protein concentration, presence of background DNA, and temperature. We additionally showed that the Cas-Transposon platform is capable of performing site-specific transposition into a plasmid in vivo in E. coli, although further optimization of the system may be necessary to effect site-specific transposition into a genomic locus. The Himar1-dCas9 protein is the first example of a transposase that inserts transposons into locations programmable by an RNA, making it a novel tool for gene insertion and knockout in potentially any organism, without relying on DNA repair by a host cell. Metagenomic Alteration of Gut microbiome by In situ Conjugation (MAGIC) is an approach to directly modify gut bacteria in their native habitat by harnessing naturally occurring horizontal gene transfer activity to deliver engineered DNA. Because many gut bacteria are difficult to cultivate and thus difficult to genetically manipulate in the laboratory, MAGIC uses donor bacteria, delivered directly into the gut environment, to conjugate mobile vectors bearing engineered genetic payloads. Using payloads with selectable markers, we identified organisms across 4 major phyla of gut bacteria that were amenable to genetic modification with libraries of conjugative vectors we created. Using a lab-adapted E. coli strain as a donor, we achieved transient expression of the engineered payload in the microbiome. We also demonstrated that engineered native gut bacteria containing conjugative vectors could be deployed back into the gut to stably recolonize and mediate secondary transfer of the payload into other microbes, potentially enabling long-term infiltration of the payload into the metagenome. The results from this study suggest that both short-term and long-term genetic alteration of the metagenome are possible by choosing different donors, and that the MAGIC platform could enable development of more diverse microbial chasses for synthetic biology applications. MAGIC could also be used to create personalized engineered probiotics for diagnostic or therapeutic applications. In Chapter 4 of this thesis, we explored the targeted use of MAGIC to genetically modify Segmented Filamentous Bacteria, a gut commensal that is important for immune regulation but recalcitrant to in vitro cultivation. The Cas-Transposon and MAGIC technologies expand our capabilities in the areas of targeted genome editing and gene delivery into bacteria, respectively. Together, they form a suite of complementary approaches to genetically engineer undomesticated gut commensal bacteria and probe the functional genetic networks in the gut microbiome, which will enhance our understanding of microbiome ecology and host-microbiome interactions. In addition, the expanded range of genetic manipulations made possible by these tools may enable production of more diverse, perhaps personalized, probiotics containing engineered functions, such as sensing disease markers or drug delivery.

Human gastrointestinal tract is complex and highly dynamic environment, harboring trillions of microbes that collectively play a pivotal role in host metabolism, immune function, and the maintenance of general health. However, studying the gut microbes and their interplay with host requires measurements at multiple dimensions is challenging since we currently lack the tools to directly measure these microorganisms at many different levels, which hinders our ability to unravel the ecology of gut microbes and their mechanistic underpinnings in human health and disease. Here, I present a set of novel techniques to study gut microbe and host-microbe interactions at unprecedented resolution, providing tremendous innovative insights to comprehensively understand the gut microbes and their environment. First, I leverage a high-throughput automation system to build a machine learning guided culturomics platform, enabling rapid isolation and culturing of personalized gut strains. Second, to better characterize the functions of these non-model microorganisms, I describe a ribonuclease-based ribosomal RNA depletion approach for microbe, paving the way for high-throughput bacterial transcriptome profiling. Next, shifting from gut microbes to the interaction with the host, I develop a technique of fecal exfoliome sequencing to robustly profile host gastrointestinal transcriptome in a non-invasive way, providing a powerful approach to study the temporal behaviors of gut cells in health and disease. Finally, to study massive genetic variants in microbiome arena, I describe a template-mediated synthesis approach for rapid and efficient generation of genetic variants sequences and demonstrate its utility in a few cases. Taken together, these techniques provide in-depth novel measurements into the ecology and physiology of human gut microbes, collectively making up an exciting set of tools for future studies.

