<<

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

ABSTRACT

Ruminant animals obtain most of their from products produced by ruminal microbiome consisting of , , protozoa and fungi. In the ruminal microbiome, bacteria are the most abundant 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 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 predominant were and , while the largest archaeal phylum was . More than 50% of all the bacterial sequences could not be classified into any known . 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, 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 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

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 1995. 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

VITA

April 1977 ...... Born-Kwangju, Republic of Korea

2002 ...... B.A., Department of Animal Sciences, Seoul

National University, Republic of Korea

2002 to February 2004 ...... M.S., Department of Animal Sciences,

Seoul National University, Republic of

Korea

March 2004 to August 2006 ...... 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., 2011. Phylogenetic diversity of bacterial communities in bovine rumen as affected by diets and microenvironments. Folia Microbiologica, DOI 10.1007/s12223-011-0066-5

Kim, M., Morrison, M., Yu, Z., 2011. Status of the phylogenetic diversity census of ruminal microbiomes. FEMS Microbiology Ecology, 76, 49-63

Kim, M., Morrison, M., Yu, Z., 2011. Evaluation of different partial 16S rRNA gene sequence regions for phylogenetic analysis of microbiomes. Journal of Microbiological Methods, 84, 81-87

vi

Nam, E.S., Kim, M.S., Lee, H.B., Ahn, J.K., 2010. β-Glycosidase of Thermus thermophilus KNOUC202: Gene and biochemical properties of the enzyme expressed in . Applied Biochemistry Microbiology, 46, 515-524

Kim, M.S., Sung, H.G., Kim, H.J., Lee, S.S., Chang, J.S., Ha, J.K., 2005. Study on rumen cellulolytic bacterial attachment and fermentation dependent on initial pH by cPCR. Journal of Animal Science and Technology (in Korean). 47, 615-624

Fields of Study

Major Field: Animal Science

Focus: Rumen Microbial Ecology

vii

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 ...... 49

Table 4.1 Primers and a TaqMan probe used in the real-time PCR assays for total bacteria, total archaea or cultured bacteria ...... 72

Table 4.2 Primers used in the real-time PCR assays for uncultured bacteria ...... 73

Table 7.1 Estimates of species-level OTUs calculated from partial and full-length archaeal 16S rRNA gene sequences ...... 132

Table 7.2 Estimates of genus- and family-level OTUs calculated from partial and full- length archaeal 16S rRNA gene sequences ...... 133

Table 7.3 Estimates of species-level OTUs calculated from partial and full-length bacterial 16S rRNA gene sequences ...... 134

Table 7.4 Estimates of genus- and family-level OTUs calculated from partial sequence regions and full length of bacterial 16S rRNA gene sequences ...... 135

Table 7.5 Estimates of OTUs calculated from partial and full-length archaeal 16S rRNA gene sequences at 0.01 distance ...... 136

Table 7.6 Estimates of OTUs calculated from partial and full-length bacterial 16S rRNA gene sequences at 0.01 distance ...... 137

viii

Table 7.7 Estimates of bacterial species-level OTUs calculated from full-length and short partial bacterial 16S rRNA gene sequences ...... 138

Table 8.1 Ingredient composition of dietary treatments ...... 155

Table 8.2 Sequence data and alpha diversity indices for the six fractions ...... 156

Table A List of bacterial primers for pyrosequencing analysis ...... 199

ix

LIST OF FIGURES

Figure 3.1 Bacterial phyla represented by the 16S rRNA gene sequences of rumen origin

...... 50

Figure 3.2 A taxonomic tree showing the genera of ruminal archaea identified by the

RDP database sequences ...... 51

Figure 3.3 A taxonomic tree showing the bacteria (grouped into genera) isolated from the rumen ...... 52

Figure 3.4 A taxonomic tree showing the archaea (grouped into genera) isolated from the rumen ...... 55

Figure 3.5 A taxonomic tree showing all the genera of ruminal bacteria identified by the

13478 16S rRNA gene sequences of rumen origin ...... 56

Figure 4.1 Populations of total archaea and total bacteria in the rumen of cattle ...... 74

Figure 4.2 Populations of three major cellulolytic bacteria and spp. in the rumen ...... 75

Figure 4.3 Populations of major non-cellulolytic cultured bacteria in the rumen ...... 76

Figure 4.4 Populations of uncultured bacteria originally identified from the rumen of sheep ...... 77

x

Figure 5.1 Clustering analysis of DGGE banding profiles based on the V3 region of 16S rRNA genes ...... 87

Figure 5.2 A taxonomic tree showing the bacterial genera represented by the 144 sequences ...... 88

Figure 5.3 A Venn diagram showing the numbers of species-level OTUs shared among the four composite samples ...... 89

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

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

Figure 6.2 A venn diagram showing the number of detected OTUs ...... 109

Figure 6.3 A hierarchical tree showing signal intensities and similarity among the fractionated samples...... 110

Figure 6.4 PCA for comparison among all the fractionated samples ...... 111

Figure 8.1 Sequence distribution at the phylum level for each fraction ...... 157

Figure 8.2 The distribution of sequences and OTUs at genus or the lowest classifiable rank in the phylum Firmicutes ...... 158

Figure 8.3 The distribution of sequences and OTUs at genus or the lowest classifiable rank in the phylum Bacteroidetes ...... 160

Figure 8.4 The distribution of sequences and OTUs at the lowest classifiable rank in minor phyla ...... 161

xi

Figure 8.5 Principal coordinates analysis for the six fractions ...... 162

Figure 8.6 Principal coordinates analysis for comparison among the three datasets ... 163

xii

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 ...... 6

2.1 Microbial communities in the rumen ...... 6

2.1.1 Bacterial communities ...... 6

2.1.1.1 Major cellulolytic bacteria ...... 6

2.1.1.2 Minor cellulolytic bacteria ...... 8

2.1.1.3 Uncultured cellulolytic bacteria ...... 9

2.1.2 Archaeal communities ...... 10

2.2 Methods to investigate microbial diversity ...... 12

2.2.1 Cultivation-dependent methods ...... 12

2.2.2 Cultivation-independent methods ...... 13

2.3 Bioinformatics analysis to investigate microbial diversity ...... 19

2.4 The use of monensin and its role in rumen fermentation...... 20

xiii

2.5 Biohydrogenation in the rumen ...... 22

2.6 Summary ...... 24

CHAPTER 3: STATUS OF THE PHYLOGENETIC DIVERSITY CENSUS OF

RUMINAL MICROBIOMES ...... 25

3.1 Abstract ...... 25

3.2 Introduction ...... 25

3.3 Materials and Methods ...... 28

3.3.1 Sequence data collection and phylogenetic analyses ...... 28

3.3.2 Diversity estimate ...... 29

3.4 Results ...... 30

3.4.1 Data summary ...... 31

3.4.2 Firmicutes ...... 31

3.4.3 Bacteroidetes...... 35

3.4.4 ...... 37

3.4.5 Minor phyla ...... 38

3.4.6 Archaea ...... 40

3.4.7 Estimates of OTU richness ...... 42

3.5 Discussion ...... 43

3.5.1 Phylogenetic diversity ...... 43

xiv

3.5.2 Diversity estimates ...... 47

CHAPTER 4: QUANTITATIVE COMPARISONS OF CULTURED AND

UNCULTURED MICROBIAL POPULATIONS IN THE RUMEN OF CATTLE FED

DIFFERENT DIETS ...... 63

4.1 Abstract ...... 63

4.2 Introduction ...... 64

4.3 Materials and Methods ...... 65

4.3.1 Sample collection, fractionation and DNA extraction ...... 65

4.3.2 Real-time PCR assays ...... 66

4.4 Results and Discussion ...... 67

4.4.1 Quantification of populations of total bacteria and total archaea ...... 67

4.4.2 Quantification of cultured bacteria ...... 67

4.4.3 Quantification of uncultured bacteria ...... 69

4.5 Conclusions ...... 71

CHAPTER 5: PHYLOGENETIC DIVERSITY OF BACTERIAL COMMUNITIES IN

BOVINE RUMEN AS AFFECTED BY DIETS AND MICROENVIRONMENTS ...... 78

5.1 Abstract ...... 78

5.2 Introduction ...... 78

5.3 Materials and Methods ...... 79

5.3.1 Sample collection, fractionation and DNA extraction ...... 79

xv

5.3.2 DGGE analysis...... 80

5.3.3 Construction of rrs clone libraries ...... 80

5.3.4 Restriction fragment length polymorphism (RFLP) analysis, DNA sequencing

and phylogenetic analysis ...... 81

5.3.5 Comparison of ruminal bacterial communities among the four composite

samples ...... 82

5.3.6 Nucleotide sequence accession numbers ...... 82

5.4 Results and Discussion ...... 82

CHAPTER 6: DEVELOPMENT OF A PHYLOGENETIC MICROARRAY FOR

COMPREHENSIVE ANALYSIS OF RUMINAL MICROBIOME ...... 91

6.1 Abstract ...... 91

6.2 Introduction ...... 92

6.3 Materials and Methods ...... 94

6.3.1 Oligonucleotide probe design and microarray fabrication...... 94

6.3.2 Sample collection, fractionation and DNA extraction ...... 95

6.3.3 Sample preparation and labeling ...... 95

6.3.4 Microarray hybridization ...... 96

6.3.5 Signal detection and data analysis ...... 96

6.3.6 Determination of the specificity and detection limit...... 97

xvi

6.3.7 Comparison between microarray and real-time PCR data ...... 98

6.4 Results and Discussion ...... 98

6.4.1 Validation of the specificity, sensitivity and detection limit ...... 98

6.4.2 Data summary ...... 99

6.4.3 Diversity of ruminal bacteria assigned to known genus ...... 100

6.4.4 Diversity of ruminal bacteria that are not assigned to any known genus ...... 103

6.4.5 PCA for comparison between fractions ...... 106

6.4.6 Comparison of RumenArray and real-time PCR data ...... 106

6.5 Conclusions ...... 107

CHAPTER 7: EVALUATION OF DIFFERENT PARTIAL 16S rRNA GENE

SEQUENCE REGIONS FOR PHYLOGENETIC ANALYSIS OF MICROBIOMES . 112

7.1 Abstract ...... 112

7.2 Introduction ...... 112

7.3 Materials and Methods ...... 116

7.3.1 Sequence collection, alignment, and clipping...... 116

7.3.2 Diversity estimates ...... 117

7.3.3 UniFrac analysis...... 118

7.3.4 Analysis of sequence datasets recovered from uncultured bacteria ...... 118

7.3.5 Analysis of short partial sequence regions...... 118

xvii

7.4 Results ...... 119

7.4.1 Analysis of partial archaeal sequences ...... 120

7.4.2 Analysis of partial bacterial sequences ...... 122

7.4.3 UniFrac analysis...... 125

7.4.4 Analysis of uncultured bacterial sequences ...... 125

7.4.5 Analysis of short partial sequence regions...... 126

7.5 Discussion ...... 127

CHAPTER 8: INVESTIGATION OF RUMINAL BACTERIAL DIVERSITY IN

CATTLE FED SUPPLEMENTARY MONENSIN OR FAT USING

PYROSEQUENCING ANALYSIS...... 139

8.1 Abstract ...... 139

8.2 Introduction ...... 140

8.3 Materials and Methods ...... 141

8.3.1 Sample collection ...... 141

8.3.2 Metagenomic DNA extraction ...... 142

8.3.3 Pyrosequencing ...... 142

8.3.4 Sequence processing and bioinformatics analysis ...... 143

8.3.5 Comparison among three datasets ...... 144

8.4 Results and Discussion ...... 145

8.4.1 Data summary ...... 145

xviii

8.4.2 Firmicutes ...... 145

8.4.3 Bacteroidetes...... 150

8.4.4 Minor phyla ...... 152

8.4.5 Comparison among the diets ...... 152

8.4.6 Comparison among three datasets ...... 153

8.5 Conclusions ...... 154

CHAPTER 9: GENERAL DISCUSSION ...... 164

WORKS CITED ...... 170

APPENDIX A: ADDITIONAL PYROSEQUENCING METHODS ...... 197

xix

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 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 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 H2-producing bacteria. As a result, monensin can reduce production in the rumen and shift fermentation towards more reduced VFA (e.g. propionate), decreasing :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 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

To support the above pyrosequencing analysis, we also evaluated different regions (500 - 700 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 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 semi- quantitative 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

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.

2.1.1 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.

2.1.1.1 Major cellulolytic bacteria

6

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 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 -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 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.

2.1.1.2 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

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 ruminicola (Cai and Dong, 2010). Future studies are needed to evaluate their significance in rumen feed digestion.

2.1.1.3 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 showed that one uncultured Acetivibrio OTU named Ad-H1-14-1 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 , unclassified Clostridiales, or unclassified (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.

2.1.2 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 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 rumen. Methanobrevibacter is the predominant genus of . 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). mobile is another predominant cultured ruminal methanogen species, reaching a population as high as >

108/ 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 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

2.2.1 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

(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 O2 free CO2.

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

(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 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.

2.2.2 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

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 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 real- time 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 novel bacteria can be detected using this method, but it is difficult to confirm non- predominant 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 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 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

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 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 non- linear 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 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 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 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 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

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 genus- level 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

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 CO2 and H2 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 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

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 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

3.3.1 Sequence data collection and phylogenetic analyses

28

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 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).

3.3.2 Diversity estimate

29

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 meta- analysis” 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 consolidated perspective of the ruminal microbiome based on most current phylogenetic diversity information.

3.4.1 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 3.3.

3.4.2 Firmicutes

Approximately 90.6% of the Firmicutes sequences were assigned to class

Clostridia, with the rest assigned to , Erysipelotrichi and Unclassified_Firmicutes

(Figure 3.5). The Firmicutes sequences were assigned to a total of 78 known genera, and

31

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 , 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 108

32 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, , 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 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 , Moryella, , ,

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 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 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.

3.4.3 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 (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

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 , and one isolate unclassified within

36

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.

3.4.4 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 semi- continuous 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

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 (Bozal et al., 2003) and that rumen is a mesophilic environment, the occurrence of Psychrobacter- like 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 succinogenes (Figure

3.3). Twenty-one of the 53 identified genera of Proteobacteria were represented by at least one isolate from the rumen.

3.4.5 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 by only one sequence. Most (95.5%) sequences within phylum 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 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). 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 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). 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). comprised 10 sequences recovered from cultured isolates named Fusobacterium sp. (Figures 3.3 and 3.5). The remaining minor phyla, including Acidobacteria, , Deferribacteres, Lentisphaerae,

Planctomycetes, , OP10, , SR1, and TM7 were represented by a very small number of sequences recovered from uncultured bacteria

39

(Figure 3.5). It remains to be determined if they are resident bacteria and have any significant role in the rumen.

3.4.6 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): (70.3% of all the archaeal sequences), (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 were represented:

Methanobrevibacter, 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

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 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).

3.4.7 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

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.

3.5. Discussion

3.5.1 Phylogenetic diversity

43

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 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 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 H2 and CO2, 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 H2 sink, such as homoacetogenesis, is needed to reduce possible negative effect on fiber digestion.

46

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 well- studied 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 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: SRA009223.8, <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

0.10

5928

2587

9042

18111

11829

34882

r

0.05

5674

4919

8652

3955

14903

31750

58168

18746

coverage

488

0.03

8149

1686

1302

7077

7252

9567

5578

22291

41951

78218

24480

99.9%

Clones needed fo needed Clones

277

131

673

550

0.10

1181

2370

419

0.05

4411

1055

1044

2384

8507

2145

ACE

66

334

182

927

0.03

1750

1833

1947

5232

8381

4017

16614

276

145

681

442

0.10

1166

2356

1

714

751

230

0.05

1884

3407

6466

1375

Chao

39

251

170

508

0.03

1132

1282

1300

3573

5826

2309

11257

207

103

536

956

307

0.10

1859

486

567

502

207

895

0.05

1399

2442

4599

Rarefaction richness Rarefaction

32

869

173

134

750

897

798

320

0.03

2344

3993

7426

1469

Maximum # of OTUs of # Maximum

)

)

)

)

)

)

96

91

97

98

91

(

94

(

(

(

(

(

0.10

97

189

518

936

278

1785

)

)

)

)

)

)

)

)

84

82

83

78

80

80

83

75

(

(

(

(

(

(

(

(

0.05

381

452

400

172

670

1162

2050

3774

)

)

)

)

)

)

)

)

)

)

)

69

74

73

77

81

70

66

76

71

65

(

(

81

(

(

(

(

(

(

(

(

(

0.03

34

26

distances. distances.

637

1

108

524

588

606

226

949

1610

2958

5271 5271 (71)

Observed # of OTUs (% coverage) #OTUs of Observed

The The number of OTUs for total bacteria, total archaea and major groups of bacteria, and their percentage coverage at three

.

Group

nospiraceae

phylogenetic phylogenetic

Table 3.1 Table

Prevotella

Firmicutes

Fibrobacter

Butyrivibrio

Unclassified Unclassified

Unclassified Unclassified

Unclassified Unclassified

Clostridiales

Bacteroidetes

Total Archaea Total

Total Bacteria Total Ruminococcus

Proteobacteria

Lach Ruminococcaceae 49

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.

