AN INTEGRATED INVESTIGATION OF RUMINAL MICROBIAL COMMUNITIES DISSERTATION

Transcription

1 AN INTEGRATED INVESTIGATION OF RUMINAL MICROBIAL COMMUNITIES USING 16S rrna GENE-BASED TECHNIQUES DISSERTATION Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University By Min Seok Kim Graduate Program in Animal Sciences The Ohio State University 2011 Dissertation Committee: Dr. Mark Morrison, Advisor Dr. Zhongtang Yu, Co-Advisor Dr. Jeffrey L. Firkins Dr. Michael A. Cotta

2

3 ABSTRACT Ruminant animals obtain most of their nutrients from fermentation products produced by ruminal microbiome consisting of bacteria, archaea, protozoa and fungi. In the ruminal microbiome, bacteria are the most abundant domain and greatly contribute to production of the fermentation products. Some studies showed that ruminal microbial populations between the liquid and adherent fraction are considerably different. Many cultivation-based studies have been conducted to investigate the ruminal microbiome, but culturable species only accounted for a small portion of the ruminal microbiome. Since the 16S rrna gene (rrs) was used as a phylogenetic marker in studies of the ruminal microbiome, the ruminal microbiome that is not culturable has been identified. Most of previous studies were dependent on sequences recovered using DGGE and construction of rrs clone libraries, but these two techniques could recover only small number of rrs sequences. Recently microarray or pyrosequencing analysis have been used to examine microbial communities in various environmental samples and greatly contributed to identifying numerous rrs sequences at the same time. However, few studies have used the microarray or pyrosequencing analysis to investigate the ruminal microbiome. The overall objective of my study was to examine ruminal microbial diversity as affected by dietary modification and to compare microbial diversity between the liquid and adherent fractions using the microarray and pyrosequencing analysis. In the first study (Chapter 3), a meta-analysis of all the rrs sequences of rumen origin deposited in the RDP database was performed. Collectively, 5,271 and 943 OTUs of bacteria and archaea, respectively, were identified at 0.03 phylogenetic distance. The ii

4 predominant bacterial phyla were Firmicutes and Bacteroidetes, while the largest archaeal phylum was Euryarchaeota. More than 50% of all the bacterial sequences could not be classified into any known genus. The bacterial OTUs identified in this study were used to develop a phylogenetic microarray as demonstrated in Chapter 6. In the second study (Chapter 4), select cultured bacteria and uncultured bacteria were quantified using specific real-time PCR assays in order to compare the abundance between the cultured bacteria and uncultured bacteria. The populations of some uncultured bacteria were as abundant as those of major cellulolytic cultured bacteria such as Fibrobacter succinogenes, Ruminococcus albus and Ruminococcus flavefaciens. In the third study (Chapter 5), the diversity of ruminal microbiome in cattle was examined using rrs clone libraries. Six known rrs clones were used to validate the phylogenetic microarray (Chapter 6). The phylogenetic data of the cloned sequences supported the predominance of Firmicutes and Bacteroidetes and the abundance of unclassified groups as described in the meta-analysis (Chapter 3). In the fourth study (Chapter 6), a phylogenetic microarray that detects 1,600 OTUs of ruminal bacteria was developed in a 6 5K format based on the OTUs identified in Chapter 3. The utility of the phylogenetic microarray (referred to as RumenArray) was tested in comparative analysis of fractionated bacterial microbiomes obtained from sheep fed two different diets. Species-level OTUs are commonly defined at 0.03 phylogenetic distance based on full-length rrs sequences. However, the current 454 pyrosequencing method is not able to produce full-length rrs sequences. Because sequence divergence is not distributed evenly along the rrs, pyrosequencing analysis of different rrs regions can lead to overestimated or underestimated species richness. To identify a region or phylogenetic distance that can iii

5 support species richness estimate as reliably as full-length rrs sequences, in the fifth study I compared datasets of partial rrs sequences corresponding to different variable regions with a dataset of nearly full-length rrs sequences (Chapter 7). The results indicated that the V1-V3 and the V1-V4 regions at 0.04 distance provide more accurate estimates than other partial regions. Based on the results obtained in Chapter 7, pyrosequencing analysis was performed to investigate bacterial diversity in the rumen of cattle as affected by supplementation of monensin, or 4% fat from distillers grains, roasted soybeans and an animal vegetable blend in my sixth study (Chapter 8). Supplementary fat resulted in significant shift of bacterial populations when compared to the control diet, but supplementary monensin did not. As shown in the previous Chapters, the pyrosequencing analysis showed that numerous rrs sequences that cannot be assigned to any characterized genus were predominant in the rumen. The overall results of the above studies provided further insights into the ruminal microbiome as affected by different diets and different fractions. Integration of RumenArray and pyrosequencing techniques will improve our understanding of the ruminal microbial microbiome and its integration with nutritional studies. iv

6 ACKNOWLEDGMENTS I first would like to thank my advisors, Drs. Mark Morrison and Zhongtang Yu, for their support and guidance during my graduate study. I would like to express special thanks to Dr. Zhongtang Yu for his support and thoughtful discussions during individual meetings. I would like to thank Drs. Jeff Firkins and Mike Cotta for their service on my dissertation committee. I wish to thank Dr. Kichoon Lee for his service on my candidacy committee. I wish to acknowledge all former and present colleagues for their friendship in the Morrison/Yu lab: Seungha Kang, Jill Stiverson, Mike Nelson, Mike Cressman, Lingling Wang, Shan Wei, Wen Lv, Katie Shaw, Yueh-Fen Li, Amanda Gutek, Amlan Patra, Phongthorn Kongmun, Gunilla Bech-Nielsen, Sally Adams, Yan Zhang, Zhenming, Mohd Saufi Bastami, Jing Chen, Premaraj, Bethany and anyone else I missed. I am especially grateful to Jill Stiverson for her help and technical support when I first joined the lab. I especially thank Mike Nelson for his help in bioinformatics analysis when I first started pyrosequencing analysis. I would like to thank all my fellow graduate students working in the Department of Animal Sciences for their friendship. I would like to thank Sangsu Shin for his friendship since I would like to thank Sunghee Park and Changsoo Lee working in the Department of Food Science & Technology for their friendship. I wish to acknowledge all my family members for their support, love and encouragement. v

7 VITA April Born-Kwangju, Republic of Korea B.A., Department of Animal Sciences, Seoul National University, Republic of Korea 2002 to February M.S., Department of Animal Sciences, Seoul National University, Republic of Korea March 2004 to August Researcher, Department of Agricultural Sciences, Korea National Open University September 2006 to present...graduate Research Associate, Department of Animal Sciences, The Ohio State University Publications Kim, M., Morrison, M., Yu, Z., Phylogenetic diversity of bacterial communities in bovine rumen as affected by diets and microenvironments. Folia Microbiologica, DOI /s Kim, M., Morrison, M., Yu, Z., Status of the phylogenetic diversity census of ruminal microbiomes. FEMS Microbiology Ecology, 76, Kim, M., Morrison, M., Yu, Z., Evaluation of different partial 16S rrna gene sequence regions for phylogenetic analysis of microbiomes. Journal of Microbiological Methods, 84, vi

8 Nam, E.S., Kim, M.S., Lee, H.B., Ahn, J.K., β-glycosidase of Thermus thermophilus KNOUC202: Gene and biochemical properties of the enzyme expressed in Escherichia coli. Applied Biochemistry Microbiology, 46, Kim, M.S., Sung, H.G., Kim, H.J., Lee, S.S., Chang, J.S., Ha, J.K., Study on rumen cellulolytic bacterial attachment and fermentation dependent on initial ph by cpcr. Journal of Animal Science and Technology (in Korean). 47, Major Field: Animal Science Focus: Rumen Microbial Ecology Fields of Study vii

9 LIST OF TABLES Table 3.1 The number of OTUs for total bacteria, total archaea and major groups of bacteria, and their percentage coverage at three phylogenetic distances Table 4.1 Primers and a TaqMan probe used in the real-time PCR assays for total bacteria, total archaea or cultured bacteria Table 4.2 Primers used in the real-time PCR assays for uncultured bacteria Table 7.1 Estimates of species-level OTUs calculated from partial and full-length archaeal 16S rrna gene sequences Table 7.2 Estimates of genus- and family-level OTUs calculated from partial and fulllength archaeal 16S rrna gene sequences Table 7.3 Estimates of species-level OTUs calculated from partial and full-length bacterial 16S rrna gene sequences Table 7.4 Estimates of genus- and family-level OTUs calculated from partial sequence regions and full length of bacterial 16S rrna gene sequences Table 7.5 Estimates of OTUs calculated from partial and full-length archaeal 16S rrna gene sequences at 0.01 distance Table 7.6 Estimates of OTUs calculated from partial and full-length bacterial 16S rrna gene sequences at 0.01 distance viii

10 Table 7.7 Estimates of bacterial species-level OTUs calculated from full-length and short partial bacterial 16S rrna gene sequences Table 8.1 Ingredient composition of dietary treatments Table 8.2 Sequence data and alpha diversity indices for the six fractions Table A List of bacterial primers for pyrosequencing analysis ix

11 LIST OF FIGURES Figure 3.1 Bacterial phyla represented by the 16S rrna gene sequences of rumen origin Figure 3.2 A taxonomic tree showing the genera of ruminal archaea identified by the RDP database sequences Figure 3.3 A taxonomic tree showing the bacteria (grouped into genera) isolated from the rumen Figure 3.4 A taxonomic tree showing the archaea (grouped into genera) isolated from the rumen Figure 3.5 A taxonomic tree showing all the genera of ruminal bacteria identified by the S rrna gene sequences of rumen origin Figure 4.1 Populations of total archaea and total bacteria in the rumen of cattle Figure 4.2 Populations of three major cellulolytic bacteria and Butyrivibrio spp. in the rumen Figure 4.3 Populations of major non-cellulolytic cultured bacteria in the rumen Figure 4.4 Populations of uncultured bacteria originally identified from the rumen of sheep x

12 Figure 5.1 Clustering analysis of DGGE banding profiles based on the V3 region of 16S rrna genes Figure 5.2 A taxonomic tree showing the bacterial genera represented by the 144 sequences Figure 5.3 A Venn diagram showing the numbers of species-level OTUs shared among the four composite samples Figure 5.4 A PCA analysis plot comparing the bacterial communities in the four composite samples Figure 6.1 Linear range of detection of the RumenArray as determined using crna pools of the 6 positive clones Figure 6.2 A venn diagram showing the number of detected OTUs Figure 6.3 A hierarchical tree showing signal intensities and similarity among the fractionated samples Figure 6.4 PCA for comparison among all the fractionated samples Figure 8.1 Sequence distribution at the phylum level for each fraction Figure 8.2 The distribution of sequences and OTUs at genus or the lowest classifiable rank in the phylum Firmicutes Figure 8.3 The distribution of sequences and OTUs at genus or the lowest classifiable rank in the phylum Bacteroidetes Figure 8.4 The distribution of sequences and OTUs at the lowest classifiable rank in minor phyla xi

13 Figure 8.5 Principal coordinates analysis for the six fractions Figure 8.6 Principal coordinates analysis for comparison among the three datasets xii

14 TABLE OF CONTENTS ABSTRACT... ii ACKNOWLEDGMENTS... v VITA... vi LIST OF TABLES... viii LIST OF FIGURES... x CHAPTER 1: INTRODUCTION... 1 CHAPTER 2: REVIEW OF LITERATURE Microbial communities in the rumen Bacterial communities Major cellulolytic bacteria Minor cellulolytic bacteria Uncultured cellulolytic bacteria Archaeal communities Methods to investigate microbial diversity Cultivation-dependent methods Cultivation-independent methods Bioinformatics analysis to investigate microbial diversity The use of monensin and its role in rumen fermentation xiii

15 2.5 Biohydrogenation in the rumen Summary CHAPTER 3: STATUS OF THE PHYLOGENETIC DIVERSITY CENSUS OF RUMINAL MICROBIOMES Abstract Introduction Materials and Methods Sequence data collection and phylogenetic analyses Diversity estimate Results Data summary Firmicutes Bacteroidetes Proteobacteria Minor phyla Archaea Estimates of OTU richness Discussion Phylogenetic diversity xiv

16 3.5.2 Diversity estimates CHAPTER 4: QUANTITATIVE COMPARISONS OF CULTURED AND UNCULTURED MICROBIAL POPULATIONS IN THE RUMEN OF CATTLE FED DIFFERENT DIETS Abstract Introduction Materials and Methods Sample collection, fractionation and DNA extraction Real-time PCR assays Results and Discussion Quantification of populations of total bacteria and total archaea Quantification of cultured bacteria Quantification of uncultured bacteria Conclusions CHAPTER 5: PHYLOGENETIC DIVERSITY OF BACTERIAL COMMUNITIES IN BOVINE RUMEN AS AFFECTED BY DIETS AND MICROENVIRONMENTS Abstract Introduction Materials and Methods Sample collection, fractionation and DNA extraction xv

17 5.3.2 DGGE analysis Construction of rrs clone libraries Restriction fragment length polymorphism (RFLP) analysis, DNA sequencing and phylogenetic analysis Comparison of ruminal bacterial communities among the four composite samples Nucleotide sequence accession numbers Results and Discussion CHAPTER 6: DEVELOPMENT OF A PHYLOGENETIC MICROARRAY FOR COMPREHENSIVE ANALYSIS OF RUMINAL MICROBIOME Abstract Introduction Materials and Methods Oligonucleotide probe design and microarray fabrication Sample collection, fractionation and DNA extraction Sample preparation and labeling Microarray hybridization Signal detection and data analysis Determination of the specificity and detection limit xvi

18 6.3.7 Comparison between microarray and real-time PCR data Results and Discussion Validation of the specificity, sensitivity and detection limit Data summary Diversity of ruminal bacteria assigned to known genus Diversity of ruminal bacteria that are not assigned to any known genus PCA for comparison between fractions Comparison of RumenArray and real-time PCR data Conclusions CHAPTER 7: EVALUATION OF DIFFERENT PARTIAL 16S rrna GENE SEQUENCE REGIONS FOR PHYLOGENETIC ANALYSIS OF MICROBIOMES Abstract Introduction Materials and Methods Sequence collection, alignment, and clipping Diversity estimates UniFrac analysis Analysis of sequence datasets recovered from uncultured bacteria Analysis of short partial sequence regions xvii

19 7.4 Results Analysis of partial archaeal sequences Analysis of partial bacterial sequences UniFrac analysis Analysis of uncultured bacterial sequences Analysis of short partial sequence regions Discussion CHAPTER 8: INVESTIGATION OF RUMINAL BACTERIAL DIVERSITY IN CATTLE FED SUPPLEMENTARY MONENSIN OR FAT USING PYROSEQUENCING ANALYSIS Abstract Introduction Materials and Methods Sample collection Metagenomic DNA extraction Pyrosequencing Sequence processing and bioinformatics analysis Comparison among three datasets Results and Discussion Data summary xviii

20 8.4.2 Firmicutes Bacteroidetes Minor phyla Comparison among the diets Comparison among three datasets Conclusions CHAPTER 9: GENERAL DISCUSSION WORKS CITED APPENDIX A: ADDITIONAL PYROSEQUENCING METHODS xix

21 CHAPTER 1 INTRODUCTION Ruminal fermentation is mediated by a complex microbiome consisting of Bacteria, Archaea, and Eukaryota. Although many studies have been reported that characterize the ruminal microbiome using cultivation-based methods, the isolated species accounted for only a small proportion of the ruminal microbiome (Kim et al., 2011b; Stevenson and Weimer, 2007). Since 16S rrna gene (rrs) sequences were applied to investigation of the diversity of ruminal bacteria and archaea (Stahl et al., 1988), the complex microbial diversity in the rumen has begun to be revealed and appreciated. The microbial diversity of ruminal microbiome has been a research focus of many studies since the late 1980 s. Cloning and sequencing of 16S rrna genes were used to identify ruminal microorganisms in primarily domesticated ruminant animals, but also wild ruminant animals (Sundset et al., 2007; Nelson et al., 2003). Because of the limited numbers of clones sequenced in individual studies, only predominant members of the ruminal microbiome were identified. In addition, individual studies were biased due to techniques used and limited scopes of sampling (e.g., small numbers of animals, diets, and geographic areas sampled). In an effort to assess the current status of species richness that has been revealed in the rumen, we performed a meta-analysis on all the rrs sequences (more than 10,000) of rumen origin found in public databases. This meta- 1

22 analysis provided a global phylogenetic framework which can be used in guiding future studies and tool development. From the above meta-analysis of 16S rrna gene sequences, many thousands of OTUs were identified and they represent about 70% of the predicted diversity in the rumen. All these sequences were determined using the Sanger DNA sequencing technology. Although the Sanger technology is not as cost-effective as pyrosequencing on a per-sequence basis, it produces sequence data that are more accurate than those obtained using pyrosequencing technologies. As researchers started to choose pyrosequencing over Sanger sequencing in characterizing microbiome, we performed a study using Sanger sequencing of clone libraries with an objective to determine if conventional clone libraries can still contribute to discovery of novel diversity in the rumen. This study also created the clones that were used in validation of the phylogenetic microarray we developed. Monensin is an ionophores and it has been fed to feedlot cattle to improve production efficiency (Russell, 2002). It has been proposed that Gram-positive bacteria are more sensitive to monensin than Gram-negative bacteria in in-vitro cultures due to lack of outer membrane (Nagaraja et al., 1997; Callaway et al., 2003). Thus, monensin can inhibit Gram-positive bacteria, including H 2 -producing bacteria. As a result, monensin can reduce methane production in the rumen and shift fermentation towards more reduced VFA (e.g. propionate), decreasing acetate:propionate ratio (Callaway et al., 2003). However, some bacteria do not display the above norm (Russell and Houlihan, 2003). Monensin resistance is thought to be mediated by extracellular polysaccharides and results from a physiological selection (Russell and Houlihan, 2003). Some in vivo 2

23 studies using rrs-based techniques (Stahl et al., 1988; Weimer et al., 2008) showed that supplementary monensin did not shift ruminal bacterial populations due to monensin resistance. In addition, some in vivo studies showed that supplementary monensin did not decrease the acetate:propionate ratio (e.g. Firkins et al., 2008; Oelker et al., 2009). Firkins et al. (2008) indicated that long-term change in ruminal bacterial communities could be attributed to other dietary factors than sensitivity to monensin. As the first comprehensive study to examine the bacterial effect of monensin supplementation in cattle, we analyzed the ruminal bacteriome in Holstein dairy cattle fed monensin to examine to what extent monensin alters bacterial populations. Lipids, especially unsaturated oils, function to modulate the ruminal microbiome and improve some aspects of rumen function, such as decreasing methanogenesis and ammonia production (Martin et al., 2010). Unsaturated fat can readily be incorporated into rations for high producing dairy cows to improve energy intake. However, findings and conclusions vary among studies with respect to the magnitude of efficacy or, conversely, their negative effect of feeding fat (e.g., decreased fiber digestibility, milk fat depression, and milk protein decrease). Conceivably, depending on the composition of the lipids, different groups of ruminal microbes could be affected by lipids to different extents. Fibrolytic bacteria can be inhibited by coating feed particles with lipids that interferes in their adhesion or activity (Calsamiglia et al., 2007). We contend that delineating and understanding the impact of lipids on the different members of the ruminal microbiome will help understand inconsistent outcomes of lipid supplementations. To this end, we examined the effects of lipids on ruminal microbiome using pyrosequencing analysis. 3