Thesis (PhD)--Stellenbosch University, 2005. / ENGLISH ABSTRACT: This study describes the use of a gastro-intestinal model to screen lactic acid bacteria isolated from the gastro-intestinal tract of post-weaned piglets (raised on six different diets) for probiotic properties. Intestinal bacteria were isolated from ,the stomach, duodenum, jejunum, caecum, ileum and colon. The highest cell numbers (6 x 107 cfulg) were isolated from the ileum. No significant differences in viable cell counts were recorded for piglets raised on the six diets. Isolates with the best overall probiotic properties were identified as members of Lactobacillus salivarius and Lactobacillus fermentum. The two strains selected for further studies were Lactobacillus plantarum 423 (originally isolated from sorghum beer) and Lactobacillus salivarius 241 (isolated from pig intestine). Enterococcus faecalis FAIR E 92 was originally isolated from pig intestine and was included in this study as a non-pathogenic challenge strain. L. plantarum 423 produces a bacteriocin plantaricin 423, active against E. faecalis FAIRE 92. L. plan/arum 423 and L. salivarius 241 were included in the gastro-intestinal model and their adhesion to the mucus of porcine ileum studied with fluorescent-in-si/u-hybridization (FISH). A decrease in viable cell numbers of L. plan/arum 423 was recorded in the duodenum, jejunum and ileum in the presence of bile and pancreatic juice. However, higher cell numbers were recorded in the caecum and anterior colon, which suggested that strain 423 recovered from these stress factors. Plantaricin 423 was detected for up to 28 hours in the duodenum, jejunum, ileum and middle colon. Lower cell numbers (one log unit) of L. salivarius 241 were recorded in the gastro-intestinal model over seven days, compared to strain 423. Piglets of one, 14 and 28-days-old were dosed with L. plan/arum 423 and L. salivarius 241, separately and in combination (1: 1). In a separate experiment, 14-day-old piglets were challenged twice with E. faecalis FAIRE 92, followed by dosage with strains 423 and 241. New-borne piglets dosed with L. plantarum 423 gained more weight (4 kg over 19 days) compared to piglets dosed with L. salivarius 241 (2.2 kg over 19 days), or a combination of the two strains (2 kg over 19 days). Piglets of 14 and 28-days-old, on the other hand, gained more weight when dosed with a combination of strains 423 and 241. The cell numbers of E. faecalis FAIR E 92 and other enterococci decreased drastically (two log units) when the piglets were dosed with the latter two strains. Overall, piglets of various ages reacted differently when administered L. plantarum 423 and L. salivarius 241, separately or in combination. Fluorescent-in-situ-hybridization (FISH) was used to study the in vivo adhesion of L. plantarum and L. salivarius to mucus in the stomach, duodenum, jejunum, ileum, caecum and colon. The highest number of L. plantarum cells was recorded in the ileum, whereas L. salivarius favoured adhesion to the duodenum. A decrease in cell numbers of E. faecalis in the ileum mucus was recorded when a combination of the probiotic strains 423 and 241 was administered. This study provided a reliable estimation of the presence and/or adhesion of L. plantarum and L. salivarius to various parts of the porcine gastro-intestinal tract, without the use of expensive cultivation techniques. Insight was gained into the co-evolution existing between probiotic bacteria and the porcine gastro-intestinal tract, emphasizing the use of gastro-intestinal models to study the dynamics of the gastro-intestinal tract. / AFRIKAANSE OPSOMMING: Hierdie studie beskryf die gebruik van 'n gastro-intestinale