50

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

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

Figure 3.3 Continued

Figure 3.3 Continued

53

Figure 3.3 Continued

54

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

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

Figure 3.5 Continued

Figure 3.5 Continued 57

Figure 3.5 Continued

Figure 3.5 Continued 58

Figure 3.5 Continued

Figure 3.5 Continued

59

Figure 3.5 Continued

Figure 3.5 Continued

60

Figure 3.5 Continued

Figure 3.5 Continued

61

Figure 3.5 Continued

62

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 108 rrs copies/

µg DNA and similar across all the fractions, while the population of total archaea was less than 105 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 107 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 106 rrs copies/ µg DNA, while the population of Ruminococcus albus was less than 106 rrs copies/ µg DNA. The

63 populations of four of the six uncultured bacteria were approximately 106 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

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

4.3.1 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 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 oC prior to DNA extraction. Metagenomic DNA was extracted from each of the fractionated samples as described previously (Yu and Morrison, 2004b).

4.3.2 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-H1-75-1 and Ad-H2-90-2 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 101 to 107 rrs copies were used in the real-time PCR assays. Each real-time PCR assay was conducted in three technical

66 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

4.4.1 Quantification of populations of total bacteria and total archaea

Total bacterial populations ranged from 1.71×108 to 5.47×108 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 (2.62×104 ~ 3.67×104) than the respective fractions recovered from cattle fed hay plus grain (7.22×102 ~ 5.34×103) 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).

4.4.2 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

(1.61×106 ~ 9.96×106) and R. flavefaciens (2.56×106 ~ 3.03×107) were more abundant

67 than the population of R. albus (7.59×104 ~ 8.64×105) 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 106 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

Butyrivirio spp. present in the adherent fraction may contribute slightly to cellulose digestion.

The population of genus Prevotella ranged from 4.40×107 to 1.88×108 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).

4.4.3 Quantification of uncultured bacteria

The populations of six different uncultured bacteria were quantified using specific real-time PCR assays. Ad-C1-74-3, Lq-C2-16-3 and Lq-C2-58-2 were originally recovered from sheep fed corn:hay, whereas Ad-H1-14-1, Ad-H1-75-1 and Ad-H2-90-2

69 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-C1-74-3 may be associated with the degradation of amino acids. Lq-C2-16-3 and Lq-C2-58-2 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-

C2-58-2 was greater than 106 rrs copies/ µg DNA across all the fractions. The populations of Ad-H1-14-1 and Ad-H2-90-2 that were assigned to Acetivibrio and

‘Unclassified Clostridia’, respectively, were about 106 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-

H1-14-1 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-H1-75-1 is also presumed to be involved in fiber degradation.

Detailed physiology can only be gained through studies of pure cultures. A reverse

70 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

a

Kobayashi, Kobayashi, 2001

Koike Koike and

Tajima et Tajima al., 2001a

Tajima et Tajima al., 2001a

2011

Stiverson et al.,

Tajima et Tajima al., 2001

Kobayashi, Kobayashi, 2001

Koike Koike and

Yu et Yu al., 2005

2000

Bernhard and Field,

2004,

Bartosch et al.,

2002

Nadkarni et al.,

Larue et al.,Larue 2005

References

295

642

513

371

446

175

273

418

467

1,535

length (bp)length

Amplicon Amplicon

C)

o

55

57

53

65

63

55

60

56

70

60

54

temperature (

Annealing

and

total bacteria, total archaea or cultured bacteria or total cultured archaea bacteria, total

and

s for

s

Archaea

multacida

amylophilus

ruminantium

.

.

.

R. flavefaciens

R

M

S

Butyrivibrio

F. succinogenes

R. albu

Total

Prevotella

Bacteroides

TotalBacteria

TotalBacteria

Target

time PCR assay

-

real

used in the used

probe

3’)

n

t al., 2011) t al., 2011)

TAMRA

-

CGT ATT GCG ACC GCT GCT

GGG AGG CAG TCAG

-

3’

-

5’

-

GGA GGA AAC GGA TGG TA

AC

FAM

-

Stiverson e

CCT TTA AGA GAG CAG TTT A ACA

TCT

CAC TACCAC TCA TGG CAA CAT

CAA CCACAA GTC TTCGCA AGA

TCC TGC ACT GAACAA AGA

TGCTAA TAC CGA ATG TTG

TGC GGC ACY GAC TCC CTA TG

GTG CCA GTGGCM GCC CCA GCG G

GCC TGCGCC CCC TGA ACT ATC

GGT ATGTGA GGA GCT TGC

CCT CCT TGCTAG GGT AAC A

CCC CCC TAA AAG TCTCAG TTCTAG G

GCC GCC ATG CAC CWC CTC T

ATT TAC CCS BGTAGA AGT CC

CAA TCGCAA GAG TTCTTC GTG

GAA GAA GGT CCC CCA CAT TG

GGC

6

GGA CTA CCA GGG TAT CTA ATC CTG TT

TCC TAC

AAG GAG GTG WTC CAR CC

AGA AGA GTTTGA TCM TGG CTC AG

Sequences (5’

f

r

K

K

-

-

-

-

Primers and a TaqMa a and Primers

900r

f

r

-

-

-

Mit

Mit

-

-

r

f

-

-

Rf425r

Rf154f

Ram

Ram

Sel

Sel

Buty

530f

Fs

Fs

Ra1439r

Ra1281f

ARC1059R

ARC787F

Bac708r

Bac303f

probe

TaqMan

806r

340f

1525r

27f

Primers

Reproduced from Reproduced

(

Table 4.1.

72

Partial sequences Sequences (5’  3’) Annealing position Amplicon (GenBank accession no.) (E. coli numbering) length (bp) Ad-C1-74 (AY816616) GAA GGG ACC GGT TAA GGT C 1013-1031 1,024

Lq-C2-16 (AY816578) GAC TTT GCT TCC CTT TGT TTT 1245-1265 1,258 G Lq-C2-58 (AY816550) AGC CTC CGA TAC ATC TCT GC 1010-1029 1,022

Ad-H1-14 (AY816508) GAT TTG CTT ACC CTC GCG 1260-1282 1,275 GGT TT

Ad-H1-75 (AY816420) CAC ACC TTG TAT CTC TAC 1006-1027 1,020 AAG C Ad-H2-90 (AY816432) CTT CGA CAG CTG CCT CCT TA 1451-1470 1,463

Table 4.2. Primers used in the real-time PCR assays for uncultured bacteria (Reproduced from Stiverson et al., 2011)

73

Total Archaea 10 9 10 8 10 7 10 6 5 10 10 4 10 3 copies/ ug DNA 10 2

rrs 10 1 10 0 Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Fraction

Total Bacteria 10 9 10 8 7 10 10 6 10 5 10 4 10 3 copies/ ug DNA 10 2 rrs 10 1 10 0 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

F. succinogenes R. albus 9 9 10 10 10 8 10 8 10 7 10 7 10 6 10 6 10 5 10 5 4 10 4 10

10 3 10 3 copies/ ug DNA ug copies/

copies/ ug DNA ug copies/ 10 2 10 2 rrs rrs 10 1 10 1

10 0 10 0 Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Fraction Fraction

R. flavefaciens Butyrivibio 10 9 10 9 10 8 10 8 10 7 10 7 10 6 10 6 10 5 10 5 4 4 10 10

10 3 10 3 copies/ ug DNA ug copies/

copies/ ug DNA ug copies/ 10 2 10 2 rrs rrs 10 1 10 1 10 0 10 0 Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C 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

P. ruminicola Prevotella 10 9 10 9 10 8 10 8 10 7 10 7 10 6 10 6 10 5 10 5 4 4 10 10

10 3 10 3 copies/ ug DNA ug copies/

DNA ug copies/ 10 2 10 2 rrs rrs 10 1 10 1 10 0 10 0 Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Fraction Fraction

S. ruminantium R. amylophilus 10 9 10 9 10 8 10 8 10 7 10 7 6 6 10 10 10 5 10 5 10 4 10 4

10 3 10 3 copies/ ug DNA ug copies/

DNA ug copies/ 10 2 10 2 rrs rrs 1 10 1 10 10 0 10 0 Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Fraction 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

Ad-C1-74-3 Lq-C2-16-3 10 9 10 9 8 10 8 10 10 7 10 7 10 6 10 6 10 5 10 5 10 4 10 4

10 3 10 3 copies/ ug DNA

copies/ ug DNA 10 2 10 2 rrs rrs 10 1 10 1 10 0 10 0 Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C

Fraction Fraction

Lq-C2-58-2 Ad-H1-14-1 10 9 10 9 10 8 10 8 7 7 10 10 10 6 10 6 10 5 10 5 10 4 10 4

10 3 10 3 copies/ ug DNA

copies/ ug DNA 10 2 10 2 rrs rrs 10 1 10 1 10 0 10 0 Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C Fraction Fraction

Ad-H1-75-1 Ad-H2-90-2 10 9 10 9 10 8 10 8 10 7 10 7 6 6 10 10 10 5 10 5 10 4 10 4

10 3 10 3 copies/ ug DNA

copies/ ug DNA 10 2 10 2 rrs rrs 10 1 10 1 10 0 10 0 Lq-H Lq-C Ad-H Ad-C Sld-H Sld-C 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

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 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

5.3.1 Sample collection, fractionation and DNA extraction

Whole rumen content was collected from four cannulated cows: two Jersey cattle

79 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).

5.3.2 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).

5.3.3 Construction of rrs clone libraries

The DNA extracts were pooled based on diets and fractions, resulting in four

80 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).

5.3.4 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 ), 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 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).

5.3.5 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.

5.3.6 Nucleotide sequence accession numbers

The rrs sequences obtained in this study have been deposited in the GenBank database (JF319298 - 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 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

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

86

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

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

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

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

90

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

1.0 × 107 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 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

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. Real- time 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 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:

6.3.1 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 oligonucleotide microarray by MYcroarray (Ann Arbor, MI), with each probe represented in triplicates.

6.3.2 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.

6.3.3 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, 95oC for 30 s; annealing, 55oC for 30 s; and extension, 72oC 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 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 37oC 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 NanoDropTM 1000 and then stored at -80oC until microarray hybridization.

6.3.4 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 65oC 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 65oC 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.

6.3.5 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, 500-600 PMT photomultiplier sensitivity, and 5 µm resolution. The GenePix Pro 6.0 program (Axon Instruments, Union

96

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-to- noise 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.

6.3.6 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 42oC, 45oC and 47oC. 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 1010 rrs copies to 106 rrs copies) and then subjected to microarray hybridization using individual microarrays.

97

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:

6.4.1 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 42oC, 45oC or 47oC. Twenty-one false positives were detected at

42oC, whereas 12 and 7 false positives were detected at 45oC or 47oC, respectively.

Although the number of false positives was reduced at 47oC compared to 45oC, one false negative was found at 47oC. 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 42oC, whereas 9 and 8 false positives were detected at 45oC and 47oC, respectively. Because one false negative in the specificity test using the 10 positive rrs clones was found at 47oC, we selected 45oC 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

RumenArray.

The Cy5-labeled cRNA synthesized from pools of six positive rrs clones were used to determine the detection limit at 45oC. The 1 × 1010 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 107 (Figure 6.1). When 2.09 × 106 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 (107 -

1010 rrs copies). The detection limit indicates that the RumenArray is not as sensitive as real-time PCR in investigating ruminal bacteria with low abundance.

6.4.2 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 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.

6.4.3 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

(S000991018, 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 sheep fed hay, and they may be associated with cellulose degradation. Ruminococcus

OTU 2 (S000607272, 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 (S000991061, 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 (S000728607, 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 (S000361669, 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

(S000650332, 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

(S000561168, 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 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 (S000843932, 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 (S000012651, unpublished sequence data) and

Selenomonas bovis (S000769745, RDP ID) (Zhang and Dong, 2009) were detected only in sheep fed corn:hay. Lachnobacterium bovis (S000390937, RDP ID) that ferments and produce primarily lactic acid (Whitford et al., 2001b) was detected only in sheep fed corn:hay. Amylolytic Ruminobacter (S001160055, RDP ID) was detected in sheep fed corn:hay but not in sheep fed hay. Megasphaera elsdenii (S000390583, 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 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.

6.4.4 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 (S000991126, 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 (S001382058, 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

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 (S000650389, 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 (S000361544, unpublished sequence data) and U_Ruminococcaceae OTU 3

(S000566650, 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 (S000888009, 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

(S000616063, 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 (S001143463, 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 (S000806419, 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

3 (S000927675, 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 (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 (S000508062, 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

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.

6.4.5 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).

6.4.6 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-H2-90-2 (S001790257, 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

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 real- time 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 semi- quantification of abundant ruminal bacteria in a comparative manner.

107

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

108

Lq-C Lq-H 25 58

11 17 36

Ad-C Ad-H 19 11 15 13

91 2 2 11 6

2

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

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

Figure 6.4: PCA for comparison among all the fractionated samples. PC2 separated sheep fed hay from sheep fed corn:hay.

111

CHAPTER 7

EVALUATION OF DIFFERENT PARTIAL 16S rRNA GENE SEQUENCE REGIONS FOR PHYLOGENETIC ANALYSIS OF MICROBIOMES

7.1 Abstract:

Operational taxonomic units (OTUs) are conventionally defined at a phylogenetic distance (0.03-species, 0.05-genus, 0.10-family) based on full-length 16S rRNA gene sequences. However, partial sequences (700bp or shorter) have been used in most studies.

This discord may affect analysis of diversity and species richness because sequence divergence is not distributed evenly along the 16S rRNA gene. In this study, we compared a set each of bacterial and archaeal 16S rRNA gene sequences of nearly full length with multiple sets of different partial 16S rRNA gene sequences derived therefrom

(approx. 440 - 700 bp), at conventional and alternative distance levels. Our objective was to identify partial sequence region(s) and distance level(s) that allow more accurate phylogenetic analysis of partial 16S rRNA genes. Our results showed that no partial sequence region could estimate OTU richness or define OTUs as reliably as nearly full- length genes. However, the V1-V4 regions can provide more accurate estimates than others. For analysis of archaea, we recommend the V1-V3 and the V4-V7 regions and clustering of species-level OTUs at 0.03 and 0.02 distance, respectively. For analysis of bacteria, the V1-V3 and the V1-V4 regions should be targeted, with species-level OTUs being clustered at 0.04 distance in both cases.

7.2 Introduction:

The difficulty in culturing most microbes present in natural or managed

112 environments forces microbiologists to use non-culture based methods to study populations. The use of 16S rRNA gene sequencing has proven to be an effective phylogenetic marker in examining microbial diversity and classifying microbes. Even though the scarcity of well-characterized microbes and the lack of a reliable prokaryotic taxonomy system often make it difficult to classify microbes to species or sub-species level with certainty solely based on 16S rRNA gene sequences, 16S rRNA gene sequences can provide more objective and reliable classification of microbes than phenotyping (Schloss and Handelsman, 2005). Since Lane et al. (1985) first described the use of 16S rRNA gene for identifying and classifying uncultured microbes in the environment, PCR amplification, cloning and sequencing have been the primary technologies used in determining 16S rRNA gene sequences from various environments.

During the past two decades, more than 1.3 million bacterial and 54,000 archaeal 16S rRNA gene sequences have been archived in RDP (as of March 20, 2010, Release 10,

Update 18) (Cole et al., 2009). These sequences are curated and include 16S rRNA genes recovered from both cultured and uncultured prokaryotes, with the latter accounting for most of the sequences. The 16S rRNA gene sequences in RDP have been classified into genera among 35 bacterial phyla and 5 archaeal phyla, but many of these phyla are composed largely or entirely of uncultured prokaryotes (Schloss and Handelsman, 2004).

The 16S rRNA gene sequences generated from microbiomes are typically clustered into operation taxonomic units (OTUs) at a few distance levels to determine species richness, diversity, composition, and community structure. Species, genus, family, and phylum are conventionally defined with distance values of 0.03, 0.05, 0.10 and 0.20, respectively, based on nearly full-length (approx. 1,540 bp) 16S rRNA gene sequences

113

(Schloss and Handelsman, 2004). However, 16S rRNA gene sequences produced in most studies are partial sequences of 700 bp or shorter due to cost (with the Sanger DNA sequencing technology) or technology limitations (with the next generation DNA sequencing technologies). Indeed, in the RDP database less than 44% of the bacterial and

15.3% of the archaeal sequences are longer than 1,200 bp. Only a very small percentage of the sequences in RDP reached nearly full length. Therefore, most researchers have used partial 16S rRNA gene sequences to make taxonomic assignments. Such a discord may create uncertainty in taxonomic placement of OTUs for the following reasons: First, divergence among different 16S rRNA gene sequences is not distributed evenly along the

16S rRNA gene but concentrated primarily in the nine hypervariable (V) regions

(Stackebrandt and Goebel, 1994). Second, some of the V regions are more variable than others (Youssef et al., 2009; Yu and Morrison, 2004a). Third, some regions of the 16S rRNA genes produce more reliable taxonomic assignments than others (Liu et al, 2007,

2008; Wang et al., 2007). We hypothesize that different V regions may produce different results with respect to estimates on species richness, diversity, and microbiome composition and structure, and some partial sequence regions may be better suited for microbiome analysis than others. A different taxonomic cutoff value, or distance level, may be required for a particular partial sequence region to give rise to similar results as nearly full-length sequences.