24 To support the above pyrosequencing analysis, we also evaluated different regions ( bp) of this phylogenetic marker to identify the best partial sequence region(s) and suitable phylogenetic distance values that can provide as reliable species richness estimates as full-length sequences. We were able to identify that the V1-V3 region at 0.04 distance is the best region to define species-level OTUs. In another study, the populations of select cultured and uncultured bacteria present in different ruminal samples were determined to assess their prevalence in the rumen. The results showed that some of the uncultured bacteria can be as important, at least numerically, as some of the previously cultured bacteria that have been perceived to be important to rumen function. Pyrosequencing has been increasingly used in analysis of microbiomes, including ruminal microbiomes (Callaway et al., 2010). Although it can support detailed diversity analysis of ruminal samples, it produces considerable amounts of artifactual sequences (Quince et al., 2009). Because these artifactual sequences are produced randomly, the true population sizes and community structure are difficult to assess. Although some of the artifactual sequences can be filtered out, diversity is still overestimated considerably (11% to 35%) (Gomez-Alvarez et al., 2009; Kunin et al., 2010). Phylogenetic microarrays do not have these limitations and have been proven to be useful in analysis of complex microbiomes in human gut (Palmer et al., 2006; Rajilic-Stojanovic et al., 2009). Using the phylogenetic framework established from the meta-analysis, we developed a phylogenetic microarray dedicated to comprehensive analysis of ruminal bacteria. The microarray, referred to as RumenArray, was validated in terms of probe specificity, detection limits, and dynamic range of detection. We also tested the utility of the RumenArray by analyzing some fractionated ruminal samples collected from sheep 4

25 fed two different diets. The RumenArray is the first phylogenetic microarray dedicated to comprehensive analysis of ruminal bacteriome, and it may be used in support of semiquantitative analysis of rumen bacteria in nutritional studies. Collectively, the series of studies described here have advanced our understanding of the ruminal microbiome, providing detailed information on the effects on monensin and dietary fat on ruminal bacteria, and developed useful tools that can help significantly improve future studies of ruminal microbiome. 5

26 CHAPTER 2 REVIEW OF LITERATURE 2.1 Microbial communities in the rumen The rumen harbors a complex microbiome consisting of diverse bacteria, archaea, fungi, protozoa, and viruses. This microbiome co-evolves with the host and forms a stable yet dynamic climax community. The very survival and well-being of ruminant animals depends on digestion and subsequent fermentation of ingested feed by a functional ruminal microbiome Bacterial communities Bacteria dominate in the rumen. Bacteria play the most important role in converting ingested feed to nutrients (e.g., acetic, propionic, and butyric acids) that can be assimilated by the ruminant hosts. Bacteria also make the greatest contribution to the nitrogen supply to the host animals. Although amylolytic and lipolytic bacteria also contribute to feed digestion, cellulolytic bacteria are the focus of many studies on the ruminal microbiome because degradation of cellulose, which is recalcitrant, is often the limiting step in feed digestion Major cellulolytic bacteria 6

27 Fibrobacter succinogenes, Ruminococcus albus, and R. flavefaciens are three major cellulolytic bacteria commonly isolated from the rumen. F. succinogenes is a Gram-negative and rod-shaped anaerobe first isolated from the rumen of cattle (Hungate, 1950). Strains tested of F. succinogenes degrade fiber and even crystalline cellulose more actively than those of R. albus or R. flavefaciens (Kobayashi et al., 2008). Because F. succinogenes is phylogenetically diverse, Kobayashi et al. (2008) divided F. succinogenes into four phylogenetic groups and indicated that one of the four groups (group 1) is more important in fiber degradation than the other three groups. The genus Fibrobacter was represented by 26 operational taxonomic units (OTUs), and 11 of the 26 OTUs were associated with F. succinogenes, supporting the diversity of F. succinogenes (Kim et al., 2011b). Six of the 11 OTUs were represented by less than 5 rrs sequences, whereas the remaining 5 OTUs were represented by more than 5 rrs sequences. The 6 OTU groups may play more important role in cellulose degradation than the 5 less prevalent OTU groups. Phylogenetic analysis of the genus Ruminococcus on the basis of 16S rrna gene (rrs) sequence comparisons showed that all species previously isolated fall within the Class Clostridia of phylum Firmicutes. However, a few species were classified to class Bacilli. The newly defined genus Ruminococcus is a monophylogenic group and contains both the Ruminococcus species isolated from the rumen: R. albus and R. flavefaciens. R. albus is a Gram-positive, non-pigmented (white) and coccus-shaped anaerobe first isolated from the rumen of dairy cattle (Hungate, 1957) and subsequently from other herbivorous animals (Hobson and Stewart, 1997). The growth of R. albus in the medium containing cellulose as a sole energy source was shown to be dependent on the addition 7

28 of rumen fluid in the growth medium (Wood et al., 1982). About the same time, a distinguishing feature of R. albus strain 8 was determined as its dependence on the provision of micromolar concentrations of phenylacetic acid (PPA) and phenylpropionic acids (PPA) for optimal growth and cellulose degradation (Hungate and Stack, 1982; Stack and Cotta, 1986; Stack and Hungate, 1984). Subsequent work with other strains of R. albus as well as other cellulolytic ruminal bacteria established the effects of PAA/PPA were species-specific, and it is now widely accepted that cellulose but not xylan degradation by R. albus strains is conditional on the availability of PAA/PPA (Morrison et al. 1990; Reveneau et al. 2003; Stack and Cotta, 1986). Ruminococcus flavefaciens, which is a Gram-variable and coccus-shaped anaerobe produces a yellow pigment. R. flavefaciens is generally less abundant than R. albus in the rumen because R. flavefaciens is inhibited by a bacteriocin produced by some strains of R. albus (Russell, 2002). However, some studies including a study in Chapter 4 showed that R. flavefaciens is more abundant than R. albus (Kongmun et al., 2011; Mosoni et al., 2011). R. flavefaciens mainly digested epidermis, schlerenchyma and phloem cells via attachment to their cut edges (Hobson and Stewart, 1997). Unlike R. albus, PAA/PPA is not required for optimum growth of R. flavefaciens Minor cellulolytic bacteria Butyrivibrio fibrisolvens is a Gram-negative and rod-shaped anaerobe first isolated from the bovine rumen (Hungate, 1950). Some strains of B. fibrisolvens were involved in cellulose degradation, although they have a limited ability to degrade cellulose compared to F. succinogenes, R. albus and R. flavefaciens (Dehority, 2003). 8

29 Therefore, B. fibrisolvens strains are thought to play a minor role in cellulose degradation in the rumen. Eubacterium cellulosolvens is a Gram-positive and rod-shaped anaerobe first isolated from the bovine rumen, and it degrades cellulose (Bryant et al., 1958). However, E. cellulosolvens is thought to play an unimportant role in cellulose degradation due to its low abundance (Bryant et al., 1958). Four cellulolytic Clostridium species, which are C. cellobioparus, C. locheadii, C. longisporum and C. polysaccharolyticum, were isolated from the rumen (Dehority, 2003). Their contribution to cellulose degradation in the rumen remains to be determined. A cellulolytic Micromonospora strain was isolated from the ovine rumen but it is thought not to be important in cellulose degradation (Maluszynska and Janota-Bassalik, 1974). Other cellulolytic bacteria isolated from the rumen include Fusobacterium polysaccharolyticum (Van Gylswyk, 1980), a Cellulomonas strain (Kim et al., 2011b), and Cellulosilyticum ruminicola (Cai and Dong, 2010). Future studies are needed to evaluate their significance in rumen feed digestion Uncultured cellulolytic bacteria Fibrobacter, Ruminococcus, and Butyrivibrio had many species-level OTUs identified from the rrs sequences of uncultured ruminal bacteria (Kim et al., 2011b). These uncultured Fibrobacter, Ruminococcus and Butyrivibrio OTUs are presumed to play an important role in cellulose degradation (Kim et al., 2011b). Cellulolytic Acetivibrio cellulolyticus and Aectivibrio cellulosolvens had been isolated from sewage sludge (Patel et al., 1980; Khan et al., 1984). Ruminal Acetivibrio had 106 species-level OTUs defined from ruminal rrs sequences (Kim et al., 2011b). The real-time PCR assay 9

30 showed that one uncultured Acetivibrio OTU named Ad-H was more abundant in sheep fed hay than in sheep fed corn:hay and as numerous as R. flavefaciens (Stiverson et al., 2011). Therefore, some strains of Acetivibrio may contribute significantly to cellulose degradation. Larue et al. (2005) reported that many rrs clones were closely related to Clostridium, and some of these uncultured Clostridium spp. may be important to cellulose degradation. This premise corroborates the finding of a diverse Clostridium spp., such as C. cellobioparus, C. locheadii, C. longisporum and C. polysaccharolyticum, in the rumen (Dehority, 2003). Numerous rrs sequences recovered from the adherent fraction of rumen contents were assigned to unclassified Ruminococcaceae, unclassified Clostridiales, or unclassified Lachnospiraceae (Kim et al., 2011b). High abundance of uncultured bacteria classified into these groups was confirmed using the real-time PCR assays (Stiverson et al., 2011). Many of the uncultured bacteria assigned to these three groups could be important to cellulose degradation in the rumen Archaeal communities Methane is produced as an end-product of rumen fermentation, and approximately 17 liters per an hour can be produced by a cow (Russell, 2002). Because of the short retention time in the rumen, almost all the methanogens found in the rumen are hydrogenotrophic methanogens, which grow much faster than acetotrophic methanogens. Smith and Hungate (1958) showed that methanogens are present in the 10-7 dilution of rumen liquid. Janssen and Kirs (2008) reported that, as of 2008, methanogens had been classified to 28 genera and 113 species of which only a few were recovered from the 10

31 rumen. Methanobrevibacter is the predominant genus of methanogen. It is not only a genus frequently reported from isolations but also the genus represented by most rrs sequences (Kim et al., 2010b). Methanobrevibacter ruminantium was the first methanogen isolated from the rumen (Smith and Hungate 1958). Two new species of Methanobrevibacter, M. millerae and M. olleyae, were isolated in recent years from cattle and sheep, respectively (Rea et al., 2007). Methanomicrobium mobile is another predominant cultured ruminal methanogen species, reaching a population as high as > 10 8 / ml (Hobson and Stewart, 1997). Methanobacterium formicicum (Gilbert et al. 2010) and Methanobacterium bryantii (Janssen and Kirs, 2008) are two species isolated from bovine rumen. Methanobacterium beijingense was represented by only one rrs sequence recovered from goat rumen (Kim et al., 2011b). Methanoculleus spp. were only recently isolated from the rumen. Janssen and Kirs (2008) isolated a Methanoculleus olentangyi strain from the rumen, but no ruminal rrs sequence corresponding to this isolate was found in the RDP database (Kim et al., 2011b). Although not published, three Methanoculleus marisnigri strains of rumen origin were also recorded in the RDP database (Kim et al., 2011b). Methanosarcina was reported (Hobson and Stewart, 1977) in the rumen, but not as a numerous methanogen (Janssen and Kirs, 2008). Only two ruminal rrs sequences recovered from unpublished studies were classified to Methanosarcina barkeri (Kim et al., 2011b). Ruminal Methanosaet spp., the obligate acetotrophic methanogens, has not been isolated or represented in rrs sequence databases. Collectively, cultured methanogens only account for a small part of diversity of ruminal methanogens (Janssen and Kirs, 2008), and based on a meta-analysis, rrs sequences 11

32 recovered from methanogen isolates account for less than 2% of all the methanogen sequences of rumen origin (Kim et al. 2011b). 2.2 Methods to investigate microbial diversity Cultivation-dependent methods Investigation of ruminal bacteria has been done with cultivation-based methods for many decades. Cultivation-based studies have helped to elucidate some of the important metabolic functions in the rumen from in vitro studies of model organisms (Hobson and Stewart, 1997). Anaerobic roll tubes from which individual colonies could be selected had been used to isolate ruminal bacteria (Dehority, 2003). After anaerobic glove boxes were developed, Petri plates could be easily used to isolate ruminal bacteria in the anaerobic glove boxes (Leedle and Hespell, 1980). Hungate (1950) primarily developed an anaerobic technique using tubes sealed with rubber stoppers under O 2 free CO 2. Cellulolytic bacteria such as F. succinogenes, R. albus and R. flavefaciens were first isolated from the bovine rumen using the anaerobic technique by Hungate (1950). Butyrivibrio fibrisolvens and Prevotella ruminicola, predominant hemicellulose degrading bacteria, were also isolated from the rumen using the anaerobic technique (Dehority, 2003). The anaerobic technique also helped isolate the following amylolytic bacteria: Streptococcus bovis, Ruminobacter amylophilus, Succinimonas amylolytica, and Selenomonas ruminantium (Dehority, 2003). In addition, lactate-utilizing bacteria and methanogens could be isolated from the rumen using the Hungate anaerobic technique 12

33 (Dehority, 2003). These culturable species have been used as models of rumen microbial ecology. However, rrs-based studies showed that culturable ruminal bacteria could explain only a small proportion of the ruminal bacteria (Deng et al., 2008; Kim et al., 2011b). Ruminal bacteria are very sensitive to oxygen because they are strict anaerobes. Also, liquid media for isolation of pure cultures may not represent the true ruminal environment. These limitations led to the use of the rrs-based techniques. However, isolation of pure cultures is still very important in our understanding of ruminal microbial ecology. New isolation methods need to be developed for future studies Cultivation-independent methods Woese et al. (1983) first suggested rrs as a phylogenetic marker because it is phylogenetically conserved and not laterally transferred (Gentry et al., 2006). All the microbes have rrs sequences consisting of hypervariable and universal regions (Yu et al., 2004a; Gentry et al., 2006). Therefore, the hypervariable regions can be used to identify individual species whereas the universal regions can be targeted to analyze broad groups of microbes. Since rrs-targeted analysis was first applied to examining ruminal microbial diversity (Stahl et al., 1998), numerous studies have attempted to investigate ruminal bacterial diversity by rrs-targeted analysis using several molecular approaches (Kim et al., 2011b). For the past three decades, construction of rrs clone libraries has been used to identify predominant bacteria in the rumen. Most studies using rrs clone libraries have focused on ruminal bacteria present in rumen fluid (e.g. Tajima et al., 2000, 2007; 13

34 Ozutsumi et al., 2005; Zhou et al., 2009), whereas other studies examined ruminal bacteria attached to plant particles (Larue et al., 2005; Yu et al., 2006; Brulc et al., 2009). A few studies focused on ruminal bacteria present on the rumen wall (Cho et al., 2006; Lukas et al., 2010). These studies showed that ruminal bacterial diversity differed among these microenvironments. All the ruminal sequences that passed quality controls are archived in the RDP database, while GenBank serves as the main depository of all the sequences (including sequences of poor quality and chimeric sequences) recovered from any source, including the rumen. Analysis of rrs clone libraries has contributed to identifying predominant and novel ruminal bacteria that are both culturable and unculturable. However, it is laborious to use this method and the results are not quantitative. Also, it is difficult to identify less numerous members of the population using this method due to the limited number of clones that researchers can afford to sequence. Nonetheless, this method is still important because a small number of rrs clones can still help find novel species-level OTUs as demonstrated recently (Kim et al., 2011c). Fluorescence in situ hybridization (FISH) has been used to visually detect target organisms using probes labeled fluorescently. The probes can hybridize to rrna within the undamaged microbial cell (Deng et al., 2008) and detect microbes in situ within its natural environment. Yanagita et al. (2000) examined the diversity of ruminal methanogens using the FISH method and showed that Methanomicrobium mobile accounts for 54% of all the methanogens. Mackie et al. (2003) identified the presence of ruminal Oscillospira species using FISH. The FISH method has helped to visualize both culturable and unculturable bacteria in situ within a ruminal environment. However it is a 14

35 laborious and non-quantitative method (McSweeney et al., 2007), and its application is limited due to the constraint of probe design for FISH (Deng et al., 2008). Numerous uncultured bacteria were identified in the ruminal adherent fraction, and they may be associated with fiber degradation (Kim et al., 2011b). Their attachment to plant biomass can be confirmed using the FISH method. Real-time PCR assays have been used to quantitatively estimate microbial populations in complex environmental samples (McSweeney et al., 2007). Tajima et al. (2001a) designed primer sets for 12 ruminal species and quantified these using a realtime PCR assay. Ruminal archaea, fungi and protozoa have also been quantified using real-time PCR (Sylvester et al., 2004; Denman and McSweeney, 2006; Jeyanathan et al., 2011). Stiverson et al. (2011) reported the first study that quantified uncultured bacteria represented by rrs sequences in the rumen using real-time PCR. This method is not suitable for discovery of novel diversity because primers and probes have to be designed from known sequences. However, it is the most suitable method to evaluate dietary effect on ruminal microbes because dietary manipulation often results in quantitative changes in population sizes, which cannot be precisely determined by other methods, such as clone libraries, DGGE, or pyrosequencing. DGGE has been used to separate rrs PCR fragments amplified from complex environmental DNA. Yu et al. (2004a) used DGGE to examine diversity of total ruminal bacteria and identified the most useful variable region based on rrs sequences. Genera Prevotella and Treponema in the rumen were analyzed using the DGGE technique with genus-specific primers (Bekele et al., 2010, 2011). Yu et al. (2008) also reported that the V3 region is the best target in DGGE profiling of ruminal archaea. The rrs sequences of 15

36 novel bacteria can be detected using this method, but it is difficult to confirm nonpredominant bacteria. As a profiling technique, DGGE can help identify represent samples that can be analyzed in detail using other molecular techniques, such as clone libraries and DNA sequencing. Terminal restriction fragment length polymorphisms (T-RFLP) were first used to rapidly screen microbial diversity in sludge, aquifer sand, and termite guts (Liu et al., 1997). Metagenomic DNA extracted from the environment sample is subjected to PCR using universal primers of which one is labeled at the 5 -end, and then PCR products are digested with a restriction enzyme. A DNA sequencer is commonly used to visualize the terminal fragment that is fluorescently labeled (Kitts, 2001). Fernando et al. (2010) analyzed microbial diversity in feedlot cattle during shift from a high-forage diet to a high-grain diet using the T-RFLP technique. Frey et al. (2010) also examined microbial diversity in rumen, duodenum, ileum and feces of cattle using the T-RFLP technique. The use of the DNA sequencer helps provide greatly reproducible data for repeated samples (Liu et al., 1997). However, rrs sequence information cannot be obtained directly from the T-RFLP profile, and two different rrs sequences representing different species can have the same peak if a terminal restriction site is shared. Kim et al. (2011b) retrieved more than 10,000 bacterial rrs sequences of rumen origin, and they can be used as a reference set in inferring the bacteria represented by individual TRFs recovered from ruminal bacterial communities. Such a ruminal dataset can reduce ambiguity that results from large generic databases. A phylogenetic microarray is a small gene chip that includes numerous oligonucleotide probes and allows comprehensive simultaneous detection of numerous 16