model, om melksuurbakterieë wat geïsoleer is uit die spysverteringskanaal (SVK) van reeds gespeende varkies (gevoed op ses verskillende diëte) vir probiotiese eienskappe te toets. Ingewandsbakterieë is uit die maag, duodenum, jejunum, caecum, ileum en kolon geïsoleer. Die hoogste aantal selle (6 x 107 kve/g) is geïsoleer uit die ileum. Geen betekenisvolle verskille in lewensvatbare seltellings, vir varkies gevoed op ses verskillende voere is aangeteken nie. Isolate met die beste algehele probiotiese eienskappe is as Lactobacillus salivarius en Lactobacillus fermentum geïdentifiseer. Vir verdere studie is twee isolate Lactobacillus plantarum (oorspronklik uit sorghum-bier geïsoleer) en Lactobacillus salivarius (uit die varkdermkanaal geïsoleer) geselekteer. Enterococcus faecalis FAIRE 92, oorspronklik uit die varkdermkanaal geïsoleer, is in hierdie studie as 'n nie-patogeniese indikator gebruik. L. plantarum 423 produseer 'n bakteriosien plantarisien 423 wat aktief is teen E. faecalis FAIR E92. L. plantarum 423 en L. sa/ivarius 241 is ingesluit in die gastro-intestinale model, en vashegting van die bakterieë aan die mukus van vark-ileum is met fluoresensie-in-si/uhibridisasie (FISH) bestudeer. 'n Afname in lewende selgetalle van L. plan/arum 423 in die duodenum, jejunum en ileum is aangetoon in reaksie tot die byvoeging van gal en pankreatiese sappe. Hoër selgetalle is nietemin aangeteken in die caecum en voorste gedeelte van die kolon, wat 'n aanduiding gee dat isolaat 423, ten spyte van hierdie stres-faktore, oorleef. Plantaricin 423 is vir 'n tydperk (28 uur) in die duodenum, jejunum, ileum en sentrale kolon gevind. Laer selgetalle (een logaritmiese eenheid) van L. salivarius 241 is in die gastro-intestinale modeloor 'n tydperk van sewe dae aangetoon, in vergelyking met isolaat 423. Een, 14 en 28 dag oud varkies is met L. plantarum 423 en L. salivarius 241 (afsonderlik en in kombinasie 1:1) twee keer gedaag met E. faecalis FAIR E 92, opgevolg met dosering van 423 en 241. Pasgebore varkies het die hoogste gewigstoename getoon (4 kg oor 19 dae) na dosering met L. plantarum 423 in vergelyking met varkies gedoseer met L. salivarius 241 (2.2 kg oor 19 dae) of 'n kombinasie van die twee isolate (2 kg oor 19 dae). Daarenteen het veertien- en 28 dag oud varkies beter gewigstoename getoon na dosering met 'n kombinasie van isolate 423 en 241. Die selgetalle van E. faecalis FAIRE 92 en ander enterococci het drasties afgeneem (twee logaritmiese eenhede) nadat die varkies met laasgenoemde twee isolate gedoseer is. Varkies van onderskeie ouderdom het verskillend gereageer na dosering met L. plantarum 423 en L. salivarius 241 afsonderlik of in kombinasie. Fluoresensie-in-situ-hibridisasie (FISH) is gebruik om die in vivo vashegting van L plantarum en L. salivarius tot die vark mukus in die maag, duodenum, jejunum, ileum, caecum en kolon te bestudeer. Die hoogste telling van L. plantarum selle is aangeteken in die ileum, terwyl L. salivarius aanhegting tot die duodenum verkies het. 'n Afname in seltellings van E. faecalis in die ileum mukus was aangeteken na toediening met 'n kombinasie van probiotiese isolate 423 en 241. Hierdie studie het 'n betroubare bepaling van die voorkoms en/ofvashegting van L. plantarum en L. sa/ivarius isolate in verskeie gedeeltes van die varkspysverteringskanaal voorsien, sonder die hulp van duur kwekings tegnieke. Probiotiese bakterieë is in 'n gastro-intestinale model, wat die natuurlike omgewing verteenwoordig, bestudeer. Insig oor die ko-evolusie tussen probiotiese bakterieë en die SVK van die vark is verkry. Die gebruik van 'n gastro-intestinale model om die dinamika van die SVK te bestudeer is met hierdie studie beklemtoon.