Recently, a number of studies used 454 pyrosequencing in comprehensive analysis of species richness and diversity present in complex microbiomes (Claesson et al., 2009;

Sogin et al., 2006; Krober et al., 2009; Youssef et al., 2009). These studies generated large numbers of partial 16S rRNA gene sequences. By necessity, these partial sequences

114 were clustered into OTUs using the same conventional distance values that were used for nearly full-length sequences. In some of these studies, two single V regions are compared between them and also against a set of nearly full-length sequences (Claesson et al.,

2009; Dethlefsen et al., 2008; Huse et al., 2008) in estimating OTU richness. It is recognized that the choice of V regions significantly affects estimates on OTU richness and diversity. One study also showed that the V1-V2 (approx. 350 bp) and the V8 regions produced different OTU evenness when a termite sample was analyzed (Engelbrektson et al., 2010). Another study compared eight V regions, either singular or dual, but the length of the partial sequence regions only ranged from 99 to 361 bp (Youssef et al., 2009).

More importantly, in these studies conclusions were drawn from comparing short partial sequences recovered from one or a few . As such, the conclusions derived from these studies may not be applied broadly to other environments. As the read length of pyrosequencing continues to increase, longer partial sequences (up to 800 currently) of

16S rRNA genes can be sequenced. Thus, there is a need to identify suitable partial sequence regions and phylogenetic distance cutoff values that can provide reliable analysis of microbiomes. In this study, we systematically compared all the partial sequence regions (approx. 450 to 700 bp) delineated by commonly used domain-specific

(bacterial or archaeal) PCR primers against nearly full-length 16S rRNA gene sequences archived in RDP that represent a broad taxonomy of both cultured bacteria and archaea.

The comparisons were focused on observed OTU richness, parametric and nonparametric estimates of maximum OTU richness, accuracy of OTU clustering, and community structure. The objective was to identify a partial region(s) of 16S rRNA gene and a distance cutoff value(s) that enable analysis of 454 pyrosequencing reads and produce

115 comparable results as nearly full-length sequences.

7.3 Materials and Methods:

7.3.1 Sequence collection, alignment, and clipping

All the sequences longer than 1,200 bp were retrieved from the RDP database

(Release 10, Update 18) in March 2010. All these sequences are of good quality as determined by RDP. For domain Bacteria, only the sequences recovered from type strains were selected, while for domain Archaea, sequences derived from both type and non-type strains were chosen because only a small number of archaeal type strains are archived in

RDP. The bacterial and archaeal sequences were downloaded separately. The sequences that do not have nearly full-length (<1,424 bp), as determined by the absence of the annealing sites of the domain-specific PCR primer pairs that anneal near the termini of

16S rRNA genes (A2Fa and U1510r for archaea, 27f and 1492r for bacteria), were removed manually from the datasets. The nearly full-length sequences were aligned using the NAST aligner according to a core set of alignment templates in the Greengenes database (DeSantis et al., 2006). Partial sequence regions that were delineated by the binding sites of domain-specific primer pairs (Tables 7.1 and 7.3) targeting different hypervariable regions (Baker et al., 2003; Yu and Morrison, 2004a; Yu et al., 2008) were clipped out from the alignment of the nearly full-length sequences using the Geneious program (Biomatters Ltd, Auckland, New Zealand), with the original alignment being retained. The alignment of each partial sequence region and the full-length sequences was analyzed as a separate ‘clone library’.

116

7.3.2 Diversity estimates

From the alignment of the nearly full-length sequences and each partial sequence region, a distance matrix at distance of 0.03, 0.05, and 0.10 was computed using the

DNADIST program of the PHYLIP package (version 3.69, http://evolution.genetics.washington.edu/phylip.html) with the Jukes-Cantor correction applied. A distance matrix was also computed at 0.01 distance, which was suggested to be a new taxonomic cutoff value for species (Stackebrandt and Ebers, 2006). The

DOTUR program (Schloss and Handelsman, 2005) was used to cluster the sequences into

OTUs (referred to as ‘observed’ OTUs) and determine the maximum number of OTUs represented by each ‘clone library’ using nonparametric Chao1 and ACE richness estimates. From the rarefaction output calculated by the DOTUR program, a parametric estimate of maximum number of OTUs in each ‘clone library’ was also performed using the non-linear models procedure (PROC NLIN) of SAS (V9.1, SAS Inst. Inc., Cary, NC) as described previously (Larue et al., 2005). The number of OTUs defined by each partial sequence ‘clone library’ (referred to as observed OTU richness) and the maximum number of OTUs predicted from each partial sequence ‘clone library’ (referred to as maximum OTU richness) were compared to those defined by the corresponding nearly full-length sequence ‘clone library’.

To identify a distance cutoff value that produces better estimate on species-level

OTUs than the commonly used 0.03 distance, all the partial sequence regions were also analyzed at 0.02 and 0.04 distances. Distance matrices at 0.02 and 0.04 were computed as described above. Each of the distance matrices was then used in clustering OTUs and estimating observed and maximum OTU richness as described above. The sequence

117 composition was compared between respective OTUs defined from the nearly full-length

‘clone library’ and from each of the partial sequence ‘clone library’. The ‘accuracy’ of

OTU clustering was introduced as a percentage of partial sequence-based OTUs that have identical sequence composition as the OTUs defined by the nearly full-length sequences.

7.3.3 UniFrac analysis

The UniFrac program (Lozupone and Knight, 2005) was used to assess differences in ‘clone libraries’ represented by individual partial sequence datasets and the nearly full- length sequence dataset. The sequences from all the ‘libraries’ were aligned against the

Greengenes database and then inserted into the ARB tree to build phylogenetic trees. The constructed trees were subjected to UniFrac significance test and P test.

7.3.4 Analysis of sequence datasets recovered from uncultured bacteria

To verify their applicability, one sequence dataset each recovered from rumen

(Brulc et al., 2009) and deep-sea surface sediment (Schauer et al., 2010) were also analyzed as described for the composite RDP sequence datasets of bacterial strains. Each of the two sequence datasets was retrieved from the RDP database, aligned, and analyzed as mentioned above for the RDP sequence datasets. Because no individual studies reported large numbers of archaea full-length archaeal sequences, this verification was not done for archaea.

7.3.5 Analysis of short partial sequence regions

Short partial sequences spanning 1 or 2 consecutive V regions (94 - 362) were also

118 evaluated using the composite RDP sequence dataset for bacteria. Due to lack of primers that anneal to individual V regions of archaeal sequences, this evaluation was not done for the composite RDP sequences of archaea. The Illumina GAIIx system, which produces short reads of about 100bp, has been used in microbiome analysis in a recent study (Caporaso et al., 2011). To evaluate the reliability of such short sequence reads in estimating richness and diversity, both the 100bp regions downstream of primer F515 and upstream of primer R806 (primers F515 and R806 were used in the study by Caporaso et al., 2011) were also analyzed and compared to the nearly full-length sequences in the bacterial RDP dataset as described above.

7.4 Results:

From the RDP database (Release 10 Update 18), 7,450 bacterial sequences longer than 1,200 bp were found that were derived from type strains. Of these sequences, 887 have a length of ≥1,458 bp and contain the annealing sites of universal or domain- specific primers near both ends of 16S rRNA gene. These nearly full-length sequences represent a broad taxonomic spectrum, including 18 phyla, 25 class, 64 order, 165 family,

361 genera, and 4 unclassified groups above genus level. In total, 8,375 archaeal sequences were found longer than 1,200 bp. Among them, only 284 were derived from type strains and have nearly full length. To increase the taxonomic representation, the nearly full-length sequences derived from non-type strains of archaea were also included.

The archaeal sequence dataset used in this study contained 1,071 nearly full-length sequences (≥1,435 bp), which represent all the 4 archaeal phyla, 9 class, 14 order, 25 family, 73 genera, and 14 unclassified groups above the phylum, class, order, family or

119 genus level. We intentionally selected and used sequences from a broad taxonomic spectrum so that the results derived can be less biased toward taxa found in particular habitats and the conclusions drawn can be better applied broadly to various environments. The lengths of the partial sequence regions ranged from 463 to 702 bp for archaea and from 446 to 652 bp for bacteria. Each of the partial sequence regions was compared to the corresponding nearly full-length sequence dataset with respect to observed OTU richness, maximum OTU richness, community structure, and accuracy of

OTU clustering.

7.4.1 Analysis of partial archaeal sequences

The different partial sequence ‘clone libraries’ (partial sequence regions) of the archaeal sequence dataset gave rise to varying observed OTU richness at 0.03 distance, and all the estimates differed from that computed from the full-length archaeal sequences

(Table 7.1). The V1-V3 region resulted in the best estimates of both observed OTU richness (5.5% overestimate) and the rarefaction-predicted (6.9% overestimate) maximum OTU richness, followed by the V1-V4 region. For both the Chao1 and ACE estimates, the V3-V5 region afforded the best predictions, which were lower than (3.9 and 12%, respectively) that estimated from the nearly full-length sequences. The partial sequence region that generated the second best estimate at species level was the V7-V9 region for Chao1 estimate and the V1-V4 region for ACE estimate. Overall, the V1-V4 region yielded overestimates, while the downstream regions underestimated all the richness estimates.

None of the partial sequence regions faithfully recaptured all the OTUs that were

120 defined by the nearly full-length archaeal sequences at any of the tested distances. At

0.04 distance, all the partial sequence regions produced worse estimates on both the observed and the maximum OTU richness measurements than at 0.03 distance. The V1-

V3 and the V1-V4 regions rendered worse estimates on OTU richness at 0.02 distance than at 0.03 distance, but the downstream partial sequence regions gave rise to better estimates on both observed and maximum richness (Table 7.1). The only exception was the Chao1 estimate from the V3-V5 region. The V4-V7 region resulted in the best estimates of observed OTU richness and rarefaction-predicted maximum OTU richness at

0.02 distance, whereas the V3-V5 and the V5-V7 regions resulted in better Chao1 and

ACE estimates at 0.03 and 0.02 distance, respectively. At 0.02 distance, the V4-V7 region produced the highest accuracy (72.2%) of OTU clustering among all the partial sequence regions, but OTU clustering based on this partial sequence region was 3.4% more accurate at 0.03 distance. At 0.02 distance, the partial sequence regions downstream of V1-V4 produced more accurate estimates on OTU richness. These results reflect the greater sequence divergence of the V1-V4 region than the downstream regions (Yu et al.,

2008).

At 0.01 distance, a new distance cutoff value recommended by Stackebrandt and

Ebers (2006) to define prokaryotic species, the V1-V4 and the V1-V3 were the first and second best partial sequence regions with respect to estimates of observed OTU richness and maximum OTU richness irrespective of prediction methods used (Table 7.5).

At 0.05 distance (equivalent to genus), the V1-V4 region produced the best estimates on both observed and maximum OTU richness (Table 7.2). The V1-V3 region yielded the second best estimate of maximum OTU richness, though the observed OTU richness was

121 overestimated. At 0.10 distance (equivalent to family), the V3-V5 region produced the best estimates on both the observed OTU richness and the rarefaction-predicted maximum OTU richness. For Chao1 estimate, the V4-V7 region gave rise to a more accurate estimate than other partial sequence regions, followed by the V3-V5 region. The

V1-V4 region supported the best ACE estimate, followed by the V4-V7 region. We note that the V1-V4 region (1-639 bp) tended to overestimate both observed and maximum

OTU richness at 0.05 and 0.10 distances, whereas the downstream regions considerably underestimated all the estimates of OTU richness (Table 7.2).

7.4.2 Analysis of partial bacterial sequences

The estimates of OTU richness of the bacterial sequence dataset differed also among the different partial sequence regions analyzed (Table 7.3). The observed species- level OTU richness estimated from the V6-V9 region delineated by primers 968f-1492r was the closest (5.9% underestimate) to that calculated from the full-length sequences, followed by the V1-V3 region delineated by primers 27f-519r. For the rarefaction estimate of maximum species-level OTU richness, the V6-V9 region delineated by primers 968f-1492r is the best region (13.8% underestimate), followed by the V1-V3 region delineated by primers 27f-519r. The V6-V9 region delineated by primers 968f-

1492r also generated the best Chao1 (3.0% underestimate) and ACE (14.3% underestimate) estimates, which was followed by the V6-V9 region delineated by primers

926f-1492r. As in the case of partial archaeal sequences, the upstream regions (V1-V4) consistently overestimated OTU richness, while the downstream regions underestimated

OTU richness.

122

When different partial sequence regions were evaluated in clustering species-level

OTUs at 0.02 and 0.04 distance levels, no partial sequence region completely recaptured the OTU estimates defined by the nearly full-length sequences either. However, the upstream regions (i.e. V1-V3 and V1-V4) produced more accurate estimates of observed and maximum OTU richness at 0.04 than at 0.03 distances, whereas the downstream partial sequence regions improved estimates at 0.02 distance, with a few exceptions

(Table 7.3). The three estimates on maximum OTU richness based on the V6-V9 region delineated by primers 968f-1492r and the Chao1 estimate from the V6-V9 region delineated by primers 926f-1492r were more accurate at 0.03 than at 0.02 distance. The

V1-V3 region delineated by primers 27f-519r afforded nearly the same observed OTU richness at 0.04 distance as the nearly full-length sequences did at 0.03 distance. The V1-

V4 region delineated by primers 63f-685r also generated a very close estimate of observed OTU richness at 0.04 distance. At this same distance, the V1-V3 region delineated by primers 63f-519r, the V1-V4 region delineated by primers 27f-685r, and the V1-V4 region delineated by primers 63f-685r also supported better estimate on maximum OTU richness by rarefaction, Chao1 and ACE, respectively, than other partial sequence regions. The V1-V4 region delineated by primers 27f-685r, however, produced the best accuracy (86.4%) of OTU clustering at 0.04 distance, while also producing rather accurate estimate (2.2% underestimate) on observed OTU richness.

At 0.01 distance, the V1-V4 region delineated by primers 27f-685r produced the best estimate on observed OTU richness, while the V1-V4 region delineated by primers

63f-685r generated the second best estimate (Table 7.6). For rarefaction estimate of maximum OTU richness, the V6-V9 region delineated by primers 968f-1492r was the

123 best region, while the V1-V4 region delineated by primers 27f-685r was the second best choice. For both Chao1 and ACE estimates, the V1-V4 region delineated by primers 63f-

685r was the best region, while the V1-V4 region delineated by primers 27f-685r generated the second best estimate. When the rumen sequence dataset (Brulc et al., 2009) was analyzed at 0.01 distance, the V1-V4 region delineated by primers 27f-685r and 63f-

685r also produced the first and second best estimates on observed and maximum OTU richness, respectively (data not shown). Therefore, the V1-V4 region is probably the best region for species richness and diversity estimates at 0.01 distance.

At genus level, the V6-V9 region delineated by primers 968f-1492r allowed the best estimates on both observed OTU richness and maximum OTU richness (Table 7.4).

However, all these predictions were underestimated. The V1-V4 region delineated by primers 27f-685r yielded the second closest estimates of observed OTU richness and the maximum OTU richness that was predicted by rarefaction and Chao1, while the V3-V5 region delineated by primers 357f-907r resulted in the second best ACE estimate of maximum OTU richness. At family level, the V6-V9 region delineated by primers 968f-

1492r again generated the closest estimates, all of which were underestimated (Table

7.4). The V3-V5 region delineated by primers 357f-907r produced the second closest estimates on both observed OTU richness and maximum OTU richness. Again, the upstream regions (V1-V4) overestimated OTU richness, while the downstream regions underestimated OTU richness (Table 7.4). These results corroborate the greater bacterial sequence divergence of the V1-V4 region than the downstream regions (Yu and

Morrison, 2004a) and are in general agreement with finding of Youssef et al. (Youssef et al., 2009).

124

7.4.3 UniFrac analysis

Both the UniFrac significance test and the P test showed that none of the

‘microbiomes’ represented by individual partial sequence regions was significantly different (P ≥ 0.25) than that represented by the full-length sequences. These results suggest that the locations of the partial sequence regions of the analyzed lengths might not significantly affect comparison of microbiomes using UniFrac. This conclusion is consistent with the finding of several previous studies from which several shorter partial sequence regions were analyzed (Huse et al., 2008; Liu et al., 2007; Wang et al., 2007).

Therefore, any of the partial sequence regions analyzed in this study can depict a comparable microbiome structure as the full-length sequences.

7.4.4 Analysis of uncultured bacterial sequences

The rumen sequence dataset contained 1,388 sequences (≥1,438 bp) of rumen origin (Brulc et al., 2009) that represented 9 bacterial phyla, but Firmicutes,

Bacteroidetes, and Proteobacteria predominated. Various partial sequence regions were compared to their corresponding nearly full-length sequences with respect to estimates on

OTU richness (data not shown). For observed OTU richness, the V1-V4 region delineated by primers 63f-685r produced better estimate (0.3% overestimate) at 0.04 distance than at other distances, whereas the V1-V4 region delineated by primers 27f-

685r gave rise to the second best estimate (2.4% underestimate) also at 0.04 distance. The

V1-V4 region delineated by primers 27f-685r and 0.04 distance also provided the best

Chao1 (1.8% overestimate) and ACE (0.3% overestimate) estimate and the second best

125 rarefaction estimate (1.7% underestimate vs. 1.5% underestimate from the V4-V6 region). All other distances and partial regions produced worse estimates on either observed OTU richness or maximum OTU richness. When the bacterial sequence dataset

(634 sequences of ≥1,421 bp) recovered from deep-sea surface sediments (Schauer et al.,

2010) was analyzed (data not shown), the V1-V4 region delineated by primers 27f-685r and 0.04 distance also produced the best estimate on OTU richness (0.3% overestimate) and the second best rarefaction estimate of maximum OTU richness (2.3% overestimate vs. 1.4% overestimate from the V6-V9 region delineated by primers 968f-1492r).

Although not the best combination, this V1-V4 region and 0.04 distance only led to 7.6% and 6.5% overestimate for Chao1 or ACE estimates, respectively. Evidently, the V1-V4 region and 0.04 distance can provide accurate estimate on OTU richness from these two sets of uncultured bacterial sequences.

7.4.5 Analysis of short partial sequence regions

Short partial sequence regions delineated by primers used in a previous study

(Youssef et al., 2009) were analyzed to evaluate their utility in estimate of species richness. These short sequences (94-362 bp) span 1 or 2 consecutive V regions (Table

7.7). For observed OTU richness, the V6 region and 0.03 distance produced the best estimate, while the V1-V2 region gave rise to best rarefaction and ACE estimate and V7-

V8 the best Chao1 estimate. Nevertheless, none of these short partial sequence region produced accurate estimate at any of the distances (0.01 to 0.05) examined. This also holds true for the rumen sequence dataset (data not shown). Because different V regions have been used in many different studies, β-diversity analysis using previously published

126 datasets can be difficult.

The 100bp region downstream of the forward primer F515 and the 100bp region upstream of reverse primer R806, which were generated by the Illumina GAIIx system and used in analysis of several samples including human feces and soil, fresh water and freshwater sediments (Caporaso et al., 2011), were compared to the nearly full-length sequences of RDP to assess the accuracy of such a short region in estimating species richness (data not shown). Except the 100bp region downstream of F515 and 0.01 distance that produced an accurate estimate on observed OTU richness (3.7% underestimate), both short regions underestimated both observed OTU richness and maximum OTU richness by 13.5 to 66% at 0.01, 0.02, or 0.03 distances. Similar results were observed when the rumen sequence dataset (Brulc et al., 2009) was subjected to this analysis (data not shown). Thus, although the Illumina GAIIx can produce sequence reads more cost-effectively, the short sequences probably do not support accurate analysis of microbiomes.

7.5 Discussion:

Defining the full diversity of microbiomes is essential for microbial ecologists to assess the functional significance of any bacterial or archaeal species or to determine if the major members have been accounted for in analysis of specific microbiomes.

Pyrosequencing recently emerged as the enabling technology to comprehensively characterize complex microbiomes in natural environments (Gilbert et al., 2008), managed (Krause et al., 2008; Liu et al., 2008; Zhang et al., 2009), or human and animal gut (Claesson et al., 2009; Dowd et al., 2008; Huse et al., 2008). It is difficult

127 to assemble individual pyrosequencing reads into full-length 16S rRNA genes because of the conserved nature of this gene. Therefore, partial 16S rRNA gene sequences are directly used in microbiome analysis. Because different regions of the 16S rRNA gene have different divergence, the choice of partial sequence regions can significantly affect the analysis results (Engelbrektson et al., 2010; Liu et al., 2007; Youssef et al., 2009).

Thus, it is important and useful to determine how a partial 16S rRNA gene sequence region can support characterization of microbiomes as ‘reliably’ as nearly full-length 16S rRNA genes.

Unlike other reported studies that compared single or dual V regions (94 - 360 bp), in this study we compared all partial sequence regions that span at least three consecutive V regions (463-702 bp for archaea, 446-652 bp for bacteria). All these partial sequence regions can be generated using domain-specific primer pairs so that they can be amplified and used in pyrosequencing analysis. In addition, instead of using uncultured bacterial sequences recovered from a particular habitat, we chose the sequences only recovered from cultured organisms (type strains for bacteria, both type and non-type strains for archaea). These sequences are better taxonomically characterized and are free of chimeric artifacts. Furthermore, because the sequences were not recovered from a particular habitat, the sequences used in this study represent a much broader taxonomy and diversity. As such, the results of this study might be applied to analysis of different microbiomes. To test this premise, two large datasets of nearly full-length sequences were analyzed in parallel to verify if the best partial sequence region(s) and distance(s) can be applied to individual sequence datasets.

128

The 454 GS FLX systems is the primary technology used in most diversity studies of microbiomes (Droege and Hill, 2008). The previous 454 GS FLX system produces sequence reads about 250 bp, a length typically spanning single V regions of 16S rRNA gene. Such a length only allows for classification of 16S rRNA gene sequences to genus in RDP (Liu et al., 2008). Most studies reported so far pyrosequenced single V regions, and when compared, two different V regions typically produce different results (Claesson et al., 2009; Dethlefsen et al., 2008; Huse et al., 2008; Sogin et al., 2006; Youssef et al.,

2009). Several of these studies also compared partial sequences to nearly full-length sequences in estimating OTU richness (Claesson et al., 2009; Huse et al., 2008; Youssef et al., 2009). However, few studies have assessed if partial sequences can be clustered into species-level OTUs as ‘reliably’ as nearly full-length sequences. Thus, this study is probably among the early studies in a continuum of research that will lead to improved analysis of 16S rRNA gene sequences.

Different partial sequence regions produced different estimates of OTU richness, both observed and predicted maximum, at all the three conventional distance levels for either archaea or bacteria. However, none of the analyzed partial sequence regions of bacteria or archaea faithfully recaptured the richness estimates (observed or predicted) that were determined by the nearly full-length sequences at conventional 0.03, 0.05, or

0.10 distances. These results corroborate the finding of a recent study (Schloss, 2010).

Therefore, estimates of OTU richness calculated from partial sequences should be interpreted with caution. As shown in this study, the V1-V4 region can provide improved estimates on species richness and accuracy of OTU clustering when clustered at 0.04

129 distance, while the downstream partial sequence regions need to be clustered at 0.02 distance (Tables 7.1 and 7.3).

As the 454 pyrosequencing technology is increasingly used in analysis of microbiomes and the sequence read length continues to increase, longer partial 16S rRNA gene sequences will be sequenced that will improve accuracy of microbiome analysis. If a common partial sequence region is targeted by different researchers, analysis results can be compared among laboratories. A common target region will also facilitate global analysis of microbial diversity in a particular type of environment of interest as well as β-diversity across multiple environments. As a beginning of this effort, for analysis of archaea we recommend the V1-V3 region to be targeted with species-level

OTUs being clustered at 0.03 distance if the current FLX Titanium system is used, or the

V4-V7 region to be targeted with species-level OTUs being clustered at 0.02 distance if the newest 454 FLX system that generates up to 800bp sequence reads is used. For analysis of bacteria, the V1-V3 region should be targeted using the FLX Titanium system, while the V1-V4 should be targeted with the newest 454 FLX system, with species-level OTUs being clustered at 0.04 distance in both cases. Additionally, these partial sequence regions also provide better analysis of richness and diversity than other partial regions if 0.01 distance is used to define OTUs. It should be pointed out that if distance is set at thousandth or below partial sequence regions may produce similar richness estimates as nearly full-length sequence. However, it will be time consuming to compare millions of richness estimates (5x107 for archaea and 5x108 for bacteria sequences). In addition, most OTU clustering algorithms do not support thousandth distance.

130

The V1-V3 or the V1-V4 regions of bacterial 16S rRNA genes provide two additional advantages: First, the V1-V3 or the V1-V4 regions are more divergent and thus can provide more phylogenetic resolution than other regions. Greater resolution is especially important in analysis of microbiomes from specialized habitats, such as intestinal tract of animals and humans, rumens, anaerobic digesters, biological wastewater treatment reactors, where great diversity exists at low taxa. Second, RDP and other databases stored more partial sequences that correspond to the V1-V4 region than the downstream regions. As such, partial sequences corresponding to this region will have more database sequences to compare to, greatly facilitating phylogenetic analysis.

131

15.9)

24.5)

27.3)

16.1)

12.0)

-

-

-

-

-

ACE

±

696 (5.5)

555 (

671 (1.7)

498 (

663 (0.5)

480 (

733 (11.1)

554 (

726 (10.0)

581 (

757 (14.7)

758 (14.8)

660 (0.0)

.

3.9)

11.9)

21.2)

23.7)

13.1)

-

-

-

-

-

length sequences. length The

Chao1

-

660 (11.9)

520 (

646 (9.5)

465 (

627 (6.3)

450 (

682 (15.6)

513 (

696 (18.0)

567 (

680 (15.3)

684 (15.9)

590 (0.0)

full

Maximum # of OTUs Maximum

2.

7.

14.7)

21.0)

6.5)

28.4)

18.4)

13.2)

-

-

-

-

-

-

(

Rarefaction

478 (3.5)

394 (

477 (3.2)

365 (

432 (

331 (

473 (2.4)

377

495 (7.1)

401 (

499 (8.0)

494 (6.9)

462 (0.0)

ained.

64.5)

OTUs (%) OTUs

# of identical # of identical

243 (

231 (70.4)

251 (66.4)

227 (73.9)

231 (67.5)

189 (67.7)

265 (72.2)

235 (75.6)

256 (66.7)

231 (69.8)

269 (69.3)

252 (65.8)

363 (100)

length archaeal 16S rRNA gene sequences 16Sgene archaeal rRNA length

-

. Same as in other Table Table . in other as Same

±

9)

9.6)

15.4)

5.8)

23.1)

14.3)

8.8)

-

-

-

-

-

-

length sequences. Positive values designate overestimates, sequences.overestimates, Positiveand designate length values

-

# of OTUs # of OTUs

377 (3.9)

328 (

378 (4.1)

307 (

342 (

279 (

367 (1.1)

311 (

384 (5.8)

331 (

388 (6.

383 (5.5)

363 (0.0)

Yu et al. (2008) Yu et

underlined and bolded are the estimates, boldedthat are best and values only are while the underlined

0.02

0.03

0.02

0.03

0.02

0.03

0.02

0.03

0.02

0.03

0.03

0.03

0.03

level

Distance Distance

re both

*

463

585

527

702

553

639

467

1,435

Sequence Sequence

length (bp) length

evel OTUs calculated full evel partial OTUs and from

l

-

length sequences and partial sequence regions at 0.03 (shaded) are listed. For some partial sequence partial some sequence listed.are For sequencesregions at 0.03 (shaded) and sequence partial length

V9

V9

V7

V7

V5

V4

V3

V9

-

-

-

-

-

-

-

-

-

V7

V6

V5

V4

V3

V1

V1

V1

V regions

U1510r

ARC915r

U1510r

UA1204r

Estimates of species Estimates

UA1204r

A693r

519r

U1510r

1.

§

7.

Primer set Primer

A1040f A1040f

ARCH915

A679r

U519f U519f

ARC344f

A2Fa A2Fa

A2Fa A2Fa

A2Fa A2Fa

Values in parenthesis show the estimates relative show relative of full to that estimates in parenthesis the Values

The reverse complementary of reported A693r primer bycomplementary reverse The

The estimates for nearly full for estimates The

Number of OTUs that contain the same sequences as the corresponding OTUs clustered nearly that the OTUs OTUs from corresponding contain sequencesclustered of as the Number the same

Table

§

values in parenthesis represent accuracy (%) of OTU clustering (same in (same as OTU tables). (%) of accuracy clustering other in parenthesis represent values

underlined are the second best estimates (same as in other tables). as the second are tables). in other (same estimates underlined best

negative values underestimates. The values that underestimates. values values The a negative

±

* Calculated from consensus sequences (same as consensusas tables). in other from (same * Calculated sequences

regions, the estimates at 0.02 or 0.04 are also listed when better estimates were obt were better estimates listed at also when 0.04 are the estimates 0.02 or regions, †

132

18.1)

27.5)

29.9)

15.7)

16.2)

20.8)

20.3)

29.7)

13.6)

15.2)

-

-

-

-

-

-

-

-

-

-

ACE

167 (

148 (

143 (

172 (

171 (

235 (15.2)

238 (16.7)

204 (0.0)

366 (

368 (

325 (

399 (

392 (

477 (3.2)

495 (7.1)

462 (0.0)

±

7.6)

2.6)

19.5)

28.6)

31.0)

8.1)

20.7)

19.4)

31.6)

14.1)

20.3)

3.0)

Chao1

-

-

-

-

-

-

-

-

-

-

-

-

169 (

150 (

145 (

194 (

193 (

235 (11.9)

265 (26.2)

210 (0.0)

372 (

378 (

321 (

403 (

374 (

457 (

455 (

469 (0.0)

Maximum # of OTUs

.0)

5.8)

length archaeal 16S rRNA gene sequences. 16Sgene archaeal rRNA length

8.5)

12.0)

19

24.6)

12.7)

1

20.8)

29.5)

16.8)

13.7)

-

-

-

-

-

-

-

-

-

-

-

(0.0)

Rarefaction

125 (

115 (

107 (

124 (

130 (

172 (21.1)

174 (22.5)

142 (0.0)

271 (

255 (

227 (

268 (

278 (

350 (8.7)

355 (10.2)

322

±

5.8)

10.2)

16.1)

23.4)

10.9)

10.6)

15.4)

24.5)

14.3)

9.9)

-

-

-

-

-

-

-

-

-

-

# of OTUs # of OTUs

123 (

115 (

105 (

122 (

129 (

165 (20.4)

168 (22.6)

137 (0.0)

244 (

231 (

206 (

234 (

246 (

300 (9.9)

303 (11.0)

273 (0.0) 273 (0.0)

0.10

0.05

level

Distance Distance

level OTUs calculated from partiallevel full OTUs from and calculated

-

463

585

527

702

553

639

467

463

585

527

702

553

639

467

1435

1,435

Sequence Sequence

length (bp)* length

and familyand

-

V9

V9

V7

V7

V5

V4

V3

V9

V9

V9

V7

V7

V5

V4

V3

V9

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

V7

V6

V5

V4

V3

V1

V1

V1

V7

V6

V5

V4

V3

V1

V1

V1

V regions

Estimates of genus Estimates

U1510r

U1510r

ARC915r

ARC915r

U1510r

U1510r

UA1204r

UA1204r

2.

UA1204r

UA1204r

A693r

519r

U1510r

A693r

519r

U1510r

7.

§

§

Primer set Primer

Table

A1040f A1040f

ARCH915

A679r

U519f U519f

ARC344f

A2Fa A2Fa

A2Fa A2Fa

A2Fa A2Fa

A1040f A1040f

ARCH915

A679r

U519f U519f

ARC344f

A2Fa A2Fa

A2Fa A2Fa A2Fa A2Fa

133

1.4)

14.3)

20.3)

22.8)

1.9)

27.4)

7.1)

-

-

-

-

-

-

-

.

±

2250 (24.4)

1551 (

2072 (14.5)

1441 (

2040 (12.8)

1396 (

1775 (

1313 (

1784 (

2420 (33.8)

1950 (7.8)

2235 (23.5)

1681 (

2385 (31.8)

1859 (2.8)

2255 (24.7)

1809 (0.0)

ACE

3.7)

3.0)

7.0)

17.0)

22.0)

-

-

-

-

-

1946 (21.6)

1554 (

1855 (15.9)

1493 (

1813 (13.3)

1335 (

1670 (4.4)

1244 (

1662 (3.9)

2133 (33.3)

1875 (17.2)

2019 (26.2)

1541 (

2085 (30.3)

1807 (12.9)

2126 (33.0)

1600 (0.0)

Chao1

Maximum # of OTUs Maximum

1.6)

1.1)

1.5)

8.1)

2.3)

-

-

-

-

13.8)

20.1)

2

25.5)

-

-

-

-

1286 (16.4)

952 (

1219 (10.3)

883 (

1169 (5.8)

866 (

1093 (

823 (

1088 (

1375 (24.4)

1112 (0.6)

1372 (24.2)

1015 (

1368 (23.8)

1080 (

1329 (20.0)

1105 (0.0)

Rarefaction

2)

length bacterial 16S rRNA gene sequences bacterial 16Sgene rRNA length

84.7)

-

462 (78.7)

432 (82.8)

461 (80.5)

416 (82.4)

468 (82.5)

408 (81.8)

459 (83.2)

402 (82.4)

472 (

482 (79.5)

456 (81.0)

459 (75.0)

469 (86.4)

478 (79.3)

459 (82.6)

459 (76.

555 (100)

OTUs (%) (%) OTUs

# of identical # of identical

±

5.9)

9.0)

10.1)

0.5)

12.1)

2.2)

-

-

-

-

-

-

om partial full and

587 (5.8)

522 (

573 (3.2)

505 (

567 (2.2)

499 (

552 (

488 (

557 (0.4)

606 (9.2)

563 (1.4)

612 (10.3)

543 (

603 (8.6)

556 (0.2)

602 (8.5)

555 (0.0)

# of OTUs # of OTUs

0.02

0.03

0.02

0.03

0.02

0.03

0.02

0.03

0.04

0.03

0.04

0.03

0.04

0.03

0.04

0.03

0.03

level

Distance Distance

level calculated OTUs fr

544

605

597

563

614

446

652

484

-

1,458

Sequence Sequence

length (bp)* length

ns

V9

V9

V6

V5

V4

V3

V4

V3

V9

-

-

-

-

-

-

-

-

-

V6

V6

V4

V3

V1

V1

V1

V1

V1

V regio

Estimates of species Estimates

3.

7.

1492r 1492r

1492r 1492r

1100r

907r 907r

685r

519r

685r

519r 519r

1492r

Table

Note: estimates at 0.03 distance are shaded. are 0.03 distance estimates at Note:

968f

926f

533f

357f

63f

63f

27f

27f

27f Primer set Primer

134

9.1)

21.7)

33.3)

37.2)

22.0)

22.9)

25.5)

21.7)

-

-

-

-

-

-

-

-

ACE

460 (0.0)

964 (

689 (49.8)

749 (62.8)

645 (40.2)

679 (47.6)

1061 (0.0)

360 (

307 (

289 (

359 (

818 (

790 (

831 (

1468 (38.4)

1588 (49.7)

1367 (28.8)

1518 (43.1)

±

0)

7.0)

2.2)

-

35.0)

40.0)

29.0)

25.0)

31.0)

21.0)

-

-

-

-

-

-

-

Chao1

490 (0.0)

384 (

766 (56.3)

732 (49.4)

690 (40.8)

679 (39.0)

1087 (0.

320 (

295 (

346 (

1007 (

812 (

749 (

863 (

1399 (28.7)

1434 (31.9)

1282 (17.9)

1452 (34.0)

Maximum # of OTUs

12.4)

11.6)

24.4)

27.5)

17.9)

23.7)

23.4)

24.3)

-

-

-

-

-

-

-

-

291 (0.0)

688 (0.0)

423 (45.4)

471 (61.9)

398 (36.8)

424 (45.7)

888 (29.1)

928 (34.9)

846 (23.0)

909 (32.1)

255 (

220 (

211 (

239 (

608 (

525 (

527 (

521 (

Rarefaction

±

6.6)

6.5)

16.8)

20.5)

14.8)

14.2)

13.5)

16.4)

-

-

-

-

-

-

-

-

ated from partial sequence regions and full length of bacterial 16S and rRNA partialregions full length ated from sequence

94 (

244 (0.0)

444 (0.0)

228 (

415 (

324 (32.8)

345 (41.4)

311 (27.5)

321 (31.6)

507 (14.2)

522 (17.6)

496 (11.7)

512 (15.3)

203 (

1

208 (

381 (

384 (

371 (

# of OTUs # of OTUs

0.10

0.05

level

Distance Distance

level OTUs calcul

-

544

605

597

563

614

446

652

484

544

605

597

563

614

446

652

484

1458

1458

Sequence Sequence

length (bp)* length

and familyand

-

V9

V9

V6

V5

V4

V3

V4

V3

V9

V9

V9

V6

V5

V4

V3

V4

V3

V9

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

V6

V6

V4

V3

V1

V1

V1

V1

V1

V6

V6

V4

V3

V1

V1

V1

V1

V1

regions

Hypervariable Hypervariable

Estimates of genus Estimates

4.

7.

1492r 1492r

1492r 1492r

1100r

907r 907r

1492r 1492r

1492r 1492r

1100r

907r 907r

685r

519r

685r

519r 519r

1492r

685r

519r

685r

519r 519r

1492r

Primer set Primer

gene sequences. sequences. gene

Table

968f

926f

533f

357f

63f

63f

27f

27f

27f

968f

926f

533f

357f

63f

63f

27f

27f 27f

135

24.8)

21.9)

15.8)

6.8)

-

-

-

-

29.5)

29.5)

-

-

ACE

±

958 (

1021 (

958 (

1061 (

1144 (

1375 (1.3)

1265 (

1358 (0.0)

2.7)

11.4)

-

-

31.4)

28.3)

32.5)

26.0)

25.6)

-

-

-

-

-

Chao1

917 (

959 (

902 (

990 (

995 (

1301 (

1184 (

1337 (0.0)

Maximum # of OTUs

)

24.6)

26.0)

30.5)

26.7)

21.1)

3.0)

6.2)

-

-

-

-

-

-

-

Rarefaction

668 (

656 (

616 (

649 (

699 (

859 (

831 (

886 (0.0

±

.9)

eal 16S rRNA gene sequences at 0.01 distance. 16S rRNA eal gene

13

15.2)

20.8)

16.8)

12.9)

0.4)

1.6)

-

-

-

-

-

-

-

# of OTUs # of OTUs

475 (

468 (

437 (

459 (

481 (

550 (

543 (

552 (0.0)

length sequences. Positive overestimates, sequences. length values designate and

-

1

length archa length

).

-

).

0.0

e level e

Distanc

*

Table 7.6 Table

Table 7.6 Table

463

585

527

702

553

639

467

1,435

Sequence Sequence

length (bp) length

V9

V9

V7

V7

V5

V4

V3

V9

-

-

-

-

-

-

-

-

V7

V6

V5

V4

V3

V1

V1

V1

V regions V

restimates. The values that are both underlined and bolded are the estimates, boldedthat are best that and values only are values while the restimates. both underlined The are

U1510r

ARC915r

A1204r

U1510r

UA1204r

U

A693r

519r

U1510r

§

Estimates of OTUs calculated from partial from full calculated and of OTUs Estimates

.

Primer set Primer

7.5

A1040f A1040f

ARCH915

A679r

U519f U519f

ARC344f

A2Fa A2Fa

A2Fa A2Fa

A2Fa A2Fa

Table

underlined are the second best estimates (same as the second are in (same estimates underlined best

negative values unde values negative

± relativeof full to that estimates in parenthesis show the Values

* Calculated from consensus sequences (same as consensusas in from (same * Calculated sequences

136

1.6)

26.2)

30.8)

31.8)

35.7)

-

-

-

-

±

3523 (

3301 (

3255 (

3068 (

5802 (2

6197 (29.9)

5837 (22.3)

6432 (34.8)

4771 (0.0)

ACE

25.5)

27.4)

24.8)

35.8)

-

-

-

-

2873 (

2799 (

2900 (

2475 (

4422 (14.6)

4621 (19.8)

4617 (19.7)

4822 (25.0)

3857 (0.0)

Chao1

Maximum # of OTUs

22.8)

27.7)

28.8)

29.5)

-

-

-

-

1819 (

1703 (

1678 (

1660 (

3002 (27.4)

3107 (31.9)

3001 (27.4)

3207 (36.1)

2356 (0.0)

Rarefaction

±

6.1)

7.3)

8.2)

9.8)

length bacterial 16S rRNA gene sequences at 0.01 distance. sequences bacteriallength 16Sgene rRNA

-

-

-

-

-

652 (

643 (

637 (

626 (

725 (4.5)

727 (4.8)

723 (4.2)

729 (5.0)

694 (0.0)

# of OTUs # of OTUs

0.01

level

Distance Distance

84

544

605

597

563

614

446

652

4

1,458

Sequence Sequence

length (bp)* length

V9

V9

V6

V5

V4

V3

V4

V3

V9

-

-

-

-

-

-

-

-

-

V6

V6

V4

V3

V1

V1

V1

V1

V1

V regions V

Estimates of OTUs calculated from partial from full calculated and of OTUs Estimates

.

1492r 1492r

1492r 1492r

1100r

907r

7.6

685r

519r

685r

519r

1492r

able

T

968f

926f

533f

357f

63f

63f

27f

27f

27f Primer set Primer

137

5.6)

18.9)

19.7)

12.3)

39.6)

30.4)

-

-

-

-

-

-

48.4)

64.2)

-

-

1708 (

933 (

1467 (

648 (

2102 (16.2)

1452 (

1937 (7.1)

2236 (23.6)

2443 (35.0)

1587 (

2069 (14.4)

1093 (

2225 (23.0)

1259 (

1855 (2.5)

2527 (39.7)

1809 (0.0)

ACE

±

terial 16S rRNA sequences 16S rRNA terial gene

# of OTUs

2.7)

35.1)

10.1)

11.3)

28.6)

28.2)

-

-

-

-

-

-

60.5)

-

1557 (

1039 (

1439 (

632 (

1877 (17.3)

1401 (12.4)

1842 (15.1)

2198 (37.4)

2233 (39.6)

1420 (

1845 (15.3)

1142 (

1763 (10.2)

1149 (

1748 (9.3)

2376 (48.5)

1600 (0.0)

Chao1

Maximum Maximum

4.9)

8.2)

3.5)

-

-

-

47.5)

25.2)

62.9)

19.6)

15.3)

37.1)

34.0)

-

-

-

-

-

-

-

1051 (

580 (

827 (

410 (

1212 (9.7)

888 (

1014 (

1163 (5.2)

1303 (17.9)

936 (

1204 (9.0)

695 (

1208 (9.3)

729 (

1066 (

1494 (35.2)

1105 (0.0)

Rarefaction

length length bac short partial and

-

±

0.5)

26.8)

13.9)

42.0)

7.9)

5.8)

6.8)

19.1)

19.8)

0.9)

-

-

-

-

-

-

-

-

-

-

11 (

552 (

406 (

478 (

322 (

578 (4.1)

5

523 (

557 (0.4)

583 (5.0)

517 (

569 (2.5)

449 (

564 (1.6)

445 (

550 (

626 (12.8)

555 (0.0)

# of OTUs # of OTUs

0.03

0.03

0.03

0.03

0.03

0.03

0.03

0.03

0.03

0.01

0.01

0.02

0.04

0.02

0.01

0.01

0.05

level

Distance Distance

level OTUs calculated full level OTUsfrom calculated

-

94

362

191

268

275

274

192

325

1,458

species

Sequence Sequence

length (bp)* length

bacterial bacterial

2

V8

V7

V6

V

V9

-

-

-

-

-

V7

V6

V4

V3

V1

V1

V7

V6

V5

V regions V

Estimates of Estimates

.

1406

7.7

1238

826

548

1238

1065

1065

-

-

355

1492

-

-

able

1046

T

Note: estimates at 0.03 distance are shaded. are 0.03 distance estimates at Note:

1046

967

967

805

530

338

27

27 Primer set Primer

138

CHAPTER 8

INVESTIGATION OF RUMINAL BACTERIAL DIVERSITY IN CATTLE FED SUPPLEMENTARY MONENSIN OR FAT USING PYROSEQUENCING ANALYSIS

8.1 Abstract:

Monensin and dietary fats have been used to improve the efficiency of feed utilization and reduce methane production. However, the effect of these dietary manipulations on the ruminal bacteriome has not been examined in detail. The objective of this study was to examine and compare the effects of monensin or monensin in combination with fat on ruminal bacterial communities in cattle fed with the following three diets: (1) a control diet without monensin or fat (C), (2) the control diet supplemented with monensin as Rumensin (CR), and (3) the control diet supplemented with both monensin and fat (CRF). The bacteriome in the liquid and adherent fractions was analyzed using 454 pyrosequencing analysis. In total, 24,792 16S rRNA gene (rrs) sequences were obtained. Most sequences were assigned to phyla Firmicutes and

Bacteroidetes irrespective of fractions. Firmicutes was more abundant in the adherent fraction than in the liquid fraction, while Bacteroidetes was less abundant in the adherent fraction than in the liquid fraction. Gram-positive Firmicutes was not affected by monensin but stimulated by monensin in combination with fats. Prevotella was dominant among known genera. However, most sequences were assigned to unclassified groups, with unclassified Lachnospiraceae and unclassified Clostridiales being predominant. In total, 9,867 OTUs at 0.04 distance were identified at species-equivalent level from the

24,792 sequences across all the fractions. Although the sequence distribution was similar between bacterial communities in the C and CR groups, OTU distribution differed

139 between these two groups. It seems that numerous monensin-resistant strains are stimulated to adapt to a new ruminal environment. Based on both sequence and OTU distribution, bacterial communities in the CRF group were different from those in C or the CR groups. Numerous strains involved in lipolysis or biohydrogenation (BH) appeared to increase in the rumen of cattle fed CRF.

8.2 Introduction:

Monensin has been fed to ruminant animals in order to improve the efficiency of feed utilization (Russell and Strobel, 1989). Monensin is thought to modulate ruminal microbiome through its selective effect on Gram-positive bacteria than Gram-negative bacteria as documented in in vitro cultures (Callaway et al., 2003). As a result, monensin can decrease methane production as well as acetate:propionate ratio in the rumen

(Callaway et al., 2003). However, in vivo studies using rrs-based techniques showed that

Gram-positive bacterial populations were not significantly affected by monensin due to monensin resistance (Stahl et al., 1988; Weimer et al., 2008). In addition, some in vivo studies showed that the acetate:propionate ratio was not altered by monensin (Firkins et al., 2008; Oelker et al., 2009). Ruminal bacterial populations could be affected by other dietary factors other than monensin (Firkins et al., 2008). Supplementary monensin often results in milk fat depression (MFD) that might be associated with specific BH reactions

(Mathew et al., 2011). However, other studies could not support this finding (Duffield et al., 2003). The inconsistent effects of monensin supplementation might be associated with interaction between monensin and other dietary factors such as unsaturated fat, effective fiber, or starch availability (Mathew et al., 2011).

140

Dietary fats are hydrolyzed by lipase produced by rumen microbes and plants. As a result, polyunsaturated fatty acids (PUFA) are released. These PUFA are converted into saturated fatty acids through BH by rumen microbes. Fats such as long-chain fatty acids are also used to reduce methane production and the ratio of acetate and propionate through manipulation of ruminal fermentation (Nagaraja et al., 1997). However, some microbes including cellulolytic bacteria can be inhibited by PUFA (Maia et al., 2007).

Supplementation of both monensin and fat might worsen MFD compared to supplementation of monensin alone (AlZahal et al., 2008). However, this finding is not supported by a recent study (Mathew et al., 2011).

The objective of this study was to examine the effects of monensin alone or monensin in combination with fat on the bacterial communities present in the liquid and the adherent fractions recovered from Holstein dairy cattle using pyrosequencing analysis.

8.3 Materials and Methods:

8.3.1 Sample collection

All rumen samples were collected in a previous study and kindly provided by Dr.

Eastridge, The Ohio State University. Briefly, whole rumen contents were obtained from six cannulated lactating Holstein cattle that were used in a 6 x 6 balanced Latin square design experiment with a 3-week period adaptation on each diet (Mathew et al., 2011).

Three of the six dietary treatments (6 cattle × 3 diets) were selected to examine ruminal bacterial diversity: (1) C = control diet containing ground corn and long alfalfa hay, (2)

CR = C plus monensin as Rumensin (12 g/ 909 kg DM), and (3) CRF = CR plus 4% fat.

141

Detailed dietary treatments were shown in Table 8.1. The rumen sample obtained from each treatment was divided into two fractions (the liquid fraction vs. the adherent fraction) as described previously (Kim et al., 2011c). Bacteria present in the liquid fraction (Lq) was obtained by centrifugation, while the remaining solid digesta was used to recover bacteria adherent to solid digesta (Ad) using a detaching buffer (Larue et al., 2005; Kim et al., 2011c).

8.3.2 Metagenomic DNA extraction

Metagenomic DNA was extracted from each fractionated sample (6 cattle × 3 diets

× 2 fractions = 36 samples) as described previously (Yu and Morrison, 2004b). The DNA extracts were pooled based on diets and fractions, resulting in the following six composite samples: the liquid fraction and the adherent fraction recovered from six cattle fed with C (Lq-C and Ad-C), CR (Lq-CR and Ad-CR), and CRF (Lq-CRF and Ad-CRF).

8.3.3 Pyrosequencing

Each composite DNA sample was amplified with universal primers 27F (5’-

Adaptor primer-AKRGTTYGATYNTGGCTCAG-3’) and 519R (5’-Adaptor primer- barcode-GTNTBACCGCDGCTGCTG-3’). All 454 sequencing primers designed are listed in Table A in the appendix. PCR reaction mixtures (50-ul) consisted of 100 ng composite DNA, 1X PCR buffer (20 mM Tris-HCl [pH 8.4] and 50 mM KCl), 200 uM deoxynucleoside triphosphates, 100 nM (each) primer, 1.75 mM MgCl2, 670 ng of bovine serum albumin/ul, and 1.25 U of Platinum Taq DNA polymerase (Invitrogen

Corporation, Carlsbad, CA). PCR amplification of the rrs V1-V3 region was achieved

142 with 30 cycles of PCR (denaturation at 95 oC for 30 s; annealing at 60 oC for 30 s; and extension at 72 oC for 60 s) using a PTC-100 thermocycler (MJ Research, Waltham,

Mass.). All amplicons were purified using a QIAquick Gel Extraction Kit (Qiagen,

Valencia, CA). The gel-purified amplicons were quantified using a NanoDrop ND-1000

UV-Vis Spectrophotometer (NanoDrop products, Wilmington, DE). Amplicons from each composite sample were diluted with the Elution buffer of the QIAquick Gel

Extraction Kit to a concentration of 20ng/ µl, and an amplicon pool was prepared by combining an equal amount of each diluted amplicons. This pool was sequenced at the

Plant-Microbe Genomics Facility, The Ohio State University using the 454 GS FLX

Titanium system (designated as “Dataset A”). The same six composite DNA was also sequenced using a 454 GS FLX Titanium system at Research and Testing Laboratories

(Lubbock, TX), resulting in “Dataset B”. The composite DNA was sequenced again by

Research and Testing Laboratories (Lubbock, TX) on a different PTP plate, resulting in

“Dataset C”. The three datasets were analyzed individually to assess the repeatability of the pyrosequencing analysis and then were combined together and analyzed to examine the effect of monensin alone or monensin plus fat on the ruminal bacteriome.

8.3.4 Sequence processing and bioinformatics analysis

The QIIME program (Caporaso et al, 2010) was used to trim off the barcodes and primers and sort the six libraries from returned sequences. Sequences with read length shorter than 200nt or longer than 650nt, mean quality score below Q25, and homopolymer stretches longer than 8nt were excluded. The sequences that passed the above quality control were aligned against the Greengenes core dataset using the NAST

143 algorithm as implemented in the Mothur program (Schloss et al., 2009). Sequences suspected to have resulted from pyrosequencing errors or to be chimeric sequences were removed from the aligned file using the pre.cluster and the chimera.slayer commands as implemented in the Mothur program (Schloss et al., 2009). The same number of sequences for each fraction treatment was selected to normalize sequence data using the sub.sample command as implemented in the Mothur program (Schloss et al., 2009).

Hierarchical taxa assignment was conducted using the RDP naïve Bayesian rRNA

Classifier as implemented in the QIIME program (Caporaso et al., 2010), and species- level OTUs were calculated at 0.04 distance as recommended previously (Kim et al.,

2011a) using the furthest neighbor algorithm as implemented in the Mothur program

(Schloss et al., 2009). A rarefaction curve and alpha diversity indices were calculated for each library using the rarefaction.single and the summary.single commands as implemented in the Mothur program (Schloss et al., 2009). A full description for sequence processing is available in Appendix A. Principal coordinate analysis (PCoA) was conducted using the unweighted UniFrac method (Lozupone and Knight, 2005) as described previously (Kim et al., 2011b).

8.3.5 Comparison among three datasets

Comparisons among the three datasets were performed using PCoA as implemented in the UniFrac program (Lozupone and Knight, 2005). A phylogenetic tree was constructed using the ARB program (Ludwing et al., 2004) and then used as an input file to run the UniFrac program as described previously (Kim et al., 2011b).

144

8.4 Results and Discussion:

8.4.1 Data summary

After removing sequences resulting from pyrosequencing errors or chimeric sequences, 24,792 sequences (4,132 sequences × 6 fractions) were obtained from the six sample fractions (Table 8.2). Approximately 53% of the 24,792 sequences were assigned to the phylum Firmicutes, while 26% of them were assigned to the phylum Bacteroidetes

(Figure 8.1). The predominance of these two phyla corroborates the previous study (Kim et al., 2011b). About 17% of the total sequences could not be classified into any known phylum (Figure 8.1). The other phyla accounted for less than 2% of the total sequences.

In total, 9,867 OTUs were identified at 0.04 distance from all the sequences. The number of OTUs for each fraction ranged from 2,395 to 2,726 (Table 8.2). More than 70% of all the OTUs were represented by singletons (Table 8.2). Firmicutes and Bacteroidetes were represented by 4,952 and 2,657 OTUs, respectively, and the OTUs of these two phyla accounted for 77% of the total OTUs. The maximum number of OTUs estimated from a rarefaction curve for each fraction ranged from 4,392 to 5,755 (Table 8.2).

8.4.2 Firmicutes

Firmicutes was the most abundant phylum across all the six fractions, accounting for 40-75% of all the sequences. The proportion of Firmicutes was higher in the Ad fractions than in the Lq fractions and the highest in the Ad-CRF fraction (Figure 8.1).

Because both cultured and uncultured fibrolytic bacteria are assigned mostly to

Firmicutes (Kim et al., 2011b), the higher proportion of Firmicutes in the Ad fractions than in the Lq fractions is not surprising. The proportion of Firmicutes was higher in

145 cattle fed CRF than in cattle fed C or CR, and the predominance of Firmicutes was not lower in the CR or the CRF samples than in the C samples (Figure 8.1). These findings are not expected because monensin was proposed to selectively inhibit Gram-positive bacteria and most members of phylum Firmicutes are Gram-positive bacteria (Wolf et al.,

2004). The mode of action of monensin in the rumen may differ from in vitro cultures and needs to revise. The high predominance of Firmicutes in the CRF samples than in the

C and the CR samples is also intriguing. It is logic to conclude that Firmicutes was stimulated, while other bacterial phyla, most of which are Gram-negative bacteria, were not affected. Alternatively, Firmicutes was not affected, while other bacterial phyla were inhibited. Phylum-specific quantitative analysis is needed to test the above notions. It should be noted that fats are degraded through lipolysis and resulting unsaturated fatty acids inevitably is subjected to biohydrogenation (BH) in the rumen. As well known examples, Anaerovibrio lipolytica and Butyrivibrio fibrisolvens, both are Gram-positive bacteria of Firmicutes, are involved in lipolysis and BH, respectively. Supplementation of fats could result in increases in these two species and related species within these two genera. Numerous uncultured bacteria including members of Anaerovibrio and

Butyrivibrio can also be involved in lipolysis and BH, resulting in high abundance of

Firmicutes in the Ad-CRF and Lq-CRF fractions. However, the lower relative abundance of Butyrivibrio (Figure 8.2) does not support this premise.

The majority of Firmicutes sequences were assigned to unclassified Clostridiales, unclassified Lachnospiraceae, or unclassified Ruminococcaceae (Figure 8.2), with

Unclassified Clostridiales, unclassified Lachnospiraceae and unclassified

Ruminococcaceae being represented by 1,856, 1,537 and 581 OTUs, respectively, in

146 combination across all the 6 fractions. This result supports the predominance of these three unclassified groups as reported previously (Kim et al., 2011b). The proportions of all the three groups were higher in the Ad fraction than in the Lq fraction (Figure 8.2), suggesting their potentially important role in fiber degradation as maintained previously

(Kim et al., 2011b). The proportion of unclassified Clostridiales or unclassified

Ruminococcaceae was higher in cattle fed CRF than in cattle fed C or CR. Ruminal bacteria involved in lipolysis or BH within these two groups might have increased due to the fat supplementation. The three unclassified groups each included many unique OTUs detected only in cattle fed C, CR or CRF. The unique OTUs detected only in cattle fed C might represent bacteria that are sensitive to monensin, while the unique OTUs detected only in cattle fed CR and in cattle fed CRF likely represent monensin-resistant bacteria.

The unique OTUs detected only in the cattle fed CRF may also be associated with lipolysis and/or BH. The use of a reverse metagenomic approach (Nichols, 2007; Pope et al., 2011) may help isolate these unique OTUs.

The proportion of Gram-positive Butyrivibrio was lower in the Ad-CR fraction than in the Ad-C fraction (Figure 8.2). Some Butyrivibrio strains involved in BH appeared to be inhibited by monensin as described previously (Weimer et al., 2008). The proportion of Butyrivibrio decreased in cattle fed CRF (Figure 8.2). The supplementation of additional fats might have inhibited Butyrivibrio strains. The proportion of another unknown bacteria rather than Butyrivibrio might increase for BH. Genus Butyrivibro was represented by 107 OTUs in combination across all the 6 fractions. Although the sequence distribution of Butyrivibrio was similar between the Lq-C and the Lq-CR fractions, the number of OTUs was greater in the Lq-CR fraction than in the Lq-C

147 fraction due to 15 unique OTUs detected only in the Lq-CR fraction (Figure 8.2). These

15 OTUs might be strains highly resistant to monensin. Although the number of sequences and OTUs was less in cattle fed CRF than in cattle fed C or CR, 21 unique

OTUs were detected only in cattle fed CRF and they may be monensin-resistant as well as associated with BH.

Sequences classified to any known genera within Firmicutes accounted for no more than 1% of total sequences, except for Gram-negative genus Succiniclasticum that ferments succinate to propionate (van Gylswyk, 1995). The proportion of

Succiniclasticum was lower in the Lq-C fraction than in the Lq-CR and the Lq-CRF fractions and the highest in the Lq-CRF fraction among all the fractions (Figure 8.2). The high abundance of Succiniclasticum in the Lq-CRF fraction might have resulted from increase of succinate, the substrate of Succiniclasticum. Succiniclasticum was represented by 101 OTUs in combination across all the 6 fractions, and 7 OTUs of them were represented by more sequences in cattle fed CR or in cattle fed CRF than in cattle fed C.

These 7 OTUs might be monensin-resistant and take up the niche left by monensin- sensitive bacteria.

Genus Anaerovibrio was represented by 10 OTUs in combination across all the fractions, and 6 of them were detected only in the Lq-CRF fraction. Gram-negative

Anaerovibrio lipolytica is a major lipolytic species and it produces succinate and propionate main fermentation products (Russell, 2002). As expected, the proportion of

Anaerovibrio sequences was the highest in the Lq-CRF fraction among all the fractions.

Anaerovibrio spp. may positively interact with Succiniclasticum in the rumen when fed monensin and fats. However, future studies are needed to verify such a positive

148 relationship.

The proportion of Mogibacterium was higher in cattle fed CR or CRF than in cattle fed C. This result indicates that Gram-positive Mogibacterium is not inhibited by monensin. Mogibacterium was found more abundant in the Ad fractions than in the Lq fractions. Mogibacterium is not fibrolytic. Its preferential occurrence in the Ad fraction cannot be explained. Because it produces phenyl propionic acid (PPA) and phenyl acetic acid (PAA) (Larue et al., 2005), both of which are needed by R. albus for to adhere and degrade fiber, there might be a mutualism between these two groups of bacteria in the rumen. Mogibacterium was represented by 33 OTUs in combination across all the 6 fractions. Five of the 33 OTUs were detected only in the Ad-CR fraction, while another

12 were detected only in the Ad-CRF fraction. All these OTUs should be monensin resistant. The latter 12 OTUs could play a role in BH.

The proportion of Syntrophococcus was the highest in the Ad-C fraction and decreased in the Ad-CR and the Ad-CRF fractions. Syntrophococcus seemed to be inhibited by monensin in terms of sequences and OTUs (Figure 8.2). Because both Gram- negative S. sucromutans and Gram-positive Eubacterium cellulosolvens were representative species of genus Syntrophococcus in the RDP database (Kim et al., 2011b),

Syntrophococcus sequences identified in this study might be phylogenetically related to S. sucromutans or E. cellulosolvens. Gram-positive Anaerovorax was not inhibited by monensin and its population increased in the Ad-CRF fraction (Figure 8.2). This result indicates that Anaerovorax may be associated with fiber degradation as well as BH.

Anaerovorax was represented by 20 OTUs, and 7 of them were detected only in the Ad-

CRF fraction. Ruminococcus was represented by less than 10 sequences in each fraction

149 and its population decreased in the rumen of cattle fed CRF, suggesting that

Ruminococcus is sensitive to monensin, which is consistent with the previous finding

(Weimer et al., 2008). Supplementation of the fats further decreased the frequency of

Ruminococcus sequences, corroborating the inhibitory effect of fats towards

Ruminococcus (Maia et al., 2007). Ruminococcus was represented by 15 OTUs in combination among all the fractions, and 5 OTUs of them were detected only in the rumen of cattle fed CR. These OTUs could be monensin-resistant members of

Ruminococcus.

Although previous studies based on pure cultures have reported that monensin inhibits Gram-positive bacteria more than Gram-negative bacteria (Nagaraja et al., 1997;

Callaway et al., 2003), some in vivo studies (Stahl et al., 1988; Weimer et al., 2008) do not support the above generalization. Our study supports provided further evidence that some Gram-positive ruminal bacteria are monensin resistant.

8.4.3 Bacteroidetes

Bacteroidetes was the second most abundant phylum across all the six fractions, accounting for 10-43% of all the sequences. The proportion of Bacteroidetes was higher in the Lq fractions than in the Ad fractions and the highest in the Lq-CR fraction (Figure

8.1), consistent with the finding of Kim et al. (2011b). The proportions of Bacteroidetes were similar between cattle fed C and cattle fed CR, whereas the proportion of

Bacteroidetes in cattle fed CRF highly decreased (Figure 8.1). All these results suggest that Bacteroidetes as a whole is not sensitive to monensin but can be inhibited by dietary fats (Figure 8.1).

150

Most sequences classified to Bacteroidetes were assigned to unclassified

Bacteroidetes, unclassified Bacteroidales, or unclassified Prevotellaceae (Figure 8.3).

The populations of these three groups were more abundant in the Lq fractions than in the

Ad fractions (Figure 8.3). The proportions of these three groups decreased in cattle fed

CRF compared to cattle fed C or CR. The supplementation of the fats seemed to inhibit ruminal bacteria assigned to these three groups. Most of the Bacteroidetes OTUs were also assigned to these three groups: unclassified Bacteroidetes (801 OTUs), unclassified

Bacteroidales (722 OTUs), and unclassified Prevotellaceae (575 OTUs). These three groups were represented by many unique OTUs detected only in cattle fed C, CR or CRF, indicating that some bacterial strains were affected by monensin or monensin in combination with fats.

Prevotella was the most abundant genus among known genera as described previously (Stevenson et al., 2007; Kim et al., 2011b). The proportion of Prevotella was higher in the Lq fractions than in the Ad fractions (Figure 8.3). The proportion of

Prevotella was higher in cattle fed CR than in cattle fed C (Figure 8.3), and this result is consistent with the previous study (Weimer et al., 2008). However, the proportion of

Prevotella decreased in cattle fed CRF compared to cattle fed CR (Figure 8.3). Prevotella was represented by 375 OTUs in combination across all the 6 fractions, and the number of unique OTUs detected only in cattle fed C, CR and CRF was 67, 124 and 75, respectively. The high number of unique OTUs in cattle fed CR indicates that many

Prevotella strains resistant to monensin are stimulated and they may play an important role in ruminal fermentation when monensin is fed to cattle.

151

8.4.4 Minor phyla

The phylum TM7 has been proposed from rrs sequence data and recognized as

Gram-positive bacteria (Hugenholtz et al., 2001). Because the proportion of TM7, which was represented by 121 OTUs, was the highest in the Ad-C fraction, it may be associated with fiber degradation. However, TM7 was inhibited by monensin or monensin in combination with fat (Figure 8.4). The proportion of the phylum SR1 was the highest in the Lq-C fraction, and the population of SR1 may be amylolytic. SR1 was also inhibited by monensin or monensin in combination with fat (Figure 8.4). SR1 was represented by

17 OTUs, and 2 of the OTUs were predominant. Unclassified within the phylum Proteobacteria was highly detected in the Lq-C fraction and greatly inhibited by monensin or monensin in combination with fats, but it was not detected in the Ad fractions (Figure 8.4). Members of the Enterobacteriaceae are Gram-negative and contains many such as Salmonella and Escherichia coli. Although E. coli

O157:H7 (ATCC 43895) and Salmonella (ATCC 6960) were not inhibited by monensin

(Edrington et al., 2003), monensin may inhibit other pathongenic strains within unclassified Enterobacteriaceae. Unclassified Enterobacteriaceae was represented by 4

OTUs, and 2 of the OTUs were predominant in the Lq-C fraction based on the number of sequences. However, the 2 predominant OTUs were highly inhibited by monensin. Small number of sequences was assigned to Fibrobacter, and this result could result from PCR bias against members of this phylum as described previously (Larue et al., 2005).

8.4.5 Comparison among the diets

To compare the bacterial diversity among the six fractions, fraction correlations

152 were determined using PCoA by the unweighted UniFrac method. P1 explained 23.56% variation and seemed to separate the liquid fractions from the adherent fractions, while P2 explained 20.81% variation and separated the six fractions based on different diets

(Figure 8.5). This result further indicates that the ruminal bacterial communities were affected by monensin or monensin in combination with fats. Although the sequence distribution was similar between cattle fed C and cattle fed CR (Figure 8.1), their separation along P2 might be attributed to different OTU distributions between these two diets. Callaway et al. (2003) indicated that highly monensin-resistant bacterial strains are stimulated to adapt to the monensin-rich environment. Therefore, the different OTU distributions between cattle fed C and cattle fed CR could result from the selection of ruminal bacterial strains that are highly resistant to monensin.

8.4.6 Comparison among three datasets

Comparison by PCoA among the three datasets was shown in Figure 8.6. P1 separated the Dataset A from the Dataset B and or C, but P1 only explained 10.81% of the variation. Axis P2 separated the liquid fractions from the adherent fractions across all the three datasets and explained 8.95% variation. These results suggest that the differences among the three datasets are relatively small (<10%) and pyrosequencing analysis can be repeatable. Slight differences between Dataset A and the other two datasets might result from difference in barcode sequences, PCR condition (including primers), or pyrosequencing procedures. The separation of the Ad fractions from the Lq fractions could be attributed to bias caused by differences in bacterial composition between these two fractions.

153

8.5 Conclusions:

Ruminal bacterial communities were affected by different diets and fractions. Many uncultured bacteria present in the adherent fraction may be associated with fiber degradation. The sequence distribution of cattle fed CR was similar to that of cattle fed C.

Most members of both Gram-positive Firmicutes and Gram-negative Bacteroidetes did not seem to be affected by monensin. Although previous in vitro studies have showed that Gram-positive bacteria are sensitive to monensin, our study did not support that conclusion. Difference in OTU distribution between cattle fed C and cattle fed CR indicated that highly monensin-resistant strains were stimulated in the presence of monensin. The proportion of Firmicutes was higher in cattle fed CRF than in cattle fed C or CR, whereas that of Bacteroidetes was lower in cattle fed CRF than in cattle fed C or

CR. Therefore, many uncultured bacteria within Firmicutes might be involved in lipolysis or biohydrogenation, whereas many uncultured bacteria within Bacteroidetes could be inhibited by dietary fats. The OTUs affected by monensin or fats will need to be isolated and characterized to further understand the mode(s) of action of monensin.

154

Diet, % of DM1 Ingredient C CR CRF Alfalfa hay 15.0 15.0 15.0 Corn silage 35.0 35.0 35.0 Corn, ground 13.6 11.6 11.6 Soybean mean, 48% CP 8.10 8.10 - Dried distillers grains 10.0 10.0 - Dried distillers grains with solubles - - 16.0 Roasted soybeans, cracked - - 12.8 Animal-vegetable blend - - 0.70 Soybean hulls 16.3 16.3 5.16 Dicalcium phosphate 0.68 0.68 0.14 Limestone 0.36 0.36 0.84 Magnesium oxide 0.20 0.20 0.14 Potassium sulfate 0.16 0.16 - Trace mineralized salt1 0.50 0.50 0.50 Vitamin A2 0.02 0.02 0.02 Vitamin D3 0.04 0.04 0.04 Vitamin E4 0.06 0.06 0.06 Rumensin premix5 - 2.00 2.00

1Included 0.10% Mg; 38.0% Na; 58.0% Cl; 0.04% S; 5,000 mg of Fe/kg; 7,500 mg of Zn/kg; 2,500 mg of Cu/kg; 6,000 mg of Mn/kg; 100 mg of I/kg; 60 mg of Se/kg; and 50 mg of Co/kg. 2Included 30,000 IU of vitamin A/g. 3Included 3,000 IU of vitamin D/g. 4Included 44 IU of vitamin E/g. 5Elanco Animal Health, Greenfield, IN (initially formulated to provide 12 g of monensin/909 kg of total dietary DM; based on analysis of the premix, provided 14 g/909 kg).

Table 8.1: Ingredient composition of dietary treatments (Reproduced from Mathew et al., 2011).

155

Sample Individual # of Singletons Doubletons Chao 1 Rarefaction Phylogenetic Sequences OTUs Estimatea distance

Ad-C 4132 2726 2090 327 9382 5755 71.24

Ad-CR 4132 2551 1870 352 7501 4895 71.00

Ad-CRF 4132 2395 1730 333 6873 4392 55.19

Lq-C 4132 2573 1913 378 7398 5182 70.40

Lq-CR 4132 2682 2047 368 8357 5764 77.92

Lq-CRF 4132 2544 1891 368 7386 5126 67.63

a: Maximum number of OTUs estimated from rarefaction curves (Kim et al., 2011b)

Table 8.2: Sequence data and alpha diversity indices for the six fractions.

156

100%

90%

80%

70%

60% Actinobacteria 50% TM7 Bacteroidetes 40% U_Bacteria Firmicutes 30%

20%

10%

0% Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF Total Fraction

Figure 8.1: Sequence distribution at the phylum level for each fraction. Phyla accounting for less than 1% of all the sequences were not included.

157

U_Clostridiales U_Lachnospiraceae 1400 800 1200

1000 600 800 400 600

400 200 200 0 0 Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF

No. of sequences No. of OTUs No. of sequences No. of OTUs

U_Ruminococcacea Butyrivibrio 500 60

e 400 50

40 300 30 200 20 100 10 0 0

Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF

No. of sequences No. of OTUs No. of sequences No. of OTUs

300 Succiniclasticum 35 Mogibacterium 250 30 200 25 20 150 15 100 10

50 5 0 0 Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF

No. of sequences No. of OTUs No. of sequences No. of OTUs

Figure 8.2: The distribution of sequences and OTUs at genus or the lowest classifiable rank in the phylum Firmicutes. Minor taxa accounting for small numbers of sequences were not included.

Figure 8.2 Continued 158

Figure 8.2 Continued

Syntrophococcus Ruminococcus 35 10 30 8 25 20 6 15 4 10 2 5 0 0 Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF

No. of sequences No. of OTUs No. of sequences No. of OTUs

Anaerovibrio Anaerovorax 8 14 12 6 10 8 4 6 2 4 2 0 0 Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF

No. of sequences No. of OTUs No. of sequences No. of OTUs

159

U_Bacteroidetes U_Bacteroidales 600 500 500 400 400 300 300 200 200 100 100 0 0 Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF

No. of sequences No. of OTUs No. of sequences No. of OTUs

U_Prevotellaceae Prevotella 500 400 400 300 300 200 200 100 100 0 0 Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF

No. of sequences No. of OTUs No. of sequences No. of OTUs

Figure 8.3: The distribution of sequences and OTUs at genus or the lowest classifiable rank in the phylum Bacteroidetes. Minor taxa accounting for small numbers of sequences were not included.

160

TM7 200

150

100

50

0 Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF

No. of sequences No. of OTUs

SR1 50 40 30 20 10 0 Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF

No. of sequences No. of OTUs

U_Enterobacteriaceae 80

60

40

20

0 Ad-C Ad-CR Ad-CRF Lq-C Lq-CR Lq-CRF

No. of sequences No. of OTUs

Figure 8.4: The distribution of sequences and OTUs at the lowest classifiable rank in minor phyla. Minor taxa accounting for small numbers of sequences were not included.

161

Figure 8.5: Principal coordinates analysis for the six fractions.

162

Ad-CRF (B)

Ad-CRF (C)

Ad-CRF (A) Ad-CR (B) Ad-C (C) Ad-CR (C) Ad-C (B)

Ad-CR (A)

Lq-CRF (B) Ad-C (A) Lq-CRF (C)

Lq-CRF (A) Lq-CR (C) Lq-CR (B)

Lq-C (C) Lq-C (B) Lq-C (A) Lq-CR (A)

Figure 8.6: Principal coordinates analysis for comparison among the three datasets

163

CHAPTER 9

GENERAL DISCUSSION

Since 16S rRNA gene (rrs) sequences were applied to investigation of the diversity of ruminal microbiome, numerous studies have been conducted to better understand the complex ruminal microbiome. Using an integrated approach consisting of conventional molecular biology techniques and emerging metagenomic and microarray technologies, we conducted a series of studies on the ruminal microbiome. The goal of these studies was to provide a new insight into the diversity of the ruminal bacteriome and the effect of diets and feed additives. Such knowledge might help improve rumen function in the future.

Most of the previous studies have examined the diversity of the ruminal microbiome using DGGE and rrs clone libraries. However, these techniques produced only small numbers of sequences, and as such only limited microbial diversity was explained. For that reason, we performed a meta-analysis of all the rrs sequences that were publicly available to better define ruminal microbial diversity. More than 55% of all the rrs sequences could not be classified into any known genus as described in Chapter 3.

A large number of operational taxonomic units (OTUs) consisting of uncultured bacterial sequences were predominant based on the number of sequences. Future studies are needed to verify the predominance of these OTUs. A representative sequence of each predominant OTU can be used to design PCR primers and probes for quantification of abundance using a real-time PCR assay. In addition, OTUs that may be associated with cellulose degradation could be selected using the meta-analysis. The numerous OTUs assigned to unclassified Clostridiales, unclassified Lachnospiraceae, or unclassified

164

Ruminococcaceae, especially those predominant OTUs found in the adhering fractions, are good candidates for isolation. The population dynamics of the OTUs putatively involved in cellulose degradation can be determined in a comparative manner in different nutritional studies so that their niche can be inferred. Further, the dataset consists of entirely sequences generated using the Sanger technology and has been stored as an ARB database. This dataset can be updated regularly as new high-quality sequences from rumen are available in the RDP database. This database can serve as a central depository of rrs sequences of rumen origin. A meta-analysis using multiple datasets of large numbers of sequences in the future may help define a “core microbiome”, which is not achievable in this study. This core microbiome might play an important role in rumen function.

Using real-time PCR assays, we showed that some uncultured bacteria can be as predominant as major cultured bacteria, and several of the analyzed uncultured bacteria are more predominant in the adhering fraction than in the liquid fraction (Chapter 4).

Further studies are needed to verify if they actually attach to plant biomass in the rumen.

Probes for fluorescence in situ hybridization (FISH) can be designed from the rrs sequences of select uncultured bacteria. Their attachment to plant biomass in the rumen can be assessed using FISH. Uncultured bacteria of interest can also be isolated using a probe-guided approach or a reverse metagenomic approach as described previously

(Nichols, 2007; Pope et al., 2011) for detailed studies of their metabolism and physiology.

More than 3,500 OTUs were identified from the bacterial rrs sequences retrieved from the RDP database, and yet these OTUs represented only approximately 70% of ruminal bacterial diversity (Chapter 3). Therefore, novel ruminal OTUs remains to be

165 identified. The study in Chapter 5 showed that conventional rrs clone libraries can still help identify novel OTUs even though small numbers of rrs clones were used. Therefore, clone library-based studies are still very useful in finding novel OTUs from the ruminal microbiome, especially from animals fed different diets in different geographic regions.

These novel OTUs can be used to design new primers or probes for rrs-based analysis such as real-time PCR assays, FISH, and phylogenetic microarray. The predominance of the novel OTUs, and thus their ecological importance, can be evaluated using specific real-time PCR assays as described in Chapter 4. The novel OTUs also can be used to update new probes for the RumenArray (Chapter 6). It should be noted that although pyrosequencing can generate sequences more cost effectively and large number of sequences can be produced from single samples, the high percentage of artifactual sequences (Kunin et al., 2010) will compound diversity analysis and prevent design of robust and accurate primers or probes. Therefore, clone library-based studies of ruminal microbiome should continue until accurate massively parallel sequencing technologies become available.

A phylogenetic microarray (RumenArray) was developed based on the OTUs identified from the meta-analysis described in Chapter 3 (Chapter 6). Numerous OTUs that could not be classified into a known genus were detected by the RumenArray, and most of the detected OTUs were assigned to unclassified Clostridiales, unclassified

Lachnospiraceae, or unclassified Ruminococcaceae within the phylum Firmicutes. The data from the RumenArray analysis corroborates the predominance of the OTUs assigned to these three groups. The RumenArray showed differential distribution of predominant

OTUs between different ruminal fractions and sheep fed different diets. These

166 preliminary results showed that the RumenArray will be a new tool in comprehensive analysis of predominant ruminal bacteria in a semi-quantitative manner. Feed efficiency is important to sustainability and economic viability of dairy and beef industries. The

RumenArray can help identify important core species or microbiome structure features that are associated with feed efficiency because several hundreds of different bacteria can be detected in a sample. As ruminal rrs sequences constantly increase in the RDP database, the specificity of the designed microarray probes will need to be regularly checked (Dugat-Bony et al., 2011). Additionally, as the RDP database is being updated on a regular basis, new microarray probes will need to be regularly designed and added to the current version of RumenArray (version 1).

OTUs are typically calculated from full-length rrs sequences at a phylogenetic distance (0.03-species, 0.05-genus, 0.10-family), but current pyrosequencing system produces partial sequences (700 bp or shorter). Many sequence reads produced from the

Sanger technology are also no longer than 500 bp. Given that sequence divergence is not distributed evenly along the rrs gene, we evaluated and identified partial rrs sequence region(s) and corresponding distance level(s) that produce comparable results as nearly- full length sequences (Chapter 7). For bacterial analysis, the V1-V3 region at 0.04 distance produced comparable results as the nearly-full length region. The use of the 0.04 distance will help reduce overestimate of OTUs richness defined at 0.03 distance in future studies. For archaeal analysis, the V1-V3 at 0.03 distance and the V4-V7 at 0.02 distance were recommended. If these recommended partial regions and distances are used in future studies, comparisons among different research groups will be more accurate and future meta-analyses will be more greatly facilitated.

167

Cattle have been typically fed diets containing monensin to improve production efficiency. Although many in vitro studies have showed that Gram-positive bacteria are more sensitive to monensin than Gram-negative bacteria, this finding was not supported by two in vivo studies (Stahl et al., 1998; Weimer et al., 2008). Therefore, we investigated the effect of monensin, either alone or in combination with dietary fats, using pyrosequencing analysis (Chapter 8). The result showed that some Gram-positive bacteria were not affected by monensin. However, the OTU distribution within a genus or an unclassified group was affected by monensin, indicating the stimulation of highly monensin-resistant bacteria. Further analysis of monensin-resistant bacteria is needed to verify their abundance using real-time PCR assays. Pure cultures of monensin-resistant bacteria will help further understand the mode(s) of action of monensin in the rumen. In addition, the examination of the effect of monensin plus fats on the diversity of ruminal bacteriome provided an opportunity for us to investigate the interaction between monensin and fats. Effects of the dietary fats were inferred from comparison between the monensin group and the monensin-fat group. Future studies will benefit from inclusion of a fats only group. Understanding the metabolism of fatty acids is important because they affect composition of meats and milk. The greater population of Firmicutes that increased corresponding to feeding monensin and fats suggests selection of some bacteria of Firmicutes in fat hydrolysis and subsequent BH. Isolation and studies of pure culture involved in fat metabolism will contribute to developing accurate modeling of lipolysis and biohydrogenation in future nutritional studies.

Manipulation of ruminal fermentation is necessary and highly sought after to improve rumen functions and enhance animal productivity. A better understanding of the

168 ruminal microbiome will help develop efficient fermentation modifiers. The integrated investigation reported in this dissertation advanced our knowledge and understanding of ruminal microbiome and developed new tools (real-time PCR approach and

RumenArray), which may be very useful in future studies.

169

WORKS CITED

Adamczyk J, Hesselsoe M, Iversen N, et al. (2003) The isotope array, a new tool that

employs substrate-mediated labeling of rRNA for determination of microbial

community structure and function. Applied and Environmental Microbiology 69:

6875-6887.

Allison MJ, Dawson KA, Mayberry WR & Foss JG (1985) Oxalobacter formigenes gen.

nov., sp. nov.: oxalate-degrading anaerobes that inhabit the gastrointestinal tract.

Archives of Microbiology 141: 1-7.

Allison MJ, Mayberry WR, Mcsweeney CS & Stahl DA (1992) Synergistes jonesii gen.

nov., sp. nov.: a rumen bacterium that degrades toxic pyridinediols. Systematic

and Applied Microbiology 15: 522-529.

AlZahal O, Odongo NE, Mutsvangwa T, et al. (2008) Effects of monensin and dietary

soybean oil on milk fat percentage and milk fatty acid profile in lactating dairy

cows. Journal of Dairy Science 91: 1166-1174.

Amann RI, Lin CH, Key R, Montgomery L & Stahl DA (1992) Diversity among

Fibrobacter Isolates - Towards a Phylogenetic Classification. Systematic and

Applied Microbiology 15: 23-31.

Ashelford KE, Chuzhanova NA, Fry JC, Jones AJ & Weightman AJ (2005) At least 1 in

20 16S rRNA sequence records currently held in public repositories is estimated

to contain substantial anomalies. Applied and Environmental Microbiology 71:

7724-7736.

Baker GC, Smith JJ & Cowan DA (2003) Review and re-analysis of domain-specific 16S

primers. Journal of Microbiological Methods 55: 541-555.

170

Bekele AZ, Koike S & Kobayashi Y (2010) Genetic diversity and diet specificity of

ruminal Prevotella revealed by 16S rRNA gene-based analysis. FEMS

Microbiology Letters 305: 49-57.

Bekele AZ, Koike S & Kobayashi Y (2011) Phylogenetic diversity and dietary

association of rumen Treponema revealed using group-specific 16S rRNA gene-

based analysis. FEMS Microbiology Letters 316:51-60.

Bera-Maillet C, Ribot Y & Forano E (2004) Fiber-degrading systems of different strains

of the genus Fibrobacter. Applied and Environmental Microbiology 70: 2172-

2179.

Boeckaert C, Morgavi DP, Jouany JP, Maignien L, Boon N & Fievez V (2009) Role of

the protozoan Isotricha prostoma, liquid-, and solid-associated bacteria in rumen

biohydrogenation of linoleic acid. Animal 3: 961-971.

Bozal N, Montes MJ, Tudela E & Guinea J (2003) Characterization of several

Psychrobacter strains isolated from Antarctic environments and description of

Psychrobacter luti sp. nov. and Psychrobacter fozii sp. nov. International Journal

of Systematic and Evolutionary Microbiology 53: 1093-1100.

Brulc JM, Antonopoulos DA, Miller MEB, et al. (2009) Gene-centric metagenomics of

the fiber-adherent bovine rumen microbiome reveals forage specific glycoside

hydrolases. Proceedings of the National Academy of Sciences of the United States

of America 106: 1948-1953.

Bryant MP, Small N, Bouma C & Robinson IM (1958) Characteristics of ruminal

anaerobic cellulolytic cocci and Cillobacterium cellulosolvens n. sp. Journal of

Bacteriology 76: 529-552.

171

Buchko SJ, Holley RA, McAllister TA, McMahon LR & Veira DM (2000) Effect of

monensin on E. coli O157: H7 and E. coli ATCC 25922. Canadian Journal of

Animal Science 80: 777.

Cai SC & Dong XZ (2010) Cellulosilyticum ruminicola gen. nov., sp. nov., isolated from

the rumen of yak, and reclassification of Clostridium lentocellum as

Cellulosilyticum lentocellum comb. nov. International Journal of Systematic and

Evolutionary Microbiology 60: 845-849.

Callaway TR & Russell JB (2000) Variations in the ability of ruminal Gram-negative

Prevotella species to resist monensin. Current Microbiology 40: 185-189.

Callaway TR, Dowd SE, Edrington TS, et al. (2010) Evaluation of bacterial diversity in

the rumen and feces of cattle fed different levels of dried distillers grains plus

solubles using bacterial tag-encoded FLX amplicon pyrosequencing. Journal of

Animal Science 88: 3977-3983.

Callaway TR, Edrington TS, Rychlik JL, et al. (2003) Ionophores: their use as ruminant

growth promotants and impact on food safety. Currunt Issues Intestinal

Microbiology 4: 43-51.

Calsamiglia S, Busquet M, Cardozo PW, Castillejos L & Ferret A (2007) Essential oils as

modifiers of rumen microbial fermentation. Journal of Dairy Science 90: 2580-

2595.

Caporaso JG, Kuczynski J, Stombaugh J, et al. (2010) QIIME allows analysis of high-

throughput community sequencing data. Nature Methods 7: 335-336.

172

Caporaso JG, Lauber CL, Walters WA, et al. (2011) Global patterns of 16S rRNA

diversity at a depth of millions of sequences per sample. Proceedings of the

National Academy of Sciences of the United States of America 108: 4516-4522.

Castiglioni B, Rizzi E, Frosini A, et al. (2004) Development of a universal microarray

based on the ligation detection reaction and 16S rRNA gene polymorphism to

target diversity of cyanobacteria. Applied and Environmental Microbiology 70:

7161-7172.

Cheng KJ, Akin DE & Costerton JW (1977) Rumen bacteria: interaction with particulate

dietary components and response to dietary variation. Federation Proceedings 36:

193-197.

Cheng YF, Mao SY, Liu JX & Zhu WY (2009) Molecular diversity analysis of rumen

methanogenic Archaea from goat in eastern China by DGGE methods using

different primer pairs. Letters in Applied Microbiology 48: 585-592.

Cho SJ, Cho KM, Shin EC, et al. (2006) 16S rDNA analysis of bacterial diversity in three

fractions of cow rumen. Journal of Microbiology and Biotechnology 16: 92-101.

Claesson MJ, O'Sullivan O, Wang Q, et al. (2009) Comparative analysis of

pyrosequencing and a phylogenetic microarray for exploring microbial

community structures in the human distal intestine. PLoS One 4:e6669.

Cole JR, Wang Q, Cardenas E, et al. (2009) The Ribosomal Database Project: improved

alignments and new tools for rRNA analysis. Nucleic Acids Research 37: D141-

D145.

Dehority BA (2003) Rumen Microbiology, 1st edn. Nottingham University Press,

Nottingham.

173

Dehority BA & Grubb JA (1980) Effect of short-term chilling of rumen contents on

viable bacterial numbers. Applied and Environmental Microbiology 39: 376-381.

Deng WD, Xi DM, Mao HM & Wanapat M (2008) The use of molecular techniques

based on ribosomal RNA and DNA for rumen microbial studies: a

review. Molecular Biology Reports 35: 265-274.

Denman SE & McSweeney CS (2006) Development of a real-time PCR assay for

monitoring anaerobic fungal and cellulolytic bacterial populations within the

rumen. FEMS Microbiology Ecology 58: 572-582.

DeSantis TZ, Hugenholtz P, Keller K, et al. (2006) NAST: a multiple sequence

alignment server for comparative analysis of 16S rRNA genes. Nucleic Acids

Research 34: W394-W399.

Dethlefsen L, Huse S, Sogin ML & Relman DA (2008) The pervasive effects of an

antibiotic on the human , as revealed by deep 16S rRNA

sequencing. PLoS Biology 6:e280.

Dewhirst FE, Paster BJ, Tzellas N, Coleman B, Downes J, Spratt DA & Wade WG

(2001) Characterization of novel human oral isolates and cloned 16S rDNA

sequences that fall in the family Coriobacteriaceae: description of Olsenella gen.

nov., reclassification of of Lactobacillus uli as Olsenella uli comb. nov. and

description of Olsenella profusa sp. nov. International Journal of Systematic and

Evolutionary Microbiology 51: 1797-1804.

Dodd D, Kiyonari S, Mackie RI & Cann IKO (2010) Functional diversity of four

glycoside hydrolase family 3 enzymes from the rumen bacterium Prevotella

bryantii B(1)4. Journal of Bacteriology 192: 2335-2345.

174

Dowd SE, Callaway TR, Wolcott RD, Sun Y, McKeehan T, Hagevoort RG & Edrington

TS (2008) Evaluation of the bacterial diversity in the feces of cattle using 16S

rDNA bacterial tag-encoded FLX amplicon pyrosequencing (bTEFAP). BMC

Microbiology 8:125.

Droege M & Hill B (2008) The genome sequencer FLX (TM) system-longer reads, more

applications, straight forward bioinformatics and more complete data sets.

Journal of Biotechnology 136: 3-10.

Duffield T, Bagg R, Kelton D, Dick P & Wilson J (2003) A field study of dietary

interactions with monensin on milk fat percentage in lactating dairy cattle.

Journal of Dairy Science 86: 4161-4166.

Dugat-Bony E, Peyretaillade E, Parisot N, et al. (2011) Detecting unknown sequences

with DNA microarrays: explorative probe design strategies. Environmental

Microbiology doi:10.1111/j.1462-2920.2011.02559.x.

Edrington TS, Callaway TR, Varey PD, et al. (2003) Effects of the antibiotic ionophores

monensin, lasalocid, laidlomycin propionate and bambermycin on Salmonella and

E. coli O157 : H7 in vitro. Journal of Applied Microbiology 94: 207-213.

Edwards JE, McEwan NR, Travis AJ & Wallace RJ (2004) 16S rDNA library-based

analysis of ruminal bacterial diversity. Antonie Van Leeuwenhoek International

Journal of General and Molecular Microbiology 86: 263-281.

Engelbrektson A, Kunin V, Wrighton KC, Zvenigorodsky N, Chen F, Ochman H &

Hugenholtz P (2010) Experimental factors affecting PCR-based estimates of

microbial species richness and evenness. ISME Journal 4: 642-647.

175

Fellner V, Sauer FD & Kramer JKG (1997) Effect of nigericin, monensin, and tetronasin

on biohydrogenation in continuous flow-through ruminal fermenters. Journal of

Dairy Science 80: 921-928.

Fernando SC, Purvis HT, Najar FZ, et al. (2010) Rumen microbial population dynamics

during adaptation to a high-grain Diet. Applied and Environmental Microbiology

76: 7482-7490.

Firkins JL & Yu Z (2006) Characterisation and quantification of the microbial

populations of the rumen. In Ruminant physiology: Digestion, metabolism and

impact of nutrition on gene expression, immunology and stress. Sejrsen K,

Hvelplund T & Nielsen MO (eds). Wageningen, The Netherlands: Wageningen

Academic Publisher, pp. 19-54.

Firkins JL, Oldick BS, Pantoja J, Reveneau C, Gilligan LE & Carver L (2008) Efficacy

of liquid feeds varying in concentration and composition of fat, nonprotein

nitrogen, and nonfiber carbohydrates for lactating dairy cows. Journal of Dairy

Science 91: 1969-1984.

Flint HJ (1997) The rumen microbial ecosystem - some recent developments. Trends in

Microbiology 5: 483-488.

Frey JC, Pell AN, Berthiaume R, et al. (2010) Comparative studies of microbial

populations in the rumen, duodenum, ileum and faeces of lactating dairy cows.

Journal of Applied Microbiology 108: 1982-1993.

Gentry TJ, Wickham GS, Schadt CW, He Z & Zhou J (2006) Microarray applications in

microbial ecology research. Microbial Ecology 52: 159-175.

176

Gilbert JA, Field D, Huang Y, Edwards R, Li WZ, Gilna P & Joint I (2008) Detection of

Large Numbers of Novel Sequences in the Metatranscriptomes of Complex

Marine Microbial Communities. PLoS One 3:e3042.

Gilbert RA, Ouwerkerk D, Zhang LH, Klieve AV & Genetic CRCB (2010) In vitro

detection and primary cultivation of bacteria producing materials inhibitory to

ruminal methanogens. Journal of Microbiological Methods 80: 217-218.

Gomez-Alvarez V, Teal TK & Schmidt TM (2009) Systematic artifacts in metagenomes

from complex microbial communities. ISME Journal 3: 1314-1317.

Grech-Mora I, Fardeau ML, Patel BKC, et al. (1996) Isolation and characterization of

Sporobacter termitidis gen. nov., sp. nov., from the digestive tract of the wood-

feeding termite Nasutitermes lujae. International Journal of Systematic

Bacteriology 46: 512-518.

He ZL, Gentry TJ, Schadt CW, et al. (2007) GeoChip: a comprehensive microarray for

investigating biogeochemical, ecological and environmental processes. ISME

Journal 1: 67-77.

Henderson C (1971) Study of lipase produced by Anaerovibrio lipolytica, a rumen

bacterium. Journal of General Microbiology 65: 81-89.

Hobson PN & Stewart CS (1997) Rumen Microbial Ecosystem, 2nd edn. Blackie

Academic and Professional, New York. .

Hong SH, Bunge J, Leslin C, Jeon S & Epstein SS (2009) Polymerase chain reaction

primers miss half of rRNA microbial diversity. ISME Journal 3: 1365-1373.

Hugenholtz P, Tyson GW, Webb RI, Wagner AM & Blackall LL (2001) Investigation of

candidate division TM7, a recently recognized major lineage of the domain

177

bacteria with no known pure-culture representatives. Applied and Environmental

Microbiology 67: 411-419.

Hughes JB, Hellmann JJ, Ricketts TH & Bohannan BJM (2001) Counting the

uncountable: Statistical approaches to estimating microbial diversity. Applied and

Environmental Microbiology 67: 4399-4406.

Hungate RE (1950) The anaerobic mesophilic cellulolytic bacteria. Bacteriological

Reviews 14: 1-49.

Hungate RE (1957) Microorganisms in the rumen of cattle fed a constant ration.

Canadian Journal of Microbiology 3: 289-311

Hungate RE & Stack RJ (1982) Phenylpropanoic acid: growth factor for Ruminococcus

albus. Applied and Environmental Microbiology 44: 79-83.

Huse SM, Dethlefsen L, Huber JA, Welch DM, Relman DA & Sogin ML (2008)

Exploring microbial diversity and taxonomy using SSU rRNA hypervariable tag

sequencing. PLoS Genetics 4:e1000255.

Itoh T, Yoshikawa N & Takashina T (2007) Thermogymnomonas acidicola gen. nov., sp.

nov., a novel thermoacidophilic, cell wall-less archaeon in the order

Thermoplasmatales, isolated from a solfataric soil in Hakone, Japan. International

Journal of Systematic and Evolutionary Microbiology 57: 2557-2561.

Janssen PH (2010) Influence of hydrogen on rumen methane formation and fermentation

balances through microbial growth kinetics and fermentation thermodynamics.

Animal Feed Science and Technology 160: 1-22.

Janssen PH & Kirs M (2008) Structure of the archaeal community of the rumen. Applied

and Environmental Microbiology 74: 3619-3625.

178

Jenkins TC, Fellner V & McGuffey RK (2003) Monensin by fat interactions on trans

fatty acids in cultures of mixed ruminal microorganisms grown in continuous

fermentors fed corn or barley. Journal of Dairy Science 86: 324-330.

Jenkins TC, Wallace RJ, Moate PJ & Mosley EE (2008) BOARD-INVITED REVIEW:

Recent advances in biohydrogenation of unsaturated fatty acids within the rumen

microbial ecosystem. Journal of Animal Science 86: 397-412.

Jeyanathan J, Kirs M, Ronimus RS, Hoskin SO & Janssen PH (2011) Methanogen

community structure in the rumens offarmed sheep, cattle and red deer fed

different diets. FEMS Microbiology Ecology Jan 21. doi: 10.1111/j.1574-

6941.2011.01056.x.

Kang S, Denman SE, Morrison M, Yu Z, Dore J, Leclerc M & McSweeney CS (2010)

Dysbiosis of faecal microbiota in Crohn's disease patients as revealed by a custom

phylogenetic microarray. Inflammatory Bowel Disease 16: 2034-2042.

Kemp P, White RW & Lander DJ (1975) The hydrogenation of unsaturated fatty acids by

five bacterial isolates from the sheep rumen, including a new species. Journal of

General Microbiology 90: 100-114.

Khan AW, Meek E, Sowden LC & Colvin JR (1984) Emendation of the genus

Acetivibrio and description of Acetivibrio cellulosolvens sp. nov., a nonmotile

cellulolytic . International Journal of Systematic Bacteriology 34: 419-

422.

Kim M, Morrison M & Yu ZT (2011a) Evaluation of different partial 16S rRNA gene

sequence regions for phylogenetic analysis of microbiomes. Journal of

Microbiological Methods 84: 81-87.

179

Kim M, Morrison M & Yu Z (2011b) Status of the phylogenetic diversity census of

ruminal microbiomes. FEMS Microbiology Ecology 76: 49-63.

Kim M, Morrison M & Yu Z (2011c) Phylogenetic diversity of bacterial communities in

bovine rumen as affected by diets and microenvironments. Folia Microbiologica

DOI 10.1007/s12223-011-0066-5.

Kitts CL (2001) Terminal restriction fragment patterns: a tool for comparing microbial

communities and assessing community dynamics. Current Issues in Intestinal

Microbiology 2: 17-25.

Klieve AV, O'Leary MN, McMillen L & Ouwerkerk D (2007) Ruminococcus bromii,

identification and isolation as a dominant community member in the rumen of

cattle fed a barley diet. Journal of Applied Microbiology 103: 2065-2073.

Kobayashi Y, Shinkai T & Koike S (2008) Ecological and physiological characterization

shows that Fibrobacter succinogenes is important in rumen fiber digestion -

Review. Folia Microbiologica 53: 195-200.

Koike S, Yoshitani S, Kobayashi Y & Tanaka K (2003a) Phylogenetic analysis of fiber-

associated rumen bacterial community and PCR detection of uncultured bacteria.

FEMS Microbiology Letters 229: 23-30.

Koike S, Pan J, Kobayashi Y & Tanaka K (2003b) Kinetics of in sacco fiber-attachment

of representative ruminal cellulolytic bacteria monitored by competitive PCR.

Journal of Dairy Science 86: 1429-1435.

Kongmun P, Wanapat M, Pakdee P, Navanukraw C & Yu Z (2011) Manipulation of

rumen fermentation and ecology of swamp buffalo by coconut oil and garlic

powder supplementation. Livestock Science 135: 84-92.

180

Kraatz M, Wallace RJ & Svensson L (2011) Olsenella umbonata sp. nov., a

microaerotolerant anaerobic lactic acid bacterium from the sheep rumen and pig

jejunum, and emended descriptions of Olsenella, Olsenella uli and Olsenella

profusa. International Journal of Systematic and Evolutionary Microbiology 61:

795-803.

Krause L, Diaz NN, Edwards RA, et al. (2008) Taxonomic composition and gene content

of a methane-producing microbial community isolated from a reactor.

Journal of Biotechnology 136: 91-101.

Krober M, Bekel T, Diaz NN, et al. (2009) Phylogenetic characterization of a biogas

plant microbial community integrating clone library 16S-rDNA sequences and

metagenome sequence data obtained by 454-pyrosequencing. Journal of

Biotechnology 142: 38-49.

Kunin V, Engelbrektson A, Ochman H & Hugenholtz P (2010) Wrinkles in the rare

biosphere: pyrosequencing errors can lead to artificial inflation of diversity

estimates. Environmental Microbiology 12: 118-123.

Lane DJ, Pace B, Olsen GJ, Stahl DA, Sogin ML & Pace NR (1985) Rapid determination

of 16S ribosomal RNA sequences for phylogenetic analyses. