ANNA SEEKATZ, PHD. RESEARCH EXPERIENCE Graduate Research Assistant

Transcription

1 ANNA SEEKATZ, PHD PhD, Program in Molecular Microbiology and Immunology Institute for Genome Sciences, University of Maryland School of Medicine EDUCATION Ph.D., Program in Molecular Microbiology and Immunology Institute for Genome Sciences University of Maryland School of Medicine, Baltimore, MD Thesis Title: The Effect of Live-attenuated and Wild-type Shigella Strains on the Gastrointestinal Microbiota in Cynomolgus Macaques B.S. in Biology, magna cum laude Western Washington University, Bellingham, WA Cellular and Molecular Biology Outstanding Student in Biology RESEARCH EXPERIENCE Graduate Research Assistant University of Maryland School of Medicine, Baltimore, MD Department of Microbiology and Immunology Institute for Genome Sciences Graduate Advisor: Claire M. Fraser, Ph.D. Thesis project: Effect of oral live-attenuated Shigella vaccine and wild-type Shigella infection on the intestinal microbiota of cynomolgus monkeys Performed DNA extraction, multiplex PCR, qpcr, 16S rrna sequence analysis, and gained knowledge in UNIX command line, R software, MySQL database management, Perl script editing Rotation project #2 Feb University of Maryland School of Medicine, Baltimore, MD Department of Microbiology and Immunology Rotation Mentor: Stefanie Vogel, Ph.D. Rotation Project: Examined the pathway of Francisella tularensis induction of IFN-b in murine macrophages Collected and cultured primary murine cells, conducted RNA extractions for RT-PCR and detected proteins using western blotting Rotation project #1 Dec University of Maryland School of Medicine, Baltimore, MD Department of Microbiology and Immunology Department of Microbial Pathogenesis Rotation Mentor: Patrik Bavoil, Ph.D. Rotation Project: Identified protein-protein interactions in Chlamydia trachomatis, Chlamydia pneumoniae, specifically chlamydiaphages Cloned open reading frames and expressed these in recombinant cells,

2 ANNA SEEKATZ PAGE 2 Undergraduate Research Western Washington University, Bellingham, WA Biology Department Undergraduate Mentor: Jeff Young, Ph.D. Research Projects: Determined function of proton pumps in Arabidopsis thaliana plasma membrane H + - ATPase gene family (AHA) using reverse genetic studies, identify unique mutants Performed random mutagenesis of Arabidopsis and conducted environmental and developmental screens of isolated AHA mutants, and validated results using PCR PUBLICATIONS Cole LE, Laird MHW, Seekatz AM, Santiago A, Jiang Z, Barry E, Shirey KA, Fitzegerald KA and Vogel SN. Phagosomal retention of francisella tularensis results in TIRAP/Malindependent TLR2 signaling. J. Leukoc. Biol., 2010; 87: Eloe-Fadrosh EA, Seekatz AM, Drabek EF, McArthur MA, Rasko DA, Sztein MB, and Fraser CM. Impact of oral typhoid vaccination on the human gut microbiota and correlations with S. Typhispecific immunological responses. PLoS One, 2013 (submitted). Seekatz AM, Eloe-Fadrosh EA, Panda A, Rasko DA, Panda A, Shipley ST, DeTolla LJ, Sztein MB, and Fraser CM. Differential response of the cynomolgus macaque gastrointestinal microbiota to shigella infection. PLoS One, 2013 (under review). ABSTRACTS / PRESENTATIONS Seekatz AM. The effect of an oral live-attenuated shigella vaccine and wild type shigella infection on the intestinal microbiota of cynomolgus monkeys. Maryland Branch ASM Annual Student Poster and Oral Presentation Meeting, University of Maryland School of Medicine, Baltimore, MD; May 3, Oral Presentation. Seekatz AM, Eloe-Fadrosh EA, Panda A, Shipley ST, DeTolla LJ, Sztein MB, and Fraser CM. The Effect of an oral live-attenuated shigella vaccine and wild type shigella infection on the intestinal microbiota of cynomolgus monkeys. Graduate Research Conference, University of Maryland School of Medicine, Baltimore, MD; April 5, Poster Presentation. Seekatz AM, Eloe-Fadrosh EA, Panda A, Shipley ST, DeTolla LJ, Sztein MB, and Fraser CM. The effect of an oral live-attenuated shigella vaccine and wild type shigella infection on the intestinal microbiota of cynomolgus monkeys. International Human Microbiome Congress, Paris, France; March 19-21, Poster Presentation. Seekatz AM, Eloe-Fadrosh EA, Panda A, Shipley ST, DeTolla LJ, Sztein MB, and Fraser CM. Effects of live-attenuated shigella vaccine and wild type shigella challenge on the intestinal microbiota of cynomolgus monkeys NIAID Cooperative Centers on Human Immunology (CCHI) Annual Meeting, Atlanta, GA: December 13-14, Oral Presentation. Seekatz AM, Liu Z, Panda A, Shipley ST, DeTolla LJ, Sztein MB, and Fraser-Liggett CM. 16S rrna-based pyrosequencing reveals fluctuation in the diversity of bacterial populations over time in the cynomolgus macaque intestinal microbiota. International Human Microbiome Congress, Vancouver, BC, Canada; March 9-11, Poster Presentation. Seekatz AM, Hsiao WWL, Khan AQ, Panda A, Shipley ST, DeTolla LJ, Levine MM, Sztein MB, and Fraser-Liggett CM. Longitudinal analysis of bacterial populations in the cynomolgus

3 ANNA SEEKATZ PAGE 3 monkey gut using 16S rrna-based pyrosequencing. 110 th General Meeting of the American Society for Microbiology, San Diego, CA; May 23-27, Poster Presentation. Seekatz AM, Hsiao WWL, Khan AQ, Panda A, Shipley ST, DeTolla LJ, Levine MM, Barry EM, Sztein MB, and Fraser-Liggett CM. The effect of immunization with an oral shigella vaccine on the gut microbiota in cynomolgus monkeys. 109 th General Meeting of the American Society for Microbiology, Philadelphia, PA; May 17-21, Poster Presentation. AWARDS J. Howard Brown Award recipient for outstanding graduate research at the Maryland Branch ASM Annual Student Poster and Oral Presentation Meeting, May Graduate Research Conference Poster Award for poster entitled The Effect of an Oral Live Attenuated Shigella Vaccine and Wild Type Shigella Infection on the Intestinal Microbiota of Cynomolgus Monkeys at the University of Maryland Graduate Research Conference in Baltimore, MD, April NIH Travel Award for the International Human Microbiome Congress in Paris, France, March 9-11, PROFESSIONAL SOCIETY MEMBERSHIPS American Society for Microbiology Member LEADERSHIP/TEACHING Graduate Student Association, University of Maryland, Baltimore, MD Microbiology and Immunology Student Representative GSA Executive Board Member: Secretary GPILS Core Course, University of Maryland, Baltimore, MD Student Discussion Leader Dec GPILS Bioinformatics Class, University of Maryland, Baltimore, MD Student Assistant 2010, 2011 Maryland Science Center After School Science Fair, Baltimore, MD Ask a Scientist Panel Leader Dec. 2012

4 Abstract Title of Dissertation: Dissertation Directed By: The Effect of Live-attenuated and Wild-type Shigella on the Gastrointestinal Microbiota in Cynomolgus Macaques Claire Fraser, PhD Professor of Medicine, Microbiology and Immunology Director, Institute for Genomic Sciences University of Maryland School of Medicine Little is known about the role of the gastrointestinal microbiota in susceptibility to infection with enteric pathogens and response to live oral vaccines. This study examined the effect of immunization with an oral live-attenuated Shigella dysenteriae 1 vaccine and challenge with wild-type S. dysenteriae 1 on the gastrointestinal microbiota of cynomolgus macaques using 16S rrna analysis. Multi-dimensional cluster analysis identified distinct bacterial community types within healthy macaques. The microbiota found in association with Mauritian macaques is distinct from and characterized by significantly higher diversity than the microbiota found in macaques from other geographic origins. Mauritian macaques also contain genetically distinct microsatellites in loci spanning the major histocompatibility complex (MHC) region, providing a possible link between the MHC repertoire and the intestinal microbiota. The intestinal microbiota in distinct macaque populations responds differently to immunization and subsequent challenge with wild-type Shigella, and is altered in fecal samples collected post-immunization and post-challenge in macaques from Indonesia, Indochina and the Philippines, but not from Mauritius. Specifically, Shigella exposure results in the appearance of a community type that is dominated by Enterococcus, a genus typically present at low abundance. While both Mauritian and non-mauritian macaques exhibit anti-shigella antibody responses upon immunization and challenge, clinical symptoms of shigellosis post-challenge are only observed in non-mauritian macaques. These studies highlight the importance of further investigation into the possible protective role of the microbiota against enteric pathogens and the importance of host genetic backgrounds in conducting vaccine studies.

5 The Effect of Live-attenuated and Wild-type Shigella on the Gastrointestinal Microbiota in Cynomolgus Macaques by Anna M. Seekatz Dissertation submitted to the faculty of the Graduate School of the University of Maryland, Baltimore in partial fulfillment of the requirements for the degree of Doctor of Philosophy 2013

6 Copyright 2013 by Anna M. Seekatz All Rights Reserved

7 Acknowledgments I would like to acknowledge the Fraser lab for all of their support and help over the years. My mentor, Dr. Claire Fraser, has been a great source of inspiration and advice throughout these studies, and without her, this project would not have been as successful. Cheron Jones, Shana Cirimotich, and Ryan Foster have all been immensely helpful in the lab, especially when the need for trouble-shooting arose. I owe much of my knowledge about analysis techniques to Dr. Will Hsiao, Dr. Brandi Cantarel, Dr. Elliot Drabek, and Dr. Emiley Eloe- Fadrosh. Without their creative ideas and help with conducting different analysis programs, I would have been lost. I would also like to acknowledge the help and data provided by our collaborators. Dave Rasko deserves my endless gratitude for helping me think of creative ways to analyze this dataset. All of the immunology data presented in this project was collected and measured by the Sztein lab. Dr. Marcelo Sztein has been instrumental in designing experiments and analyses for this project, and I thank him for all his help in combining the data from different disciplines. In particular, Dr. Franklin Toapanta s help was crucial for understanding and presenting the immunological aspect of this project. These data would not have been possible were it not for the experiments and sample collections provided by the veterinary resources. Specifically, I would like to thank Dr. Steven Shipley and Dr. Aruna Panda, who were always very accommodating with any questions or concerns I had about monkeys. Both the Microbiology & Immunology Program and my committee have been crucial throughout graduate school. I express thanks to my committee for their advice: Dr. Marcelo Sztein, Dr. Eileen Barry, Dr. James Kaper, Dr. Alessio Fasano, and Dr. Shiladitya DasSarma. I also greatly thank Dr. Carbonetti and June Green from the Microbiology & Immunology program for their support over the years. Finally, I am grateful for the support from my friends and family. I have had the opportunity to meet numerous amazing people in graduate school, and it would have been difficult without their encouragement during tough times. Carly Page and Caitlin Castro have been there since the beginning, and I am immensely happy we started this journey together. My family and friends back home have always encouraged me from afar, and I thank you for that. I am lucky to have the support from my parents, Eeva and Tim. Finally, I thank Mark for patiently listening to my complaints and his constant encouragement, and Snuggles for always putting a smile on my face. This study was supported by the National Institute of Health (NIH) grant U19 AI (CCHI). iii

8 Table of Contents List of Tables... vi List of Figures... vii List of Abbreviations... ix Chapter 1: Background and Specific Aims...1 Introduction...1 The Intestinal Microbiota...5 Composition of the human intestinal microbiota...5 The role of the intestinal microbiota in nutrition and metabolism...7 The intestinal microbiota in immune development and regulation...9 Factors that Shape the Intestinal Microbiota...13 Development of the intestinal microbiota...13 Host vs. environmental impact on the microbiota...14 Dysbiosis of the Microbiota and its Relationship to Disease...16 Importance of maintaining microbial diversity...16 Perturbing the microbiota: consequences of antibiotic use...17 Correlations between the microbiota and disease...18 Shigella, an Enteric Pathogen...20 Shigella invasion and infection...21 Shigella vaccine development...22 Specific Aims...24 Chapter 2: Materials and Methods...26 Ethics Statement...26 Animal Screening and Handling...26 Study Design...28 Preparation of the Bacterial Inocula...29 Intragastric Inoculation of Bacteria...30 Clinical Monitoring and Treatment...31 Anti-Shigella Antibody Determination...31 Animal Geographic Origin...32 Stool Collection for 16S rrna Analysis...33 DNA Extraction and Pyrosequencing...33 Sample Preparation (Primer sequences and PCR conditions)...35 Data Processing...36 Genotype Analysis of MHC Microsatellites...38 Chapter 3: Aim 1- To establish robust methods for analysis of the gut microbiota in cynomolgus macaques...39 Introduction...39 Results...43 Composition of the cynomolgus macaque intestinal microbiota...43 Comparison of differential sampling...47 iv

9 Comparison of DNA extraction methods...49 Comparison of two sequencing platforms: 454 GS FLX vs. GS Titanium...54 Summary...58 Chapter 4: Aim 2- To determine the effect of oral vaccination with liveattenuated S. dysenteriae 1 strains and subsequent challenge with wild-type S. dysenteriae 1 on the GI microbiota in cynomolgus macaques...59 Introduction...59 Results...61 Core intestinal microbiota in cynomolgus macaques...61 Community Types of the Cynomolgus Monkey Intestinal Microbiota...67 Host genetic influence on the gastrointestinal microbiota...71 Impact of live-attenuated and wild type S. dysenteriae 1 strains on microbiota composition...79 Summary...88 Chapter 5: Aim 3-To determine correlations between the gut microbiota and immune responses following oral immunization with live-attenuated S. dysenteriae 1 strains and challenge with wild-type S. dysenteriae 1 in cynomolgus macaques...91 Introduction...91 Results...93 Induction of effector immune responses following oral administration of live-attenuated and wild-type S. dysenteriae Correlation between effector immune responses and the microbiota community over time Summary Chapter 6: Discussion References v

10 List of Tables Table 4.1. Community types within each study...69 Table 4.2. Probability of geographic origin for cynomolgus macaques...73 Table 4.3. Correlation of microsatellite regions to community type persistence using ANOVA...77 vi

11 List of Figures Figure 1.1. The human microbiome varies by body site...3 Figure 1.2. Variation of the human microbiome within each body site...4 Figure 1.3. Homeostasis between the microbiota and the immune system...10 Figure 1.4. Development of the human gut microbiota...13 Figure 3.1. Composition of the cynomolgus macaque intestinal microbiota...45 Figure 3.2. Intra- and inter-individual similarity of the microbiota composition in cynomolgus macaques...46 Figure 3.3. Relative abundance of genera in different aliquots of the same sample...48 Figure 3.4. Multidimensional compositional comparison of different aliquots of the sample...49 Figure 3.5. Relative abundance of genera in samples processed by different DNA extraction methods...51 Figure 3.6. Comparison of sample composition by extraction method...53 Figure 3.7. Effect of read length and number of reads per sample on taxonomic composition for all samples analyzed...56 Figure 3.8. Multidimensional comparison of samples sequenced by different sequencing platforms...57 Figure 4.1. Experimental design for cynomolgus monkey studies...62 Figure 4.2. Rank abundance plots of phyla and top 35 most abundant genera...63 Figure 4.3. Comparison of the genera found in both humans and cynomolgus macaques...64 Figure 4.4. Core gastrointestinal microbiota profiles in cynomolgus macaques...65 Figure 4.5. Rank abundance plots for studies Figure 4.6. Community types within the gastrointestinal microbiota of cynomolgus macaques...68 Figure 4.7. Rank abundance plots (left) and correlation networks (right) for community types I, II, III, and IV...71 Figure 4.8. Analysis of genetic variability of cynomolgus macaques in studies Figure 4.9. Localization of seven microsatellite markers tested within the MHC region...74 Figure (A to G) Allele frequencies for different geographic populations for seven MHC loci...75 Figure Phylogenetic tree calculated from indicated MHC microsatellite data for all macaques...76 Figure Estimates of diversity and similarity over time...80 Figure Community type composition, clinical disease symptoms, and elicited immune response over time for studies Figure Changes in the Shannon diversity index following immunization or PBS administration and wild type challenge in all macaques from studies 1 and 2 compared to control study 4 macaques...82 vii

12 Figure Increase in relative abundance of normally rare genera over time in study 1 macaques (n=12)...85 Figure Relative read abundance of less abundant organisms and clinical symptoms of Shigella infection in study 1 macaques (n=12)...86 Figure 5.1. Immune responses following wild-type S. dysenteriae 1 challenge in study 3 macaques...94 Figure 5.2. Immune responses following immunization and wild-type challenge with S. dysentariae 1 strains challenge in study 1 macaques...95 Figure 5.3. Mean fold increases following immunization and wild-type challenge with S. dysentariae 1 strains challenge in study 1 macaques...96 Figure 5.4. Immune responses following immunization and wild-type challenge with S. dysenteriae 1 strains challenge in study 2 macaques...98 Figure 5.5. Mean fold increases following immunization and wild-type challenge with S. dysentariae 1 strains challenge in study 2 macaques...99 Figure 5.6. Total IgA anti-shigella protein antibody-secreting cells (ASC) counts measured for study 2 macaques following immunization and challenge with S. dysenteriae 1 strains Figure 5.7. Total IgG anti-shigella protein antibody-secreting cells (ASC) counts measured for study 2 macaques following immunization and challenge with S. dysenteriae 1 strains Figure 5.8. Local similarity analysis between genera and clinical and/or immunological measurements in cynomolgus macaque Shigella vaccine studies Figure 6.1. Phylogeny of shared operational taxonomic units (OTUs) in studies 1-4 s viii

13 List of Abbreviations 16S ribosomal RNA (16S rrna) gastrointestinal tract (GI tract) nonhuman primate (NHP) Human Microbiome Project (HMP) Metagenomics of the Human Intestinal Tract (MetaHIT) Whole genome shotgun (WGS) operational taxonomic units (OTUs) Body mass index (BMI) gut-associated lymphoid tissue (GALT) Toll-like receptors (TLRs) isolated lymphoid follicles (ILFs) lymphoid tissue inducer (LTi) pattern-recognition receptors (PRRs) nucleotide-binding oligomerization domain (NOD)-like receptors Dendritic cells (DCs) Immunoglobulin (Ig) IL-17-producing T-helper (Th 17 ) cells Regulatory T (T reg ) cells quantitative trait loci (QTL) major histocompatibility complex (MHC) antibiotic-associated diarrhea (AAD) lipopolysaccharide (LPS) antibody-secreting cells (ASCs) modified Zymo (modzymo) phenol chloroform (PCl) base pair (bp) principal coordinate analysis (PCoA) partitioning around medoids (pam) local similarity alignment (LSA) ix

14 Chapter 1 Background and Aims Introduction Mammalian species are colonized by an enormous number of microorganisms, collectively termed the microbiota. These microbes, which include bacteria, archaea, fungi, and viruses at various locations in the human anatomy, are estimated to outnumber host cells by several orders of magnitude, and have coevolved alongside the host to provide necessary functions otherwise unavailable to the host (1, 2). In other words, the microbiota is essential to host survival. Metabolic function, nutrient acquisition, energy harvest, immune development, and pathogen protection all depend on a resident microbiota. As with other complex microbial ecosystems, such as in soil or ocean environments, the human microbiota likely adapt in response to external perturbations over the course of the host s lifetime. The human host can be thought of as a composite of both host and microbial cells, or a super-organism (3), with its biology determined by the genes encoded in both the human genome and the genomes of its microbial partners. In this context, research concerning human health and disease cannot be understood without studying both humans and their microbial commensals. The Human Microbiome Project (HMP), a five-year, large-scale project, was launched in 2007 to investigate the role of the microbiota in human health and disease (4, 5). Many species in complex microbial communities, including the human microbiota, have not yet been cultured, and their studies have been facilitated by the development of DNA-based approaches. These methods include 1

15 16S ribosomal RNA (16S rrna) surveys, which focus on the ubiquitous gene found in all bacteria, and metagenomic shotgun sequencing approaches that analyze the gene content (microbiome) present within these organisms. Improvements in culture-independent methods have allowed for a shift from the study of single microbes grown in culture to investigations on how microbes interact together within an ecosystem. One of the main goals of the HMP was to establish what a normal microbiome is at five major body sites (gastrointestinal (GI) tract, vagina, skin, nasal, and oral cavity) in order to understand its role in health and disease. The largest HMP project included 16S rrna taxonomic profiles and whole genome shotgun (WGS) or metagenomic sequencing of full genomes from 250 subjects over time from various body sites, totaling over 5,000 samples overall (5, 6). The data produced by the HMP has been made available to the public for use as a baseline reference for research of the microbiome in disease states. In conjunction with the HMP s international counterpart, MetaHIT, Metagenomics of the Human Intestinal Tract, a vast number of reference sequences are now available for the investigation of the microbiota and its role in health and disease. One of the specific questions that the HMP sought to answer was whether all humans contain a core microbiome, or shared set of microbes, genes and/or functions. Composition of the microbiota is strongly determined by body site (Fig. 1.1) (7-9), and varies between individuals (10-12). Thus, identification of a shared set of organisms among a large population of healthy subjects can be difficult. 2

16 Figure 1.1. The human microbiome varies by body site. Principal coordinates plot showing clustering of the human microbiota at the genus level by body site. (Data taken from Fig. 1 in (8)) At the phylum level, Bacteroidetes, Firmicutes, Actinobacteria, Proteobacteria, Fusobacteria, and Tenericutes are found throughout the human body, albeit at different ratios for each anatomical location. 3

17 Figure 1.2. Variation of the human microbiome within each body site. Phylumlevel description of microbiota is dependent on anatomical body site based on high-throughput sequencing from multiple studies. (Data taken from Fig. 1 in (1)) However, at the species level, the composition of the microbiota in each body site varies greatly (10, 11, 13). One of the most diverse microbial communities within the human body is in the gastrointestinal tract, where trillions of microorganisms carry out vital processes such as amino acid synthesis and plant polysaccharide digestion (1). This complex community, which is also essential in the development of a healthy immune system (14), is the focus for the remainder of this project, which investigates the role of the microbiota in enteric pathogen infection. 4

18 The intestinal microbiota Composition of the human intestinal microbiota. The gastrointestinal (GI) tract represents one of the most diverse environments in the human body, both in number of total organisms present and the estimated number of species (10). It is estimated that up to organisms exists within the gastrointestinal tract of adult humans (15). Most of this diversity has been discovered using fecal samples as starting material, and comparison of the microbiota along the GI tract has been accomplished using mucosal biopsies. The results of these studies demonstrate that although community composition differs according to location, fecal samples can be used as surrogates of the gut microbiota, clustering most closely to colonic samples (10, 14). The composition of the intestinal microbiota varies across individuals, especially at lower taxonomic levels. In general, the community is conserved at the phylum level and includes Bacteroidetes, Firmicutes, Actinobacteria, Fusobacteria, Cyanobacteria, and Proteobacteria (1, 16). Of these, Bacteroidetes and Firmicutes dominate within the human gut microbiota (7, 10). Common genera within the human gut microbiota include Bacteroides, Prevotella, Clostridium, Fusobacterium, Streptococcus, Eubacterium, Peptococcus, and Bifidobacterium, although the abundance of these genera varies greatly over time and between individuals (11, 12). Most species in the microbiota are still unculturable, and next generation sequencing technology utilizing the 16S rrna gene has provided methods of estimating species-level diversity that rely on either clustering phylogenetically similar sequences into operational taxonomic units 5

19 (OTUs) or binning sequences based on reference databases (phylotyping) (17). These methods have identified up to 1,000 species, of which about 15% can be found within the abundant set of bacteria in an individual (18, 19). Identification of a core microbiota for the intestine has been challenging due to its diversity and the observed intra-individual variation present among the human population. A study conducted by Jalanka-Tuovinen et al. identified a 41.9% overlap of shared phylotypes among 9 subjects, which included mainly Firmicutes such as Clostridium and Ruminococcus species (20). However, the gut microbiota in some individuals in this study was dominated by Bacteroides species (phylum Bacteroidetes) that were not present within all individuals, suggesting that abundance does not always correlate with prevalence. To identify similarities in the gut microbiota among individuals, recent analysis by the MetaHIT group utilized a clustering algorithm on metagenomes from 39 individuals of European, American, and Japanese heritage (21). Samples from individuals clustered into three distinct community types, termed enterotypes, each characterized by the presence of a dominant genus: Bacteroides, Prevotella, or Ruminococcus. These clusters appeared to be independent of nationality, gender, age and health status as measured by body mass index (BMI). Similar clusters were identified in a larger cohort consisting of 310 Amish individuals, in which farmers were found to be overrepresented in the Prevotella-dominated community type (22). Two of these community types (Bacteroides- or Prevotelladominated) were identified in another cohort consisting of 98 individuals, which were strongly associated with long-term eating patterns (23). 6

20 The full significance of enterotypes is not yet well understood; however, preliminary data suggests that they may differ in terms of their functional properties, as exemplified by the dominance of specific genera reflected by certain occupations or eating habits. The studies described above represent only a snapshot of the compositional differences in the gut microbiota in the world s populations since all volunteers have been recruited from developed countries. Much of the developed world s dietary consumption and environment is different in comparison to the developing world, and it is hypothesized that the genetic, nutritional, and environmental differences in these populations could all influence the microbiome (24). Increased sampling of all world populations needs to be carried out before robust conclusions about how external factors influence the composition of the gut microbiota can be fully drawn. It is not unrealistic to hypothesize that certain types of bacterial communities may place individuals at greater risk for various diseases. The role of the intestinal microbiota in nutrition and metabolism. The mammalian host and its gastrointestinal microbiota have co-evolved to provide each other mutually beneficial conditions (25). Microbes benefit from a nutrient-rich environment, and the host obtains metabolic functions not encoded by its own genome. Genes contained only in the microbiome encode for proteins capable of breaking down indigestible complex carbohydrates to monosaccharides that can be used as a source of energy (18, 19), such as the carbohydrate enzymes in Bacteroides thetaiotaomicron and related species (26, 27) or enzymes involved in 7

21 the production of short-chain fatty acids (19, 28, 29). Evidence for increased energy harvest by the microbiota came from a study by Backhed et al., where germ-free (GF) mice experienced weight gain upon colonization with microbes from the distal gut of conventional mice, accompanied by metabolic changes and fat storage (30). Because of the microbiota s potential for energy harvest, it has even been hypothesized that a specific microbiota composition could lead to obesity or related diseases, such as metabolic syndrome or diabetes (31). While studies in mice have provided evidence supporting a microbial community prone to obesity (32-34), the impact of genetics and diet in humans complicates how much is attributed to the microbiota composition alone (12, 35). The gut microbiome may partially determine the nutritional uptake of the host, but is also in turn shaped by host diet. Indeed, 16S surveys of the fecal microbiota of different mammals show clustering by diet (herbivore, omnivore, carnivore) based on differences in community composition (25). More specifically in humans, both Bacteroides- and Prevotella-dominated enterotypes have been associated with long-term diets of protein/animal products and carbohydrates, respectively (23). The interplay between diet, nutrition, and the microbiome is of particular interest when comparing different world populations, especially in malnourished populations where disease risk may be high due to a combination of factors (29). The few studies that have been published analyzing the gut microbiota of populations from developing countries have observed distinct differences in the microbiota that may in part reflect dietary intake. De Filippo et al. observed an increase in Bacteroidetes, specifically genera capable of 8

22 breaking down cellulose and xylans (Prevotella and Xylanibacter), in African children when compared to European children (36). A more recent study comparing the gut microbiota of rural Malawian, Venezuelan Amazonian, and US metropolitan populations found that both age and geography/culture explained most of the variance observed (37). Furthermore, the microbiome of both the Malawian and Venezuelan adults was enriched with genes reflective of a diet rich in starch compared to the gene content of Americans, reflective of a diet rich in protein. Nutrition s capacity to affect the gut microbiota and host biology, including signaling and maintenance of the immune system, necessitate the inclusion of all geographic populations in compositional and functional comparisons of the microbiota. The intestinal microbiota in immune development and regulation. The increased nutritional benefits provided by the microbiota was likely a major evolutionary factor in shaping the symbiotic relationship between the microbiota and its host (38). Considering the close proximity between the intestinal microbiota and mucosa, it is not surprising that the gastrointestinal microbiota and the immune system have also co-evolved to tolerate and influence each other. Within the lower GI tract, a barrier of epithelial cells and a mucus layer separate the microbiota from the rest of the host (39). Throughout this intestinal mucosa, specific host immune responses have developed that both limit and shape the microbiota, which in turn plays an integral role in the development and 9

23 maintenance of the immune system (40). Figure 1.3. Homeostasis between the microbiota and the immune system. Depiction of the branches of the immune system and the physiology surrounding it. Antimicrobial peptides, the mucus layer, and IgA help minimize contact of the microbiota to epithelial cells. DCs sample the microbiota constantly, and induce B and T cells specific for microbiota components. (Data taken from Fig. 1 in (39)) 10

24 Much of our knowledge about the role of the microbiota in immune development comes from gnotobiotic mouse models. Colonization of GF mice results in an increase in nutrient absorption, intestinal cell development, and angiogenesis within the intestines (41, 42). The initiation and development of gutassociated lymphoid tissue (GALT), such as isolated lymphoid follicles (ILFs) and Peyer s patches, is initially dependent on special innate lymphoid tissue inducer (LTi) cells (43, 44). However, sensing of the microbiota is required to complete maturation of ILFs (45). This occurs through innate pattern-recognition receptors (PRRs) via epithelial cells in the intestine called Paneth cells (46, 47). A variety of PRRs, such as Toll-like receptors (TLRs) and nucleotide-binding oligomerization domain (NOD)-like receptors, are induced by specific microbial patterns on commensals and pathogens alike (48, 49). Development of the adaptive immune system is also dependent on the microbiota in the gut. Mice lacking TLR9 activation were observed to exhibit decreased conversion of regulatory T cells (T reg ), and were more susceptible to oral infection by Encephalitozoon cuniculi (50). Similarly, differentiation of proinflammatory IL-17-producing T-helper (Th 17 ) cells was inhibited in mice lacking a microbiota, and was accompanied by increases in T reg cells (51). Both of these processes required functional TLR activation induced for bacterial sensing of the microbiota, demonstrating the connection between the innate and adaptive immune branches. Homeostasis between the microbiota and immune system is maintained by constant reciprocal communication between both sides. For instance, even the 11

25 mucus barrier, which provides a physical barrier between commensals and epithelial cells, is increased by the presence of the microbiota (52). PRR activation is also induced by the microbiota but conversely limits the microbiota. Antimicrobial peptides, such as α-defensins or C-type lectins, are continually secreted into the mucus layer upon signaling of the innate immune system (53, 54). For example, RegIIIy, a C-type lectin induced by TLR activation, has been observed to limit bacterial numbers at the epithelial surface of the intestinal mucosa, suggesting that spatial segregation of the microbiota and body tissues is important in regulating homeostasis (55). Another important molecule that maintains homeostasis between the microbiota and the immune system is IgA, which also plays a central role in protection against pathogens (56). IgA can be secreted into the intestinal lumen as secretory IgA. Before colonization, plasma cell populations and IgA levels are low, and even minor amounts of bacteria can increase the presence of plasma cells (57). Dendritic cells (DCs) regularly pick up and present antigens from various commensal regions to plasma cells, which can secrete a diverse IgA repertoire capable of binding to specific surface antigens on the microbiota and limiting its access to the intestinal mucosa (58, 59). Secretory IgA binds to specific surface antigens of the microbiota and limits its access to the intestinal mucosa. In this fashion, IgA has the potential to selectively modulate the composition of the microbiota, depending on the existing IgA repertoire of the host. 12

26 Factors that shape the intestinal microbiota Development of the intestinal microbiota. In addition to diet and the immune system, external factors such as aging, gender, genetics, and the environment add to the diversity observed in the intestinal microbiota in humans. Throughout the human lifespan, changes within the microbial populations occur with age. Colonization of the gastrointestinal tract occurs at birth, with maternal colonization influencing the composition of the microbiota initially. Not until the first year of life does stabilization of community numbers occur (60, 61). Figure 1.4. Development of the human gut microbiota. A) Spatial arrangement of the gut microbiota in the GI tract. B) Spatial arrangement of the microbiota in the intestine. C) Development of the microbiota composition over time in an individual. (Figure taken from Fig. 2 in (14)) The initial composition of the infant gut microbiome has been observed to be dependent on the mode of delivery (62, 63). For instance, infants delivered via vaginal birth exhibit a microbial community similar to that of the mother s vagina (64), while the microbiota in infants born via caesarean delivery are observed to 13

27 be less diverse, and specifically lack a diversity of Bifidobacterium species (62, 65). Feeding practices further shape the infant intestinal microbiota. Infants that are breastfed exhibit a more diverse microbiota, particularly within the Bifidobacterium sp. (66, 67), than infants that are given formula. However, these differences do not appear to affect the development of the microbiota into adulthood (68). During the first two years, the largest changes in the microbiome correlate with changes in the diet, such as an increase in Bacteroidetes following the addition of solid foods to the diet (61). During this time, the microbiota begins to stabilize and resemble that of an adult (60, 61, 69). Later in life, deterioration in digestive functions, accompanied by dietary changes, may again influence compositional changes in the microbiota related to ageing. In older persons (>65 years), the microbiota appears less diverse compared to adults, and more interindividual variation has been observed in older cohorts (70, 71). Furthermore, some studies point to alterations of the immune system and increased inflammation of the gut that may be associated with changes in the microbiota (70, 72, 73). Host vs. environmental impact on the microbiota. Both host genotype and environmental factors have been observed to impact the composition of the microbiota, although the contribution of each is difficult to determine. Most studies investigating the effect of genotype on the microbiota have been conducted in animal models (74-77). Kovacs et al. observed a higher similarity of microbiota composition within mice from the same genetic background than 14

28 within the same sex using an overall similarity comparison between the two groups (74). A more specific test was conducted by Benson et al. who used quantitative trait loci (QTL) to investigate whether genomic markers are associated with specific core microbial taxa (75). They observed significant QTLs linked to core taxa, which included Lactobacilli and Helicobacter, both of which are intimately associated with host tissues and likely to be modulated by host properties. Specific host genes linked to both dietary and immune functions are also reported to affect the microbiota (34, 78, 79). Mice or humans deficient in these genes tend to exhibit a different community structure or lack a specific taxon, such as APOA1-deficient mice, which present a different community profile compared to wild-type mice (34). Due to the close proximity of both the gut microbiota and the immune system at the gastric mucosa, it is possible that differences in immune genes, such as those of the major histocompatibility complex (MHC), could impact the selection of specific commensal microbes. Six strains of mice from the same background, but congenic for the MHC, were found to cluster based on the compositional profile determined by fatty acid content in the stool, suggesting that HLA allele type could impact the microbial composition of the GI tract (80). The impact of genotype and environment is more difficult to delineate due to lack of environmental control, homogeneous background, and easy intervention. It is generally accepted that family members are inclined to have more similar gut microbiota than unrelated individuals, as evidenced by the analysis of the microbiota in twin studies (12, 68, 81). Both monozygotic and 15

29 dizygotic twins living in the same household were observed to share the same degree of similarity, arguing that both genotype and environment shape an individual s microbiota. A shared household rather than genotype alone may thus account for the increased similarity observed among families. As methods in genomic analysis improve, the interplay between host genetics, the environment, and microbiota composition will hopefully be elaborated. Dysbiosis of the microbiota and its relationship to disease. Importance of maintaining microbial diversity. The microbes that inhabit the intestinal mucosa serve as an integral physical barrier to pathogenic colonization. Ecological diversity-stability theory maintains that higher-diversity ecosystems confer stability and resistance to perturbation, and this is no different for the microbial ecosystem in the GI tract (82, 83). Increasingly, ecological theory is applied to research on the microbiota to explain the variation observed among healthy individuals (84). The concept of a healthy microbiome, while difficult to define in terms of specific organisms, may be the maintenance of a diverse number of organisms. For instance, removing a portion of the microbiota or of a specific community member could lead to opportunistic colonization by new or rare organisms, thus upsetting a previous stable state. Depending on the resilience of the community, or ability to return to its prior state, disruption of the microbiota can lead to instability and possible susceptibility to disease states (85). 16

30 Perturbing the microbiota: consequences of antibiotic use. An example highlighting the importance of diversity is observed following antibiotic use. Antibiotics greatly reduce the diversity of the gut microbiota within days of administration (11, 86). Ciprofloxacin treatment in human subjects reduced the gut microbiota diversity by one third in a study by Dethlefson et al (11), although the effect was variable among subjects. Additionally, subjects in whom diversity was decreased were able to recover within a four-week time period, suggesting a natural resilience of the community to return to a state of normal. Some taxa, however, never recovered post-treatment, an observation made in subsequent studies in human subjects (86). Although a stable, diverse state was observed eventually, it is unclear how the disappearance of certain microbes and the appearance of new microbes could affect overall function or future compositional disturbance. A case where the intestinal microbiota does not return to a healthy, stable state is observed in antibiotic-associated diarrhea (AAD), caused by the pathogen Clostridium difficile. A decrease in the microbial diversity of the gut following antibiotic use leads to an increase in C. difficile in some patients (87-89). While the reasons for the selectivity of C. difficile infection in patients is still unknown, it is generally accepted that disruption of the microbiota is directly related to C. difficile proliferation, followed by AAD. Interestingly, treatment of the most severe cases of C. difficile involves reconstitution of the microbiota. Fecal transplantation of the colon from a donor, generally a close relative or household member, is an effective treatment in severe cases (90), emphasizing the 17