In this thesis, we develop and apply top-down, quantitative approaches to gain novel insights into various complex biological systems. Beginning at the multicellular level, we study human gut microbiome dynamics from an ecological perspective. We develop computational frameworks to enable a global understanding of the spatiotemporal variability of gut bacterial abundances. We demonstrate the utility of our frameworks to elucidate the ecological processes governing abundance changes of gut microbiota. We then shift our focus to the intracellular level by investigating the metabolic limitations of cancer cell growth. We use coarse-grained mathematical modeling to identify a major growth limitation of cancer cells associated with electron acceptor deficiency, which we then experimentally validate. Collectively, these set of approaches help to decipher the organizing principles of complex biological systems at both the individual and multicellular levels.

Bacteria in the gastrointestinal tract play an important role in intestinal motility, inflammation, homeostasis, and behavior. Bacteria, through the natural synthesis of neuroactive compounds and secondary metabolites, can modulate the host immune system and communicate with the host along the signaling pathway along the gut-brain axis. Here, we functionally design, develop, test, and characterize a platform for the study of microbial-host interactions using advancements in the field of synthetic biology. First, we describe the engineering of Escherichia coli Nissle to biosynthesize serotonin within the mammalian gut using a native-plasmid optimized approach. Serotonin is crucial for neurotransmission throughout the body and may be playing a role in microbial gut-brain communication. In the gastrointestinal tract, serotonin regulates intestinal motility, cell turnover, intestinal inflammation, and gastrointestinal homeostasis. Upon serial daily oral gavages, our engineered bacterium populates a murine colon to produce serotonin locally in the mucosa layers along the epithelial lining. Changes in host physiology were observed including decreased gastrointestinal motility, increased colonic Muc2 expression, induction of host TPH2, responsible for serotonin biosynthesis in enteric neurons, and upregulation of serotonin receptors HTR3, HTR4, and HTR7 in the colon. Behavioral tests revealed a statistically significant decrease in anxiety and depression in stress-induced environments in mice treated with the engineered bacterium. This work suggests that gut bacteria engineered to modulate host gut-brain axis may have both scientific and clinical uses to study microbial-host interactions and treat gastrointestinal and behavioral mood disorders in humans. Second, we engineered bacteria to produce exogenous butyrate and other SCFAs in the murine gut. Short chain fatty acids (SCFAs) play an important role in intestinal homeostasis, fluid dynamics, inflammation, oxidative stress, and intestinal hypersensitivity and motility. With this development, we characterized the effects of our butyrate-producing bacteria on a high-fat diet and DSS-induced colitis model within the colon. Although energetically burdensome to produce, our strains produced butyrate in the colon at higher density in an actively inflamed colitis model. After 14 days of oral administration, our engineered strain (EcN:B) increased the colon length of normal wild-type mice, in high fat fed mice, and in mice with recovering and actively inflamed DSS-induced colitis. EcN:B increased mucosal barrier thickness, upregulated gene expression of the barrier integrity markers Cldn1, Ocln, Zo1, and altered crypt and villus height during inflammation recovery. Furthermore, as butyrate is known to induce Foxp3+ Regulatory T cells, we saw a 13.01% percent increase in Foxp3+ cells in the colon of mice fed our engineered bacteria. This work suggests that synthetic gut bacteria engineered to produce short chain fatty acids may have future clinical uses to treat patients with inflammatory bowel disease including Crohn’s and Colitis with future potential to serve as a therapeutic for irritable bowel syndrome, idiopathic constipation, obesity, and colorectal cancer. This platform, with the use of synthetic biology to natively engineer Escherichia coli Nissle to produce bioactive compounds in the distal gastrointestinal tract, creates a framework for future characterization of bacterial-host communication and future microbial-based therapeutics.

Bibliography: leaves 133-148 / x, 148 leaves, [3] leaves of plates : ill ; 30 cm. / Title page, contents and abstract only. The complete thesis in print form is available from the University Library. / Thesis (Ph.D.)--University of Adelaide, 1984

Ample studies have shown that perturbation of metabolic phenotype is correlated with gene mutation and pathogenesis of colorectal cancer (CRC) and rheumatoid arthritis (RA). Mass spectrometry (MS)-based metabolomics as a powerful and stable approach is widely applied to bridge the gap from genotype/metabolites to phenotype. In CRC suffers, KRAS mutation accounts for 35%-45%. In previous study, SLC25A22 that encodes the mitochondrial glutamate transporter was found to be overexpressed in CRC tumor and thus to be essential for the proliferation of CRC cells harboring KRAS mutations. However, the role of SLC25A22 on metabolic regulation in KRAS-mutant CRC cells has not been comprehensively characterized. We performed non-targeted metabolomics, targeted metabolomics and isotope kinetic analysis of KRAS-mutant DLD1 cells with or without SLC25A22 knockdown using ultra-high performance liquid chromatography (UHPLC) coupled to Orbitrap MS and tandem MS (MS/MS). In global metabolomics analysis, 35 differentially regulated metabolites were identified, which were primarily involved in alanine, aspartate and glutamate metabolism, urea cycle and polyamine metabolism. Then targeted metabolomics analysis on intracellular metabolites, including tricarboxylic acid (TCA) cycle intermediates, amino acids and polyamines, was established by using LC-MS/MS coupled with an Amide BEH column. Targeted metabolomics analysis revealed that most TCA cycle intermediates, aspartate (Asp)-derived asparagine, alanine and ornithine (Orn)-derived polyamines were strongly down-regulated in SLC25A22 knockdown cells. Moreover, the targeted kinetic isotope analysis using [U-13C5]-glutamine as isotope tracer showed that most of the 13C-labeled TCA cycle intermediates were down-regulated in SLC25A22-silencing cells. Orn-derived polyamines were significantly decreased in SLC25A22 knockdown cells and culture medium. Meanwhile, accumulation of Asp in knockdown of GOT1 cells indicated that oxaloacetate (OAA) was majorly converted from Asp through GOT1. Exogenous addition of polyamines could significantly promote cell proliferation in DLD1 cells, highlighting their potential role as oncogenic metabolites that function downstream of SLC25A22-mediated glutamine metabolism. SLC25A22 acts as an essential metabolic regulator during CRC progression as promotes the synthesis of TCA cycle intermediates, Asp-derived amino acids and polyamines in KRAS-mutant CRC cells. Moreover, OAA and polyamine could promote KRAS-mutant CRC cell growth and survival. Rheumatoid arthritis (RA) is a chronic, inflammatory and symmetric autoimmune disease and a major cause of disability. However, there is insufficient pathological evidence in term of metabolic signatures of rheumatoid arthritis, especially the metabolic perturbation associated with gut microbiota (GM). Based on consistent criteria without special diet and therapeutic intervention to GM, we enrolled 50 RA patients and 50 healthy controls. On basis of the platform of UHPLC-MS and GC-MS, were performed for the non-targeted metabolomics to investigate alterations of endogenous metabolites in response to RA inflammation and interaction with GM. 32 and 34 significantly changed metabolites were identified in urine and serum of patients with RA, respectively. The altered metabolites were identified by HMDB, METLIN database or authentic standards, and mostly metabolites were attributed into tryptophan and phenylalanine metabolism, valine, leucine and isoleucine biosynthesis, aminoacyl-tRNA biosynthesis and citrate cycle. To obtain alterations of more components in tryptophan and phenylalanine metabolism, we developed and validated a targeted metabolomics method of 19 metabolites by using LC-QqQ MS. Combining the results of targeted metabolomics with global metabolomics, significantly up-regulated kynurenine (KYN), anthranilic acid (AA) and 5-hydroxylindoleacetic acid (HIAA) simultaneously in urine and serum was found to implicate the activation of tryptophan metabolism under the condition of RA, which acted pro-inflammatory roles in inflammation and was closely correlated with GM. IDO/TDO functioned as a pro-inflammation mediator was overexpressed in RA patients. Urinary kynurenic acid and serum serotonin that have impacts on anti-inflammation in immune system were down-regulated in RA patients. The levels of phenylacetic acid and phenyllactic acid serving as a pro-inflammatory and an anti-inflammatory agent, respectively, increased in serum of patients with RA. Moreover, certain essential amino acids (EAAs), and mostly conditional EAAs were decreased in RA patients, which have been reported to inhibit cell proliferation of immune cells. In particular, deficiency of branched chain amino acids (BCAAs, valine and isoleucine) was observed in serum of patients with RA, which may lead to muscle loss and cartilage damage. The specificity of all altered metabolites resulted from RA was considerably contributed through the GM-derived metabolites. The findings revealed that GM-modulated RA inflammation was mainly resulted from tryptophan and phenylalanine metabolism, and amino acid biosynthesis, which may provide more information for better understanding the RA mechanism.

Trillions of microbes line the gastrointestinal tract to form the gut microbiome, a symbiotic organ whose supportive functions include energy production, immune homeostasis, and defense against pathogens. Disturbances to gut microbial composition, in turn, drive the pathogenesis of various metabolic, inflammatory, and carcinogenic diseases. Much effort has been dedicated to elucidating environmental triggers of gut dysbiosis, not the least of which is the consumption of medications. Antibiotics eradicate keystone commensals and enhance pathogenic behaviors of persisting pathobionts, whose resistance mechanisms can have off-target effects on human physiology and treatment response. Recent evidence indicates that the spectrum of antimicrobial compounds that disturb the gut microbiome extends far beyond traditional antibiotics, and includes commonly prescribed cardiovascular, neuropsychiatric, metabolic, and cancer medications. Although the capacity of non-antibiotic pharmaceuticals to induce gut dysbiosis is well appreciated, their impact on gut microbial function has not been studied systematically. Bacterial multi-omic profiling offers a cost-effective, high-throughput approach to understanding bacterial genetic responses to chemical perturbations, and how these functional changes might reciprocally impact relevant human phenotypes. Our laboratory, which houses a personal strain biobank of over 30,000 gut bacterial isolates spanning over 400 taxa, has established scalable pipelines for bacterial genomic and transcriptomic profiling that are readily applicable to diverse non-model gut microbes. We applied these methodologies to healthy fecal samples and bacterial isolates to elucidate strain-level responses to common pharmaceuticals with known gut microbiome associations. We first performed a gut microbiota transcriptomic screen of 19 representative fecal isolates against 20 top-prescribed orally delivered medications. Computational analysis using the Kyoto Encyclopedia of Genes and Genomes (KEGG) revealed induction of pathways associated with metabolism and multidrug resistance, including upregulation of efflux machinery by lipid-lowering drugs, antidepressants and cardiovascular medications. We discovered many bacterial responses with clinical significance, which we computationally validated using clinical metagenomic datasets. Most importantly, we showed that statin-mediated overexpression of the AcrAB-TolC efflux pump generates collateral toxicity in dietary retinol and secondary bile acids, resulting in depletion of pump-containing Bacteroidales species from patient microbiomes. We next performed the first comprehensive screen for antimicrobial activity in cancer drugs by exposing three healthy fecal samples to a panel of 41 first-line cancer therapeutics. Using 16S-genomic profiling, we identified several members of the targeted kinase inhibitor (TKI) class that induced gut dysbiosis, including first-line hepatocellular carcinoma (HCC) treatment sorafenib. We profiled natural bacterial isolates exposed to different TKI HCC treatments, and again observed transcriptional induction of conserved multidrug efflux pumps. Adaptive evolution assays identified Resistance-Nodulation-Division (RND) efflux pumps as effectors of TKI resistance. Remarkably, we demonstrated that acquired TKI resistance in evolved Bacteroidales lineages generated strain-specific cross-resistances and collateral sensitivities to several unrelated antibiotics. Collectively, our work demonstrates the importance of profiling xenobiotic impacts on the gut microbial resistome, as bacterial adaptations to pharmaceutical toxicities can feed back onto microbiome communities and the human host to affect health outcomes.

The Attwater’s prairie-chicken (Tympanuchus cupido attwateri; APC) is a species of grouse native to Texas coastal prairies and is on the critically endangered species list as a result of habitat destruction and overhunting. All of the current populations were captively bred and released into the wild. Survivorship for released APCs is very low, and individuals seldom survive to reproduce in the wild. One factor contributing to this may be an alteration in the gut microbiota as a result of captivity. Factors potentially influencing the gut microbial composition in captivity include antibiotic therapy, stress, and a predominantly commercially formulated diet. Recent studies have begun to shed light on the importance of the host microbial endosymbionts. Antibiotic administration, stress, diet, age, genotype and other factors have been shown to influence microbial populations in the gastrointestinal tracts of many different vertebrates. Sequencing of 16S rRNA gene amplicons on the Ion Torrent™ platform was used in this study to identify groups of bacteria in the cloacas as a surrogate for the gut microbiota in the APC. Antibiotic-treated and untreated birds, wild-hatched and captive-bred birds, and individuals sampled before and after release to the wild were examined. Significant differences were found between wild-hatched and captive raised birds both pre- and post release. In addition, there was extensive variation among the populations at the lower taxonomic ranks between individuals for each group of APCs. Principal coordinate analysis based on the weighted UniFrac distance metric further exhibited some clustering of individuals by treatment. These data suggest that captive breeding may have long-term effects on the cloacal microbiota of APCs with unknown consequences to their long-term health and survivorship.

Obesity is a common disease among all ages that has threatened human health and has become a global concern. Gut microbiota can affect human metabolism and thus may modulate obesity. Certain mixes of gut microbiota can protect the host to be healthy or predispose the host to obesity. Modern next-generation sequencing technique allows accessing huge amount of genetic information underlying microbiota and thus provides new insights into the functionality of these micro-organisms and their interactions with the host. Multiple previous studies have demonstrated that the microbiome might contribute to obesity by increasing dietary energy harvest, promoting fat deposition and triggering systemic inflammation. However, these researches are either based on lab cultivation studies or basic statistical analysis. In order to further explore how gut microbiota affect obesity, this thesis utilize a series of machine learning methods to analyze large amount of metagenomics data from human gut microbiome. The publicly available HMP (Human Microbiome Project) metagenomic sequencing data, contain microbiome data for healthy adults, including overweight and obese individuals, were used for this study. HMP gut data were organized based on two different feature definitions: taxonomic information and metabolic reconstruction information. Several widely used classification algorithms: namely Naive Bayes, Random Forest, SVM and elastic net logistic regression were applied to predict healthy or obese status of the subjects based on the cross-validation accuracy. Furthermore, the corresponding feature selection algorithms were used to identify signature features in each dataset that lead to the differences between healthy and obese samples. The results showed that these algorithms perform poorly on taxonomic data than metabolic pathway data though lots of selected taxa are still supported by literature. Among all the combinations between different algorithms and data, elastic net logistic regression has the best cross-validation performance and thus becomes the best model. In this model, several important features are found and some of these are consistent with the previous studies. Rerunning classifiers by using features selected by elastic net logistic regression again further improved the performance of the classifiers. On the other hand, this study uncovered some new features that haven't been supported by previous studies. The new features could also be the potential target to distinguish obese and healthy subjects. The present thesis work compares the strengths and weaknesses of different machine learning techniques with different types of features originating from the same metagenomics data. The features selected by these models could provide a deep understanding of the metabolic mechanisms of micro-organisms. It is therefore worth to comprehensively understand the differences of gut microbiota between healthy and obese subjects, and particularly how gut microbiome affects obesity.