37 rrs targets. Phylogenetic microarrays have been used to examine microbial communities in various environments such as soil (Small et al., 2001; Liles et al., 2010), human gut (Palmer et al., 2006; Kang et al., 2010), human feces (Rajilic-Stojanovic et al., 2009), activated sludge (Adamczyk et al., 2003), and lake (Castiglioni et al., 2004). Although a phylogenetic microarray can rapidly and simultaneously detect numerous microbial populations, it is only semi-quantitative and has lower sensitivity and specificity than real-time PCR (McSweeney et al., 2007). However, it is still very useful in evaluating dietary effect because microarray allows simultaneous semi-quantification of multiple species of bacteria. No ruminal phylogenetic microarray has been reported until this study. Since the 454 Genome Sequencer (GS) (Roche, Branford, Connecticut) was developed for massively parallel sequencing applications (Margulies et al., 2005), it has been applied to analyze microbiomes in various samples (Turnbaugh et al., 2006; Roesch et al., 2007; Qu et al., 2008; Pope et al., 2010), including ruminal samples (Callaway et al., 2010). The original 454 GS FLX system could read approximately 250 bp spanning one or two variable rrs regions. As the sequencing technology continues to improve, the current 454 GS FLX Titanium system is able to produce read lengths of about 500 bp. Recently 454 Life Sciences has announced a new GS FLX system that will produce 1,000 bp reads. It is expected that sequencing of full-length rrs (1.5 kb) will be achieved in the near future. For analysis of ruminal microbiomes, Brulc et al. (2009) examined microbiomes in liquid and adherent fractions of rumen digesta using the GS20 system and showed that ruminal microbial diversity of three cattle fed the same diet were noticeably different. Pitta et al. (2010) compared ruminal microbiomes between cattle fed 17

38 bermudagrass hay diet and cattle grazing winter wheat. Callaway et al. (2010) compared ruminal microbiomes among cattle fed diets containing 0, 25, and 50% dried distillers grain plus solubles (DDGS), and they reported that dietary supplementation with DDGS altered the ruminal microbiomes. Another study assessed in vitro fermentation dynamics of corn products and examined differences in ruminal microbiomes between two different fermentation periods of time (Williams et al., 2010). However, the number of rare OTUs could be overestimated due to pyrosequencing errors. Kunin et al. (2010) evaluated the pyrosequencing errors using only Escherichia coli MG1655 as a reference rrs sequence, and they found that the number of OTUs was greatly overestimated. Kunin et al. (2010) concluded that the use of stringent quality-based trimming and the calculation of OTUs at 97% identity could reduce, but not eliminate, the pyrosequencing errors. It remains to be a challenge to distinguish real sequences from artifactual sequences produced during pyrosequencing. Recently, Huse has developed a pseudo-single linkage algorithm that can remove rrs sequences that seem to result from the pyrosequencing errors (unpublished study), and it was added in the Mothur program (Schloss et al., 2009). These approaches can contribute to improved diversity analysis. However, until sequencing accuracy is improved to a level comparable to that of the Sanger sequencing, pyrosequencing data need to be interpreted with caution. Additionally, prevalence of individual sequence reads has been used to calculate the relative abundance of the bacteria or archaea represented. However, it should be kept in mind that pyrosequencing subjects to PCR bias because two rounds of PCR are involved in pyrosequencing: generation of PCR amplicons and emulsion PCR. Therefore, relative abundance and community structure calculated from sequence prevalence can be questionable. 18

39 2.3 Bioinformatics analysis to investigate microbial diversity Rarefaction has been used to analyze and compare species richness based on OTUs among samples of different sizes (Hughes et al., 2001). Rarefaction curves can be constructed by computing species-level OTUs for the number of rrs sequences sampled as described previously (Schloss et al., 2004). Rarefaction curves increase quickly at first and then approach an asymptote where few new OTUs are found with increased sampling. However, asymptotic richness cannot be read off rarefaction curves directly because rarefaction curves rarely reach plateau. The asymptotic richness can be estimated using the non-linear model procedure (PROC NLIN) of SAS (V9.1, SAS Inst. Inc., Cary, NC) as described previously (Larue et al., 2005). Nonparametric estimators, such as Chao1 and ACE, can also be used to examine microbial species richness. These two nonparametric estimators use mark-release-recapture (MRR) statistics that uses the ratio of OTUs that have been observed and singleton OTUs (Hughes et al., 2001). Chao1 and ACE tend to underestimate richness when sample sizes are small (Hughes et al., 2001). Because the morphology of most microbes cannot be distinguished under a microscope, cultivation-independent analysis using rrs sequences has been used to estimate microbial phylogenetic diversity. To evaluate the microbial phylogenetic diversity from numerous rrs sequences, systematic analysis has been developed using various bioinformatics programs (Schloss and Handelsman, 2004). The rrs sequences need to be aligned prior to microbial phylogenetic analysis. The alignment format can be directly downloaded from the RDP database (Cole et al., 2009) or created using bioinformatic programs such as ClustalW (Larkin et al., 2007), Greengenes (Desantis et 19

40 al., 2006) and Silva (Pruesse et al., 2007), and then distance matrices can be created using the DNADIST program in the PHYLIP package (Schloss and Handelsman, 2005). OTUs can be generated from the distance matrices using the DOTUR program, and then rarefaction curves can be constructed (Schloss and Handelsman, 2005). OTUs at 0.03, 0.05, and 0.10 phylogenetic distances are conventionally used to represent species, genus, and family, respectively, based on the full-length rrs sequences (Kim et al., 2011a). The maximum number of OTUs can be estimated from the rarefaction curves using the nonlinear model, and the percent coverage can be computed from observed and maximum numbers of OTUs as described previously (Larue et al., 2005; Kim et al., 2011b). Since the Mothur program containing various bioinformatic tools was recently developed, alignment, distance matrices, and OTUs can be generated directly (Schloss et al., 2009). Caporaso et al. (2010) also developed a bioinformatics program, quantitative insights into microbial ecology (QIIME), to analyze microbial diversity. The QIIME program also includes a variety of bioinformatics tools to analyze and visualize microbial diversity, and it rapidly calculates OTUs using Uclust that is a clustering, alignment and search algorithm for the analysis of large sequence datasets (Caporaso et al., 2010). 2.4 The use of monensin and its role in rumen fermentation Monensin, an ionophore, has been used to improve feed efficiency in ruminant animals (Russell and Strobel, 1989). Monensin has been reported to decrease methane production, reducing energy loss, and increase propionate at the expense of acetate (Russell and Strobel, 1989). Monensin also can decrease urinary ammonia excretion, ruminal acidosis and liver abscesses (Callaway et al., 2003). As a result, efficient feed 20

41 utilization results in improvement of production in ruminant animals (Callaway et al., 2003). However, some studies did not show that monensin supplementation affects the acetate:propionate ratio (e.g. Firkins et al., 2008; Oelker et al., 2009; Mathew et al., 2011). Monensin can replace H + with Na + or K + as a metal/proton antiporter (Callaway et al., 2003). Intracellular K + is replaced with extracellular H +, whereas extracellular Na + is substituted for intracellular H + (Callaway et al., 2003). Because the gradient of K + is higher than that of Na +, accumulated H + decreases ph (Callaway et al., 2003). The increased H + activates a reversible ATPase and then pumps the intracellular H + out of the cell (Callaway et al., 2003). Other ATP-dependent pumps were also activated to restore Na+/K+ gradients, resulting in cell death by reduction in intracellular ATP pools (Russell and Strobel, 1989). Lipophilic monensin inhibits Gram-positive bacteria more than Gram-negative, because Gram-positive bacteria lack an outer membrane present in Gram-negative bacteria (Nagaraja et al., 1997; Callaway et al., 2003). In pure cultures, the growth of Gram-positive bacteria was inhibited by monensin, while the growth of Gram-negative Escherichia coli O157:H7 or K12 was not reduced by monensin (Buchko et al., 2000). However, the outer membrane may not be the only factor for monensin sensitivity (Russell and Strobel, 1989). This is exemplified by the finding that some Gram-negative bacteria were susceptible to monensin, and monensin tolerance was developed in both Gram-positive and Gram-negative bacteria (Callaway and Russell, 2000). Therefore, supplementary monensin should result in shift towards monensin-resistant microbial populations (Callaway et al., 2003). However, a conflicting finding was reported from an early rrs-based study where no significant change in microbial population was detected 21

42 after dietary monensin supplement (Stahl et al., 1988). Other dietary and intraruminal environmental factors other than supplementary monensin may be associated with long term change of ruminal microbial communities (Firkins et al., 2008). The use of pyrosequencing analyses will help provide deep insight into the shift of microbial populations as affected by supplementary monensin. 2.5 Biohydrogenation in the rumen Lipids in the rumen are metabolized through lipolysis and subsequent biohydrogenation (BH) of unsaturated fatty acids. Dietary lipids are hydrolyzed by ruminal microbial and plant lipase, releasing polyunsaturated fatty acids (PUFA) that are commonly contained in dietary grass and feedstuff for ruminant animals. These PUFA, such as linoleic acid (cis-9, cis-12 18:2) and α-linolenic acid (cis-9, cis-12, cis-15 18:3), are converted to saturated fatty acids through a BH process carried out by ruminal bacteria. As intermediates of the BH, conjugated linoleic acid (CLA) is formed from dietary linoleic acid in the rumen. The final end-product of this BH is stearic acid (Hobson and Stewart, 1997). In BH of PUFA, the initial isomerization step results in conversion of linoleic acid (cis-9, cis-12 18:2) to cis-9, trans-11 CLA and trans-10, cis-12 CLA. These CLA isomers are hydrogenated to trans-11 18:1 and trans-10 18:1 and then finally reduced to stearic acid (18:0). The cis-9, trans-11 CLA also can be formed from endogenous conversion of vaccenic acid (trans-11 18:1) in the host mammary glands through oxidation by 9-desaturase (Jenkins et al., 2008; Or-Rashid et al., 2009). The BH of 20:5 and 22:6 in fish oil has not been clearly identified. Isomerization, hydrogenation, or chain 22

43 shortening might be involved in the BH of these two PUFA (Jenkins et al., 2008). In addition, the Moate model described that 16:1 is hydrogenated to 16:0 (Jenkins et al., 2008). Two groups of ruminal microorganisms, termed group A and group B, are involved in the BH process (Hobson and Stewart, 1997). Group A microorganisms have the ability to convert linoleic acid to trans 18:1 isomers but cannot hydrogenate trans 18:1 isomers further. Butyrivibrio fibrisolvens is involved in this BH process (Jenkins et al., 2008). Because trans 18:1 isomers are formed via cis-9, trans-11 CLA, Group A microorganisms are thought to be cis-9, trans-11 CLA producing microorganisms. Group B microorganisms can convert not only 18:1 isomers but also 18:2 isomers to stearic acid. Consequently, both group A and B microorganisms are required for complete conversion to stearic acid (Hobson and Stewart, 1997). Kemp et al. (1975) identified that Fusocillus spp. are Group B microorganisms. A recent study (Boeckaert et al., 2009) showed that ruminal BH in the solid fraction of rumen contents is primarily responsible for complete conversion of linoleic acid (cis-9, cis-12 18:2) to stearic acid (18:0), but in the liquid fraction, BH is associated with the conversion of linoleic acid (cis-9, cis-12 18:2) to trans-10 18:1/ trans-11 18:1. Few studies have been conducted on ruminal BH of α-linolenic acid (cis-9, cis-12, cis-15 18:3). Or-Rashid et al. (2009) presumed that α- linolenic acid is converted to cis-9, trans-11, cis-15 18:3, trans-9, cis-11, cis-15 18:3, trans-10, cis-12, cis-15 18:3, trans-9, trans-11, cis-15 18:3, and trans-10, trans-12, cis-15 18:3 isomers by certain rumen microorganisms. Fellner et al. (1997) described that 18:1 isomers are increased by the inhibition of the BH process of linoleic acid by supplementary monensin. It seems that monensin 23

44 hinders the terminal step of the BH process at which 18:1 isomers are converted to stearic acid (Jenkins et al., 2003). The concentrations of trans-12 18:1, trans-15 18:1 and trans- 16/cis-14 18:1 were increased by supplementary monensin, whereas those of trans-10 and trans-11 18:1 were not affected (Mathew et al., 2011). However, Oelker et al. (2009) reported that trans-11 18:1 tended to be increased by supplementary monensin. The concentration of trans-11, cis-15 18:2 also tended to increase when monensin was fed (Lourenço et al., 2008; Mathew et al., 2011). The concentrations of cis-9, trans-11 CLA or total CLA were not affected by supplementary monensin (Oelker et al., 2011), but this result is contradictory with a previous study (Odongo et al., 2007). 2.6 Summary The ruminal microbiome is complex and diverse. The phylogenetic microarray allows for detection and semi-quantification of abundant microbes in a comparative manner. Less abundant microbes could be detected and quantified by real-time PCR with species- or genus-specific primers. Also, the construction of rrs clone libraries can still contribute to finding novel ruminal OTUs, which can be used to design new microarray probes. Further, pyrosequencing can help detect numerous rrs sequences, including those representing minor species. Integrated investigation of ruminal microbial communities using these rrs-based techniques will help elucidate the ruminal microbial communities more precisely than ever before and contribute to understanding rumen function. 24

45 CHAPTER 3 STATUS OF THE PHYLOGENETIC DIVERSITY CENSUS OF RUMINAL MICROBIOMES 3.1 Abstract In this study, the collective microbial diversity in rumen was examined by performing a meta-analysis of all the curated 16S rrna gene (rrs) sequences deposited in the RDP database. As of November 2010, 13,478 bacterial and 3,516 archaeal rrs sequences were found. The bacterial sequences were assigned to 5,271 operational taxonomic units (OTUs) at species level (0.03 phylogenetic distance) representing 19 existing phyla, of which the Firmicutes (2,958 OTUs), Bacteroidetes (1,610 OTUs), and Proteobacteria (226 OTUs) were the most predominant. These bacterial sequences were grouped into more than 3500 OTUs at genus level (0.05 distance), but only 180 existing genera were represented. Nearly all of the archaeal sequences were assigned to 943 species-level OTUs in the phylum Euryarchaeota. Although clustered into 670 genuslevel OTUs, only 12 existing archaeal genera were represented. Based on rarefaction analysis, the current percent coverage at species level reached 71% for bacteria and 65% for archaea. At least 78,218 bacterial and 24,480 archaeal sequences would be needed to reach 99.9% coverage. The results of this study may serve as a framework to assess the significance of individual populations to rumen functions and to guide future studies to identify the alpha and global diversity of ruminal microbiomes. 3.2 Introduction 25

46 The rumen has evolved to digest various plant materials by a complex microbiome consisting of bacteria, archaea, protozoa, and fungi. Within this microbiome, bacteria are the most abundant domain and make the greatest contribution to digestion and conversion of feeds to short chain fatty acids (SCFA) and microbial proteins (Hobson and Stewart, 1997). Ruminal archaea are mostly methanogens that belong to the phylum Euryarchaeota. Utilizing the CO 2 and H 2 produced from bacterial fermentation, these methanogens produce methane, a potent greenhouse gas that is implicated in global warming (Janssen and Kirs, 2008). Both bacterial and archaeal populations can be affected by many factors, such as species and age of hosts, diets, seasons, and geographic regions (Tajima et al., 2001a; Zhou et al., 2009). Numerous efforts have been attempted to optimize rumen functions by enhancing feed digestion, improving conversion of dietary nitrogen to microbial proteins, and reducing methane emission and nitrogen excretion by manipulating the ruminal microbiome through dietary means. Although limited success has been achieved, few of these dietary manipulations achieved persistent effect without negatively affecting the overall rumen function (van Nevel and Demeyer, 1996; Calsamiglia et al., 2007; Patra and Saxena, 2009). The lack of sufficient understanding of the ruminal microbiome is considered one of the major knowledge gaps that hinder effective enhancement of rumen function (Firkins and Yu, 2006). The ruminal microbiome, as with other microbiomes, has been investigated primarily using cultivation-based methods for many decades until the 1980 s when 16S rrna gene-targeted analysis was applied (Stahl et al., 1988). Cultured bacteria and archaea helped in defining some of the important metabolic activities underpinning rumen functions (Hobson and Stewart, 1997); however, it soon became evident that most 26

47 of the rumen microbes escaped laboratory cultivation (Whitford et al., 1998). Thereafter, most studies attempted to characterize the ruminal microbiome by phylogenetic analysis of 16S rrna gene (rrs) sequences recovered in clone libraries by direct PCR amplification (reviewed by Edwards et al., 2004; Deng et al., 2008). Some DNA-based studies focused on the microbes present in rumen fluid (e.g., Tajima et al., 2000, 2007; Ozutsumi et al., 2005; Zhou et al., 2009), while other studies analyzed the microbes partitioned in rumen fluid or embedded in the biofilm adhering to feed particles separately (e.g., Larue et al., 2005; Yu et al., 2006; Brulc et al., 2009). Because of practical considerations, most of the studies reported hitherto focused on the ruminal microbiome of domesticated ruminant animals, but a few studies examined the ruminal microbiome of wild ruminant species, including reindeer (Sundset et al., 2007) and several species of wild African ruminants (Nelson et al., 2003). These studies showed that the ruminal microbiome of wild ruminant animals contains ruminal microbiome distinct from that of domesticated ruminant animals. Rarefaction analysis (Hughes et al., 2001) is typically used to estimate the depth of coverage of diversity in most studies on microbiomes, including the ruminal microbiome. Because a limited number of rrs sequences were sequenced in individual studies reported hitherto, the prokaryotic diversity resident in the rumen has only been partially uncovered. Even the two studies that sequenced thousands of rrs sequences (Yu et al., 2006; Brulc et al., 2009) failed to achieved complete coverage. Additionally, some of the rrs sequences recovered from the rumen were deposited in public databases but have not been reported in the literature, contributing little to characterization and understanding of ruminal microbial diversity. 27

48 Besides the limited depth of coverage of diversity, the scope of these studies was also narrow with respect to species and number of sampled animals, diets, and geographic regions. Typically, rumen digesta of only a few animals of the same species (primarily cattle and sheep) fed one or few diets were sampled and analyzed in a small number of countries. Edwards et al. (2004) noted that individual 16S clone libraries might have been biased towards certain microbial phyla due to PCR biases and that the prokaryotic diversity in the rumen might be much greater than that indicated by individual studies. The limited scope of sampling in individual studies might be a major factor contributing to such bias. Thus, we hypothesize that the general prokaryotic diversity of ruminal microbiome can be better defined by a meta-analysis of rrs sequences (both published and unpublished) recovered from all the rumens that have been analyzed. Edwards et al. (2004) performed a collective analysis, but only the data from 3 studies (Whitford et al., 1998; Tajima et al., 1999, 2000) were analyzed. Sequences shorter than 1 kb were also excluded. In this study, we performed a global phylogenetic analysis of all publicly available rrs sequences of rumen origin in an effort to provide a collective and progressive census of the ruminal prokaryotes and to gain further insight into the ruminal microbiome. Finally, we estimated the current coverage of the prokaryotic diversity already identified and the number of sequences that would be needed to fully describe the prokaryotic diversity in the rumen in general. 3.3 Materials and Methods Sequence data collection and phylogenetic analyses 28

49 As of November 2010, the RDP database (Release 10, Update 22) was searched for rrs sequences of rumen origin using the search terms rumen and ruminal to collect sequences of both ruminal bacteria and archaea. Sequences of suspect quality were excluded using the option of Quality in the RDP database. To ensure the rrs sequences of cultured ruminal bacteria and archaea were included in our analysis, databases of ATCC, CCUG, DSMZ, and JCM culture collections were searched using the same search terms rumen and ruminal. The rrs sequences of those isolates were added to the above sequence selections if they were not found in the initial search in RDP. The sequences for total bacteria, bacterial isolates, total archaea, and archaeal isolates were downloaded separately from the RDP database in aligned format with common gaps removed. The navigation tree for each of these sequence datasets was also downloaded. Each of the sequence datasets and the associated navigation tree were imported into ARB, a software environment that can store, manage and analyze rrs sequences (Ludwig et al., 2004). Taxonomic trees with the Bergey s taxonomy applied were generated from the imported navigation trees using the ARB program. All the rrs sequences of both ruminal bacteria ( 274bp) and archaea ( 200bp) were aligned against the rrs Greengenes database (DeSantis et al., 2006). The resultant aligned sequences were inserted into the Greengenes database ARB tree to generate a detailed phylogenetic tree using the positional variance by parsimony method (Ludwig et al., 2004) Diversity estimate 29

50 Numbers of OTUs were calculated for total bacteria, total archaea, and major groups of bacteria using the Mothur program (Schloss et al., 2009). Briefly, the aligned sequences for each group of bacteria or archaea were separated from the rest of the sequences in the Greengenes alignment. Distance matrix was constructed at 0.03 (equivalent to species), 0.05 (genus) and 0.10 (family) phylogenetic distances. The number of OTUs observed within each group was calculated based on rarefaction analysis implemented in the Mothur program at each of the distances. From each of the rarefaction curves, the asymptote that indicates the maximum number of OTUs represented by each group of sequences was estimated at species, genus and family levels using the non-linear model procedure (PROC NLIN) of SAS (V9.1, SAS Inst. Inc., Cary, NC) as described previously (Larue et al., 2005). The percent coverage was calculated by dividing the observed number of OTUs by the maximum number of OTUs estimated. The number of sequences that would be required to provide 99.9% coverage at these 3 levels was predicted using the same non-linear model (Larue et al., 2005). The Chao1 and ACE estimates were also calculated using the Mothur program. 3.4 Results A naïve meta-analysis of ruminal microbiome was conducted using all publicly available rrs sequences that have been recovered worldwide from both domesticated and wild ruminant animals using the Sanger DNA sequencing technology. By naïve metaanalysis we mean we collectively analyzed all the sequences irrespective of the taxonomic associations reported in the respective studies. Thus, this analysis enabled a fresh, broad view on the global diversity of ruminal microbiome. The result is an updated 30

51 consolidated perspective of the ruminal microbiome based on most current phylogenetic diversity information Data summary In total, 13,478 bacterial and 3,516 archaeal sequences of rumen origin were analyzed (Figs. 1 and 2). The bacterial and archaeal sequences accounted for approximately 79% and 21%, respectively, of the sequence dataset analyzed. Nineteen bacterial phyla were represented, with Firmicutes, Bacteroidetes, and Proteobacteria being the most numerous phyla, accounting for 57.8, 26.7, and 6.9% of the total bacterial sequences, respectively (Figure 3.1). The remaining 16 phyla, collectively referred to as minor phyla, each was represented by <3% of the total bacterial sequences. About 99.6% of the archaeal sequences were assigned to phylum Euryarchaeota, and only 11 sequences (0.3%) were assigned to phylum Crenarchaeota (Figure 3.2). The sequences recovered from cultured bacteria and archaea accounted for only 6.5% and 1.7% of all the respective sequences, respectively, and represented 88 existing bacterial (Figure 3.3) and 6 known archaeal (Figure 3.4) genera. The generic views depicted by all the bacterial and the archaeal sequences are provided as Figure 3.5 and Figure 3.2, respectively. The cultured species within each bacterial genus were also listed in Figure Firmicutes Approximately 90.6% of the Firmicutes sequences were assigned to class Clostridia, with the rest assigned to Bacilli, Erysipelotrichi and Unclassified_Firmicutes (Figure 3.5). The Firmicutes sequences were assigned to a total of 78 known genera, and 31

52 54 of these genera are within class Clostridia (Figure 3.5). Within class Bacilli, lactic acid-producing genera Streptococcus and Carnobacterium were most predominant. Some genera (e.g., Streptococcus and Lactobacillus) were represented mostly by rumen isolates, while others (e.g., Carnobacterium and Planococcus) comprised entirely uncultured bacteria. Except for Bacillus, Planococcus, Planomicrobium, Enterococcus and Lactobacillus, other genera were each represented by <10 sequences. It seems likely that these minor genera are probably not residents in rumen. Within class Clostridia, Lachnospiraceae and Ruminococcaceae were the largest families, accounting for 23.8% and 25.8% of the Clostridia sequences, respectively, followed by Veillonellaceae (7.7%) (Figure 3.5). The predominant genera included Butyrivibrio (4.8% of the Clostridia sequences), Acetivibrio (4.5%), Ruminococcus (4.1%), Succiniclasticum (3.7%), Pseudobutyrivbrio (2.3%) and Mogibacterium (2.3%). The Butyrivibrio sequences were assigned to 134 species-level OTUs. The largest OTU contained 17 sequences recovered from cattle, buffaloes, or sheep in at least 10 studies conducted in 7 countries, reflecting its wide distribution. Cultured isolates contributed 42 sequences within 27 OTUs, and 25 of these sequences were derived from bacteria named B. fibrisolvens, while 4 sequences from bacteria named B. hungatei (Figure 3.3). The second largest genus, Acetivibrio, had 106 species-level OTUs, but only 2 sequences came from isolates including a Clostridium cellobioparum strain. The most abundant Acetivbrio OTU contained 47 sequences recovered from 15 studies on cattle, sheep, buffaloes, camels, yaks, or reindeer in several continents (Tajima et al., 1999, 2007; Koike et al., 2003a; Ozutsumi et al., 2005; Sundset et al., 2007; Brulc et al., 2009; Yang et al., 2010a, b). Ruminococcus was the third largest genus represented by

53 species-level OTUs, with the two most abundant OTUs containing 16 sequences each. Fifty-three sequences were recovered from isolates named R. albus, R. flavefaciens, R. bromii, or Ruminococcus sp. (Figure 3.5). These cultured Ruminococcus sequences were assigned to 24 OTUs and 7 of them were each represented by one isolate. The fourth largest genus, Succiniclasticum, was represented by 58 species-level OTUs. The largest OTU comprised 37 sequences recovered in 11 studies. Except one sequence recovered from a Succiniclasticum ruminis strain initially isolated from cattle (van Gylswyk, 1995), all the Succiniclasticum sequences were recovered from uncultured bacteria from cattle, sheep, camels, gayals, or yaks (Tajima et al., 2007; Brulc et al., 2009; Yang et al., 2010a). Pseudobutyrivibrio, the fifth largest genus closely related to Butyrivibrio, was represented by 27 species-level OTUs. This genus was well represented by cultured isolates (66 in total), primarily isolates previously named Butyrivibrio fibrisolvens (Figure 3.3). Mogibacterium was the sixth largest genus represented by 35 species-level OTUs. However, this genus was not represented by any cultured isolate. The most abundant OTU comprised 40 sequences recovered from rumen epithelium in an unpublished study (GenBank record). Other genera represented by >1% but <2% of the bacterial sequences included Anaerovorax, Coprococcus, Oscillibacter, and Selenomonas. Selenomonas and Oscillibacter are common genera well represented by rumen isolates, but Coprococcus and Anaerovorax were primarily represented by uncultured bacteria (Figure 3.5). The Coprococcus sequences were assigned to 42 species-level OTUs. The most abundant OTU had 21 sequences, including one recovered from Lachnospira multipara ATCC 19207, an isolate from bovine rumen. Anaerovorax comprised 67 species-level OTUs, 33

54 with the most abundant OTU containing 16 sequences, 14 of which were recovered from rumen epithelium in the same unpublished study. Besides the predominant or familiar genera mentioned above, quite a few genera that have been rarely reported in the rumen were also represented by the sequence dataset, including Lachnobacterium, Moryella, Oribacterium, Roseburia, Syntrophococcus, Papillibacter, and Dialster. Most of these uncommon genera had no cultured isolates (Figure 3.5). Within the smallest class Erysipelotrichi, Sharpea, a new genus established in 2008 (Morita et al., 2008), was the largest genus (Figure 3.5). Although bacteria of this new genus were initially isolated from horse feces, 22 of all the 50 Sharpea sequences found in RDP were recovered from the rumen, suggesting the rumen as a common habitat for bacteria of this genus. A large number of sequences (approx. 60% of the Firmicutes sequences) have not been classified to any existing family, order, or genus within class Clostridia (Figure 3.5). The largest group (1,858 sequences) of related sequences remains to be classified to a family in the order Clostridiales. These Unclassified_Clostridiales sequences represented 606 species-level OTUs, with the largest OTU containing 64 sequences recovered from cattle in the USA (Brulc et al., 2009) and Japan (Tajima et al., 2007), and buffalos in China (Yang et al., 2010a). Twelve isolates were found within this group, and they represented 9 OTUs, with 6 of them being represented by one isolate each: 3 Clostridium sp., one C. sticklandii, one C. clostridioforme, and one unnamed isolate (Figure 3.3). The second largest unclassified sequence group (1,177 sequences) was found within family Ruminococcaceae. This group of sequences represented 524 species- 34

55 level OTUs, with two major OTUs containing 19 sequences each. Sequences in one of these two OTUs were recovered from cattle or sheep in 6 studies conducted in Canada, Japan, and the USA (Tajima et al., 1999; Brulc et al., 2009). The other major OTU contained sequences that were recovered from cattle in the USA (Brulc et al., 2009). This unclassified group was represented by 4 isolates, including one named R. albus, which represented 4 OTUs (Figure 3.3). The third largest group, Unclassified_Lachnospiraceae, contained 1090 sequences that represent 588 species-level OTUs. The largest OTU had 15 sequences recovered from sheep in Belgium (unpublished data). Twenty-five of the 588 OTUs had cultured representatives: Cellulosilyticum ruminicola, Clostridium sp., C. aminovalericum, C. polysaccharolyticum, C. aminophilum, C. aerotolerans, C. symbiosum, R. gnavus, E. uniforme, E. rangiferina, L. multipara, P. xylanivorans, and 14 unnamed isolates Bacteroidetes Bacteroidetes was the second most numerous phylum (3,605 sequences). Most (88.5%) of these sequences were assigned to class Bacteroidia (Figure 3.5), and the rest of the sequences were assigned to class Sphingobacteria (38 sequences) and Unclassified_Bacteroidetes (378 sequences). The Bacteroidetes sequences were assigned to 15 genera within Bacteroidia and 2 genera within Sphingobacteria (Figure 3.5). Of these genera, 6 were represented by rumen isolates. Prevotella was the most numerous genus, accounting for 41.5% of the Bacteroidetes sequences or 11.1% of all the bacterial sequences. The large number of Prevotella sequences reflects the predominance of this genus in the rumen in general (Stevenson and Weimer 2007; Bekele et al., 2010). The 35

56 1,496 Prevotella sequences were assigned to 637 OTUs, of which 20 OTUs contained cultured sequences (58 in total). Four OTUs were represented by only one cultured bacterium. Most of the Prevotella isolates were classified as P. ruminicola, followed by P. brevis, P. bryantii, and P. albensis (Figure 3.3). Thirteen isolates remain to be classified to a Prevotella species. The remaining 617 OTUs were represented only by uncultured bacteria. The most abundant OTU had 34 sequences recovered in two unpublished studies. Nine of the 15 represented genera within Bacteroidia had no cultured representative, including the second most abundant genus Paraprevotella (represented by 44 OTUs) and the third most abundant genus Barnesiella (24 OTUs) (Figure 3.5). Except for Barnesiella, Rikenella, and Paraprevotella, other genera were only represented by <1% of the Bacteroidetes sequences. Genera Hallella and Rikenella might be adapted to the rumen-like environment because 21.4 and 17.6% of all the respective sequences found in RDP were of rumen origin. More than half (53%) of the Bacteroidetes sequences has not been classified to any existing genus, family, or class (Figure 3.5). These unclassified sequences were mostly placed into 4 groups in the RDP database: Unclassified_Bacteroidales (27.5%, 395 species-level OTUs), Unclassified_Bacteroidetes (10.5%, 212 species-level OTUs), Unclassified_Prevotellaceae (8.0%, 236 species-level OTUs), and Unclassified Porphyromonadaceae (5.7%) (Figure 3.5). All these unclassified sequences were recovered from uncultured ruminal bacteria (Lodge-Ivey et al., 2005; Brulc et al., 2009; Yang et al., 2010a), except one isolate unclassified within Porphyromonadaceae, three isolates unclassified within Bacteroidales, and one isolate unclassified within 36

57 Bacteroidetes. All the 38 sequences of Sphingobacteria were recovered from uncultured bacteria. The Unclassified_Bacteroidales sequences accounted for about 22% of all the respective rumen sequences found in RDP Proteobacteria All the 5 classes of Proteobacteria were represented in the dataset (Figure 3.5). Class -Proteobacteria was entirely represented by a small number of sequences recovered from uncultured bacteria. Phenylobacterium was the largest genus (32.8% of the -Proteobacteria sequences) but all its sequences were recovered from semicontinuous RUSITEC. Within class β-proteobacteria, 17 genera were recognized, most of which were represented by only several sequences. Aquabacterium was the most predominant genus (32.4% of the β-proteobacteria sequences); however, all its sequences were recovered from uncultured bacteria from semi-continuous RUSITEC or ruminal protozoal cultures. Four sequences were recovered from cultured isolates: one Comamonas terrigena strain, two Lampropedia hyalina strains, and one Oxalobacter formigenes strain. γ-proteobacteria was the most predominant class, which was represented by about 73% of all the proteobacterial sequences. Twenty-three existing genera within this class were represented, with Ruminobacter, Succinivibrio, Escherichia/Shigella, and Pseudomonas each accounting for 2% of the γ-proteobacterial sequences (Figure 3.5). Most of the genera within γ-proteobacteria were represented by at least one rumen isolate (Figure 3.3), including well characterized species, such as Ruminobacter amylophilus, Succinimonas amylolytica, Succinivibrio dextrinosolvens, Actinobacillus succinogenes, Mannheimia ruminalis, Mannheimia succiniciproducens, 37

58 Acinetobacter lwoffii, Pseudomonas putida, and Pseudomonas oryzihabitans. Again, most of these genera were represented by sequences recovered in a few studies, suggesting paucity of the bacteria represented by these sequences. It should be noted that genus Psychrobacter comprised >58% of the γ-proteobacteria sequences, but all its sequences were recovered from 3 steers in a single study (Brulc et al., 2009). Given that all the characterized bacteria within this genus are strictly aerobic psychrophiles (Bozal et al., 2003) and that rumen is a mesophilic environment, the occurrence of Psychrobacterlike bacteria needs to be replicated. Class δ-proteobacteria was represented by 47 sequences, with Desulfovibrio being the most predominant genus. Except 5 sequences from isolates of Desulfovibrio, most of the δ-proteobacterial sequences were recovered from uncultured bacteria unclassified within Desulfovibrionaceae. ε-proteobacteria was the smallest class represented by only 20 sequences, but 5 of them were recovered from cultured strains: one from Campylobacter sp. and 4 from Wolinella succinogenes (Figure 3.3). Twenty-one of the 53 identified genera of Proteobacteria were represented by at least one isolate from the rumen Minor phyla Approximately 7% (949 sequences) of the bacterial sequences were assigned to 16 minor phyla (Figure 3.1). These phyla varied in number of sequences, with Synergistetes being represented by 382 sequences, while Acidobacteria by only one sequence. Most (95.5%) sequences within phylum Synergistetes were recovered from rumen epithelium (an unpublished study). Except genera Pyramidobacter and Thermovirga, each of which accounted for 14% of the sequences, most sequences remain 38

59 unclassified within family Synergistaceae (Figure 3.5). These results suggest that Synergistetes might be predominant inhabitants of rumen epithelial surface. Two ruminal isolates were found within this phylum, and one of them belonged to Synergistes jonesii, a species isolated from the rumen and capable of degrading toxic pyridinediols (Allison et al., 1992). Spirochaetes was the second largest minor phylum, and most of its sequences (79.2%) were assigned to genus Treponema (Figure 3.5). However, Spirochaetes is a very small phylum, accounting for only 1.1% of all the bacterial sequences. All the Fibrobacteres sequences belonged to the cellulolytic genus Fibrobacter, and >50% of its sequences were assigned to F. succinogenes (Figure 3.5). This species contained 58 of the 59 Fibrobacter isolates (Figure 3.3), reflecting the predominance of this species within Fibrobacter and the justified effort made to isolate and characterize it (Bera-Maillet et al., 2004; Shinkai et al., 2010). Actinobacteria was represented by 107 sequences, and most of its sequences were assigned to Olsenella and Bifidobacterium (Figure 3.5). The former genus has one species (i.e., Olsenella oviles) initially isolated from the rumen (Dewhirst et al., 2001), while the latter genus contained sequences derived from cultured isolates that belonged to several species. The 16 Tenericutes sequences were assigned to genus Asteroleplasma, Acholeplasma or Anaeroplasma, the last of which was represented by the type species and strains isolated from the rumen (Figures 3.3 and 3.5). Fusobacteria comprised 10 sequences recovered from cultured isolates named Fusobacterium sp. (Figures 3.3 and 3.5). The remaining minor phyla, including Acidobacteria, Chloroflexi, Deferribacteres, Lentisphaerae, Planctomycetes, Verrucomicrobia, OP10, Cyanobacteria, SR1, and TM7 were represented by a very small number of sequences recovered from uncultured bacteria 39

60 (Figure 3.5). It remains to be determined if they are resident bacteria and have any significant role in the rumen Archaea Both phyla Crenarchaeota and Euryarchaeota were represented by the sequence dataset, but the former comprised only 11 sequences from uncultured or unclassified Thermoprotei (Figure 3.2). About 94% of all the archaeal sequences were assigned to 4 classes within phylum Euryarchaeota (Figure 3.2): Methanobacteria (70.3% of all the archaeal sequences), Methanomicrobia (16.4%), Thermoplasmata (7.4%), and Methanopyri (0.03%). Within these classes, 12 genera were represented (Figure 3.2). Within Methanobacteria, 3 genera of family Methanobacteriaceae were represented: Methanobrevibacter, Methanosphaera and Methanobacterium. There were 239 sequences unclassified to any existing genus. The Methanobrevibacter sequences represented 201 species-level OTUs, with the largest OTU containing 456 sequences. This OTU contained sequences recovered from cattle and sheep (Tajima et al., 2001b; Whitford et al., 2001a; Wright et al., 2004, 2007, 2008; Skillman et al., 2006; Rea et al., 2007; Zhou et al., 2009) and reindeer (Sundset et al., 2009) in 13 studies. This OTU also contained 4 isolates named Methanobrevibacter sp.. Genus Methanobrevibacter was represented by 31 isolate sequences that were assigned to 16 OTUs and classified as M. ruminantium, M. olleyae and M. millerae, or designated as Methanobrevibacter sp.. These results further confirmed the predominance and ubiquity of hydrogenotrophic Methanobrevibacter spp. in the rumen (Janssen and Kirs, 2008). 40

61 The Methanosphaera sequences were assigned to 225 species-level OTUs with no cultured representative, suggesting its predominance in the rumen but poor culturability in the laboratory. The most abundant OTU had 25 sequences recovered from cattle (Whitford et al., 2001a), goat (Cheng et al., 2009), reindeer (Sundset et al., 2009), and muskoxen (an unpublished study). On the contrary, Methanobacterium was a minor genus, but well represented by isolates, including M. beijingenes, M. bryantii, M. formicicum, and 7 other Methanobacterium sp. isolates. Within class Methanomicrobia, Methanomicrobium was the largest genus (92.3% of the Methanomicrobia sequences), but except for 6 sequences recovered from 2 species and one unnamed isolate (Figure 3.2), most of the sequences were recovered from uncultured methanomicrobial archaea. In total, 263 species-level OTUs were recognized, with the most abundant OTU containing 86 sequences recovered from cattle in Korea (an unpublished study). Six OTUs were represented by one cultured isolate each, named as Methanomicrobium mobile, Methanoculleus marisnigri, or Methanobacterium sp.. Methanimicrococcus was the second most abundant genus within Methanomicrobia, and it was represented only by uncultured methanogens. On the contrary, Methanoculleus, Methanofollis, and Methanosarcina were represented by a few sequences that were recovered from isolates (Figure 3.2). These sequences may represent archaeal organisms that are readily culturable but scarce in the rumen. Methanosaeta, the obligate acetoclastic methanogen genus, was only represented by one uncultured sequence, while the facultative acetoclastic Methanosarcina was represented by 4 rumen isolates. The class Thermoplasmata was only represented by uncultured archaea in the genus Thermogymnomonas or Unclassified Thermoplasmatales. The Thermogymnomonas 41

62 sequences were assigned to 31 species-level OTUs, with the most abundant OTU containing 18 sequences recovered from cattle (Wright et al., 2007, and unpublished studies), reindeer (Sundset et al., 2009), sheep and buffaloes (unpublished studies). All the isolates in this genus were recovered from rice field or deep-sea hydrothermal vents (based on GenBank records). Further studies are needed to examine their occurrence in the rumen. Three groups of archaeal sequences remain to be classified (Figure 3.2). The Unclassified_Euryarchaeota group was assigned to 86 species-level OTUs, with the most abundant one containing 18 sequences recovered from cattle (Wright et al., 2007), sheep (Wright et al., 2004; Nicholson et al., 2007), reindeer (Sundset et al., 2009), and buffaloes (an unpublished study). The Unclassified_Methanobacteriaceae sequences represented 92 species-level OTUs, and the most abundant OTU contained 29 sequences recovered from cattle and muskoxen (unpublished studies). Finally, the Unclassified_Thermoplasmatales sequences were assigned to 57 species-level OTUs, with the largest OTU containing 27 sequences recovered from sheep (Wright et al., 2006; Ohene-Adjei et al., 2007), goats (Cheng et al., 2009), and semi-continuous RUSITEC (an unpublished study) Estimates of OTU richness The numbers of OTUs of Bacteria, Archaea, the major bacterial phyla, 4 predominant genera, and 3 unclassified groups were estimated (Table 3.1). Nearly 5,300 bacterial and 950 archaeal species-level OTUs were defined by the current dataset. Firmicutes had approximately twice as many OTUs as Bacteroidetes. A large number of 42

63 OTUs were defined in the unclassified groups that contained numerous sequences recovered from the biofilm adhering to feed particles in the rumen of sheep (Larue et al., 2005) and cattle (Brulc et al., 2009) fed hay. The percentage coverage reached 65% and 71% at species levels for archaea and bacteria, respectively. For the 4 major known bacterial genera (Fibrobacter, Ruminococcus, Butyrivibrio, and Prevotella), coverage at species level ranged from 73 to 81% (Table 3.1). The coverage for the unclassified groups was comparable to that of classified groups. As expected, higher coverage was noted at genus and family levels (Table 3.1). The maximum numbers of OTUs predicted with rarefaction were lower than the Chao1 or ACE estimates (Table 3.1). Based on rarefaction estimate, the rumen might contain >7400 and 1400 species-level OTUs of bacteria and archaea, respectively. Firmicutes and Bacteroidetes were estimated to have >3900 and >2300 species-level OTUs, respectively. The 3 unclassified groups might have >700 species-level OTUs each. Ruminococcus, Butyrivibrio, and Prevotella were predicted to have a large number of species-level OTUs. At least 32 species-level OTUs could be found in genus Fibrobacter in the rumen. To reach nearly complete (99.9%) coverage at species level, at least 78,000 and 24,000 new sequences would be required for bacteria and archaea, respectively. For bacteria, most of these additional sequences are expected to be from species within Firmicutes and Bacteroidetes Discussion Phylogenetic diversity 43

64 A few phyla were typically reported by individual studies (Whitford et al., 1998; Tajima et al., 2000; Larue et al., 2005; Brulc et al., 2009), but collectively 19 phyla of bacteria were represented by the consolidated rrs sequence dataset analyzed in this study. At low taxonomic ranks (genus and species), the collected global diversity also exceeded the -diversity, which refers to the biodiversity within a particular habitat, reported in individual rumens (Table 3.1). This discrepancy might be attributed to the limited number of rrs sequenced, diets and animals used, the geographic regions and seasons sampled, and bias associated with the methods used in individual studies. Firmicutes and Bacteroidetes were the predominant phyla with respect to numbers of both sequences and species-level OTUs. Although numbers of sequences or OTUs within rrs sequence datasets may not necessarily reflect distribution or abundance, these two phyla are recognized to be omnipresent and dominant in the rumen (Edwards et al., 2004; Larue et al., 2005; Brulc et al., 2009). Based on the dataset used in this study, the phylum Firmicutes is much more predominant and diverse than Bacteroidetes. However, this might not be the case in individual studies, and the relative abundance of these two important phyla varied among studies (e.g., Whitford et al., 1998; Larue et al., 2005; Ozutsumi et al., 2005; Tajima et al., 2007; Brulc et al., 2009). The variations might result from differences in PCR primers used (Edwards et al., 2004), fractions (liquid vs. solid) of rumen samples analyzed, DNA extraction methods used, and coverage achieved. Indeed, Leser et al. (2002) noted that relative abundance of phylogenetic groups could differ greatly between separate libraries generated from the same sample due to small numbers of clones analyzed, or factors that affect PCR amplification. Therefore, the 44

65 meta-analysis reported in this study might have helped to account for among study effects and might offer new insight into ruminal microbial diversity. Ten of the 19 bacterial phyla represented by the sequence dataset did not contain any isolates from the rumen. This might be due to their dearth in the rumen and/or their recalcitrance to laboratory cultivation. As mentioned above, several phyla (e.g., Acidobacteria, Chloroflexi, Deferribacteres, Lentisphaerae, OP10, and Cyanobacteria) might not be resident bacteria in the rumen. However, some minor phyla are likely resident ruminal bacteria. For example, phylum Synergistetes was only represented by 17 sequences until January of 2010, but >360 sequences classified into this phylum were recently recovered from rumen epithelium (NCBI, unpublished data) that was rarely analyzed. Thus, caution needs to be exercised when ruling out certain bacteria or archaea as resident members of the rumen. Three large groups of unclassified bacteria: the Unclassified_Clostridiales, Unclassified_Lachnospiraceae, and Unclassified_Ruminococcaceae, are probably predominant ruminal bacteria. A quantitative analysis of 6 uncultured bacteria within these groups using respective specific real-time PCR assays showed that these uncultured bacteria were as abundant as R. albus, R. flavefaciens, and F. succinogenes in the rumen of both sheep and dairy cattle (unpublished data). These 3 unclassified bacterial groups are likely competitive in the rumen and some of their species might have an important role in ruminal feed digestion. Yet, except a few isolates orphaned from other genera or species, all 3 groups are primarily represented by uncultured bacteria, especially bacteria from adherent fraction (Larue et al., 2005; Brulc et al., 2009). Isolation and 45

66 characterization of representative strains belonging to these groups will greatly assist in better understand their ecology, physiology, and contributions to rumen function. A large number of species-level OTUs was identified by the consolidated sequence dataset (Table 3.1), but only 10 archaeal species have been described within five genera (Figure 3.4). The poor representations by cultured archaea reflect the general difficulties to isolate rumen methanogens and the limited numbers of cultivation-based studies conducted hitherto. New species, genera, and families are needed to accommodate the archaeal diversity represented by the unclassified sequences. We note the many archaeal sequences that were assigned to Thermogymnomonas, which is a new genus described in 2007 (Itoh et al., 2007). Based on the type species, Thermogymnomonas acidicola, this genus is acidophilic, strictly aerobic, and moderately thermophilic. Many sequences recovered from the rumen, an anaerobic and mesophilic environment suggest adaptation to the rumen. Methanobrevibacter, Methanomicrobium and Methanosphaera accounted for 50, 15, and 13% of all the archaeal sequences. The overall predominance of Methanobrevibacter was in general agreement with its abundance noted in individual studies (Janssen and Kirs, 2008). The predominance of these hydrogenotrophic methanogens is likely attributed to their ability to grow relatively rapidly (so avoiding washout) and to competitively utilize H 2 and CO 2, major fermentation products in the rumen. Future mitigation of rumen methane emission might be directed towards inhibition of these three hydrogenotrophic genera. Of course, alternative H 2 sink, such as homoacetogenesis, is needed to reduce possible negative effect on fiber digestion. 46

67 3.5.2 Diversity estimates Although large numbers of bacterial and archaeal species-level OTUs have been identified by the rrs sequences, more species-level OTUs remain to be identified. The diversity estimates for all the bacterial groups (Table 3.1) are much greater than previously suggested (Edwards et al., 2004; Yu et al., 2006). As noted earlier, the dataset used in this study included sequences recovered from different animals fed different diets in different countries, and thus might help reach this expanded phylogenetic view of ruminal microbiome. For the well-studied genera Fibrobacter, Ruminococcus, Butyrivibrio, and Prevotella, the numbers of OTUs estimated and predicted for each genus were also much greater than seeming possible. This surprise might be attributable to, among other factors, a lack of a reliable taxonomy that can classify sequences to species. For example, nearly all the Fibrobacter sequences were assigned to F. succinogenes, yet F. succinogenes contains a diverse group of bacteria (Amann et al., 1992). Based on a 0.03 phylogenetic distance, the Fibrobacter sequences can be assigned to 26 species-level OTUs. The genus Ruminococcus was also shown to contain two distinct and unrelated clusters (Rainey and Janssen, 1995). Defined reclassification of species within these genera will provide a taxonomic framework to support future studies of these important groups of ruminal bacteria. The current coverage at species level is still incomplete, even with the wellstudied genera (Table 3.1). Therefore, novel bacteria and archaea remain to be identified. According to the estimates from the collected sequence datasets, to achieve 99.9% coverage of ruminal global diversity at species level, >70,000 bacterial and >20,000 archaeal rrs sequences would need to be recovered from multiple animals fed different 47

68 diets across broad geographic regions. It should be noted, however, that the current coverage might be an underestimate because with increasing number of sequences, the predicted maximum richness tended to increase (Yu et al., 2006; Roesch et al., 2007). Therefore, more sequences than that estimated here might be required. Although identifying the full diversity in ruminal microbiome has loomed as technically challenging, the recent advancement of next- or third-generation DNA sequencing technologies coupled with coordinated international efforts can help achieve this goal. Indeed, three large datasets of rrs sequences have been generated recently using the Solexa 1G Genome Analyzer (an unpublished study, GenBank accession numbers: SRX007415, SRX007414, approx. 70 bp) and the 454 FLX system (GenBank accession number: SRA , <300 bp) (Pitta et al., 2010). These sequence datasets are very large, but they only contain very short sequences, especially those generated by the Solexa 1G Genome Analyzer. These short sequences were not included or analyzed in this study because they could not be reliably analyzed together with the longer sequences of our sequence dataset. With the most recent 454 FLX Titanium system, rrs sequences up to 800 bp can be massively sequenced. Partial rrs sequences of this length can be added to the consolidate sequence dataset used in this study. Eventually, the full diversity of ruminal microbiome can be defined, which may serve as a guideline in design of future studies on rumen nutrition and provide a framework to assess the significance of individual population in the rumen. 48

69 Group Observed # of OTUs (% coverage) Maximum # of OTUs Clones needed for Rarefaction richness Chao1 ACE 99.9% coverage Total Archaea 949 (65) 670 (75) 278 (91) Total Bacteria 5271 (71) 3774 (82) 1785 (96) Firmicutes 2958 (74) 2050 (84) 936 (98) Bacteroidetes 1610 (69) 1162 (83) 518 (97) Proteobacteria 226 (71) 172 (83) 97 (94) Unclassified Clostridiales 606 (76) 400 (80) 189 (91) Unclassified Lachnospiraceae 588 (66) 452 (80) Unclassified Ruminococcaceae 524 (70) 381 (78) Fibrobacter 26 (81) Ruminococcus 108 (81) Butyrivibrio 134 (77) Prevotella 637 (73) Table 3.1. The number of OTUs for total bacteria, total archaea and major groups of bacteria, and their percentage coverage at three phylogenetic distances. 49

70 50 Figure 3.1. Bacterial phyla represented by the 16S rrna gene sequences of rumen origin. The taxonomic tree was created using the ARB program. In total 19 existing bacterial phyla were represented by the 13,478 bacterial sequences. All the sequences in rectangle bars were classified down to the same taxonomic rank, whereas sequences in sloped bars were classified down to different taxonomic ranks.

71 Figure 3.2. A taxonomic tree showing the genera of ruminal archaea identified by the RDP database sequences. The lineage at class level is labeled: Mb, class Methanobacteria; Mm, class Methanomicrobia; Tp, class Thermoplasmata. In total, 12 known genera of archaea were represented by the 3,516 archaeal sequences. The number in parentheses indicate the number of sequences recovered from isolates. 51

72 Figure 3.3. A taxonomic tree showing the bacteria (grouped into genera) isolated from the rumen. In total, 882 bacterial isolates were classified into 88 known bacterial genera, accounting for 6.5% of all the bacterial sequences. The most predominant species or isolates were indicated in brackets. The numbers in parentheses indicate the total numbers of isolates for each species. Figure 3.3 Continued 52

73 Figure 3.3 Continued Figure 3.3 Continued 53

74 Figure 3.3 Continued 54

75 Figure 3.4. A taxonomic tree showing the archaea (grouped into genera) isolated from the rumen. In total, 68 archaeal isolates were classified into 6 known archaeal genera, accounting for only 1.9% of all the archaeal sequences. The most predominant species or isolates were indicated in brackets. The numbers in parentheses indicate the total numbers of isolates for each species. 55

76 Figure 3.5. A taxonomic tree showing all the genera of ruminal bacteria identified by the 13,478 16S rrna gene sequences of rumen origin. The taxonomic tree was constructed as described in Figure 3.1. Of all the 13,478 sequences, 5,845 sequences were assigned to 179 existing genera within 19 known phyla. The remaining 7,633 sequences could not be assigned to any known genus. The numbers in parentheses indicate the number of sequences recovered from cultured isolates. Figure 3.5 Continued 56

77 Figure 3.5 Continued Figure 3.5 Continued 57

78 Figure 3.5 Continued Figure 3.5 Continued 58

79 Figure 3.5 Continued Figure 3.5 Continued 59

80 Figure 3.5 Continued Figure 3.5 Continued 60

81 Figure 3.5 Continued Figure 3.5 Continued 61

82 Figure 3.5 Continued 62

83 CHAPTER 4 QUANTITATIVE COMPARISONS OF CULTURED AND UNCULTURED MICROBIAL POPULATIONS IN THE RUMEN OF CATTLE FED DIFFERENT DIETS 4.1 Abstract Sequencing analysis of 16S rrna genes (rrs) amplified by PCR is the primary method used in examining diversity and populations of bacteria in various samples, including ruminal samples. However, PCR amplification has bias, and it is often difficult to infer population functions from rrs sequence data. Thus, sequence frequencies and sequence comparison often do not provide adequate information on the abundance and function of the bacteria represented. In this study we used real-time PCR to quantify the populations of select uncultured bacteria to assess their distribution as affected by diets and microenvironments within the rumen. The liquid, adherent and solid fractions were obtained from the rumen of cattle fed two different diets (hay alone vs. hay plus grain). Specific real-time PCR assays were used to quantify the populations of six uncultured bacteria present in each fraction. The abundance of major cultured bacteria was also quantified for comparison. The population of total bacteria was more than 10 8 rrs copies/ µg DNA and similar across all the fractions, while the population of total archaea was less than 10 5 rrs copies/ µg DNA and approximately 10 times higher in cattle fed hay than in cattle fed hay plus grain. The population of Prevotella spp. was more than 10 7 rrs copies/ µg DNA, being the most abundant among all the cultured and the uncultured bacteria quantified. The populations of Fibrobacter succinogenes, Ruminococcus flavefaciens and Butyrivibrio spp. were more than 10 6 rrs copies/ µg DNA, while the population of Ruminococcus albus was less than 10 6 rrs copies/ µg DNA. The 63

84 populations of four of the six uncultured bacteria were approximately 10 6 rrs copies/ µg DNA across all the fractions. These four uncultured bacteria were similar in abundance to F. succinogenes, R. flavefaciens and Butyrivibrio spp. In addition, the populations of the six uncultured bacteria were slightly higher in the adherent and solid fractions than in the liquid fraction. These uncultured bacteria may be associated with fiber degradation. 4.2 Introduction A complex ruminal microbiome mediates hydrolysis of polymeric feedstuffs and subsequent fermentation of hydrolytic products to volatile fatty acids (VFA) that are used as the energy source for ruminant animals. Microbial biomass also constitutes the sources of major protein and B vitamins for the host animals. Being the major contributors to rumen functions, bacteria have been the focus of microbiological studies of the rumen microbiome. Cultivation-based methods were used to investigate ruminal bacteria until the 1980s. As a result, various cultured bacteria were identified, and their functions were determined through physiological studies of model species or strains. Since rrs sequences were used to investigate diversity of ruminal bacteria, it became evident that cultured ruminal bacteria represent only a small portion of the ruminal bacteriome (Stevenson and Weimer, 2007). Kim et al. (2011b) reported that rrs sequences obtained from cultured bacteria represent only 7% of all the bacterial sequences of rumen origin. More than 55% of all the bacterial sequences were assigned to unclassified groups that could not be classified into any known genus (Kim et al., 2011b). Therefore, uncultured members of the ruminal bacteriome probably play a greater role in rumen functions than the cultured peers. 64

85 Frequencies of rrs sequences are often used to infer the abundance of the uncultured bacteria represented. However, PCR is well documented to have amplification bias with universal primers. As such, sequence frequency does not necessarily reflect the relative abundance of the bacterium represented, or the importance or weight to rumen function. In a previous study (Stiverson et al., 2011), specific real-time PCR assays were shown to accurately determine the population sizes and distribution of both cultured and uncultured bacteria in the rumen of sheep. Some uncultured bacteria had abundance comparable to that of several cultured bacteria that are perceived as major bacteria in the rumen. We hypothesize that this holds true for the rumen of cattle. To test this hypothesis, real-time PCR assay quantified the populations of select cultured and uncultured bacteria in the rumen of cattle fed different diets. 4.3 Materials and Methods Sample collection, fractionation and DNA extraction Rumen contents were collected from four cannulated cattle: two Jersey cattle fed with only forage composed predominantly of Timothy grass (designated as hay, H) and two Holstein cattle fed a typical dairy diet consisting of 14% alfalfa forage, 42% corn silage, 6% cottonseed, and 38% grains (designated as including concentrate, C). The two groups of cattle were fed twice daily (early morning and late afternoon) and adapted to their respective diets for more than 3 weeks before rumen sampling, which took place approximately 6 hours after the morning feeding. The liquid and adherent fractions were obtained as described previously (Larue et al., 2005). Bacteria present in the liquid fraction (Lq) were recovered by centrifugation, whereas bacteria adherent to 65

86 the solid digesta were recovered using a detaching buffer (Dehority and Grubb, 1980) and designated as the adherent fraction (Ad). To recover bacteria that might fail to detach, the remaining solid digesta was subjected to DNA extraction and designated as the solid fraction (Sld). Twelve fraction samples (2 cattle 2 diets 3 fractions) were stored at -80 o C prior to DNA extraction. Metagenomic DNA was extracted from each of the fractionated samples as described previously (Yu and Morrison, 2004b) Real-time PCR assays Standards for Fibrobacter succinogenes, Ruminococcus albus and Prevotella ruminicola were amplified from genomic DNA of respective strains using 27F and 1525R primers. A composite sample of the 12 metagenomic DNAs at equal amount was used to prepare sample-derived standards for total bacteria, total archaea, Butyrivibrio, Prevotella, Ruminobacter amylophilus, Ruminococcus flavefaciens, Selenomonas ruminantium and six uncultured bacteria by a regular PCR reaction using specific primers as described previously (Stiverson et al., 2011). The sample-derived standards are thought to reduce bias that may result from sequence variation within total bacteria, total archaea, Butyrivibrio, or Prevotella. On the other hand, the sample-derived standards for R. amylophilus, R. flavefaciens and S. ruminantium were used because their purified genomic DNAs were not available. The six uncultured bacteria named Ad-C1-74-3, Lq- C2-16-3, Lq-C2-58-2, Ad-H1-14-1, Ad-H and Ad-H were recovered from sheep fed two different diets (Larue et al., 2005; Stiverson et al., 2011). Each standard was serially diluted and the concentration from 10 1 to 10 7 rrs copies were used in the real-time PCR assays. Each real-time PCR assay was conducted in three technical 66

87 replicates (three PCR reactions from the same template) from which the mean was calculated. The mean was also calculated from the two biological replicates (two cattle fed the same diet) of each fraction recovered from each diet. The primers (Table 4.1 and 4.2) and PCR condition used to quantify each target were the same as those used by Stiverson et al. (2011). 4.4 Results and Discussion Quantification of populations of total bacteria and total archaea Total bacterial populations ranged from to rrs copies/ µg DNA across all the fractions and were slightly higher in cattle fed hay plus grain than in cattle fed hay only (Figure 4.1). The increased digestibility concomitant with grain supplementation is a major factor that increases the total bacterial population. Total archaeal populations were much higher in the three fractions recovered from cattle fed hay ( ~ ) than the respective fractions recovered from cattle fed hay plus grain ( ~ ) as shown in Figure 4.1. It seems that the abundance of total archaea is affected by the amount of forage in the diet. This result corroborates the previous finding that more methane is produced by animals fed diets high in forage than by animals fed diet high in grain (Janssen, 2010) Quantification of cultured bacteria Populations of three major cellulolytic bacteria and Butyrivibrio spp. were shown in Figure 4.2. Among the three cellulolytic bacteria, the populations of F. succinogenes ( ~ ) and R. flavefaciens ( ~ ) were more abundant 67

88 than the population of R. albus ( ~ ) in any of the fractionated samples. This result supports the previous findings that the population of F. succinogenes is higher than that of R. albus (Koike et al., 2003b; Stevenson and Weimer, 2007). However, some studies showed contradictory results (Martin et al., 2001; Stiverson et al., 2011). The two studies (Koike et al., 2003b; Stiverson et al., 2011) that used sheep showed that R. albus is more predominant among the three cellulolytic species. More studies are needed to verify the predominance of R. albus in the rumen of sheep while F. succinogenes is the predominant cellulolytic in the rumen of cattle. Although real-time PCR assays showed the abundance of Fibrobacter succinogenes, few Fibrobacter-like rrs sequences were identified from rrs clone libraries, the microarray analysis and the pyrosequencing analysis as described in Chapter 5, 6 and 8. The lack of Fibrobacter-like rrs sequences seems to be due to the poor efficiency of PCR amplication with universal primers as demonstrated previously (Larue et al., 2005). The populations of the three cellulolytic bacteria have a tendency to be higher in the Sld fractions than in the Ad fractions (Figure 4.2). This result indicates that cellulolytic bacteria are tightly attached to plant particles, and bacterial detachment by the detaching buffer (Larue et al., 2005) is incomplete. Therefore, Sld fractions should be included in future studies to account for the total population of ruminal bacteria attached to plant particles. The population of Butyrivibrio spp. was greater than 10 6 rrs copies/ µg DNA and did not differ among all the fractions (Figure 4.2). Primary niches of Butyrivibrio are utilization of hemicellulose, starch, pectin, xylan, pentose and hexose (Russell, 2002), and some strains can even degrade cellulose (Hungate, 1950). Therefore, 68

89 Butyrivirio spp. present in the adherent fraction may contribute slightly to cellulose digestion. The population of genus Prevotella ranged from to rrs copies/ µg DNA across all the fractions and was slightly higher in cattle fed hay plus grain than in cattle fed hay (Figure 4.3). The population of Prevotella spp. was the most abundant among known bacteria. This result supports that Prevotella is the most predominant genus in the rumen (Kim et al., 2011b; Stevenson and Weimer, 2007). The population of Prevotella ruminicola was also higher in cattle fed hay plus grain than in cattle fed hay (Figure 4.3). The population difference between cattle fed the two different diets was greater for P. ruminicola than for Prevotella spp.. This result indicates that the populations of some unknown Prevotella spp. are high in the Ad or the Sld fractions than in the Lq fraction. The abundance of Prevotella spp. in the Ad or the Sld fraction might suggest their involvement in fiber degradation as described previously (Koike et al., 2003a; Dodd et al., 2010) and the presence of numerous uncultured Prevotella spp. (Bekele et al., 2010). Isolation and characterization of uncultured Prevotella spp. would need to be attempted in future studies. As expected, both lactate-utilizing Selenomonas ruminantium and starch-utilizing Ruminobacter amylophilus were more abundant in cattle fed hay plus grain than in cattle fed hay (Figure 4.3) Quantification of uncultured bacteria The populations of six different uncultured bacteria were quantified using specific real-time PCR assays. Ad-C1-74-3, Lq-C and Lq-C were originally recovered from sheep fed corn:hay, whereas Ad-H1-14-1, Ad-H and Ad-H

90 were recovered from sheep fed hay (Larue et al., 2005; Stiverson et al., 2011). The population of Ad-C1-74-3, which was assigned to Anaerovorax (Stiverson et al. 2011), was slightly higher in the Ad and the Sld fractions than in the Lq fractions, but it was similar between the Lq fractions, or between the Ad fractions. Because Matthies et al. (2000) reported that Anaerovorax of non-rumen origin frequently metabolizes amino acids, Ad-C may be associated with the degradation of amino acids. Lq-C and Lq-C were assigned to Unclassified Ruminococcaceae and Unclassified Erysipelotrichaceae, respectively (Stiverson et al. 2011). These two uncultured bacteria were slightly more abundant in cattle fed hay plus grain than in cattle fed hay and more abundant in the Ad fractions than in the Lq fractions (Figure 4.4). The population of Lq- C was greater than 10 6 rrs copies/ µg DNA across all the fractions. The populations of Ad-H and Ad-H that were assigned to Acetivibrio and Unclassified Clostridia, respectively, were about 10 6 rrs copies/ µg DNA. The populations of these two bacteria were slightly higher in the Ad fractions than in the Lq fractions (Figure 4.4). Because Acetivibrio includes cellulolytic species such as A. cellulolyticus and A. cellulosolvens as described previously (Stiverson et al., 2011), Ad- H might represent an Acetivibrio bacterium that participates in fiber degradation in the rumen. Future studies targeting Acetivibrio can help further assess the importance of this genus to cellulose degradation in the rumen. The population of Ad-H1-75-1, which was assigned to Unclassified Clostridiales, was much higher in the Ad and the Sld fractions than the Lq fractions (Figure 4.4). Ad-H is also presumed to be involved in fiber degradation. Detailed physiology can only be gained through studies of pure cultures. A reverse 70

91 metagenomic approach, as demonstrated previously (Nichols, 2007; Pope et al., 2011), may be used to help isolate these uncultured bacteria. The metagenomic data recovered in previous studies of ruminal samples can also be used to design selective media for uncultured bacteria through its metabolic reconstruction, resulting in the isolation and characterization of the uncultured bacterium. 4.5 Conclusions In this study, the populations of uncultured bacteria were as great as those of major cultured bacteria except for Prevotella. They are also ubiquitous in the rumen. Uncultured bacteria may play as an important role as the cultured bacteria, if not more. Comparative dynamic studies of uncultured bacteria in response to dietary treatments might help further reveal their ecological niche and roles in the rumen. Isolation and characterization of uncultured bacteria in the rumen would need to be attempted to define the function of uncultured bacteria. 71

92 Primers Sequences (5 3 ) Target 27f 1525r 340f 806r TaqMan probe Bac303f Bac708r ARC787F ARC1059R Ra1281f Ra1439r Fs-f Fs-r 530f Buty-900r Sel-Mit-f Sel-Mit-r Ram-f Ram-r Rf154f-K Rf425r-K AGA GTT TGA TCM TGG CTC AG AAG GAG GTG WTC CAR CC TCC TAC GGG AGG CAG CAG T GGA CTA CCA GGG TAT CTA ATC CTG TT Annealing temperature ( o C) 6-FAM-5 -CGT ATT ACC GCG GCT GCT Total Bacteria GGCAC-3 -TAMRA 70 GAA GGT CCC CCA CAT TG CAA TCG GAG TTC TTC GTG ATT AGA TAC CCS BGT AGT CC GCC ATG CAC CWC CTC T CCC TAA AAG CAG TCT TAG TTC G CCT CCT TGC GGT TAG AAC A GGT ATG GGA TGA GCT TGC GCC TGC CCC TGA ACT ATC GTG CCA GCM GCC GCG G TGC GGC ACY GAC TCC CTA TG TGC TAA TAC CGA ATG TTG TCC TGC ACT CAA GAA AGA CAA CCA GTC GCA TTC AGA CAC TAC TCA TGG CAA CAT TCT GGA AAC GGA TGG TA CCT TTA AGA CAG GAG TTT ACA A Amplicon length (bp) References Total Bacteria 54 1,535 Larue et al., 2005 Bacteroides and Prevotella Nadkarni et al., 2002 Bartosch et al., 2004, Bernhard and Field, 2000 Total Archaea Yu et al., 2005 R. albus Koike and Kobayashi, 2001 F. succinogenes Tajima et al., 2001a Butyrivibrio S. ruminantium and M. multacida Stiverson et al., Tajima et al., 2001a R. amylophilus Tajima et al., 2001a R. flavefaciens Koike and Kobayashi, 2001 Table 4.1. Primers and a TaqMan probe used in the real-time PCR assays for total bacteria, total archaea or cultured bacteria (Reproduced from Stiverson et al., 2011) 72

93 Partial sequences (GenBank accession no.) Sequences (5 3 ) Annealing position (E. coli numbering) Amplicon length (bp) Ad-C1-74 (AY816616) GAA GGG ACC GGT TAA GGT C ,024 Lq-C2-16 (AY816578) GAC TTT GCT TCC CTT TGT TTT G ,258 Lq-C2-58 (AY816550) AGC CTC CGA TAC ATC TCT GC ,022 Ad-H1-14 (AY816508) GAT TTG CTT ACC CTC GCG GGT TT ,275 Ad-H1-75 (AY816420) CAC ACC TTG TAT CTC TAC AAG C ,020 Ad-H2-90 (AY816432) CTT CGA CAG CTG CCT CCT TA ,463 Table 4.2. Primers used in the real-time PCR assays for uncultured bacteria (Reproduced from Stiverson et al., 2011) 73

94 rrs copies/ ug DNA rrs copies/ ug DNA Total Archaea Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Fraction Total Bacteria Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Fraction Figure 4.1. Populations of total archaea and total bacteria in the rumen of cattle. Liquid (Lq), adherent (Ad) and solid (Sld) fractions from the cattle were combined based on the diet. C, 42% corn silage, 14% alfalfa hay, 6% cotton seed and 38% grain; H, mixed grass hay including mostly timothy hay. The error bars indicate the standard error of the means (n=2). 74

95 rrs copies/ ug DNA rrs copies/ ug DNA rrs copies/ ug DNA rrs copies/ ug DNA F. succinogenes R. albus Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C 10 0 Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Fraction Fraction R. flavefaciens Butyrivibio Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C 10 0 Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Fraction Fraction Figure 4.2. Populations of three major cellulolytic bacteria and Butyrivibrio spp. in the rumen. The sample labeling is the same as those in Figure 4.1. The error bars indicate the standard error of the means (n=2). 75

96 rrs copies/ ug DNA rrs copies/ ug DNA rrs copies/ ug DNA rrs copies/ ug DNA P. ruminicola Prevotella Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C 10 0 Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Fraction Fraction S. ruminantium R. amylophilus Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Fraction 10 0 Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Fraction Figure 4.3. Populations of major non-cellulolytic cultured bacteria in the rumen. The sample labeling is the same as those in Figure 4.1. The error bars indicate the standard error of the means (n=2). 76

97 rrs copies/ ug DNA rrs copies/ ug DNA rrs copies/ ug DNA rrs copies/ ug DNA rrs copies/ ug DNA rrs copies/ ug DNA Ad-C Lq-C Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C 10 0 Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Fraction Fraction Lq-C Ad-H Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C 10 0 Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Fraction Fraction Ad-H Ad-H Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C 10 0 Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Fraction Fraction Figure 4.4. Populations of uncultured bacteria originally identified from the rumen of sheep (Larue et al., 2005; Stiverson et al., 2011). The sample labeling is the same as those in Figure 4.1. The error bars indicate the standard error of the means (n=2). 77

98 CHAPTER 5 PHYLOGENETIC DIVERSITY OF BACTERIAL COMMUNITIES IN BOVINE RUMEN AS AFFECTED BY DIETS AND MICROENVIRONMENTS 5.1 Abstract Phylogenetic analysis was conducted to examine ruminal bacteria in two ruminal fractions (adherent fraction vs. liquid fraction) recovered from cattle fed with two different diets: forage alone vs. forage plus concentrate. One hundred forty-four 16S rrna gene (rrs) sequences were obtained from clone libraries constructed from the four samples. These rrs sequences were assigned to 116 different operational taxonomic units (OTUs) defined at 0.03 phylogenetic distance. Most of these OTUs could not be assigned to any known genus. The phylum Firmicutes was represented by approximately 70% of all the sequences. By comparing to the OTUs already documented in the rumen, 52 new OTUs were identified. UniFrac, SONS, and denaturing gradient gel electrophoresis (DGGE) analyses revealed difference in diversity between the two fractions and between the two diets. This study showed that rrs sequences recovered from small clone libraries can still help identify novel species-level OTUs. 5.2 Introduction A complex microbiome consisting of bacteria, archaea, fungi and protozoa has evolved to efficiently degrade various types of forages in the rumen (Flint, 1997), and a biofilm formed on the surface of the ruminal forages is especially important to feed digestion (Cheng et al., 1977). Defining the phylogenetic diversity of ruminal microbiome, particularly the bacterial community, is intriguingly interesting to many 78

99 microbiologists because it is essential to functional analysis of this microbiome. Stahl et al. (1998) primarily used rrs sequence-based analysis in identifying the phylogenetic diversity of ruminal bacterial community. Most studies focused on the microbes collected from rumen fluid (e.g. Tajima et al., 2000, 2007; Ozutsumi et al., 2005), though some studies examined the microbes present in separated liquid and solid fractions (Larue et al., 2005; Yu et al., 2006; Brulc et al., 2009) and on the rumen wall (Cho et al., 2006; Lukas et al., 2010). Based on a recent meta-analysis (Kim et al., 2011b), the rrs sequences of rumen origin that have been archived in the RDP database represent more than 3500 species-level OTUs. These OTUs represent approximately 70% of the global bacterial diversity estimated to be present in the rumen. Therefore, more studies are needed to further discover the phylogenetic diversity of ruminal bacterial communities. However, it is uncertain if typical clone library-based analysis can still identify new OTUs. In this study, we analyzed four rrs clone libraries constructed from liquid and adherent fractions recovered from two cows fed with different diets (forage alone vs. forage plus concentrate), defined species-level OTUs and compared the defined OTUs with existing ruminal OTUs reported recently (Kim et al., 2011b). This study enabled us to identify many novel ruminal OTUs. The difference in phylogenetic diversity between the two fractions and the two diets were also assessed using UniFrac, SONS and DGGE analyses. 5.3 Materials and Methods Sample collection, fractionation and DNA extraction Whole rumen content was collected from four cannulated cows: two Jersey cattle 79

100 fed with only forage composed predominantly of Timothy grass and two Holstein cattle fed with a typical dairy diet consisting of 14% alfalfa forage, 42% corn silage, 6% cottonseed, and 38% grains. The two groups of cows were fed twice daily (early morning and late afternoon) and adapted to their respective diets for more than 3 weeks before rumen sampling, which took place approximately 6 hours after the morning feeding. Both the liquid fraction and the adherent fraction of each rumen digesta were separated as reported previously (Larue et al., 2005). Bacteria present in the liquid fraction (Lq) were recovered by centrifugation, while bacteria adherent to the solid digesta were recovered by using a detaching buffer containing 0.15% (v/v) Tween-80 (Dehority and Grubb, 1980) and centrifugation. Metagenomic DNA was extracted from each of the fractionated samples as described previously (Yu and Morrison, 2004b). To recover bacteria that might fail to detach, the remaining solid digesta was also subjected to DNA extraction. The resultant metagenomic DNA extracts from the adhering fraction and the solid particles were combined and designated as the adherent fraction (Ad) DGGE analysis DGGE analysis was conducted to profile and compare the bacterial communities among the four samples using the GC-357f and 519r primer set targeting the V3 region, and band patterns were analyzed using the BioNumerics program as described previously (Yu and Morrison, 2004a) Construction of rrs clone libraries The DNA extracts were pooled based on diets and fractions, resulting in four 80

101 composite samples representing the liquid fraction and the adhering fraction recovered from the two cattle fed with forage alone (Lq-H, and Ad-H, respectively) and from the two cattle fed with forage plus concentrate (Lq-C, and Ad-C, respectively). The nearly full-length rrs gene was amplified by PCR from each composite DNA sample using universal bacterial primers 27F (5 -AGAGTTTGATCMTGGCTCAG-3 ) and 1525R (5 - AAGGAGGTGWTCCARCC-3 ) and cloned using a TOPO-TA cloning kit (Invitrogen, Carlsbad, CA). Three hundred and eighty four random clones from the four libraries (96 clones per library) were subjected to screening for the presence of the insert using PCR (Yu and Mohn, 2001) Restriction fragment length polymorphism (RFLP) analysis, DNA sequencing and phylogenetic analysis To reduce sequencing redundancy of similar clones, the PCR products of the aforementioned screening were digested using both HaeIII and AluI as described previously (Larue et al., 2005). The RFLP patterns were compared using BioNumerics (BioSystematica, Tavistock, Devon, United Kingdom), and the clones with a unique RFLP pattern were sequenced using the 27F primer at High-Throughput Genomics Unit (University of Washington, Seattle, WA). Low-quality sequence regions at both ends of each sequence read were trimmed off using the FinchTV program V1.4 (Geospiza, Inc., Seattle, WA). All the sequences were then subjected to chimera check using the Pintail program (Ashelford et al., 2005), and suspected chimeric sequences were excluded from further analysis. Classification, alignment, and construction of a taxonomic tree at genus level were performed using the 81

102 bioinformatic programs as described previously (Kim et al., 2011b). Using the Mothur program (Schloss et al., 2009), the species-level OTUs (defined at 0.03 phylogenetic distance) identified in this study were compared to all the OTUs recognized in the recent meta-analysis of global diversity of ruminal bacteria (Kim et al., 2011b) Comparison of ruminal bacterial communities among the four composite samples Principal coordinate analysis (PCA) was conducted to examine relationship among the four composite samples using the UniFrac program (Lozupone and Knight, 2005). The SONS program (Schloss and Handelsman, 2006) was used to determine the numbers of shared OTUs across the four samples Nucleotide sequence accession numbers The rrs sequences obtained in this study have been deposited in the GenBank database (JF JF319441). 5.4 Results and Discussion The four composite samples were first compared using DGGE to visualize the impact on the ruminal bacterial communities from the diets and the fractions. As shown in Figure 5.1, the four samples shared a number of DGGE bands, but bands distinct to each sample were also evident. Clustering analysis showed that both the diets and the microenvironments (liquid vs. solid fractions) had affected the ruminal bacterial communities, with the diets having a greater impact than the fractions. These results are consistent with a previous study where both corn supplementation and fractions were 82

103 found to affect the ruminal bacterial communities in sheep (Larue et al., 2005). The bacterial communities in the composite samples were further examined and compared using the rrs clone libraries. From the 384 clones screened for the presence of insert, 288 clones were found to produce a unique RFLP pattern, which were then sequenced. After removing the sequences of low quality and suspected chimeric sequences, 144 high-quality rrs sequences were obtained and analyzed phylogenetically. The phyla Firmicutes, Bacteroidetes, Proteobacteria, Spirochaetes, and Verrucomicrobia were represented by 140 sequences, and the remaining four sequences could not be classified into any existing phylum (Figure 5.2). Firmicutes was the most predominant phylum and accounted for 69.4% of all the 144 sequences. Approximately 60.2% of the Firmicutes sequences could not be classified into any known family or genus, with Unclassified_Lachnospiraceae being the most abundant (27 sequences) group. Genera Butyrivibrio, Ruminococcus, and Succiniclasticum each were represented by more than nine sequences, while the remaining genera identified were represented by no more than five sequences each. The 144 sequences were assigned to 116 species-level OTUs, of which 52 appeared to be novel when compared to the existing ruminal OTUs reported previously (Kim et al., 2011b), indicating new ruminal species. This is the first study that compared OTUs identified in individual studies to those already documented. The results of this study also demonstrated that small numbers of rrs sequences could still contribute towards identifying the full phylogenetic diversity of ruminal bacteria. Future individual studies using different PCR primers, sampling methods, and DNA extraction techniques would help improve discovery of novel OTUs and eventually lead to full coverage of phylogenetic diversity (Edwards et al., 2004; Hong et al., 2009). 83

104 A similar number of species-level OTUs was identified from all the four samples, with Ad-H, Ad-C, Lq-H, and Lq-C fractions having 31, 29, 34, and 31 OTUs, respectively. However, based on comparison using the SONS program, three or fewer OTUs were shared between any two of the four composite samples, and no OTU was shared by all the four samples (Figure 5.3). Ten of the 31 Ad-H OTUs were classified into known genera, with Ruminococcus (4 OTUs) being the most abundant. The Ad-C fraction also included 10 OTUs classified into existing genera, with each genus being represented by one OTU except Ruminococcus (2 OTUs) and Succiniclasticum (2 OTUs). As expected, Ruminococcus was predominant in the adherent fraction recovered from the rumen of cattle fed with only forage. However, most of the OTUs from both the Ad-H and the Ad-C fractions could not be assigned to any known genus (21 and 19 OTUs, respectively). Eight OTUs could only be classified to the order Clostridiales in the Ad-H fraction, while another seven OTUs were only assigned to the family Lachnospiraceae in the Ad-C fraction. These OTUs might represent new families or genera. The distinct distribution of these two groups of bacteria is probably attributed to dietary effect. Future studies using quantitative analysis are needed to confirm and help explain this finding. The Lq-H and the Lq-C fractions contained 8 and 10 OTUs (of the 34 and 31 OTUs) assigned to known genera, respectively. The most numerous single genus in the Lq-H sample was Butyrivibrio (4 OTUs), while each genus in the Lq-C sample was represented by only one of the 10 OTUs except Butyrivibrio (2 OTUs) and Treponema (2 OTUs). It appeared that versatile Butyrivibrio species that can degrade several types of polysaccharides (hemicellulose, starch, and pectin) are abundant in the liquid fraction irrespective of the diets. The two OTUs assigned to Treponema are thought to be 84

105 associated with diets rich in concentrate as described previously (Bekele et al., 2011). Unclassified_Bacteroidetes (7 OTUs) and Unclassified_Clostridiales (6 OTUs) were the first and second most abundant unclassified groups in the Lq-H sample, whereas Unclassified_Lachnospiraceae (8 OTUs) was the most abundant unclassified group in the Lq-C sample. Again, Unclassified_Clostridiales was predominant in the adhering fraction of the forage-fed cattle, whereas Unclassified_Lachnospiraceae was predominant in both fractions of cattle fed with both forage and concentrate. As noted previously (Kim et al., 2011b), Unclassified_Clostridiales rather than Unclassified_Lachnospiraceae seems to be associated with fiber digestion. Based on the PCA analysis, the P1 separated the bacterial communities based on the diets, whereas the P2 separated the ruminal bacterial communities based on the fractions (Figure 5.4). This result agrees with the DGGE data and the sequence-based comparison by SONS. However, the DGGE data showed a greater similarity among the samples than the Venn diagram. This discrepancy is probably due to the limited resolution of DGGE. These results suggest significant impact on the ruminal bacterial community from diets and microenvironments (liquid vs. solid surface), although difference between the two breeds may also have some effect on the ruminal bacterial community. The effects of diets on the ruminal bacterial community have been investigated in several studies (e.g. Tajima et al., 2000; Larue et al., 2005). The partition of the bacterial populations between the solid and the liquid fractions have also been examined (Michalet-Doreau et al., 2001; Larue et al., 2005). In all the studies reported (including this present study), the same OTUs were not commonly found between diets or fractions. More studies are needed to define the core and variable bacterial 85

106 communities in the rumen. As demonstrated recently (Pitta et al., 2010), studies using comprehensive analysis may help achieve such a goal. 86

107 Figure 5.1. Clustering analysis of DGGE banding profiles based on the V3 region of 16S rrna genes. Lq-C and Ad-C represent the liquid fraction and the adhering fraction recovered from cattle fed with forage plus concentrate, while Lq-H and Ad-H represent the liquid fraction and the adhering fraction recovered from cattle fed with forage alone. The dendrogram was constructed using the BioNumerics program. 87

108 Figure 5.2. A taxonomic tree showing the bacterial genera represented by the 144 sequences. The lineage at phylum level is labeled: V, phylum Verrucomicrobia; B, phylum Bacteroidetes; S, phylum Spirochaetes; P, phylum Proteobacteria; F, phylum Firmicutes. In total, 14 known genera were represented by 49 sequences, while the remaining 95 sequences could not be assigned to any existing genus. Numbers in parenthesis indicate the number of sequences from respective fraction. 88

109 Figure 5.3. A Venn diagram showing the numbers of species-level OTUs shared among the four composite samples. The four samples had 116 species-level OTUs in combination. 89

110 Figure 5.4. A PCA analysis plot comparing the bacterial communities in the four composite samples. 90

111 CHAPTER 6 DEVELOPMENT OF A PHYLOGENETIC MICROARRAY FOR COMPREHENSIVE ANALYSIS OF RUMINAL MICROBIOME 6.1 Abstract: Phylogenetic microarray is a powerful tool that enables simultaneous detection and semi-quantitation of thousands of different members of a microbiome. The objective of this study was to develop a microarray to support comprehensive analysis of a complex ruminal bacteriome. All the 16S rrna gene (rrs) sequences of rumen origin collected worldwide were retrieved from the RDP database and subjected to phylogenetic analysis. The rrs sequences assigned to each genus based on the new Bergey s Taxonomy were aligned and assigned to species-equivalent operational taxonomic units (OTUs) at a 0.03 phylogenetic distance. One representative sequence was selected from each OTU, and one specific GoArray probe was designed for each OTU. The specificity of the probes was verified in silico using the Probe Match function in the RDP database. The specificity, sensitivity, and linear range of detection are determined using pools of rrs clones of known sequences. Of approximately 2,500 OTUs identified, 1,664 OTUs that include 10 OTUs obtained from the known clones were targeted by a specific GoArray probe on the ruminal microarray (referred to as RumenArray). In addition, 2 GoArray probes specific to the bovine mitochondrial rrs sequence were added to the RumenArray as internal controls. The RumenArray has been custom fabricated in a 6x5k format with each probe being represented in triplicates. The RumenArray can detect approximately copies of targets and had a linear dynamic range of >3 orders of magnitude. Fractionated rumen samples (liquid fraction vs. adherent fraction) obtained from sheep 91

112 fed two different diets (hay alone vs. hay plus corn) were used to test the utility of the RumanArray. More than 300 different OTUs were detected in combination across the four fractions. Prevotella was the most numerous among known genera. Unclassified groups represented more than 50% of all the OTUs detected. This is the first phylochip dedicated to analysis of ruminal bacteria. It enables comprehensive semi-quantitative analysis of ruminal bacteria in support of nutritional studies of ruminant animals. 6.2 Introduction: A complex microbiome consisting of bacteria, archaea, protozoa and fungi in the rumen degrades various feedstuffs ingested by ruminant animals. Within the ruminal microbiome, Bacteria are the most abundant domain and have a predominant role in degrading ingested feedstuffs. Dietary manipulations have attempted to optimize ruminal bacterial fermentation (Calsamiglia et al., 2007; Weimer et al., 2008). Cultivation-based methods dominated attempts to investigate ruminal bacteria and elucidated important metabolic functions in the rumen until the 1980 s. However, most ruminal bacteria could not be isolated due to limitations of cultivation-based methods (Whitford et al., 1998). Contemporary molecular biology techniques overcome the limitations of cultivation-based methods and now use rrs sequence-based methods in examining diversity of ruminal bacteria. Construction of rrs clone libraries has contributed greatly to understanding the diversity of ruminal bacteria (Tajima et al., 2000; Ozutsumi et al., 2005; Zhou et al., 2009). Some studies used fractionated rumen contents (liquid fraction vs. adherent fraction) to investigate diversity of ruminal bacteria as affected by microenvironments (Larue et al., 2005; Yu et al., 2006; Brulc et al., 2009; Kim et al., 92

113 2011c), and this fractionation method contributed to understanding the niches of bacterial species within the rumen (Larue et al., 2005; Kim et al., 2011c). The PCR-DGGE technique was also used to examine bacterial diversity in the rumen and contributed to assessment of dietary effects in nutritional studies. This technique, however, provides limited information on the microbes that are affected by the dietary manipulations. Realtime PCR assays have been used in quantitatively determine the effects of diets on selected species or genera of bacteria or methanogens, including uncultured strains represented by novel rrs sequences (Stiverson et al., 2011). However, only a small number of species or genera of microbes can be practically analyzed in this manner. This limitation hinders comprehensive assessment of any dietary manipulation on the ruminal microbiome. As shown in a recent study by Kim et al. (2011b), the ruminal microbiome likely contains hundreds of species, and the cultured microbes only account for less than 7% of the entire ruminal microbiome. Therefore, new techniques are needed that support comprehensive and quantitative analysis of ruminal microbiome, which is required for simultaneous analysis of large numbers of ruminal samples collected in nutritional studies. Phylogenetic microarrays have been used as a powerful tool that enables simultaneous detection and semi-quantitation of thousands of different rrs. Although the phylogenetic microarray technique has been applied to investigation of bacteria present in various environments such as soil, human gut, human feces, sludge and lake (Adamczyk et al., 2003; Castiglioni et al., 2004; Kang et al., 2010; Palmer et al., 2006; Small et al., 2001), no phylogenetic microarray has been developed to examine ruminal bacteria. The objective of this study was to develop a phylochip dedicated to analysis of ruminal 93

114 bacteria. The utility of the microarray we developed (referred to RumenArray) was tested in examining the ruminal bacteria in the adherent and the liquid fractions recovered from sheep fed two different diets: hay alone vs. corn plus hay. 6.3 Materials and Methods: Oligonucleotide probe design and microarray fabrication: Approximately 10,000 rrs sequences of rumen origin collected worldwide were retrieved from the RDP database (Release 10, Update 5) and subjected to phylogenetic analysis as described previously (Kim et al., 2011b). The sequences were classified into respective genera based on the Bergey s Taxonomy implemented in the RDP database. Sequences within each genus were aligned using the Geneious program (Auckland, New Zealand) and assigned to species-level OTUs at 0.03 phylogenetic distance. Approximately 2,500 OTUs were identified. One specific probe was designed from the representative sequence selected from each OTU using the GoArray program (Rimour et al., 2005). Another ten GoArray probes were designed from sequenced clones (Kim et al. 2011c) and included into the RumenArray to assist in determination of specificity, sensitivity and detection limit. In addition, two probes designed from the bovine mitochondrial rrs sequences were added into the RumenArray to serve as internal controls. Each GoArray probe consisted of two short separated sub-probes (17 mers) and a short linker (6 mers) inserted between the two sub-probes. The resultant GoArray probe length was 40 nt. The specificity of the probes was verified in silico using the Probe Match function in the RDP database. The final 1,666 GoArray probes including the 10 clone probes and the 2 internal control probes were synthesized onto a 6 5K custom 94

115 oligonucleotide microarray by MYcroarray (Ann Arbor, MI), with each probe represented in triplicates Sample collection, fractionation and DNA extraction: Metagenomic DNA samples recovered from sheep fed with two different diets (100% orchard grass hay vs. a combination of 70% orchard grass hay and 30% corn) were provided by Stiverson et al. (2011). Briefly, four sheep were assigned to two groups of two sheep, and a repeated switchover design was used. The liquid (Lq) and the adherent (Ad) fractions of each sample were recovered as described previously (Larue et al., 2005). Metagenomic DNA was extracted from each of the sixteen samples (2 sheep x 2 diet x 2 fraction in a repeated switchover design = 16) using the RBB+C method (Yu and Morrison, 2004b). Fifteen of the 16 metagenomic DNAs were used for RumenArray analysis due to lack of one metagenomic DNA recovered from one sheep fed with hay Sample preparation and labeling: The nearly full-length 16S rrna genes were amplified from each metagenomic DNA sample using the universal primer set 27F (5 -AGA GTT TGA TCM TGG CTC AG-3 ) and T7/1492R (5 -TCT AAT ACG ACT CAC TAT AGG GGG YTA CCT TGT TAC GAC TT-3 ) as described previously (Palmer et al., 2006; Kang et al., 2010). PCR was performed with 30 cycles (denaturation, 95 o C for 30 s; annealing, 55 o C for 30 s; and extension, 72 o C for 90 s) using a PTC-100 thermocycler (MJ Research, Waltham, Mass). All amplicons were purified using a PCR purification Kit (Qiagen, Valencia, CA, USA), and then the purified amplicons were used as templates to synthesize single-stranded 95

116 complementary RNA (crna) using a MEGAScript T7 in vitro transcription kit (Ambion, Austin, TX, USA). After crna was purified, they were then labeled with Cy5 at 37 o C for 1 hour using the IT uarray Cy5 reagent (Mirus, Madison, WI, USA). The labeled crna was purified to remove the free Cy5 dye using the MEGAclear kit. The labeled crna was quantified using the NanoDrop TM 1000 and then stored at -80 o C until microarray hybridization Microarray hybridization: Microarray hybridization was performed using Agilent Technologies Hybridization gasket slides that are compatible with the MYcroarray slides. The hybridization solution containing 6X SSPE, 0.05% Tween-20, 0.01 mg/ml acetylated BSA, 10% formamide and 1.2 µg labeled crna was incubated at 65 o C for 5 min and then placed on ice for at least 5 min. The Agilent Hybridization Cassette, the Agilent gasket slide and the MYcroarray slide were preheated at 65 o C while the hybridization solution was prepared. The hybridization solution was added to the center of the Agilent gasket slide, and then the MYcroarray slide was placed over the gasket slide. The cassette was placed in HB-1000 hybridization oven (UVP, LLC), and then hybridization was performed for 18 hours at preselected temperatures with rotation set at 10 rpm Signal detection and data analysis: An Axon Genepix 4000B scanner (Axon Instruments, Union City, CA) was used to scan the microarray slides at 100% laser power, PMT photomultiplier sensitivity, and 5 µm resolution. The GenePix Pro 6.0 program (Axon Instruments, Union 96

117 City, CA) was used to analyze the images and create GenePix Results Format (GPR) files. Median signal intensity that was transformed to log2 was used for microarray analysis, and normalization was performed based on the signal of internal control probes targeting the bovine mitochondrial rrs region. Spots that have less than the signal-tonoise ratio (SNR) value of 3.0 were removed for accurate quantification as described previously (He et al., 2007). Significant differences between samples were examined by one-way ANOVA using the MeV program within the TM4 microarray software suite (Saeed et al., 2006). Principal component analysis (PCA) using the MeV program (Saeed et al., 2006) was conducted to compare the fractionated samples Determination of the specificity and detection limit: Ten rrs clones corresponding to the 10 clone probes (referred to as positive clones ) and another sixteen rrs clones corresponding to none of the probe (referred to as negative clones ) were also used to evaluate the specificity. PCR products were amplified from each of the clones and then used to synthesize crna as described above. A pool of the 10 positive crna and a pool of the 16 negative crnas were labeled with Cy5 in separate labeling reactions and then subjected to microarray hybridization using separate microarrays. Positive signals for respective crna pools were analyzed and compared at hybridization temperatures of 42 o C, 45 o C and 47 o C. In addition, 6 of the 10 positive clones were selected to determine the detection limit and dynamic range of detection. The pools of the 6 positive crnas were serially diluted (from rrs copies to 10 6 rrs copies) and then subjected to microarray hybridization using individual microarrays. 97

118 6.3.7 Comparison between microarray and real-time PCR data: Real-time PCR data obtained by using the same sheep samples (Stiverson et al., 2011) were compared to RumenArray data. Stiverson et al. (2011) used one-way analysis of variance (ANOVA) to analyze the fraction and diet-based data. Our study used the analysis, as implemented in the MeV program (Saeed et al., 2006), in analyzing the microarray data. 6.4 Results and Discussion: Validation of the specificity, sensitivity and detection limit: Cy5-labeled crna obtained from the 10 positive rrs clones was hybridized to the RumenArray slide at 42 o C, 45 o C or 47 o C. Twenty-one false positives were detected at 42 o C, whereas 12 and 7 false positives were detected at 45 o C or 47 o C, respectively. Although the number of false positives was reduced at 47 o C compared to 45 o C, one false negative was found at 47 o C. Another crna sample prepared from the 16 negative clones was also used to examine the specificity at the three different temperatures. Twenty false positives were detected at 42 o C, whereas 9 and 8 false positives were detected at 45 o C and 47 o C, respectively. Because one false negative in the specificity test using the 10 positive rrs clones was found at 47 o C, we selected 45 o C as the optimal hybridization temperature. However, the RumenArray is not completely specific based on the high number of false positive signals in the two specificity tests. Because only approximately 500 bp of the 1500 bp full-length sequence was sequenced for the clones used in the specificity test, the region that was not sequenced is probably hybridized to the 98

119 RumenArray. The Cy5-labeled crna synthesized from pools of six positive rrs clones were used to determine the detection limit at 45 o C. The rrs copies of Cy5-labeled crna were serially diluted, and then each diluted Cy5-labeled crna was hybridized to the RumenArray. The signal intensity was normalized using the signal intensity of the crna of bovine mitochondrial rrs. The lowest rrs copies that showed positive signals were approximately 10 7 (Figure 6.1). When rrs copies were hybridized to the RumenArray slide, some of the 6 positive rrs clones did not show the positive signal. Thus, the RumenArray has a dynamics range of at least three orders of magnitude ( rrs copies). The detection limit indicates that the RumenArray is not as sensitive as real-time PCR in investigating ruminal bacteria with low abundance Data summary In total, 319 OTUs were detected in combination across the four sampled fractions (Figure 6.2). The Lq-C, Lq-H, Ad-C, and Ad-H fractions had 186, 243, 159, and 176 detected OTUs, respectively. The number of total shared OTUs was 91, and each fraction had unique OTUs which ranged from 11 to 58 (Figure 6.2). These unique OTUs indicate that different diets or fractions affect ruminal bacterial populations as described previously (Larue et al., 2005; Kim et al., 2011c). Even though four sheep were fed the same diet, signal intensities were different among the four sheep (Figure 6.3). It seems that the genetic difference between individual sheep affects the populations of predominant bacteria. However, the similarity was higher between the sheep fed the same diet than between the sheep fed different diets or fractions (Figure 6.3). The number of 99

120 detected OTUs was greater in the Lq fractions than in the Ad fractions probably because the majority of sequence dataset used for microarray analysis were recovered from the liquid fraction. Novel sequences will need to be recovered from the Ad fraction for future studies Diversity of ruminal bacteria assigned to known genus: The RumenArray analysis showed that Prevotella was the most abundant genus among known genera, and it was represented by 47 OTUs (of all the 319 OTUs detected) in combination across all the fractions. However, all the 47 OTUs were represented by only uncultured bacterial sequences. This result agrees with the previous findings with respect to the predominance and diversity of Prevotella (Bekele et al., 2010; Kim et al., 2011b; Stevenson and Weimer, 2007). Thirty-one of the 47 OTUs did not show any significant difference among the four fractions but slightly more abundant in sheep fed corn:hay than in sheep fed hay alone. Another twelve OTUs were detected only in sheep fed hay, whereas another four OTUs were detected only in sheep fed corn:hay. Ruminococcus is one of the frequently cultured cellulolytic ruminal bacteria, and it can rapidly attach to the surface of plant materials to digest cellulose in the rumen (Koike et al., 2003b). Ruminococcus was the second most numerous genus among the detected genera, and 12 OTUs within this genus were detected. Three of the 12 OTUs showed significant difference among the fractions (Figure 6.3). Ruminococcus OTU 1 (S , RDP ID) recovered originally from Holstein cattle (unpublished sequence data, GenBank records) was more abundant (P < 0.05) in the Ad-H and the Lq-H fractions than in the Ad-C and the Lq-C fractions. Another 3 OTUs were detected only in 100

121 sheep fed hay, and they may be associated with cellulose degradation. Ruminococcus OTU 2 (S , RDP ID) recovered originally from sheep (unpublished sequence data, GenBank records) was more abundant (P <0.05) in the Lq-C fraction than in the other fractions. Ruminococcus OTU 3 (S , RDP ID) recovered originally from Holstein cattle (unpublished sequence data, GenBank records) was more abundant (P <0.05) in the Lq-C fraction than in the Ad-C and the Lq-H fractions. The Ruminococcus OTU 2 and 3 may be associated with amylolytic Ruminococcus spp. such as R. bromii, R. callidus, R. hydrogenotrophicus, R. obeum and R. schinkii (Larue et al., 2005). Our study also showed that R. bromii (S , RDP ID) recovered originally from cattle (Klieve et al., 2007) was detected only in sheep fed corn:hay. Another 3 OTUs were also detected only in sheep fed corn:hay, and they are presumed to be amylolytic. The other genera were each represented by less than 8 OTUs. Seven Butyrivibrio OTUs were detected, and one of them (S , RDP ID) recovered originally from cattle (unpublished sequence data, GenBank records) was less abundant (P < 0.05) in the Lq-H fraction than in the other fractions (Figure 6.3). Another one Butyrivibrio OTU (S , RDP ID) recovered first from reindeer grazing on natural summer pasture (Sundset et al., 2007) was detected only in sheep fed hay. On the other hand, another two Butyrivibrio OTUs were detected only in sheep fed corn:hay, and one of the two OTUs (S , RDP ID) was recovered from acidosis Holstein cattle (Tajima et al., 2000). Three Acetivibrio OTUs were detected and two of them were detected only in the Ad-H fraction. Acetivibrio is largely a genus consisting of cellulolytic bacterial species that were formerly classified as Clostridium spp. These include Clostridium thermocellus, C. hungatei, C. cellulolyticum, C. termitidis, C. stercorarium, A. cellulolyticus, and A. 101

122 cellulosolvens. The greater abundance of Acetivibrio in the Ad-H fraction suggests a potentially major role which this genus may play in fiber degradation in the rumen. Oxalobacter formigenes (S , RDP ID) that can degrade oxalate, which is contained in plants (Allison et al., 1985), was detected only in the Ad-H fraction. Three Sporobacter OTUs were detected and two of them were detected only in sheep fed hay. Because Sporobacter isolated from wood-feeding termites is involved in degradation of aromatic compounds of lignocellulose (Grech-Mora et al., 1996), ruminal Sporobacter may be beneficial to fiber degradation. Henderson (1971) reported that Anaerovibrio lipolytica isolated from rumen produces lipase. Two ruminal Anaerovibrio OTUs detected in this study are thought to be a major source of lipase. Olsenella umbonata A2, also a lactic acid bacterium (Kraatz et al., 2011), was detected only in sheep fed corn:hay. Selenomonas ruminantium K2 (S , unpublished sequence data) and Selenomonas bovis (S , RDP ID) (Zhang and Dong, 2009) were detected only in sheep fed corn:hay. Lachnobacterium bovis (S , RDP ID) that ferments glucose and produce primarily lactic acid (Whitford et al., 2001b) was detected only in sheep fed corn:hay. Amylolytic Ruminobacter (S , RDP ID) was detected in sheep fed corn:hay but not in sheep fed hay. Megasphaera elsdenii (S , RDP ID) is a lactate-utilizing bacterium and mainly found in ruminant animals fed high grain diets (Ouwerkerk et al., 2002), and it was more abundant (P <0.05) in sheep fed corn:hay than in sheep fed hay. The RumenArray analysis did not detect Fibrobacter OTUs. This finding agrees with the results obtained by using rrs clone libraries and pyrosequencing analysis as described in Chapter 5 and 8. As shown in Chapter 4, real-time PCR assay showed the 102

123 abundance of Fibrobacter. This discrepancy indicates that commonly used universal primers cannot efficiently amplify Fibrobacter rrs sequences (Larue et al., 2005). Except for Fibrobacter, it is not known if other genera also fail to be detected due to the same reason. Direct labeling and hybridization of rrna will eliminate this limitation Diversity of ruminal bacteria that are not assigned to any known genus: Unclassified Clostridiales (U_Clostridiales), unclassified Ruminococcaceae (U_Ruminococcaceae) and unclassified Lachnospiraceae (U_Lachnospiraceae) within the phylum Firmicutes were detected in high abundance in the RumenArray analysis. This result supports the prevalence of sequences assigned to these unclassified groups as discussed previously (Kim et al., 2011b). In total, 28 OTUs classified to U_Clostridiales were detected, and two of them showed significant difference among the fractions (Figure 6.3). U_Clostridiales OTU 1 (S , RDP ID) recovered first from cattle (unpublished sequence data, GenBank records) was more abundant (P < 0.05) in sheep fed hay than in sheep fed corn:hay. U_Clostridiales OTU 2 (S , RDP ID) recovered first from semi-continuous RUSITEC (unpublished sequence data, GenBank records) was also more abundant (P < 0.05) in sheep fed hay than in sheep fed corn:hay. Another 13 OTUs were detected only in sheep fed hay, and they were recovered from cattle fed with hay (Brulc et al., 2009), swamp buffaloes fed with rice straw (Yang et al., 2010a), yaks fed with pelleted lucerne (Yang et al., 2010b), or cattle fed with 66% forage (Ozutsumi et al., 2005), suggesting that these 13 OTUs are ubiquitous and play important role in fiber degradation. On the other hand, another 4 OTUs were detected only in sheep fed corn:hay. 103

124 Thirty-four OTUs assigned to U_Ruminococcaceae were detected in the RumenArray analysis, and five of them showed significant difference among the fractions (Figure 6.3). U_Ruminococcaceae OTU 1 (S , RDP ID), recovered first from reindeer grazing on natural summer pasture (Sundset et al., 2007), was higher (P <0.05) in abundance in sheep fed hay than in sheep fed corn:hay. U_Ruminococcaceae OTU 2 (S , unpublished sequence data) and U_Ruminococcaceae OTU 3 (S , RDP ID), both recovered from cattle (Ozutsumi et al., 2005), were more abundant (P <0.05) in sheep fed hay than in sheep fed corn:hay. U_Ruminococcaceae OTU 4 (S , RDP ID), recovered from Yunnan yellow cattle in China (unpublished sequence data, GenBank records), was more abundant (P <0.05) in the Lq- H and the Lq-C fractions than in the Ad-C fraction. U_Ruminococcaceae OTU 5 (S , RDP ID), recovered first from a goat (unpublished sequence data, GenBank records), had a higher abundance (P <0.05) in sheep fed corn:hay than in sheep fed hay. Another 11 OTUs were detected only in the sheep fed hay, of which 9 OTUs were also recovered from cattle fed hay (Brulc et al., 2009). These organisms may be associated with fiber degradation in both sheep and cattle. Forty-three OTUs representing U_Lachnospiraceae were detected in the RumenArray analysis, and one of them showed significant difference among the fractions (Figure 6.3). U_Lachnospiraceae OTU 1 (S , RDP ID), first recovered from cattle fed hay (Brulc et al., 2009), had a higher abundance (P <0.05) in the sheep fed hay than in the sheep fed corn:hay. U_Lachnospiraceae OTU 2 (S , RDP ID) named Cellulosilyticum ruminicola, which is a cellulolytic bacterium isolated from a yak (Cai and Dong, 2010), was detected only in the sheep fed hay. U_Lachnospiraceae OTU 104

125 3 (S , RDP ID) named Eubacterium rangiferina, which is an usnic acid resistant bacterium isolated from the reindeer rumen (Sundset et al., 2008), was detected only in the sheep fed corn:hay. Another 11 OTUs were detected only in the sheep fed hay and recovered from cattle fed hay (Brulc et al., 2009), reindeers grazing on natural summer pasture (Sundset et al., 2007), cattle fed 66% chopped Sudangrass hay (Ozutsumi et al., 2005), or cattle fed TMR including 65% forage (Whitford et al., 1998). These OTUs may represent ubiquitous cellulolytic bacteria. Another 9 OTUs were detected only in the sheep fed corn:hay and first recovered from cattle fed hay (Brulc et al., 2009), cattle fed 66% hay (Tajima et al., 2001a), swamp buffaloes fed rice straws (Yang et al., 2010a), or identified from the rumen of cattle or sheep (unpublished sequence data, GenBank records). The RumenArray analysis also detected two unclassified groups: unclassified Bacteroidales (U_Bacteroidales) and unclassified Prevotellaceae (U_Prevotellaceae) within the phylum Bacteroidetes. We detected 27 OTUs classified to U_Bacteroidales of which most were detected in a slightly higher abundance in the sheep fed corn:hay than in the sheep fed hay, although they showed no significant difference among the fractions. These OTUs might have amylolytic activity, or at least be stimulated by corn in the diet. U_Prevotellaceae had 18 OTUs detected, and one of them showed significant difference among the fractions. U_ Prevotellaceae OTU 1 (S , RDP ID) recovered from sheep fed hay (Larue et al., 2005) was more abundant (P <0.05) in the sheep fed hay than in the sheep fed corn:hay, and another 6 U_Prevotellaceae OTUs were detected only in sheep fed hay. These 7 OTUs may be associated with fiber degradation. Another one U_Prevotellaceae OTU was detected only in sheep fed corn:hay. 105

126 The abundance of unclassified Clostridiales, unclassified Ruminococcaceae, and unclassified Lachnospiraceae agrees with the previous meta-analysis (Kim et al., 2011b). Numerous uncultured bacteria assigned to these three groups are thought to play an important role in the fermentation of the rumen. These uncultured bacteria will need to be isolated and characterized. A reverse metagenomic approach, as described previously (Nichols, 2007; Pope et al., 2011), may help successfully isolate these uncultured bacteria PCA for comparison between fractions PCA, as implemented in the MeV program, showed that PC2 separates the ruminal bacteria of sheep fed hay from sheep fed hay:corn as shown in Figure 6.4. However, bacterial communities did not seem to be affected greatly by the diets because PC2 explained only 9.8% variance. On the other hand, PC1 explaining 51.7% variance is thought to separate the 15 fractions based on the number of detected OTUs observed in each fractionated samples. No clear separation was seen based on fractions (Figure 6.4) Comparison of RumenArray and real-time PCR data: The rumen samples analyzed in the RumenArray analysis previously had been analyzed by real-time PCR (Stiverson et al., 2011). In that study (Stiverson et al., 2011), the abundance of select cultured and uncultured bacteria were quantified. However, only one OTU was analyzed by both methods. This OTU, termed Unclassified Clostridia 1 in the RumenArray, is the uncultured ruminal bacterium Ad-H (S , RDP ID) (Stiverson et al., 2011). It was recovered from the adherent fraction of sheep fed hay and assigned to class Clostridia. The abundance of this OTU was higher (P <0.05) in the 106

127 Lq-H fraction than in the other fractions (Figure 6.3), consistent with the finding by Stiverson et al. (2011). The relative abundance of this OTU was similar between the realtime PCR and the RumenArray data except for the Ad-C fractions. The relative abundance of this OTU in the Ad-C fraction was 2.5-times greater in the RumenArray data than in real-time PCR data. 6.5 Conclusions: The RumenArray developed in this study supports simultaneous and rapid analysis of many predominant ruminal bacteria, both cultured and uncultured. The preliminary analysis showed that some OTUs are primarily associated with hay-fed animals, while others are associated with animals that received corn. Some OTUs also partition between the liquid and the solid fractions. The RumenArray analysis showed that unclassified Lachnospiraceae, unclassified Ruminococcaceae, and unclassified Clostridiales were more abundant than the others, supporting the meta-analysis data (Kim et al., 2011b). The RumenArray can be an alternative method for detection and semiquantification of abundant ruminal bacteria in a comparative manner. 107

128 Figure 6.1: Linear range of detection of the RumenArray as determined using crna pools of the 6 positive clones. 108

129 Lq-C 25 Lq-H Ad-C Ad-H Figure 6.2: A venn diagram showing the number of detected OTUs. The number of the total detected OTUs in combination across all the four fractions was 319, and 91 of them were shared irrespective of diets or fractions. 109

130 Figure 6.3: A hierarchical tree showing signal intensities and similarity among the fractionated samples. The hierarchical tree was constructed using the MeV program (Saeed et al., 2006). Fifteen of all the 319 OTUs showed significant difference among the four fractions. Signal intensities were also different among four sheep fed the same diet. 110