31 importance of maintaining a diverse community. Antibiotic use in mice also decreases microbial diversity and renders mice susceptible to C. difficile infection (91, 92). Recently, Reeves et al. observed that the presence of Lachnospiraceae isolates in mice is protective against C. difficile infection, suggesting a possible protective role of specific taxa against infection (93). Antibiotic treatment in mice increases susceptibility to pathogens other than C. difficile, including the enterics Salmonella enterica serovar typhimurium (94) and Citrobacter rodentium (95, 96). The severity of disease is correlated with specific mouse strains (97, 98) and suggests a genetic component to disease infection. However, transplantation of the microbiota from C. rodentium resistant mouse strains to susceptible mouse strains resulted in a less severe disease form (99). Reduced abundance of Prevotella was correlated with increased IL-22 production, which in turn mediated protection. Microbial composition, modulated by genotype, thus appears to affect disease susceptibility in C. rodentium infection in mice. Although variability in resistance to enteric infection is also observed in humans (98), specific factors involving genotype and microbiota composition that confer resistance remain unresolved. Correlations between the microbiota and disease. It is difficult to define a healthy microbiome due to the variation of the microbiota observed among the human population. Additionally, it is unclear whether differences observed between affected and unaffected individuals are causative or a result of the 18

32 disease. Regardless, multiple studies have detected correlations between distinct taxa and complex disease states. Due to its involvement in the maintenance of the immune system, the microbiota has been hypothesized to play a role in autoimmunity. This concept also aligns with the hygiene hypothesis, which suggests that the observed increase in autoimmune disease in developed countries is due to the decreased exposure to surrounding microbes, thus interrupting the natural development of the immune system. Inflammatory bowel disease (IBD), which includes both Crohn s disease and ulcerative colitis, is an autoimmune disease of the GI tract defined by mucosal inflammation and lesions. Reduced diversity of both Firmicutes (100) and general numbers of bacteria (101) were observed in IBD patients. A study in twins indicated that increased abundances of Escherichia coli and increased levels of Faecalibacterium prausnitzii were present in IBD patients compared to healthy individuals (102). Other potentially linked autoimmune conditions include celiac disease (103), multiple sclerosis (104), and rheumatoid arthritis (105). Likely, a combination of microbiota composition, genetics, and the environment impact the development of these complex diseases. Further research on how the intestinal microbiota affects disease development, and vice versa, is still necessary. The importance of a diverse, resilient microbiota has been shown to be important (88), but a direct functional role for it is still lacking. The development of simpler animal models and increasing knowledge about the interactions between host and microbes will 19

33 hopefully resolve the unknown link between the microbiota and disease susceptibility. Shigella, an enteric pathogen Shigella species represent a group of mucosally invasive bacteria that cause bacillary dysentery, or shigellosis, in humans and nonhuman primates (NHPs), spread via fecal-oral contact. Over 14,000 cases of Shigella are reported yearly in the US alone, but Shigella infection is most common in communities in the developing world, with a worldwide estimate of 90 million cases per year (106, 107). Although Shigella can infect all age groups, infants and young children suffer from most infections. Four different Shigella species cause human disease: Shigella sonnei, Shigella flexneri, Shigella boydii, and Shigella dysenteriae. Each species can also be classified into different serotypes, based on its O-antigen composition, the polysaccharide portion of the lipopolysaccharide (LPS) molecule on the bacterial surface (108). S. sonnei accounts for the majority of the cases in the developed world, and S. flexneri causes most infections in endemic regions. However, S. dysenteriae type 1, is a unique serotype capable of producing Shiga toxin and carrying multiple antibiotic resistance genes, has been the cause of several large pandemic outbreaks in developing countries ( ). Emerging antibiotic resistance and new serotypes have renewed interest in developing a safe and effective Shigella vaccine. 20

34 Shigella invasion and infection. Shigella invades the host through the large intestines and quickly disseminate via cell-to-cell contact. Only a small inoculum of Shigella ( bacteria) is required for disease development in humans, which can result in rapid dissemination within the environment. Upon ingestion, Shigella invade the M cells of the colon, are exocytosed into the basolateral side of the intestinal epithelium, and are taken up by resident macrophages (113). Shigella multiply within macrophages where it induces apoptosis and the release of proinflammatory cytokines, and can invade other enterocytes once released. Infection causes a variety of symptoms, ranging from watery to bloody diarrhea, accompanied by high fever and cramps (109). Symptoms of shigellosis can be self-resolving within 5-7 days in a healthy individual, although life-threatening complications can arise in immunodeficient or otherwise unhealthy individuals (114). A substantial host innate response is induced upon cellular invasion by Shigella, leading to an inflammatory state. Part of Shigella s invasion tactics includes the use of a type III secretion system that facilitates Shigella phagocytosis and can activate proinflammatory responses (115). PRRs, such as TLRs and NOD-like receptors, recognize Shigella LPS, nucleic acids and effectors following invasion, and lead to the induction of the inflammasome and subsequent cell death. Shigella s ability to modulate the host s inflammatory response, such as its ability to dampen NF-kB activation via the IpaH effector protein (113), contributes to the survival of Shigella within the host. 21

35 The adaptive immune system is also activated upon Shigella infection. Both humoral and cell-mediated immunity is induced by Shigella. Secretory IgA and serum IgG responses against Shigella LPS and other effector proteins have been observed following infection (109, 116). In humans, antibody-secreting cells (ASCs) secreting anti-lps IgA are believed to best correlate with protection from disease (117). Additionally, local recruitment and activation of T-cells has been observed, although their role in the protective response is unclear. Shigella vaccine development. The ideal Shigella vaccine would be highly effective without adverse side effects, and have the potential to protect against all epidemiologically relevant strains. One factor hampering the development of a successful Shigella vaccine is the lack of a Shigella animal model that replicates human pathogenesis and shigellosis (109). Rabbit models of Shigella infection have been successful in reproducing the inflammatory responses of bacterial invasion (118). However, this model requires manipulation of the animal to permit Shigella infection. The keratoconjunctivitis model (the Sereny test) (119) and newer pathogenesis models (120) in the guinea pig have been useful in establishing correlates of protection, and can be utilized to evaluate the safety and efficacy of candidate vaccines. The use of nonhuman primates (NHPs) is perhaps the most clinically relevant models for vaccine development. Vaccine models have been developed using S. flexneri in rhesus macaques, Macaca mulatta, (121) and S. dysenteriae 1 in cynomolgus macaques, Macaca fascicularis (122). 22

36 However, a large inoculum (~10 10 CFU) is necessary to cause bacillary dysentery and costs are impedients in conducting large-scale protection studies. Within the last decade, the identification of new virulence factors and the development of new vaccine technology, has provided novel insights into Shigella vaccine development (109, 114). Both live-attenuated Shigella strains and parenteral conjugate vaccine candidates have shown promise in eliciting a protective response against infection. Live-attenuated strains containing deletions in various Shigella virulence factors have been well tolerated in human Phase III trials, and have been shown to confer significant protection against wild-type infection (109). However, correlation of protection against wild-type infection has been inconsistent among different world populations. A trial of a live-attenuated oral S. flexneri 2a vaccine strain SC602 conducted in North American volunteers demonstrated efficacy, but failed to confer protection in volunteers in Bangladesh (117, 123, 124). Vaccine studies with other oral vaccines, including polio, cholera, and rotavirus, have revealed differences when conducted in developed and less developed global populations (109, 125, 126). While many external factors, such as diet, nutrition, and host genetics, can influence the host s response to oral vaccines and infection alike, an additional contribution to this variance could be explained by the resident intestinal microbiota (127). The microbiota, shaped by external factors, is a diverse microbial community that in turn can shape the type and/or extent of the immune response. The background presented here supports the hypothesis that the composition of the microbiota is capable of modifying the health status of the 23

37 host, including its immune system and ability to resist disease. To date, studies pertaining to vaccine development have not analyzed the role of the microbiota in the elicitation of an effective response to oral live-attenuated strains of enteric pathogens. Specific Aims The interplay between the microbiota, the immune system, and pathogen infection remains a complex relationship in terms of health and disease susceptibility. This project focused on elucidating connections between these factors to better understand the role of the microbiota in the immunization and infection of one enteric pathogen, S. dysenteriae 1. Here we used 16S rrna sequencing methods to investigate the gut microbiota in cynomolgus macaques following oral vaccination with a live-attenuated S. dysenteriae 1 vaccine and subsequent wild-type challenge. We hypothesized that the gastrointestinal microbiota composition will change following oral vaccination and/or wild-type (WT) infection of S. dysenteriae 1. We also hypothesized that pre-existing microbiota profiles would influence the immune response elicited following these events. We first developed robust analysis methods to characterize the gut microbiota of cynomolgus macaques. We then identified compositional and structural changes in the microbiota following both immunization and challenge. Finally, we correlated compositional changes in the microbiota to the effector immune responses measured following these events. These studies greatly add to 24

38 the knowledge of the complexity of the intestinal microbiota and its involvement in pathogen infection, as well as direct us towards new considerations in vaccine development. The following aims addressed the specific questions involved in this longitudinal study: Aim 1: To establish robust methods for analysis of the gut microbiota in cynomolgus macaques Aim 2: To determine the effect of oral vaccination with live-attenuated S. dysenteriae 1 strains and subsequent challenge with wild-type S. dysenteriae 1 on the GI microbiota in cynomolgus macaques Aim 3: To determine correlations between the gut microbiota and immune responses following oral immunization with live-attenuated S. dysenteriae 1 strains and challenge with wild-type S. dysenteriae 1 in cynomolgus macaques 25

39 Chapter 2 Materials and Methods Ethics Statement The study took place in the animal facility of the Program of Comparative Medicine (University of Maryland School of Medicine), an AAALAC-accredited facility. All animals in this study fully participated in the UMSOM "PLAN TO PROMOTE THE PSYCHOLOGICAL WELL-BEING OF NONHUMAN PRIMATES AT THE UNIVERSITY OF MARYLAND SCHOOL OF MEDICINE", which includes aspects such as social housing when appropriate, enrichment of physical environment, tactile stimulation, and positive interactions with both conspecifics and humans. This enrichment plan and the IACUC protocol describing this study were specifically approved by the UMSOM IACUC prior to study conduct. Any animal that developed clinical signs referable to dysentery was immediately treated with both fluid support and antibiotics to alleviate discomfort or distress. No animals were euthanized to collect the data presented in this study during the course of the study described. Animal screening and handling Male and female cynomolgus macaques (Macaca fascicularis; age, 2 to 5 y, n=33) were purchased from approved vendors as described for each study: study 1 (n=12) and study 3 macaques (n=3) from Sierra Biomedical Research (Reno, NV), study 2 macaques (n=12) from Charles River Laboratories (Houston, TX), 26

40 and study 4 macaques (n=6) from Harlan Laboratories, Inc (Indianapolis, IN). Genetic profiling for geographic origin for macaques in studies 1-3 was conducted by Molecular Anthropology Laboratory (Davis, CA). Upon purchase from described vendors, animals screened negative for Macacine herpesvirus 1, SIV, simian retrovirus, and simian T-lymphotrophic virus. Animals were quarantined for 3 months before the study began, and tested for intestinal parasites, Campylobacter, Salmonella, Shigella, and Yersinia spp. prior to study. Only macaques that tested negative for IgG and IgA antibodies to S. dysenteriae 1 LPS or titers < 1:50 were used. Macaques were housed in a Biosafety Level 2 containment facility in individual stainless steel primate caging for the study duration, and safe handling practices were conducted while working with S. dysenteriae 1. Animals received a commercial primate diet free of any bacterial contamination (Teklad 2050, Harlan Laboratories, Indianapolis, IN), with a daily supplement of fresh fruits and vegetables, and given supplemental food enrichment (fruit and nut mix, popcorn, peanuts, granola bars) once or twice weekly. Municipal drinking water was regularly tested and free of any bacterial growth and was provided through an automatic watering system, and was available ab libitum to all macaques throughout the studies. All housing and handling procedures conformed to the Guide for the Care and Use of Laboratory Animals and the Animal Welfare Act, and complied with the recommendations in the Biosafety in Microbiological and Biomedical Laboratories Guide. 27

41 Study Design Fig. 4.1 illustrates each study timeline (studies 1-4). Macaques in study 4 received no bacterial challenge throughout sample collection. Stool sample collections from this group were carried out during the quarantine period (91days) at the University of Maryland School of Medicine. Macaques from studies 1, 2 and 3 were involved in live-attenuated and/or wild-type S. dysenteriae 1 immunization studies to assess the efficacy of live attenuated S. dysenteriae 1 vaccine candidates. For macaques in these studies, the optimal dose to induce shigellosis was established previously (122). The 27 macaques in these studies were fasted for 12 h prior to bacterial challenge. Two S. dysenteriae 1 strain 1617 vaccine candidates were used in immunization studies 1 and 2: CVD1255 (strain 1617 guaba sen stxab) and CVD1256 (strain 1617 guaba, sen stxa::mlppstxb) (128). Macaques in studies 1 and 2 were intragastrically inoculated with either phosphate buffered saline (PBS) or colony forming units (CFU) of one of the vaccine candidates, CVD1255 or CVD1256, either two inoculations separated by 28 days (study 1: day 0 and 28), or four times within one week (study 2: days 0, 2, 4, and 7). Both groups were subsequently challenged with S. dysenteriae 1 wild-type bacteria strain 1617 (10 11 CFU via the orogastric route) and monitored on a daily basis for development of clinical signs of disease (study 1 challenged on day 56; study 2 challenged on day 28). Each macaque received 15 meq sodium bicarbonate intragastrically immediately prior to inoculation to prevent killing of the bacteria by gastric acid. 28

42 Each animal was monitored twice daily for signs of shigellosis following challenge. Stool specimens were collected from each animal twice daily until 4 days after inoculation, as described for each study (Fig. 4.1). Study 1: Animals (n=12) were given CVD1256 (n=4), CVD1255 (n=4), or phosphate buffered saline (PBS) (n=4) on days 0 and 28 (1 dose per animal). Each animal received wild-type S. dysenteriae 1 strain 1617 on day 56. Study 2: Animals (n=12) received CVD1255 (n=6) or PBS (n=6) on days 0, 2, 4, and 7 (4 doses per animal). Each animal received wild-type S. dysenteriae 1 strain 1617 on day 28. Study 3: Animals (n=3) received wild-type S. dysenteriae 1 strain 1617 on day 0. Plasma IgA and IgG antibody titers to S. dysenteriae 1 LPS were measured by end-point dilution ELISA on days 0, 7, 14, 28, 35, 56, 63, 70, 84 and 98 (study 1) or days 0, 2, 4, 7, 10, 15, 28, 35, and final collection day (36-48) (study 2). Studies 1, 2, and 3 macaques were monitored daily for clinical symptoms of shigellosis ranging from a score of 1 (soft stool) to 4 (severe, bloody diarrhea) as previously described (122). Study 3 macaques received only wildtype S. dysenteriae 1 bacteria with the same protocol. Study 4 received no treatment throughout the course of sample collection. Preparation of the bacterial inocula Two vaccine candidates, CVD1255 (S. dysenteriae 1 strain 1617 guaba sen stxab) and CVD1256 (S. dysenteriae 1 strain 1617 s guaba sen stxa::mlpp-stxb) (128), were used in immunization studies 1 and 2. Liveattenuated S. dysenteriae 1 CVD1255 and CVD1256 and challenge inocula S. 29

43 dysenteriae 1 strain 1617 were prepared from frozen master stocks, which were plated onto trypticase soy agar (Becton Dickinson, Franklin Lakes, NJ) containing 0.01% Congo red dye (Sigma Chemical, St Louis, MO). After an incubation period of 18 to 24 hours at 37 C, S. dysenteriae 1 antiserum (Denka Seiken, Tokyo, Japan) was used to confirm the identity of single-well isolated Congo-red positive colonies that exhibited Shigella morphologic characteristics, and picked and suspended in sterile saline. These colonies were used to inoculate trypticase soy agar plates for heavy growth, then incubated overnight at 37 C. Overnight growth was harvested, washed, and standardized turbidmetrically, and appropriately diluted in PBS for inoculation. These inocula were used within 4 hours of preparation. To establish the actual inoculum that was administered to the cynomolgus macaques, diluted colony counts in PBS were performed in 2 to 3 replicate plates after overnight incubation. Intragastric inoculation of bacteria Macaques were sedated with 10 mg/kg ketamine IM prior to bacterial inoculation. After onset of sedation, an 8- to 14-French orogastric tube (Tyco Healthcare Group LP, Mansfield, MA) was used to administer 15 meq sodium bicarbonate (Neogen, Lexington, KY) through the mouth into the stomach of the animal, followed by intragastric inoculation of the S. dysenteriae 1 strain. To confirm proper location of the tube, gas sounds in the stomach was verified by gently injecting 5 to 20 ml air while auscultating the upper abdomen using a stethoscope prior to inoculation, at which point the inoculum was injected down the tube. The 30

44 orogastric tube was flushed with 10 to 20 ml of sterile saline to ensure full delivery of the bacterial inoculum. Clinical monitoring and treatment Daily monitoring for diarrhea, dysentery, fever, signs of respiratory illness, changes in food intake, or any other abnormal behavior was recorded for each macaque, including rectal temperature (Sure Temp Plus Thermometer, Welch Allyn, New York, NY) in sedated macaques (Ketamine HCl given intramuscularly at 10mg/kg) for the first 7 days and then on days 14, 28, and 56 after inoculation. Fecal samples were plated on selective media to confirm presence of S. dysentariae 1 strains post-immunization and post-challenge. Macaques exhibiting symptoms of Shigella infection (diarrhea, fever in excess of 103 F, dysentery) were treated with antibiotics (enrofloxacin at 5 mg/kg PO or IM twice daily for 5 days, Bayer Animal Health, Shawnee Mission, KS, or ceftriaxone at 50 mg/kg IM once daily for 5 days, Apotex, Weston, FL) within 24 h of the time at which symptoms were observed. Sick macaques received intravenous fluid therapy (10 to 20 ml/kg Ringers Lactate Solution once or twice daily) if determined necessary by a facility veterinarian. Anti-Shigella antibody determination Plasma IgA and IgG antibody titers to S. dysenteriae 1 LPS were determined by end-point dilution ELISA. Briefly, ELISA plates (Immulon 2HB, Thermo, Milford, MA) were coated with S. dysenteriae 1 LPS (5 µg/ml in 0.1M Sodium 31

45 Carbonate, ph 9.6) for 3 hours at 37 C. Following coating, as well as after each subsequent incubation, plates were washed 6 times with PBS containing 0.05% Tween 20 (PBS/Tween). Plates were then blocked overnight (4 C) using 10% dried milk in PBS. Plasma samples were evaluated in 2 fold-dilutions in 10% dried milk in PBS/Tween. IgA and IgG anti-s. dysenteriae 1-LPS antibodies were detected using horseradish peroxidase (HRP)-labeled goat anti-monkey Fc α- (1:2000) and γ-(1:5000) chains, respectively, (KPL, Gaithersburgh, MD) in 10% dried milk in PBS/Tween. Plasma samples, as well as secondary antibodies, were incubated for 1 hour at 37 C. A tetramethylbenzidine substrate solution was added (KPL, Gaithersburgh, MD) and incubated at room temperature for 15 minutes. The reaction was stopped by adding 100 µl of 1M H 3 PO 4 and the O.D. (450 nm) was determined in an ELISA microplate reader. All samples were run in duplicate. Additionally, in each assay negative and positive (pooled sera from high-titer immunized macaques) controls were included. Linear regression curves were plotted for each for each sample and the titers were calculated as the inverse of the serum dilution that produced an O.D. of 0.2 above the blank. Animal geographic origin DNA extracted from archived peripheral blood lymphocytes was analyzed by the Molecular Anthropology Laboratory, (Davis, CA) and Primate Products, Inc. (Miami, FL) for geographic region of origin (studies 1-3, n=27). Twenty-four short tandem repeat (STR) genotypes were generated for each sample. The loci evaluated are: D1s548, D2s1333, D3s1768, D4s1626, D5s1457, D6s501, D7s794, 32

46 D7s1826, D8s1106, D8s1466, D9s921, D10s1432, D18s536, D8s1106, D9s921, D9s934, D11s2002, D11s1975, D13s318, D13s765, D18s537, D16s750, DXs2506, and AGAT007. Twenty of the 24 loci listed above were used in a linear discriminate analysis to determine the most likely region of origin for each sample. Samples were analyzed in conjunction with 13 known Sumatran (Indonesian), 33 Mauritian, 78 Philippine and 20 Vietnamese longtail macaques to determine their probability of origin from these four regions (129). Probabilities of geographic origin are shown in Table 4.2. Stool collection for 16S rrna analysis Stool samples were aliquoted into gram portions and immediately frozen at -20ºC, then stored at -80ºC until further processing. Stool samples collected from the following days were used for 16S rrna processing. Study 1: days 0, 1, 2, 7, 14, 28, 29, 30, 31, 32, 35, 56, 57, 58, 59, 60, 61, 70, and 84 (am samples). Study 2: days 0, 2, 4, 7, 10, 28, 30, and 35+ (final time point) (am samples). Study 3: days 0, 1, 2, 3, 4, 7, 14, 2and 8 (am samples). Study 4: days 0, 1, 2, 6, 7, and 14. DNA Extraction and Pyrosequencing Different DNA extraction methods were compared in Chapter 3, aim 1. We extracted DNA from six samples (macaque C6, d0-14) using four methods (PCl, modzymo, Zymo, and Qiagen) and compared compositional results. Two methods utilized standard kit directions: the standard rapid-methods ZR Fecal 33

47 DNA Isolation kit (Zymo Research, Inc.) and the QIAamp DNA Stool mini kit (Qiagen, Inc). We compared both of these standard kits to two customized extraction methods: a modified Zymo protocol (modzymo), and a phenol chloroform extraction method (PCl). Both of these customized protocols included an enzymatic pre-treatment and mechanical lysing pretreatment step, described below. For fecal samples collected on days 0-14 in study 1 and all study 4 samples, including for comparison of extraction protocols, total DNA was extracted using a modified phenol-chloroform (PCl) protocol. Samples were resuspended in 1M KPO 4, followed by 10 µl Proteinase K, 50 µl 10% SDS, 5000 units ml -1 mutanolysin, and 100 mg ml -1 lysozyme. 50 µl Phenol (ph 7.5) was added in conjunction with aggressive bead beating using Lysing Matrix tubes and FastPrep FP120 instrument (QBiogene), followed by repeated PCl extractions and ethanol washes. For the remainder of the samples in study 1 and all samples in studies 2 and 3, total DNA was extracted using a modified ZR Fecal DNA isolation kit (ZYMO Research Corp.). The same pre-treatment steps were used as with the phenol-chloroform methods (resuspension in 1M KPO 4, pre-treatment with 10 µl Proteinase K, 50 µl 10% SDS, 5000 units ml -1 mutanolysin, and 100 mg ml -1 lysozyme, and aggressive bead beating as described before). 50 ng of DNA was used to PCR amplify the V1-V2 region of the 16S rdna gene using the universal primers 27F and 338R, with unique barcodes for sample identification (130). Primer sequences and PCR cycling parameters are described in detail below. Samples were sequenced using 454 GS FLX or FLX Titanium 34

48 pyrosequencing chemistry via Life Sciences primer A at the Genomics Resource Center at the Institute for Genome Sciences, University of Maryland School of Medicine according to manufacturer recommendations. Sequence data can be accessed at SRA accession number SRA057090, study SRP Sample Preparation (Primer sequences and PCR conditions) All samples regardless of extraction protocol were prepared with the following conditions. The universal primer sequences were as follows, with underlined sections indicating 454 Life Sciences primers B and A, respectively: 27F, 5 - GCCTTGCCAGCCCGCTCAG-TCAGAGTTTGATCCTGGCTCAG-3 and 338R, 5 -GCCTCCCTCG- CGCCATCAGNNNNNNNNCATTACCGCGGCTGCTGGCA-3. PCR amplification measures are as follows for a starting volume of 50 µl: 50 ng starting DNA material, 1.0 µl of each primer (10 mm), 1.0 µl deoxyribonucleoside triphosphates (Invitrogen; 10mM), and either 0.3 µl of AmpliTaq Gold DNA polymerase with 5 µl 10x PCR buffer II and 1.5 µl MgCl2 (50mM) or AccuPrime DNA polymerase with 5 µl 10x PCR buffer II (both Invitrogen). Negative controls without DNA template were included with each PCR reaction. The following cycling parameters were used: 5 minutes (min) denaturing at 94 C followed by 30 cycles of denaturing for 30 seconds (s) at 94 C, annealing for 30 s at 55 C, and elongation for 90 s at 68 C, with a final extension period for 5 min at 72 C. Failed PCR reactions were retried at different 35

49 template and PCR conditions, and if these did not work, the sample was excluded from analysis. 100 ng of the PCR product from each sample was quantified using the Quant-iT PicoGreen dsdna assay (Invitrogen) and used for sequencing. Data Processing Mothur v was used to process the generated 16S rdna gene amplicons (131). Sequences were denoised, trimmed and quality-checked before clustering into operational taxonomic units (OTUs) at 97% pairwise identity, and aligned to the Silva reference alignment (132). The Greengenes database (downloaded March 2011) was used to assign taxonomic classification to filtered sequences (133). Genus-level classified sequence abundances were used to calculate the Shannon diversity index over time for each sample. In-house scripts, R, and mysql were used to parse data and perform analyses. The Jensen-Shannon divergence was calculated using an in-house R script, and partitioning around medoids (pam) clustering algorithm was used to identify community types, or enterotypes (21). Principle coordinate analysis was visualized using the R package scatterplot3d (134). Heatmaps were created using the R package gplots (135). Correlations between clinical symptoms of disease and normally rare organisms were calculated using JMP v Local similarity alignment (LSA) was used to calculate significant associations among genera and immunological measurements (136). Cytoscape v (137) was used to view network analyses. 36

50 Network analyses for community types were created by computing the pairwise Spearman rank correlation coefficient between pivotal genera with P < and a higher relative abundance of at least 0.1 in their corresponding community (Streptococcus, Lactobacillus, Prevotella, and Enterococcus for community types I, II, III, and IV, respectively), finding the high-order partial correlation by computing the spare inverse of the correlation matrix, and constructing a community network with the high-order partial correlation (138). Networks were visualized using Cytoscape (137). Network analyses for longitudinal analysis between microbial members and immunological measurements were analyzed using the program Local Similarity Analysis (LSA) (136). Results were filtered by a LSA cutoff of p<0.01 and q > 0.40 and viewed in Cytoscape Correlation between microsatellite loci and community type were calculated using an in-house ANOVA R script. For each macaque, community type relative abundance (calculated as the percent time spent in each community type) was normalized using arcsin(sqrt()) transformation, and an ANOVA was used to calculate statistical significance between each community type and the variation present within each locus. Macaques homozygous for an allele were calculated as one, and alleles occurring less than three times within the macaque populations tested were removed from analysis. Phylogenetic trees of shared OTUs for studies 1-4 were constructed using the filtered sequence results from Mothur. A random sampling of all representative OTUs using a similarity cutoff of 0.03 was concatenated to produce 37

51 a sequence file with 32,402 sequences total and aligned. The FastTree program was used to construct an approximately-maximum-likelihood tree, based on the generalized time-reversible (GTR) nucleotide evolution model (139). FigTree v1.3.1 ( was used to visualize and colorcode tree nodes based on taxonomic classification and study source. Genotype Analysis of MHC Microsatellites Genomic DNA was extracted from peripheral blood mononuclear cells of all cynomolgus macaques (n=33) using the Qiagen Dneasy blood and Tissue Kit (catalogue #QGN-69504). We amplified seven previously established MHCspanning microsatellites, D6S1691, D6S2741, D6S291, DRACA, DQcar, MICA, MOGc, and MICA, using previously described methods (140). 10 ng of starting DNA (2 µl of 5ng/ µl DNA) was amplified the Qiagen Multiplex PCR Reaction Kit (cat. # ) using 36 cycles of cycling parameters described previously (140, 141). The PCR product was diluted 50 times following PCR reaction. Genotypes were determined at the Biopolymer Core Facility at the University of Maryland School of Medicine using the Genscan Internal Lane Standard (cat #G9530, ILS-600 Promega-StemElite ID System). A distance matrix of the microsatellite data based on the Lynch distance was calculated using the R package polysat (142), and phylogenetic trees were constructed using Phylip v3.69 (143). The excel package GenAlEx was used to calculate MHC microsatellite allele frequencies (144). 38

52 Chapter 3 Aim 1: To establish robust methods for analysis of the gut microbiota in cynomolgus macaques Introduction The development of DNA-based, culture-independent methods in the past decades has led to improved methodology and the availability of multiple protocols to analyze bacterial populations in complex environments. Application of these approaches to the study of humans and other mammalian species has generated a massive amount of information available on the microbiota (5). Traditional methods for analyzing complex microbial communities were previously limited to the study of those organisms that could be grown in culture, and only a minor proportion of microbes in natural settings have been cultured. New DNA sequencing technology has made it possible to characterize microbial communities to a fuller extent, making use of the bacterial small-subunit (16S) ribosomal RNA (rrna) gene as a molecular marker (145, 146). The 16S rrna gene, which contains conserved regions that can be exploited in the design of universal PCR primers, is ubiquitous in all bacteria and archaea, and it has been shown that a small region of the gene can be used as a proxy for molecular identification (146). DNA sequence data from hypervariable regions of the 16S rrna gene can be used to identify bacterial taxa using an existing reference database (phylotyping), or define phylogenetic relationships by grouping the sequences into operational taxonomic units (OTUs), i. e. closely related sequences that represent a hypothetical taxonomic unit. Estimates of community diversity 39

53 and clustering algorithms from these measurements can then be used to compare different samples (131, 147). Taxonomic identification depends on the isolation of high-quality DNA from the sample, followed by PCR amplification, sequencing, and database matching of the 16S rrna hypervariable region. No single standardized protocol exists for 16S rrna analysis, and multiple steps during the process of DNA isolation can potentially affect the quality of starting material, thus impacting results for downstream analysis (148, 149). In microbial community profiling, an accurate representation of community composition is important since results are generally reported as relative abundances of taxonomic units. It is unknown how much heterogeneity is present in large samples, and differential sampling, i. e. sampling two different sites within the same sample, may not always produce the same results. Large apparent differences in community composition have also been observed when comparing different DNA extraction methods, which likely reflect differential lysis of organisms present in the starting material ( ). For example, cell lysis of many gram-positive organisms, such as Staphylococcus aureus, is more difficult than gram-negatives due to their thick cell wall. For fecal samples containing many Firmicutes, studies report that the addition of mechanical disruption, such as bead beating or enzymatic treatment, results in a proportional increase in gram-positive organisms in fecal samples (151, 152). It is thus critical to ensure consistency of the DNA extraction method within a study, and to be aware of possible inconsistencies when comparing data retrieved from 40

54 publicly available data where multiple extraction methods may have been utilized by different laboratories. Protocols for 16S rrna gene amplification also impact downstream results (145). Initial 16S rrna surveys utilized Sanger sequencing to sequence the full-length of the 16S rrna gene (1.5kb), but was limited by sequencing cost and throughput. Next-gen sequencing technology allows for an increased number of sequences per sample (up to 50,000 reads/sample), but produces shorter read lengths ( bp). Earlier 16S rrna surveys established that sequencing hypervariable regions ( bp) was comparable to surveys using the fulllength gene, although certain regions of the 16S rrna gene may be biased towards specific phylotypes (145, 153). Different information is contained in each of the 16S rrna hypervariable regions (V1-V9), and studies comparing their amplification have suggested different compositional results depending on the region used (153, 154). In comparison to classification using the full-length gene, higher classification rates have been reported for regions V1-V5. Next-generation sequencing technologies are continually improving and becoming more cost-effective. Roche 454 sequencing, a parallel pyrosequencing method that utilizes small DNA-capture beads in a water-in-oil emulsion to amplify single sequence fragments, has been used extensively in 16S rrna surveys (1). The first Roche 454 GS FLX sequencing platform produced up to 600, base pair (bp) reads per 96-sample set, with an average error rate of 0.5% overall and 82% of sequences containing no errors (155). Improvements in chemistry, data capture (CCD cameras), and sequence-filtering algorithms lead to 41

55 longer and increased reads per sample in the GS FLX Titanium platform (up to one million reads of 400 bp length, with an error rate of 0.2%) (156). Newer Illumina sequencing technology, which amplifies single DNA fragments in small DNA clusters, can sequence up to two million sequences for a 96-well plate, although read length is shorter than those produced by the FLX Titanium platform ( bp). Homopolymer regions are not problematic in Illumina sequencing, and a lower error rate of 0.5% is reported. Studies comparing different platforms of 454 sequencing (157) or different technologies (156, 158) indicate similarity between samples sequenced using different techniques, although changing technologies in large-scale studies conducted over the course of long collection periods should still warrant benchmarking studies to ensure comparability of data between samples. This project presents studies to describe changes in the intestinal microbiota of cynomolgus macaques in response to oral live-attenuated Shigella vaccines. In this section, we compared different methods of sampling, DNA extraction, and sequencing protocols to establish robust methods of analysis. We initially described the composition of the microbiota from longitudinal samples collected from a control group of macaques. Using these samples, we investigated the impact of differential sampling by comparing different aliquots of the same sample. We next compared different extraction protocols to establish whether a high-throughput protocol was sufficient to identify organisms within the cynomolgus macaque microbiota. Lastly, we investigated whether the results from samples sequenced by two different sequencing protocols were still 42

56 comparable. The results of these analyses determined which methods were used to investigate the changes occurring within the intestinal microbiota of a larger population of cynomolgus macaques. Results Composition of the cynomolgus macaque intestinal microbiota The human intestinal microbiota has been the subject of considerable study during the past 8 years. Since 2007, over 1,300 papers have been published describing the intestinal microbiota in humans (PubMed search: human gut microbiota ). To date, descriptions of the intestinal microbiota of nonhuman primates are lacking. As has been described in humans, both Bacteroidetes and Firmicutes are the dominant phyla in the intestinal microbiota of rhesus macaques (159), chimpanzees (160, 161), old world monkeys (162), and great apes (163). However, differences in the genera found in the gut microbiota of humans and nonhuman primates have been reported, such as a large proportion of the Bacteroidetes member Prevotella rather than the human inhabitant Bacteroides, and a larger abundance of the phylum Spirochaetes compared to humans. The extent of diversity present in the intestinal microbiota in cynomolgus macaques, an animal model frequently utilized in pathogenesis and vaccine research, compared to nonhuman primates and humans remains undescribed. We collected fecal samples from six untreated, control cynomolgus macaques in quarantine for use in establishing robust methods (study 4, macaques 1-6) for characterization of the 16S rrna profiles of the gut microbiota. Our 43

57 initial approach involved isolation of DNA using a phenol chloroform protocol that includes pre-treatment with proteinase K, mutanolysin, and lysozyme, followed by harsh, mechanical bead-beating using a FastPrep-24 cell homogenizer. The V1-V3 hypervariable region of the 16S rrna region was chosen based on previous studies confirming taxonomic consensus comparable to that observed using full-length 16S rrna genes (149, 153). Samples were sequenced using the Life Sciences 454 GS FLX sequencing platform, and Mothur (v1.19.1) was used for downstream analysis of sequences (see methods). Firmicutes, Bacteroidetes, Tenericutes, Actinobacteria, Proteobacteria and Spirochaetes were observed in the intestinal microbiota of cynomolgus macaques (Fig. 3.1A). Firmicutes (77.8%) and Bacteroidetes (17.1%) represented the two dominant phyla present in the cynomolgous intestinal microbiota. At the genus level, Lactobacillus (41.4%), Streptococcus (8.7%), Prevotella (12.8%), otu 2159 (7.7%) (an unidentified Ruminococcaceae family member), and Oscillospira (4.0%), were among the most abundant genera present (Fig. 3.1B). 44

58 A 100 Relative abundance (%) Firmicutes Bacteroidetes Tenericutes Actinobacteria Proteobacteria Spirochaetes 20 0 Day: B 100 Relative abundance (%) Macaque: Day: Macaque: 0 12 C1 C2 C3 C4 C Firmicutes Lactobacillus Streptococcus otu2159 (Ruminococcaceae) Oscillospira otu2087 (Lachnospiraceae) Blautia Coprococcus Bulleidia Incertae sedis, Erysipelotrichaceae Faecalibacterium Ruminococcus (clade 1) Dialister Ruminococcus (clade 2) Clostridium (clade 1) Incertae sedis, Lachnospiraceae Peptococcus Eubacterium (clade 1) Clostridium (clade 2) otu1998 (catabacteriaceae) Roseburia Sarcina Eubacterium (clade 3) Megasphaera Weissella Bacteroidetes Prevotella Parabacteroides otu1033 (Prevotellaceae) Figure 3.1. Composition of the cynomolgus macaque intestinal microbiota. Relative abundance of phyla (A) and genera present at < 2% (B) is indicated on the y-axis. Each horizontal bar represents a different fecal sample, collected from six untreated cynomolgus macaques. Day and macaque ID are labeled on the x- axis, and phyla or genera are color-coded as indicated in inset box. To compare overall compositional similarity between samples, we performed a multidimensional analysis at the genus level based on the Jaccard similarity coefficient. A cluster dendrogram of the distance matrix calculated using the Jaccard distance (1 - Jaccard coefficient similarity) suggested partial 0 12 C1 C2 C3 C4 C C C6 45

59 clustering of samples from the same individual, particularly in samples collected on consecutive days (Fig. 3.2A). Further statistical analysis using a student t-test indicated that the intra-individual distance (distance calculated between samples from the same macaque) was less than inter-individual distance (distance calculated between samples from different macaques) (Fig. 3.2B). A Height C3_2 C2_1 C2_2 C1_2 C1_0 C1_1 C4_2 C4_14 C6_7 C4_7 C1_6 C1_7 C4_1 C4_0 C4_6 C6_2 C6_6 C2_7 C5_7 C6_1 C5_0 C5_6 C5_1 C5_2 C2_6 C3_7 C6_0 C6_14 C1_14 C3_1 C3_14 C5_14 C2_14 C3_6 C2_0 C3_0 C1 C2 C3 C4 C5 C6 B Jaccard distance (1 - Jaccard similarity coefficient) Within monkey (intra-) p < Between monkeys (Inter-) Figure 3.2. Intra- and inter-individual similarity of the microbiota composition in cynomolgus macaques. A) Tree showing complete clustering based on Jaccard distance (1 Jaccard similarity coefficient). Tree labels colorcoded by individual macaque (C1-C6) as indicated in inset box. B) Mean Jaccard distance within samples from the same macaque (intra-individual variation) and between different macaques (inter-individual variation). Statistical significance calculated using a student t-test. In comparison to the intestinal microbiota of rhesus macaques, we observed a higher relative abundance of Firmicutes, particularly Lactobacillus, 46

60 and a lower abundance of Prevotella in the intestinal microbiota of cynomolgus macaques (159). While these differences may be species-specific, DNA extraction methods also varied between these two studies (we used harsh bead-beating in our extraction method and McKenna et al. did not). For future sampling protocols, we aimed to establish a reliable high-throughput extraction method to use in largescale projects and, therefore, compared different DNA sampling and sequencing methods. Using samples from the untreated group of cynomolgus macaques, we analyzed different protocols to establish the best methodology before further describing the microbial composition of a larger study sample set from vaccinated and challenged macaques. Comparison of differential sampling The types of bacteria found along the gastrointestinal tract vary according to segment. Fecal samples are extensively used as a surrogate of the intestinal microbiota, and the species identified in fecal samples have been observed to be most similar to those in the colon (10). Only a small amount of fecal sample is necessary to extract sufficient DNA for the analysis of the microbiota (~25 mg), even though a large amount of sample may be collected and frozen at a specific sampling time. It is possible that distribution of bacterial species across a large sample may be heterogeneous. To investigate potential variability in different aliquots of the same fecal sample, we assessed 16S rrna composition and similarity between three different aliquots from the six samples. DNA was isolated from six samples 47

61 collected from a single macaque using the phenol chloroform method previously described (days 0,1,2,6,7,14). The same phyla (Fig. 3.2A) and genera (Fig. 3.2B) were observed in all samples from all macaques (Fig. 3.1). Similar relative abundance of Firmicutes, particularly Lactobacillus, unclassified Lachnospiraceae, and otu 2159, were observed among the three aliquots of the same sample, with the exception of aliquot A from sample C6_1. Similarly, the relative abundance of Prevotella was comparable within each sample. Relative abundance (%) Aliquot: A B C A B C A B C A B C A B C A B C Sample: C6_0 C6_1 C6_2 C6_6 C6_7 C6_14 Firmicutes Lactobacillus unclassified Lachnospiraceae otu2159 (Ruminococcaceae) Oscillospira Coprococcus unclassified Ruminococcaceae Blautia Ruminococcus Clostridium Faecalibacterium Incertae sedis Lachnospiraceae unclassified (Clostridiales) otu1995 (Clostridiales) Incertae sedis, Erysipelotrichaceae Bulleidia otu2087 (Lachnospiraceae) Dialister Peptococcus Sarcina Eubacterium unclassified Clostridiales Family XIII Roseburia otu1998 (Catabacteriaceae) Bacteroidetes Prevotella otu991 (Bacteroidales) unclassified Bacteroidetes unclassified Bacteroidales Parabacteroides Tenericutes p-75-a5 unclassified Mollicutes Unclassified unclassified Figure 3.3. Relative abundance of genera in different aliquots of the same sample. The relative abundance (%) of genera is shown on the y-axis, and each group of three bars on the x-axis represents a different sample: A, B and C indicate different aliquots of the same sample for samples C6_0, C6_1, C6_2, C6_6, C6_7, and C6_14. Genera are color-coded as indicated by inset box. To establish whether the composition between aliquots from the same sample was more similar than the composition between different samples, we calculated a Jaccard distance matrix and built a dendrogram as previously conducted for a multidimensional analysis of all genera observed within the samples (Fig. 3.3A). Overall, the observed variation between aliquots of the same 48

62 sample was significantly less than that found between consecutive sampling days (>2 days) and distant sample days (<4 days) using a student t-test (Fig. 3.3B). A Height C6_1A C6_0B C6_0 C6_1 C6_2 C6_6 C6_7 C6_14 C6_14A C6_14C C6_0A C6_1C C6_14B C6_0C C6_1B C6_2B C6_2A C6_2C C6_7B C6_6B C6_6A C6_6C C6_7A C6_7C B Jaccard distance (1 - Jaccard similarity coefficient) within sample (0 days) p = p < consecutive samples (2> days) distant samples (4< days) Figure 3.4. Multidimensional compositional comparison of different aliquots of the sample. A) Cluster dendrogram based on the Jaccard distance of sample aliquots processed from six samples (macaque C6, days 0-14), three aliquots each (A, B, and C) (n=18). Samples are color-coded in inset box. B) Mean Jaccard distance (y-axis) between aliquots of the same sample (within sample), aliquots of samples collected on consecutive days (consecutive samples), and samples collected on distant days (distant samples). A student t-test was used for statistical significance. As observed in samples collected from each of the six untreated macaques, the intra-individual variability is less than inter-individual variability (Fig. 3.1). The analysis here indicates that there is even less variation observed in aliquots from the same sample (intra-sample variation) than the variation observed within a macaque (intra-individual variation). Comparison of DNA extraction methods To obtain an accurate measure of the composition of a microbial community using 16S rrna analysis, high-quality DNA that is representative of 49

63 the community present within that sample must be isolated. Several protocols and prepared kits are available for DNA extraction, but they differ in their efficacy to lyse Gram-positive and Gram-negative cells. Firmicutes, in particular, require harsher lysing conditions due to their thick cell wall (151). In initial sample preparations, we used a phenol chloroform (PCl) DNA extraction method that included a triple enzyme pre-treatment and a mechanical bead-beating step. However, this method is labor-intensive and not practical for large-scale, high throughput projects. Therefore, we modified the protocol in the standard rapid-method ZR Fecal DNA Isolation kit (Zymo Research, Inc.) to include an enzymatic pre-treatment and mechanical lysing step as utilized in our previous phenol chloroform method (a modified Zymo protocol). To assess results of the modified protocol, we compared both the PCl and modified Zymo protocols to two standard DNA isolation kits, the Zymo kit (using the directed protocol) and QIAamp DNA Stool mini kit (Qiagen, Inc). We extracted DNA from six samples (macaque C6, d0-14) using these four methods (PCl, modzymo, Zymo, and Qiagen) and compared compositional results. No significant difference was observed in total number of reads for a particular extraction method (Table 2.1). All methods produced 16S rrna profiles that contained the same phyla and genera as previously described (Fig. 3.5). However, the relative abundance of Firmicutes was increased in samples processed with either the PCl or modified Zymo methods in comparison to either the standard Zymo or Qiagen methods. 50

64 A 100 B 100 Relative abundance (%) Mean relative abundance (%) Extraction: A mz Z Q A mz Z Q A mz Z Q A mz Z Q A mz Z Q A mz Z Q 0 phenol modzymo Zymo Qiagen Sample: C6_0 C6_1 C6_2 C6_6 C6_7 C6_14 Firmicutes Lactobacillus otu2159 (Ruminococcaceae) unclassified Lachnospiraceae Oscillospira unclassified Ruminococcaceae Blautia Coprococcus Clostridium Ruminococcus otu1995 (Clostridiales) unclassified Clostridiales Faecalibacterium Incertae sedis, Lachnospiraceae otu2087 (Lachnospiraceae) Dialister Incertae sedis, Erysipelotrichaceae Bulleidia Sarcina Eubacterium Peptococcus Roseburia Megasphaera otu1998 (Catabacteriaceae) Butyrivibrio Adlercreutzia Dorea Bacteroidetes Prevotella otu991 (Bacteroidales) unclassified Bacteroidetes unclassified Bacteroidales Parabacteroides Collinsella Spirochaetes Treponema Tenericutes p-75-a5 unclassified Mollicutes Unclassified unclassified Figure 3.5. Relative abundance of genera in samples processed by different DNA extraction methods. A) Relative abundance of genera present at <2% (y-axis) for samples collected from macaque C6, days 0-14, each processed by four different extraction methods (x-axis, n=24)): phenol chloroform (A),modified Zymo (mz), standard zymo (Z), and Qiagen (Q) protocols. B) Mean relative abundance of genera (y-axis) by extraction method (x-axis, n=6/extraction method). Genera for both (A) and (B) are color-coded as indicated in inset box. We utilized multidimensional analysis based on the Jaccard distance to compare similarity between samples extracted by the different methods. With the exception of sample C6_1A, two clear clusters were apparent: (i) samples extracted by PCl and modified Zymo clustered together, and (ii) samples 51

65 extracted with the two standard kits (Zymo and Qiagen kits) (Fig. 3.6A). We next investigated which specific components within the community contributed to the separation between extraction methods. No significant difference was observed in the mean Shannon diversity index calculated for each sample, which reflects both the number of organisms and abundance of those organisms, indicating that all methods lysed a high diversity of species (Fig. 3.6B). However, a significant difference was observed in the mean relative abundance of Prevotella and Lactobacillus between both the PCl and modified Zymo methods in comparison to the standard kits (p < , Fig. 3.6C). A less significant difference in the relative abundance was observed between the standard Zymo or Qiagen methods and either method that utilized mechanical bead-beating (PCl and modified Zymo, (p < 0.05)). Other prominent taxa did not differ significantly between the four extraction methods. 52

66 A Height C Mean relative Abundance (%) C6_1A p < PCl p < p < modzymo p < Zymo DNA extraction method PCl (A) modi. Zymo (mz) Zymo (Z) Qiagen (Q) C6_1Z C6_0Z C6_2Z C6_2Q C6_0Q C6_1Q C6_6Q C6_14Z C6_6Z C6_7Z C6_7Q C6_14Q C6_0A C6_1mZ C6_14A C6_14mZ C6_6A C6_7A C6_6mZ C6_7mZ C6_2A C6_0mZ C6_2mZ Qiagen B Mean Shannon diversity index phenol modzymo Zymo Qiagen Extraction method Prevotella Lactobacillus otu991(bacteroidales) otu2159 (Ruminococcaceae) unclassified Lachnospiraceae Oscillospira unclassified Ruminococcus Blautia Coprococcus Clostridium Figure 3.6. Comparison of sample composition by extraction method. A) Cluster dendrogram of samples by extraction method calculated using the Jaccard distance. Samples are color-coded by extraction method according to inset box. B) Mean Shannon diversity index (y-axis) for each DNA extraction method (xaxis). C) The mean relative abundance of ten most dominant genera (y-axis) in samples for each extraction method (x-axis). Genera are color-coded as indicated in inset box, and a student t-test was used to calculate statistical significance. We observed similar results in samples processed with methods that utilized mechanical bead-beating (PCl and modified Zymo) compared to methods that did not (standard Zymo and Qiagen). A greater proportion of Firmicutes, especially Lactobacillus, was recovered with bead-beating protocols. These results are consistent with the observations made in previous studies that have 53

67 compared DNA extraction methods, which report a higher recovery of Firmicutes using protocols with mechanical disruption or enzymatic pretreatment (149). We observed no significant differences in comparison of PCl and modified Zymo DNA extraction methods, suggesting that these methods may be used interchangeably for the purposes of fecal DNA preparation. Comparison of two sequencing platforms: 454 GS FLX vs. GS Titanium There have been considerable changes in DNA sequencing technologies over the past 10 years that have increased throughput and decreased sequencing costs. One of the challenges during the past decade has been the validation of new sequencing platforms and chemistries and the development of metrics for robust comparison of sequences generated with these different approaches. Each DNA sequencing platform and chemistry has its own set of biases and errors, often confounding the comparisons of different datasets. Molecular classification of 16S rrna sequencing relies on assignment of sequences to an existing database, and sequencing errors are always a concern in large-scale comparisons of samples sequenced by different sequencing platforms. For complete analysis of the cynomolgus macaque intestinal microbiota across multiple vaccine trials, we sequenced a total of 374 samples from four different studies of macaques. During sample collection in these longitudinal studies, improvements in chemistry, camera technology, and sequence filtering algorithms in the GS FLX 454 pyrosequencing platform lead to increases in both read lengths and the number of reads produced per sample (FLX Titanium 54

68 sequencing technology). We assessed how changes in the 454 sequencing platform affected the comparability of our results by comparing all 374 samples using various multidimensional analyses. GS FLX sequences accounted for 26% of all samples (Table 2.1). Although sequences from these samples were significantly shorter (Fig. 3.7A) and samples had a reduced number of sequences overall (Fig. 3.7B), compositional differences were not significant. The estimated number of species (phylotypes) within a sample was not affected by an increase in the number of reads per sample, as observed by similar rarefaction curves (an estimation of species richness and appropriate sequencing depth) for both FLX and Titanium samples (Fig. 3.7C). The rarefaction curves for samples sequenced by both platforms plateau at approximately 3500 sample size, indicating that the depth of sequencing coverage was sufficient to provide robust estimates of diversity. Furthermore, no correlation was observed between the Shannon diversity index and the number of reads per sample, suggesting that estimates of diversity within a sample were not impacted by an increased total amount of reads (Fig. 3.7D). Similarly, the relative abundance of the dominant genera Lactobacillus, Streptococcus, and Prevotella was not significantly correlated with an increase in total read counts per sample (Fig. 3.7E-G), suggesting that sequence coverage did not impact estimates of the relative proportions of taxa observed between the two sequencing platforms. 55

69 P-Value 2.18e-07 R-Square 0.01 P-Value 4.46e-02 R-Square 0.01 P-Value 4.94e R-Square Lactobacillus Prevotella 3 2 Streptococcus P-Value 3.75e-04 4 R-Square 0.03 G Relative read abundance (%) D Sample Size Relative read abundance (%) F FLX Tit. FLX Titanium 0 Relative read abundance (%) Mean # reads/sample E Sequencing platform 0 10 Estimated species # *** FLX Tit. C # reads/sample 0 Shannon diversity index B *** Mean read length (bp) 400 A # reads/sample Figure 3.7. Effect of read length and number of reads per sample on taxonomic composition for all samples analyzed (n=374). A) Mean read length of samples and B) mean number of sequence reads per sample by either FLX or Titanium sequencing platforms. Three stars indicates p < by student t-test. C) Rarefaction curves for each sample of the number of estimated species as a function of sample size (reads per sample). D) Shannon diversity index (y-axis) as a function of the number of reads per sample (x-axis). D-F) Relative abundance of Lactobacillus (D), Prevotella (E), and Streptococcus (F) as a function of the number of reads per sample (x-axis). The Pearson correlation coefficient R2 is indicated in the topleft corner and statistical significance calculated using the student t-test in the topright corner for C-F. Sequencing platform (FLX or Titanium) is color-coded as indicated in inset box in (C). Multidimensional analysis of the composition of samples sequenced by either method supported these results. Principal coordinate analysis (PCoA) based on both the Jaccard and Jensen-Shannon distance (Fig. 3.8) indicated only slight 56

70 variation when comparing the multicomponent axes. A slight bias in the PCoA axis 2 using the Jaccard index (12% variance explained) and axis 3 using the Jensen-Shannon index (13% variance explained) is visible. However, neither of these differences was significant (student t-test). Jaccard distance A B C Jensen-Shannon distance D E F FLX Titanium PCoA 2 (15%) PCoA 3 (13%) PCoA 3 (13%) PCoA 2 (12%) PCoA 3 (9%) PCoA 3 (9%) PCoA 1 (28%) PCoA 1 (28%) PCoA 2 (12%) PCoA 1 (42%) PCoA 1 (42%) PCoA 2 (15%) Figure 3.8. Multidimensional comparison of samples sequenced by different sequencing platforms. PCoA of the Jaccard distance (A-C) and Jensen-Shannon distance (D-F) calculated for each sample, color-coded by sequencing platform as indicated by inset box. X- and y-axes are labeled by PCoA dimension and the percent variance explained by that axis. Our results suggest that the use of GS FLX vs. Titanium platforms does not significantly impact estimates of community composition. Although minor differences are observed at the multidimensional level, actual composition and 57

71 taxonomic assignment remain unbiased regarding both sequence length and total number of sequences. We are confident that both methods can be used interchangeably to compare samples. Summary The variety of methods available for sample preparation, amplification, and sequencing produces concerns about the comparability of studies to one another. In studies where communities will be compared, methods should not be changed without benchmarking the effect of such a change on the comparability between samples. Our analyses suggested that differential sampling and changes in the sequencing platform had a minor impact on the comparability of samples and community composition within this study (Fig. 3.3, 3.4, 3.7, 3.8). However, the DNA extraction method had a significant effect on both community composition and overall similarity between samples (Fig. 3.5, 3.6). Specifically, methods that incorporated mechanical bead-beating were more effective at lysing Firmicutes, particularly Lactobacillus, compared to methods that did not. Due to the observed high prevalence of Firmicutes in the cynomolgus intestinal microbiota (Fig. 3.1), mechanical bead-beating should be incorporated into future protocols. 58

72 Chapter 4 Aim 2: To determine the effect of oral vaccination with live-attenuated S. dysenteriae 1 strains and subsequent challenge with wild-type S. dysenteriae 1 on the GI microbiota in cynomolgus macaques Introduction The world-wide burden of Shigella is estimated to be near 90 million cases of illness and over 100,000 deaths each year (106). The small inoculum of Shigella required to cause disease in humans ( bacteria), and the recent appearance of antibiotic-resistant strains and new serotypes is a concern. As discussed in the introduction, a successful vaccine against Shigella would have high immunogenicity against all epidemiologically relevant species without adverse side effects. A licensed vaccine is not yet available despite progress in developing promising vaccine strategies during the past 5 years (109). Vaccine development against Shigella has been slowed by variations observed in the efficacy of vaccines evaluated worldwide (109). For example, the attenuated S. flexneri 2a vaccine strain SC602 demonstrated strong immunogenicity and conferred protection in North American volunteers, but it was found to be associated with minimal vaccine shedding and low immune stimulation in volunteers in Bangladesh (117, 123, 124). Similar conflicting results between developed and developing global populations have been observed in oral vaccine studies against polio, cholera, and rotavirus (109, 125, 126). Development of a successful vaccine against Shigella has also been hampered by the lack of a small animal model that mimics human disease (109). Mice, guinea 59

73 pigs, and rabbits have been used to assess virulence and screen for vaccine candidates, but these model systems lack the ability to directly predict protection in humans (107). A more clinically relevant model has been developed with S. flexneri in rhesus macaques, Macaca mulatta, (121) and S. dysenteriae 1 in cynomolgus macaques, Macaca fascicularis (122), although a large inoculum (~10 10 CFU) is necessary to cause bacillary dysentery. Multiple factors such as diet, nutrition, and host genetics may impact vaccine efficacy (127). An additional important factor that may contribute to the variability in immunogenicity observed between different geographic populations with regard to the efficacy of orally administered vaccines is the composition of the gastrointestinal microbiota. A diverse and rich microbial community within the gastrointestinal tract plays an essential function in human health through the acquisition of nutrients, the maintenance of the immune system, and protection from pathogens (1, 2). To date, the potential role of the microbiota in vaccine development against enteric pathogens remains unclear. In this aim, we investigated the interaction between the intestinal microbiota and live-attenuated or wild type strains of S. dysenteriae 1 in cynomolgus macaques. We further characterized the intestinal microbiota in cynomolgus macaques from different geographic origins to expand on the results described in the first aim of this project. We then examined the effects of liveattenuated and wild-type S. dysenteriae 1 strains on the composition of the intestinal microbiota. Finally, we examined correlations between distinct microbial community profiles and clinical symptoms of infection following 60

74 challenge with wild type S. dysenteriae 1. Aim 3 (Chapter 5) investigated these results further by analyzing correlations between the microbial community profiles characterized in Aim 2 and the immune response mounted following immunization and challenge. Results Core intestinal microbiota in cynomolgus macaques We preliminarily described the intestinal microbiota of six control, untreated cynomolgus macaques in Chapter 3 (Fig. 3.1, 3.2). These untreated, control macaques, collectively referred to as the study 4 group from this point on, received no treatment throughout the course of sample collection, but provided an important measurement of the consistency of the microbiota over time. In addition to this group, we collected fecal samples from cynomolgus macaques participating in immunization studies over the course of different treatment and sampling timelines (studies 1-3, Fig. 4.1) at the Center for Vaccine Development (CVD) at the University of Maryland School of Medicine. Studies 1-3 examined the impact of different immunization protocols with live-attenuated S. dysenteriae 1 vaccines and/or challenge with wild type S. dysenteriae 1. To further evaluate the intestinal microbiota in cynomolgus macaques, the composition of the macaque intestinal microbiota was characterized from a total of 374 fecal samples (see methods). A total of 2,779,766 high quality sequences, with an average sequence length of 324 bp and an average of 7,433 reads/sample, were produced using the Roche 454 pyrosequencing platform. 61

75 Figure 4.1. Experimental design for cynomolgus monkey studies. (A) Study 1. Macaques were immunized with live-attenuated S. dysenteriae 1 strain (CVD1256 or CVD1255) or mock-immunized with PBS on days 0 and 28, followed by wild type S. dysenteriae 1 strain 1617 challenge on day 56. (B) Study 2. Macaques were immunized with CVD1256 or mock-immunized with PBS on days 0, 2, 4, and 7, followed by wild type challenge with S. dysenteriae 1 strain 1617 on day 28. (C) Study 3. Macaques were challenged with wild type S. dysenteriae 1 strain (D) Study 4. Macaques received no intervention. For all studies, samples were collected at labeled time points (days) for 16S rrna pyrosequencing. Immunization types occurred on indicated days (inset box). To describe the composition of the bacterial communities of the intestinal microbiota in cynomolgus macaques, the Greengenes 16S rrna database was utilized to identify 247 different genera (133). The two most abundant phyla identified in the cynomolgus macaque gastrointestinal microbiota were 62

76 Bacteroidetes (10.63% of classified sequence reads) and Firmicutes (78.1%) (Fig. 4.2A). While these are the same dominant phyla as those found in the human intestinal microbiota, the cynomolgus macaque microbiota differs from that in humans by its almost 8-fold greater abundance of Firmicutes, the most abundant phyla in cynomolgus macaques, as compared to Bacteroidetes (1). Major genera within the Firmicutes phylum included Lactobacillus (38.8%), Streptococcus (11.7%), Clostridium (clade 1, 6.6%), Enterococcus (4.7%), and a Ruminococcaceae family member designated as otu 2159 (5.7%) (Fig.4.2B). Typically, one or two of these genera represented the majority of genera in these samples at any given time, with other organisms each comprising 2% of the classified sequences. A B Relative abundance (%) Firmicutes Bacteroidetes Tenericutes Actinobacteria Proteobacteria Spirochaetes Relative abundance (%) Firmicutes Bacteroidetes Other Lactobacillus Streptococcus otu2159 (Ruminococcaceae) Clostridium (clade 1) Enterococcus otu2087 (Lachnospiraceae) Oscillospira Bulleidia Ruminococcus (clade 1) Incertae sedis (Erysipelotrichaceae) Sarcina Coprococcus Blautia Leuconostoc Dialister otu1998 (Catabacteriaceae) Butyrivibrio Weissella Incertae sedis (Lachnospiraceae) Eubacterium (clade 1) Ruminococcus (clade 2) Aerococcus Peptococcus Faecalibacterium Prevotella Flavobacterium Pedobacter Parabacteroides Paludibacter otu1033 (Prevotellaceae) Comamonas Yersinia Treponema p-75-a5 Figure 4.2. Rank abundance plots of (A) phyla and (B) top 35 most abundant genera. Relative read abundance (within total reads per study) is shown on the y- axis, and genera are listed on the x-axis. 63

77 We compared the genera observed in cynomolgus monkeys to those found in a subset of four humans (see methods). While the same dominant phyla are found in the intestinal microbiota of both primate species, the cynomolgus macaque microbiota differs from that in humans by its almost 8-fold greater abundance of Firmicutes (most notably Lactobacillus and Streptococcus), the most abundant phyla in cynomolgus macaques, as compared to Bacteroidetes (1). There is a significantly greater percentage of the Bacteroidetes, Bacteroides clade 1 and Prevotella, in humans Mean relative abundance Human Cynomolgus 0.00 Lactobacillus Bacteroides clade 1 Streptococcus Prevotella otu2159 (Ruminococcaceae) Clostridium clade 1 Coprococcus Faecalibacterium Enterococcus Blautia Ruminococcus clade 1 Oscillospira Roseburia otu2087 (Lachnospiraceae) family Erysipelotrichaceae Ruminococcus clade 2 Treponema Bulleidia Sarcina Dialister Collinsella Catenibacterium Parabacteroides otu1998 (Catabacteriaceae) Leuconostoc Clostridium clade 2 Clostridium clade 3 Alistipes Pseudomonas family Lachnospiraceae Weissella Butyrivibrio p.75.a5 otu4381 (F16) Comamonas Flavobacterium Eubacterium clade 4 Lachnospira Yersinia Taxa Figure 4.3. Comparison of the genera found in both humans and cynomolgus macaques. Mean relative abundance (y-axis) of most dominant genera (x-axis) shared by humans (blue) and cynomolgus macaques (red). 64

78 We defined a core bacterial community within the cynomolgus macaque intestinal microbiota as the genera present in at least 85% of all samples at abundance greater than 0.1% of the total community, regardless of the study group examined (Fig. 4.4). Lactobacillus, Streptococcus, Prevotella, Clostridium, and the Ruminococcaceae clade otu 2159 represented the most abundant organisms across all macaque samples examined, and are considered to be members of the core, being present in >90% of all samples collected. The core microbiota also contained several low abundance genera including Oscillospira (1.4%) and otu 1998 (family Catabacteriaceae) (0.55%) % prevalence: Treponema Sarcina Unclassified Erysipelotrichaceae otu 1998 (Catabacteriaceae) Adlercreutzia 90-95% prevalence: Core genera Streptococcus Bulleidia Coprococcus Blautia % prevalence: Lactobacillus Prevotella Clostridium (clade 1) otu 2159 (Ruminococcaceae) otu 2087 (Lachnospiraceae) Oscillospira Ruminococcus (clade 1) Mean relative abundance Study 1 - Indochina/Indonesia/ Philippines or Admixed Study 2 - Mauritius Study 3 - Philippines Study 4 - Indochina Figure 4.4. Core gastrointestinal microbiota profiles in cynomolgus macaques. Core genera (genus-level bacterial groups identified at 85-90%, 90-95%, and % prevalence rates within the complete sample set of n=374) are shown on the y-axis. The relative mean abundance for each of these genera within each study group (studies 1-4) is shown on the x-axis, color-coded as indicated in the boxed inset, with the vendor-reported geographic origin. 65

79 While these core genera were present within samples from all four studies, the mean relative abundance of these dominant organisms varied between the studies, especially within study 2 (Fig. 4.5, 4.4). The relative abundances of genera in study 2 were more evenly distributed compared to the relative abundances of genera in studies 1, 3, and 4, where Lactobacillus accounted for 40% or more of the sequences. Lactobacillus accounted for less than 20% of the sequences in study 4. A B Lactobacillus Clostridium (clade 1) Streptococcus Enterococcus Comamonas Sarcina Treponema Leuconostoc otu2087 (Lachnospiraceae) Prevotella Weissella Ruminococcus (clade 1) otu2159 (Ruminococcaceae) Oscillospira Coprococcus Bulleidia Blautia Paludibacter Eubacterium (clade 1) Parabacteroides Incertae sedis (Erysipelotrichaceae) Yersinia otu1998 (Catabacteriaceae) Faecalibacterium otu1033 (Prevotellaceae) Incertae sedis (Lachnospiraceae) p-75-a5 Butyrivibrio Peptococcus Dialister Ruminococcus (clade 2) Flavobacterium Aerococcus Pedobacter Relative abundance (%) Lactobacillus Streptococcus Clostridium (clade 1) Enterococcus Prevotella Treponema otu2159 (Ruminococcaceae) Sarcina otu2087 (Lachnospiraceae) Leuconostoc Oscillospira otu1998 (Catabacteriaceae) Ruminococcus (clade 1) Bulleidia p-75-a5 Flavobacterium Yersinia Incertae sedis (Erysipelotrichaceae) Pedobacter Coprococcus Blautia Comamonas Weissella Eubacterium (clade 1) Dialister Butyrivibrio Parabacteroides Incertae sedis (Lachnospiraceae) Paludibacter Ruminococcus (clade 2) Faecalibacterium otu1033 (Prevotellaceae) Peptococcus Aerococcus Relative abundance (%) Lactobacillus Prevotella Streptococcus otu2159 (Ruminococcaceae) Oscillospira otu2087 (Lachnospiraceae) Blautia Coprococcus Bulleidia Incertae sedis (Erysipelotrichaceae) Faecalibacterium Ruminococcus (clade 1) Dialister Ruminococcus (clade 2) Clostridium (clade 1) Incertae sedis (Lachnospiraceae) Peptococcus Eubacterium_(clade_1) Parabacteroides p-75-a5 otu1998 (Catabacteriaceae) Sarcina Butyrivibrio Treponema otu1033 (Prevotellaceae) Weissella Paludibacter Leuconostoc Enterococcus Aerococcus Flavobacterium Pedobacter Comamonas Yersinia Relative abundance (%) Lactobacillus Prevotella Streptococcus Bulleidia Oscillospira Coprococcus Dialister Blautia Butyrivibrio Treponema Aerococcus Sarcina otu1998 (Catabacteriaceae) Eubacterium (clade 1) Weissella Ruminococcus (clade 2) Peptococcus Leuconostoc Faecalibacterium Enterococcus p-75-a5 otu1033 (Prevotellaceae) Parabacteroides Paludibacter Comamonas Flavobacterium Pedobacter Yersinia Relative abundance (%) otu2159 (Ruminococcaceae) otu2087 (Lachnospiraceae) Clostridium (clade 1) Ruminococcus (clade 1) Incertae sedis (Erysipelotrichaceae) Incertae sedis (Lachnospiraceae) C D 66

80 Figure 4.5. Rank abundance plots for all samples in (A) study 1, (B) study 2, (C) study 3, and (D) study 4. Relative read abundance (within total reads per study) is shown on the y-axis, and genera are listed on the x-axis. Community Types of the Cynomolgus Monkey Intestinal Microbiota To further characterize the cynomolgus gastrointestinal microbiota, we investigated whether the microbial communities clustered into specific enterotypes (distinct community types within the gastrointestinal tract), as has previously been described in human fecal samples (21, 23). We used multidimensional cluster analysis and principal coordinate analysis (PCoA) to compare the calculated Jensen-Shannon divergence for each pairwise comparison of samples in this study. Using the partitioning around medoids (pam) method (21), we identified four distinct clusters, or community types, analogous to the human enterotypes, designated as community types I - IV (Fig. 4.6). The average silhouette width suggests a gradient effect for our data set (S(i) = 0.32; 4 clusters), rather than distinct clusters observed in human studies (21). The majority of samples (87.9%) clustered within community types I - III, and these three community types were found in macaques in all studies, with the exception that the high diversity community type III was not observed in study 3 (Table 4.1). 67

81 A PCoA 3 (13%) PCoA 1 (42%) PCoA 2 (15%) I II III IV B Streptococcus Prevotella + otu2159 Lactobacillus Enterococcus C Shannon Diversity % Relative abundance Shannon Diversity Index I II III IV I II III IV Community Type I II III IV I II III IV I II III IV Community Type Figure 4.6. Community types within the gastrointestinal microbiota of cynomolgus macaques. (A) PCoA based on Jensen-Shannon divergence of all samples (n=374) color-coded by community type clusters identified by the partitioning around medoids (pam) clustering algorithm. The percent variance explained is shown in parentheses. (B) Boxplots of the median relative abundance (y-axis) of dominant genera observed in each community type (xaxis), as determined by read abundance. Error bars indicate the interquartile range between the first and third quartiles. (C) Boxplots of the median Shannon diversity index calculated within each community type. 68

82 Community Type: I II III IV Number of samples in study 1: Number of samples in study 2: Number of samples in study 3: Number of samples in study 4: 79 (36%) 21 (22%) 3 (12%) 16 (44%) 5 (2%) 66 (70%) 0 13 (36%) 98 (45%) 38 (17%) Total number of samples: (7%) (58%) 7 (19%) 7 (29%) Total number of samples: Table 4.1. Community types within each study. Number and percentage of samples within each study belonging to community type I, II, III, or IV. Total numbers within each community type are indicated on the bottom row, and total number of samples on the right end row. Firmicutes members dominated each of these community types, with the exception of community type II, which was characterized by a similar abundance of both a Firmicutes (otu 2159, 19.2%) and Bacteroidetes member (Prevotella, 16.4%), and a lower abundance of Lactobacillus (4.1%) compared to other community types (Fig. 4.6B, Fig. 4.7). In contrast, Lactobacillus and Streptococcus accounted for approximately equal numbers of reads in community type I (31.8% and 29.8%), and Lactobacillus accounted for the majority of the reads in community type III (69.8%). Samples clustering into the less abundant community type IV were dominated by Enterococcus (28.2%), a prevalent but less abundant member of the core microbiota. These four community types also differed in diversity as measured by the Shannon diversity index (Fig. 4.6C), which is influenced by both richness of genera (number of organisms present) and 69

83 their relative proportions (evenness). In particular, community type III, which was dominated by a very high percentage of Lactobacillus (> 60%), exhibited the lowest Shannon diversity, and community type II, which contained several genera in approximately equal proportions, displayed the highest diversity (Fig. 4.6C). Interactions among genera in a given community type may be essential determinants of functional capabilities and may also influence the stability of a community in response to environmental perturbations. To identify specific relationships between genera within each community type, we constructed correlation networks surrounding the most abundant genera (Streptococcus, Lactobacillus, Prevotella, and Enterococcus for community types I, II, III, and IV, respectively) based on Spearman correlation coefficients (P value < 0.001) within each community type (138). Both dominant and lower abundance genera were significantly correlated within each community type (Fig. 4.7). Increases in the dominant genus within community types I, III and IV were negatively correlated with less abundant community members (Fig. 4.7, A, C, and D). The correlation network for the high diversity community type II was more sparse, suggesting that changes in the relative abundance of specific taxa occur independently. 70

84 A Community Type I Relative read abundance B Community Type II Relative read abundance Lactobacillus Streptococcus Prevotella otu2159 (Ruminococ.) Clostridium (clade 1) otu2087 (Lachnospir.) Treponema Oscillospira Bulleidia Erysipelotrichaceae fam. Pedobacter Enterococcus Comamonas Coprococcus Ruminococcus (clade 1) Blautia Flavobacterium otu1998 Weissella Sarcina C unclassified Lachnospiraceae 0.54 Oscillospira Relative read abundance Streptococcus Clostridium (clade 2) Community Type III otu Ruminococcus (clade 2) 0.54 Blautia otu2159 (Ruminococ.) Prevotella Bulleidia Oscillospira otu2087 (Lachnospir.) Lactobacillus Erysipelotrichaceae family Ruminococcus (clade 1) Clostridium (clade 1) Coprococcus Dialister Blautia otu4381 Treponema Aerococcus Butyrivibrio Streptococcus Lachnospiraceae family Weissella Sarcina D Parabacteroides 0.35 Prevotella Faecalibacterium Community Type IV Relative read abundance Lactobacillus Clostridium (clade 1) Streptococcus Sarcina Enterococcus Prevotella Treponema otu2159 (Ruminococ.) Flavobacterium otu2087 (Lachnospir.) Leuconostoc Ruminococcus (clade 1) otu1998 Bulleidia Coprococcus Weissella Oscillospira otu2066 Pseudomonas Blautia Flavobacterium Streptococcus Lactobacillus Leuconostoc Pseudomonas Clostridium Sarcina 0.41 (clade 1) 0.31 Weissella Enterococcus Lactobacillus Clostridium (clade 1) Treponema Prevotella Leuconostoc Streptococcus Yersinia Sarcina otu2159 (Ruminococ.) Comamonas Ruminococcus (clade 1) p-75-a5 Erysipelothrix Oscillospira otu2087 (Lachnospir.) Lactococcus otu1998 Pseudomonas Dysgonomonas Ruminococcus (clade 1) Adlercreutzia Enterococcus Oscillospira 0.63 Paludibacter Treponema Figure 4.7. (A to D) Rank abundance plots (left) and correlation networks (right) for community types I, II, III, and IV. For rank abundance plots, the most abundant genera are shown on the y-axis and its relative read abundance is shown on the x-axis, calculated within each community group. Correlation networks were calculated using a Spearman rank correlation coefficient (P < 0.001) for the most abundant (pivotal) genera within each community type. Sub-networks (modules) for directly connective genera were identified for each pivotal genera. For each module, the orange diamond vertex indicates the main contributor and yellow ellipse vertices co-occurring genera. Blue and red lines indicate positive and negative correlations, respectively. Host genetic influence on the gastrointestinal microbiota Several reports have revealed that cynomolgus macaques obtained from different geographic origin represent different genetic backgrounds, including differences within the MHC (140, 164, 165). In particular, cynomolgus macaques originating from the geographically isolated island of Mauritius are reported to exhibit a 71

85 restricted MHC allele repertoire and higher frequency of the same haplotypes in comparison to the alleles found in cynomolgus macaques from other geographic regions (Indonesia, Vietnam, and China) ( ). Many previous studies have also reported that differences in cynomolgus macaque MHC haplotypes are associated with differential susceptibility to and resolution of infectious diseases such as malaria and Simian Immunodeficiency Virus (SIV) (168, 169). We conducted analysis of the host genotype to investigate the influence of the host genotype on microbiota composition. To confirm the geographic origin of macaques in these studies, genotype analysis on 24 non-mhc, short tandem repeats (STR) from archived peripheral blood lymphocytes from cynomolgus macaques in studies 1, 2, and 3 was performed by the Molecular Anthropology Laboratory (Davis, CA) and Primate Products, Inc., Miami, FL (Materials and Methods). Results were compared to reference STR data obtained from known Sumatran (Indonesian), Mauritian, Philippine and Vietnamese macaques (129). The data revealed that the macaques in these studies originated from Indochina, Philippines, Indonesia or Mauritius (Table 4.2). 72

86 Macaque Study Prob. Mauritian Origin Prob. Philippine Origin Prob. Indochinese Origin Prob. Indonesian Origin Most likely assignmen t Philippines Admixed (Philippines / Indonesia) Philippines Philippines Admixed (Indochina / Indonesia) Philippines Philippines Indochina Indochina Admixed (Philippines / Indonesia) Admixed (Philippines / Indonesia) Indochina Philippines Philippines Philippines Mauritius Mauritius Mauritius Mauritius Mauritius Mauritius Mauritius Mauritius Mauritius Mauritius Mauritius Mauritius Table 4.2. Probability of geographic origin for cynomolgus macaques. Probabilities based on STR data genotypes analyzed in conjunction with known Sumatran (Indonesian), Mauritian, Philippine and Vietnamese (Indochinese) cynomolgus macaques (n=27). Multivariate analysis of the STR data using PCoA indicated separation of Mauritius macaques (study 2) from macaques of other geographic origin (Fig. 4.8). No clustering was identified among the other populations. Genotype was not 73

87 analyzed for study 4 macaques (n=6); however, according to the vendor (Harlan Laboratories, Inc.; Indianapolis, IN), these macaques originated from Indochina. PCoA Philippines Mauritius Indochina Admixed PCoA1 Figure 4.8. Analysis of genetic variability of cynomolgus macaques in studies 1-3. PCoA based on Lynch pairwise distance calculated from 24 STR data derived from archived peripheral blood lymphocytes from each cynomolgus macaque in studies 1, 2, and 3. Each individual dot represents an individual macaque, colorcoded by geographic origin (boxed inset). We next assessed the genetic differences within the MHC region among the cynomolgus macaques in all four studies (studies 1-3). We genotyped seven previously reported MHC microsatellites spanning both class I and class II regions: D6S1691, MOGc, MICA, DRACA, DQcar, D6S2741, and D6S291 (140, 170) (Fig. 4.9) (see methods). D6S1691 MOGc MICA DRACA D6S2741 D6S291 0 Mamu-A Mamu-B 5000kb DQcar telomere class I class II centromere 74

88 Figure 4.9. Localization of seven microsatellite markers tested within the MHC region (adapted from rhesus macaque MHC map) (140). The allele frequencies for each geographically distinct population (Philippines, Mauritius, Indochina or Admixed) within each loci indicated a broad range of MHC allele types within each population (Fig. 4.10). A D6S B D6S C D6S D DGcar E DRACA F MICA G MOGc Pop1 - Philippines Pop2 - Mauritius Pop3 - Indochina Pop4 - Admixed Figure (A to G) Allele frequencies for different geographic populations for seven MHC loci. Allele identity and its frequency per geographic population are indicated on the y-axis, and the nucleotide length of microsatellite alleles on the x-axis. Frequencies are color-coded by geographic population (boxed inset). We next conducted a multivariate analysis for all MHC microsatellite alleles for each of the individual cynomolgus macaques. A phylogenetic tree calculated using the Lynch distance from the MHC microsatellite data indicated distinct clustering of all macaques originating from Mauritius (Fig. 4.11), 75

89 confirming that divergent MHC allele repertoires are represented in geographically distinct macaques. Monkey 13 Monkey 15 Monkey 02 Monkey 01 Philippines Mauritius Indochina Monkey 14 Admixed Monkey 07 Monkey C6 Monkey 12 Monkey 08 Monkey 05 Monkey C3 Monkey 06 Monkey 04 Monkey C4 Monkey C1 Monkey 09 Monkey C5 Monkey 10 Monkey C2 Monkey 26 Monkey 31 Monkey 30 Monkey 28 Monkey 29 Monkey 24 Monkey 22 Monkey 20 Monkey 27 Monkey 23 Monkey 25 Monkey 21 Monkey 03 Monkey Figure Phylogenetic tree calculated from indicated MHC microsatellite data for all macaques (studies 1-4), using the Lynch pairwise distance measure. Individual monkeys are color-coded as indicated in the boxed inset. Using ANOVA, we identified significant correlations to at least one community type within four of the seven loci (D6S1691, D6S2741, DQCar, and DRACA), suggesting that allelic differences in these loci may contribute to community type structure (p<0.01, Table 4.3). 76

90 Microsatellite Region Community Type p-value Expected false positives (at p-value threshold) False Discovery Rate (FDR) D6S1691 I * D6S1691 II * D6S1691 III * D6S1691 IV * D6S2741 I D6S2741 II * D6S2741 III * D6S2741 IV * D6S291 I D6S291 II D6S291 III D6S291 IV DQcar I DQcar II * DQcar III * DQcar IV * DRACA I DRACA II * DRACA III * DRACA IV * MICA I MICA II MICA III * MICA IV MOGc I MOGc II MOGc III MOGc IV Table 4.3. Correlation of microsatellite regions to community type persistence using ANOVA. For each macaque, microsatellite alleles and community type relative abundance (measured as percent time spent in each community type) were analyzed using ANOVA. Microsatellite alleles that occurred less than three times were excluded, and homozygous alleles were counted as one. Community type relative abundance was normalized using arcsin(sqrt()) transformation. Significant correlations between microsatellite regions and community types (pvalue < 0.01) are indicated by the symbol (*). 77

91 Each cynomolgus macaque study displayed a unique distribution of genotypes for both non-mhc and MHC loci. Study 1 macaques were of Indochinese, Philippine and/or Indonesian origin (n=12), all study 2 macaques were of Mauritian origin (n=12), and all study 3 macaques were of Philippine origin (n=3). Furthermore, macaques of Mauritian origin (study 2) exhibited a unique MHC allele repertoire separate from that of macaques from Indochinese, Philippine and/or Indonesian origin. Combining the data regarding composition of the gastrointestinal microbiota with the genotype analysis, we observed that cynomolgus macaques from different geographic origins harbored different communities of gastrointestinal microbiota. Most notably, the gastrointestinal microbiota in the majority of macaques from Mauritius (study 2) represent a high diversity community type II which is characterized by an abundance of taxa from both the Firmicutes and the Bacteroidetes, and a significantly lower level of Lactobacillus as compared to the other community types (Fig. 4.4, 4.5, 4.6, and Table 4.1). The gastrointestinal microbiota in macaques from Indochina, Indonesia, and the Philippines were less diverse, and contained very low levels of any genera from the phylum Bacteroidetes. These data suggest that differences in the intestinal microbiota in cynomolgus macaques may correlate with differences in host genotype, including those of MHC alleles. 78

92 Impact of live-attenuated and wild type S. dysenteriae 1 strains on microbiota composition Longitudinal fecal samples collected from each macaque in their respective studies were analyzed with regard to the stability of the microbiota over time and to evaluate the impact of immunization and wild type challenge on community composition (Fig. 4.1). In macaques from studies 1 and 2, we assessed the efficacy of two live-attenuated S. dysenteriae 1 vaccine candidates, CVD1255 and CVD1256 (128), to protect against subsequent challenge with wild type S. dysenteriae 1 strain 1617, using macaques that received phosphate buffered saline (PBS) as controls. Macaques in study 3 were only challenged with wild type S. dysenteriae 1 strain All groups were monitored on a daily basis for development of clinical signs of shigellosis and serum was collected to measure IgA and IgG antibody titers against S. dysenteriae 1 LPS. Macaques in study 4 received no treatment throughout the course of sample collection, and served as a control group to assess microbiota composition and stability compared to the treated study groups 1, 2, and 3. To assess the effect of immunization and subsequent wild-type challenge on the composition of the gastrointestinal microbiota over time, we compared measures of diversity (using the Shannon diversity index), measures of similarity (using the Jaccard distance to calculate similarity between samples collected on consecutive days), and community type variation (based on the Jensen-Shannon distance) over time in each macaque study group. 79

93 As expected, the gastrointestinal microbiota in samples from the control group (study 4, no intervention) was stable over a period of 14 days as indicated by a consistent Shannon diversity index and Jaccard similarity coefficient over time (Fig. 4.12D). No significant changes in either measure were observed over time within these macaques (Wilcoxon test). A Shannon diversity index = Treatment (live-attenuated S. dysenteriae 1 or PBS) = Challenge (wild-type S. dysenteriae 1) Jaccard similarity coefficient B Time (days) C D Jaccard similarity coefficient 1 Shannon diversity index Time (days) Time (days) Time (days) Figure Estimates of diversity and similarity over time. Boxplots of the median Shannon diversity index (top panel, blue) and Jaccard similarity index (bottom panel, red) over time for samples from all macaques in (A) study 1 (n=12), (B) study 2 (n=12), (C) study 3 (n=3), and (D) study 4 (n=6). Measures are indicated on the y-axis, and time (in days) on the x-axis. Error bars within boxplots indicate the interquartile range between the first and third quartiles. Yellow arrows indicate treatment with either live-attenuated vaccine strains or PBS, and red arrows indicate challenge with wild type S. dysentariae 1 as indicated in the boxed inset. Stability was further evidenced by the persistence of a given community type (Fig. 4.13D). In the absence of intervention, relatively few community type 80

94 changes occurred within an individual macaque. With the exception of macaque C5, both the high diversity community type II and the moderately diverse community type I were found. A B M 2 In/P 27 M 28 M 3 P 29 M 4 P 30 M 31 M 20 M 21 M 22 M CVD In/Ic 6 P 7 P 8 Ic 9 Ic 23 M 10 In/P 24 M 11 In/P 25 M 12 Ic Layers: D P 14 P 15 P untreated PBS 4 P C challenge 2 1 PBS CVD1255 CVD C1 C2 C3 C4 C5 C Ic Ic Ic Ic Ic Ic Community Type Clinical Stool Score IgA anti-lps IgG anti-lps Community Type: Treatment: live-attenuated Shigella I II vaccination or PBS III wild-type Shigella IV No Sample challenge IgA/IgG Fold Increase: Geographic Origin: <4 Ic = Indochina 0 Normal Loose In = Indochina Mild Diarrhea P = Philippines Mild Diarrhea, w/ blood M = Mauritius >25 4 Strong Diarrhea, w/ blood No Sample Clinical Score: Figure Community type composition, clinical disease symptoms, and elicited immune response over time for (A) study 1, (B) study 2, (C), study 3, and (D) study 4. Individual monkey identification and treatment administered (CVD1255, CVD1256 or PBS) are indicated on the y-axis (left), with geographic origin of each macaque on the right. Sampling day is indicated on the x-axis, with color-coded arrows (immunization or challenge) indicating time of treatment (boxed inset). Community type (A-D) is indicated on the first top bar, stool severity score ranging from 1 (mild) to 4 (severe) (A to C) is indicated on the second bar, and anti-lps IgA and anti-lps IgG fold differences over antibody levels before immunization and challenge (day 0) (A-C) are indicated on the third 81

95 and fourth (bottom) bars, respectively, each color-coded according to the boxed inset, with day 0 antibody titers shown the same as <4 fold increases. Similar measures of community stability over time were observed in samples from study 2 compared to the control group, despite the fact that these animals received four immunizations with CVD1256, or mock immunizations with PBS over a seven day period (Fig. 4.12B). Using the Wilcoxon nonparametric test, no significant differences in the Shannon diversity index were observed for pre-treatment, post-vaccination or post-challenge samples when compared to each other (Fig. 4.12B) or to study 4 samples (Fig. 4.14A). A Shannon diversity index B Shannon diversity index pretreatment (d0) pretreatment (d0) posttreatments (d2-28) p<0.001 post-1st treatment (d1-28) postchallenge (d28-final) Time periods p<0.001 p< post-2nd treatment (d29-56) Time periods study 4 (all days) postchallenge (d57-84) study 4 (all days) Figure Changes in the Shannon diversity index following immunization or PBS administration and wild type challenge in all macaques from studies 1 and 2 compared to control study 4 macaques. (A) Study 2 (n=12): Boxplots of median 82

96 Shannon diversity of pre-treatment time period samples (day 0), post-treatment samples (days 2-28, CVD1256 or PBS), and post-challenge samples (days 30-35, wild type S. dysenteriae 1) from all macaques in study 2 (n=12) compared to samples from control study 4 macaques (n=6). (B) Study 1: Boxplots of median Shannon diversity index of pre-treatment time period samples (day 0), post-1 st - treatment samples (days 1-28, CVD1255, CVD1256, or PBS), post-2 nd -treatment samples (days 29-56, 2 nd dose of same treatment), post-challenge samples (days 57-84) for all macaques (n=12) in study 1 compared to samples from control study 4 macaques (n=6). The nonparametric wilcoxon test was used for all statistics, and error bars indicate the interquartile range between the first and third quartiles. No significant changes in the Jaccard similarity coefficient were observed in samples from study 2 over time, although it was higher overall compared to the Jaccard distance from study 4 samples, indicating a compositional difference compared to the control group (Fig. 4.12B, D). This observation supports other data in this study that suggest a unique, high-diversity composition in macaques originating from Mauritius compared to macaques from other geographic origins. More than 90% of the macaques in study 2, both naïve and immunized, harbored the high diversity gastrointestinal community type II at one or more times point (Fig. 4.13B) compared to 50% of the macaques in study 4. Little or no change in community type or the Shannon diversity index was observed following administration of PBS or CVD1256 and subsequent wild type challenge in samples from study 2 (Fig. 4.13B, 4.12B). Greater variation was observed across time in fecal samples obtained from macaques in study 1 (two doses of live-attenuated Shigella vaccines followed by wild type challenge). The Shannon diversity index for study 1 samples collected from both immunized and PBS control macaques changed following treatment (Fig. 4.12A), and the Wilcoxon test indicated that the diversity was significantly 83

97 reduced following these two treatments compared to pre-treatment time points (day 0) or samples from the untreated study 4 macaques (Fig. 4.14B). Changes in the Jaccard similarity coefficient calculated for these samples were also observed over time, indicating compositional changes in these samples (Fig. 4.12A). Changes in the community type are also observed in study 1 samples following both immunizations. The gastrointestinal microbiota of the macaques in study 1 was initially represented by community type I, which is dominated by both Streptococcus and Lactobacillus (Fig. 4.13A). Following the first treatment, either CVD1255, CVD1256 or PBS administration, there was a shift in the structure of the gastrointestinal microbiota in 58% of the macaques to the lower diversity community type III, dominated by Lactobacillus. Following the second immunization (day 28), the gastrointestinal microbiota in all macaques shifted to community type III. Of importance is the finding that changes in community composition were observed in all macaques, even those receiving PBS. This suggests that the immunization protocol (i.e., anesthesia and handling), rather than immunization with attenuated strains of Shigella, was responsible for the observed changes. Subsequent challenge with wild type S. dysenteriae 1 on day 56 in study 1 resulted in a shift to community type IV, which is dominated by Enterococcus, at one or more time points, in 91% of the macaques (Fig. 4.13A). We also observed an increase in the proportion of previously identified but less abundant genera, including Pedobacter, Flavobacterium, Comamonas, and Yersinia, in fecal samples collected within two days post-challenge (Fig. 4.15). In post-challenge 84

98 samples from 40% of the macaques, the relative abundance of these less prevalent genera was as high as 80%. We also observed an increase in the Shannon diversity index in post-challenge samples (Fig. 4.12A, 4.14B), which likely reflects the increase in the relative abundance of these rare genera. Relative abundance Enterococcus Clostridium Comamonas Pedobacter Flavobacterium Yersinia Time (days) Figure Increase in relative abundance of normally rare genera over time in study 1 macaques (n=12). Relative read abundance is on the y-axis and time (days) on the x-axis. Each point represents an individual sample and is colorcoded as indicated in the boxed inset. Throughout studies 1-3, stool consistency was monitored for clinical shigellosis. A stool score ranging from 1 (loose stool) to 4 (severe bloody diarrhea) was recorded using previously established measures (Materials and Methods) (122). No macaques experienced severe disease symptoms following vaccination (< 2 days post-vaccination), but 83% of the macaques in study 1, both naïve and immunized, exhibited clinical symptoms of S. dysenteriae 1 infection (stool score of grades 2 or higher at one or more time points) following wild type challenge (Fig. 4.13A). Using the non-parametric Wilcoxon test between the severity of the clinical score and the abundance of specific genera, increases in 85

99 Yersinia, Comamonas, Pedobacter, and Flavobacterium genera were found to correlate with a stool score of 2 or more (compared to abundance when score was 0) (Fig. 4.16). A Pedobacter clinical scores 3 and 4: p< Clinical score: 0 Normal 1 Loose 2 Mild Diarrhea 3 Mild Diarrhea, w/ blood 4 Strong Diarrhea, w/ blood B Yersinia clinical score 2: p< C Flavobacterium D Comamonas clinical scores 2 and 4: p< clinical score 2: p< Time (days) Figure Relative read abundance of less abundant organisms and clinical symptoms of Shigella infection in study 1 macaques (n=12). The relative read abundance for Pedobacter, Yersinia, Flavobacterium, and Comamonas (y-axis) over time in days (x-axis), color-coded by clinical score of stool symptom severity. Description of severity is listed in boxed inset. Significant correlations were calculated using a one-way nonparametric Wilcoxon test, comparing the indicated genus abundance in stool specimens with a clinical stool score of 2 compared to a null clinical score of 0. 86

100 Interestingly, no macaques in study 2 exhibited clinical disease symptoms associated with wild type challenge, and there was no detectable increase in the relative abundance of rare genera in any of these samples (Fig. 4.13B and table). In study 3, unvaccinated macaques were challenged on day 0 with wild type S. dysenteriae 1. As observed in study 1, the composition of the gastrointestinal microbiota was altered following exposure to wild type Shigella, and the Enterococcus-dominated community type IV emerged (Fig. 4.13C). All challenged macaques (studies 1-3) mounted immune responses against Shigella antigens following administration of live-attenuated and wild-type S. dysenteriae 1 strains, as measured by IgA and IgG antibodies against S. dysenteriae 1 LPS (Fig. 4.13, A-C). Antibody responses were not measured in study 4 macaques since they were neither immunized nor challenged. Most immunized macaques in studies 1 and 2 (macaques 1-8, 26-31) mounted a Shigella-specific immune response following immunization. Compared to immunized macaques, macaques receiving PBS prior to challenge in both studies (macaques 9-12, 20-25) elicited a less robust immune response post-challenge, which is to be expected if challenge after immunization is viewed as an immune boost. These results demonstrate that differences in microbiota composition did not impair the ability of the host to mount IgA and IgG anti-lps antibody responses following intragastric vaccination or challenge. However, it markedly affected the susceptibility to disease following wild type challenge. 87

101 Summary This aim further characterized the intestinal microbiota in cynomolgus macaques. Compared to the gastrointestinal microbial consortia in humans, the gastrointestinal microbiota in cynomolgus macaques is dominated by Firmicutes, particularly the genus Lactobacillus (21). We observed distinct community types in cynomolgus macaques, suggesting that enterotypes may also exist in nonhuman primates, although confirmation of this will require examination of a larger number of animals. Two community types (I and II) were consistently found in healthy macaques, and two community types (III and IV) were associated with inteventions and/or clinical shigellosis. In the absence of immunization or challenge, the microbial community remains stable over time, as observed in the untreated control group of macaques (study 4). Differential responses to S. dysenteriae 1 immunization and subsequent wild-type challenge were observed for macaques of Mauritian origin compared to those of Indonesian/Indochinese/Philippine origin. Mauritian macaques harbored a high diversity community type II, which was more stable, compared to other community types following immunization and challenge. Additionally, these macaques did not develop any clinical shigellosis signs following challenge with wild-type S. dysenteriae 1 (study 2). These results contrast with the significant changes in diversity, community type, and clinical outcome observed in macaques from Indochina/Indonesia/Philippines. Taken together, these findings suggest that a more diverse intestinal microbiota may play a protective role against enteric pathogens. This idea is consistent with ecological stability-diversity theory, which 88

102 suggests that higher diversity within an ecosystem confers stability, or resistance to environmental perturbations (82, 83). Although differences in genetics could also impact immune responses and disease susceptibility, the strength of the immune effectors measured here did not differ significantly between macaques of different geographic origin. It is also notable that within the clinically susceptible macaques (study 1), changes in the diversity of the intestinal microbiota were seen in macaques receiving both PBS and live-attenuated bacteria, suggesting that the protocol for immunization, which involved handling, anesthesia, and endoscopy (and associated stress), was responsible for the observed changes. Both human and animal studies have shown that stress can reduce gastrointestinal microbial diversity ( ) and affect immune function (62, 174), possibly leading to increased susceptibility to pathogen infection. In this study, Mauritian cynomolgus macaques were more resistant to S. dysenteriae 1 infection than other macaques regardless of whether the animals were immunized and mounted an immune response. Our genotype analysis of seven microsatellites spanning the MHC region in these macaques supports previous observations that MHC alleles within cynomolgus macaques vary depending on their geographic origin. Previous studies suggest that due to the close interaction of the immune system with the gastrointestinal environment, the highly polymorphic MHC alleles may contribute to the observed variability of the microbiota within the host (80). Specific MHC alleles, frequently observed in cynomolgus macaque populations originating from Mauritius, have been reported to increase susceptibility to SIV infection (166, 169). However, the factors that 89

103 shape MHC variability, such as pathogen-mediated or environmental selection, remain unclear. While the impact of specific MHC alleles on the intestinal microbiota has not been reported, our data suggests a possible correlation between gastrointestinal microbial communities and MHC genetic variability. Together, these data highlight the importance of considering host genotype, environmental factors, and the resident gastrointestinal microbiota in the context of vaccine development. 90

104 Chapter 5 Aim 3: To determine correlations between the gut microbiota and immune responses following oral immunization with live-attenuated S. dysenteriae 1 strains and challenge with wild-type S. dysenteriae 1 in cynomolgus macaques Introduction As discussed in Chapter 4, vaccine development against Shigella has been hampered by the appearance of new Shigella serotypes, the lack of clinically relevant animal models, and variations in the observed immunogenicity across different world populations (109). In particular, lower efficacy rates of a liveattenuated S. flexneri 2a vaccine were observed in Bangladesh (124) compared to the efficacies initially reported in studies conducted in the US (117). Similar discrepancies have been noted in other orally administered live-attenuated vaccine studies against other intestinal pathogens, in particular for rotavirus (175), polio (176), and cholera (125, 126). These differences are often attributed to socioeconomic, nutritional, and environmental differences, but also suggest a possible role for resident microbiota in modulating immune responses (107, 177). Another difficulty in the development of a Shigella vaccine is the lack of specific correlates to immune responses that best predict protection against Shigella. Prior exposure to wild-type Shigella alone has been reported to mediate protection against secondary infection in both humans and NHP studies (109). For humans, convincing evidence for protection from Shigella infection comes from a challenge study conducted by Coster et al., where anti-lps IgA antibodysecreting cells (ASCs) were observed to best correlate with protection (117). This 91

105 supports observations in other studies of increased levels of gut derived IgA anti- O antigen ASCs following wild-type exposure ( ). In another cohort study following young children in an endemic area where S. sonnei, S. flexneri 2a and S. flexneri 6 were mainly responsible for endemic shigellosis, a protection rate of 72% against secondary infection with a homologous serotype was observed (181). A low rate of protection was observed against other serotypes, suggesting that antibodies to the O-antigen of Shigella are important. Although the studies here support a protective role for serum IgG or IgA anti-o antigen responses, other studies report no apparent correlation between these parameters and the development of clinical symptoms following infection (182), further challenging the identification of the key immune responses indicative of protection in humans. Similar results have been observed in NHP models. In rhesus macaques, it was observed that S. flexneri 2a infection was capable of inducing both local and systemic anti-shigella immune responses, a significant step in the development of a Shigella vaccine model (183). This infection did not mediate protection against S. sonnei, a non-homologous serotype. Another NHP challenge study using rhesus macaques conducted by Formal et al. demonstrated that an S. flexneri 2a infection resulting in clinical symptomology conferred 100% protection against subsequent re-challenge with S. flexneri 2a, suggesting that wild-type exposure is sufficient to confer homologous protection (121). A previously established cynomolgus macaque model has reported increased anti-lps IgA and IgG antibodies and ASCs following wild-type challenge, similar to humans (122, 184). In addition, studies reported an increase in antibodies against the Shigella-specific proteins 92

106 IpaB, IpaC, IpaD, VirG and MxiH, which are involved in Shigella pathogenesis (184). However, increases in these responses were not observed in all studies and did not consistently correlate with protection. In this aim, we investigated the relationship between the microbiota and immune responses following the administration of live-attenuated or wild-type S. dysenteriae 1 in cynomolgus macaques to assess whether specific genera or community profiles correlated with the measured immune response. We measured Shigella-specific immune responses throughout the course of immunization and challenge in studies 1-3. Using statistical correlation measures, we then evaluated whether distinct microbial genera or community types correlated with different effector responses. Results Induction of effector immune responses following oral administration of liveattenuated and wild-type S. dysenteriae 1 To evaluate the induction and efficacy of immune responses following exposure to either live-attenuated or wild-type S. dysenteriae 1 strains in the cynomolgus macaque vaccine model, Shigella-specific immune responses were measured from blood samples collected in studies 1-3. Macaques in studies 1 and 2 were immunized with live-attenuated S. dysenteriae 1 strains, or PBS as a control, and subsequently challenged with wild-type S. dysenteriae 1 (Fig. 3.1). Macaques in study 3 were only challenged with wild-type S. dysenteriae 1. Cross-study comparison is difficult to interpret because of the different immunization 93

107 protocols. However, within-study comparisons of effector immune responses following immunization and challenge were analyzed, and subsequent correlations to the microbial communities were conducted. Macaques in study 3 that were challenged with wild-type S. dysenteriae 1 mounted immune responses by day 14 as measured by both serum IgA and IgG antibodies against S. dysenteriae 1 LPS and the S. dysenteriae 1 proteins IpaB, IpaC, IpaD, VirG, and MxiH, confirming a robust immune response to wild-type S. dysenteriae. All animals mounted strong immune responses following challenge (4-fold increase or more), and two of three macaques mounted a 25-fold increase or more against all antigens except MxiH, in accordance with previous results observed in cynomolgus macaques that mounted strong anti-lps and anti- IpaB effector responses following wild-type S. dysenteriae 1 exposure (184). Both macaques 14 and 15 exhibited severe clinical symptoms following challenge (Fig. 4.13), consistent with previous studies of S. dysenteriae 1 infection in cynomolgus macaques (122). Figure 5.1. Immune responses following wild-type S. dysenteriae 1 challenge in study 3 macaques. Both IgA (left panel) and IgG (right panel) antibody fold differences over antibody levels before challenge (day 0) are reported, colorcoded according to the boxed inset. Challenge events are indicated by the yellow arrow on the horizontal axis. 94

108 Macaques in study 1 mounted immune responses following immunization and challenge with S. dysenteriae 1 strains, as measured by serum IgA and IgG antibodies against S. dysenteriae 1 LPS and IgA antibodies against IpaB (Fig. 5.1). Higher fold increases of IgA responses compared to IgG responses were observed for most macaques, particularly anti-ipab IgA, consistent with prior cynomolgus macaque model data in which IgA responses were observed to be most dominant (184). Figure 5.2. Immune responses following immunization and wild-type challenge with S. dysentariae 1 strains challenge in study 1 macaques. Both IgA (left panel) and IgG (right panel) antibody fold differences over antibody levels before challenge (day 0) are reported, color-coded according to the boxed inset. Days and challenge events are indicated on the horizontal axis. Robust increases in anti-lps IgA and IgG antibodies following immunization (over 4-fold) were observed only in macaques immunized with CVD1255 and not CVD1256, whereas anti-ipab IgA antibodies were observed in macaques immunized with both vaccines, suggesting a difference in immunogenicity between the two vaccine strains. Overall, CVD1255 induced higher fold increases 95

109 for each antibody measured compared to CVD1256 (Fig. 5.3). However, the small number of animals prevents definite conclusions about differences in the immunogenicity elicited by these two vaccine two strains. anti-lps IgA anti-lps IgG anti-ipab IgA Fold increase *** ** *** * Vaccine Group Non-immunized CVD1256 CVD1255 pre pre postvacc postchal postvacc postchal pre postvacc postchal Time period Figure 5.3. Mean fold increases following immunization and wild-type challenge with S. dysentariae 1 strains in study 1 macaques. Measured serum antibody is labeled for each graph (top), with mean fold increase on y-axis for timepoints collected pre-treatment (pre, day 0), post-immunization (post-vacc, days 1-56), and post-challenge (post-chal, days 63-98) on x-axis. Vaccine group (nonimmunized, CVD1255-immunized, and CVD1256-immunized) is colorcoded as indicated by the inset box. Statistical significance was calculated using a student t-test (* = p-value < 0.05, ** = p-value < 0.005, *** = p-value < ). All macaques mounted immune responses following challenge with wild-type S. dysenteriae 1, although immunized macaques mounted higher IgA responses in comparison to naïve macaques, particularly as measured by anti-ipab IgA, suggesting that challenge resulted in boosting of the immune responses following immunization (Fig. 5.3), as expected following mucosal administration of an oral vaccine. 96

110 Similar to study 1, macaques in study 2, which followed a different immunization protocol, mounted immune responses following both immunization and challenge, measured by serum IgA and IgG antibodies against S. dysenteriae 1 LPS and IpaB, IpaC, IpaD, VirG, and MxiH (Fig. 5.4). IgG antibody responses following both immunization and challenge were robust, with the exception of detection of anti-mxih antibodies. Lower fold increases were observed in IgA levels following immunization and challenge, and the levels of both anti-virg and anti-mxih antibodies were low following exposure to S. dysenteriae 1 strains (Fig. 5.5). This contrasts to the results for study 1 and 3 macaques where, in general, IgA responses were more robust compared to IgG responses. Following challenge with wild-type S. dysenteriae 1, naïve macaques in study 2 exhibited lower antibody responses as measured by anti-lps IgA and IgG antibodies compared to immunized macaques, similar to the results observed in study 1 (Fig. 5.3, 5.5). However, measurements for other antibodies in study 2 were comparable or higher in naïve macaques compared to immunized macaques. It is reasonable to hypothesize that immunization decreased the ability of wild-type challenge to colonize, thereby resulting in decreased host s immunity compared to that elicited in PBS-treated controls. Additionally, macaques in study 2 exhibited higher fold-differences in antibody titers following both immunization and challenge compared to macaques in study 1, suggesting that the immunization protocol used in study 2 was more effective in inducing effector immune responses. 97

111 Figure 5.4. Immune responses following immunization and wild-type challenge with S. dysenteriae 1 strains in study 2 macaques. Both IgA (left panel) and IgG (right panel) antibody fold differences over antibody levels before challenge (day 0) are reported, color-coded according to the boxed inset. Days and challenge events are indicated on the horizontal axis. 98

112 Fold increase anti-lps IgA anti-ipab IgA *** anti-ipac IgA *** anti-ipad IgA anti-virg IgA 60 * * anti-mxih IgA pre postvacc postchal pre postvacc postchal pre postvacc postchal pre postvacc postchal pre postvacc postchal pre postvacc postchal 200 anti-lps IgG anti-ipab IgG anti-ipac IgG anti-ipad IgG anti-virg IgG anti-mxih IgG * *** Fold increase *** *** * pre postvacc postchal pre postvacc postchal pre postvacc postchal pre postvacc postchal pre postvacc postchal pre postvacc postchal Vaccine Group Time period Non-immunized CVD1256 Figure 5.5. Mean fold increases following immunization and wild-type challenge with S. dysentariae 1 strains challenge in study 2 macaques. Measured serum antibody is labeled for each graph (top), with mean fold increase on y-axis for time points collected pre-treatment (pre, day 0), post-immunization (post-vacc, days 2-28), and post-challenge (post-chal, days 36-final day) on x-axis. Vaccine group (non-immunized and CVD1256-immunized) is colorcoded as indicated by the inset box. Statistical significance was calculated using a student t-test (* = p- value < 0.05, ** = p-value < 0.005, *** = p-value < ). In addition to IgA and IgG antibody levels, ASCs producing both IgA and IgG antibodies against S. dysenteriae 1 LPS and IpaB, IpaC, IpaD, VirG and MxiH were measured in study 2 (Fig. 5.6 and 5.7). Large increases in IgAproducing ASC responses were observed for macaques following immunization with CVD1256, and increases in IgA-producing ASCs were 99

113 observed for all macaques following challenge with wild-type S. dysenteriae 1 (Fig. 5.6). The responses for IgG ASC were more heterogeneous. Large increases in anti-lps and anti-ipab IgG-producing ASCs and lower increases in anti-ipac and anti-ipad IgG-producing ASCs were observed for immunized macaques, and only macaque 26 produced anti-virg and anti-mxih IgG-producing ASCs (Fig.5.7). Similar to the IgA-producing ASC counts, IgG-producing ASC responses increased following challenge for macaques receiving either PBS or CVD1256. While robust increases in ASCs were observed for some individual monkeys, the overall measured response was inconsistent among the groups. 100

114 Figure 5.6. Anti-LPS IgA (A), anti-ipab IgA (B), anti-ipac IgA (C), anti-ipac IgA (D), anti-virg IgA (E), and anti-mxih IgA (F) antibody-secreting cells (ASC) counts measured for study 2 macaques following immunization and challenge with S. dysenteriae 1 strains. Macaques administered PBS (left panel) 101

115 or CVD1255 (right panel) are each color-coded as indicated in the inset box, and days are indicated on the x-axis. Figure 5.7. Total IgG (A), anti-lps IgG (B), anti-ipab IgG (C), anti-ipac IgG (D), anti-virg IgG (E), and anti-mxih IgG (F) antibody-secreting cells (ASC) counts measured for study 2 macaques following immunization and challenge 102

116 with S. dysenteriae 1 strains. Macaques administered PBS (left panel) or CVD1255 (right panel) are each color-coded as indicated in the inset box, and days are indicated on the x-axis. As observed in the clinical symptomatology following challenge, discussed in Chapter 4, macaques in study 1 mounting effector immune responses as measured by antibody production following immunization were still susceptible to clinical disease following challenge, suggesting no correlation between the measured IgA and IgG induction and disease resistance in this model (Fig. 4.13). A strong immune effector response, as measured by these antibodies, did not confer protection against challenge (Fig. 4.13). In particular, low response levels (less than a 4-fold increase) were observed for macaque 12, which exhibited no disease symptoms. Similarly, high response levels (25-fold increases or higher) were observed for macaques that exhibited clinical symptoms (in particular, macaques 1-5, 10). The observed inconsistency between immune responses and protection may be an effect of the immunization model in which immunization was insufficient to produce strong immune responses. Macaques in study 2, including macaques administered PBS prior to challenge, did not exhibit clinical shigellosis post-challenge (Fig. 4.13). Correlations between immune responses and disease are not possible since unimmunized macaques did not exhibit clinical disease following wild-type infection, and, therefore, efficacy cannot be calculated on a 0% attack rate in controls. However, robust effector immune responses were observed following immunization and challenge in immunized macaques and following challenge in naïve macaques. The fact that unimmunized controls did not exhibit disease 103

117 symptoms following challenge suggests that a factor other than antibody response is responsible for resistance to S. dysenteriae 1 infection. We hypothesize that a diverse microbiota community in these macaques may play a protective role. Correlation between effector immune responses and the microbiota community over time To investigate whether specific components of the microbiota within the two vaccine study populations (study 1 and 2) were significantly correlated with serum antibody responses, or the clinical stool score for study 1 macaques, we applied a statistical model, local similarity alignment (LSA), which explores timedependent correlations. Compared to Pearson correlation analysis, which is typically utilized to identify linear relationships, the LSA ecological time-series model is able to identify complex, time-dependent interactions among microbial members as well as various environmental parameters. From the LSA, we then constructed a network for each study using only significant associations calculated by this method (LSA correlation coefficient > 0.4, p<0.01). Due to the different collection periods, vaccine regimen, and significant differences in community structure of the microbiota observed in the two studies, it is difficult to directly compare the antibody responses and microbiota between studies 1 and 2. However, within-study associations revealed associations between the microbiota and immunological parameters. The network for study 1 was dense, suggesting a complex relationship among genera observed in the microbiota, antibody response, and clinical stool score (Fig. 5.8). The clinical 104

118 stool score for all study 1 macaques, both naïve and immunized, was positively associated with many genera, including the genera Flavobacterium, Pedobacter, Yersinia, and Enterococcus, consistent with our observations of increased abundance of rare genera following challenge. Network analysis in study 1 also revealed increases in other rare genera in some monkeys following challenge, including Psychrobacter, Erysipelothrix, Paludibacter, and Dysgomonas. Figure 5.8. Local similarity analysis between genera and clinical and/or immunological measurements in cynomolgus macaque Shigella vaccine studies. Networks of significant correlations (p<0.01; q > 0.40) using local similarity analysis between genera and the immunological or clinical measurements in macaques in (A) study 1 and (B) study 2. Nodes are colored as described in boxed inset (genera sized for relative abundance). Correlations among non-core genera and genera not directly related to immunological or clinical measurements have been greyed out. Compared to study 1, the resulting network for study 2 was less dense despite a robust antibody response following challenge (Figure 5.8). This is likely a reflection of the community stability observed over time in these macaques despite multiple immunizations and subsequent challenge. In study 2, anti- 105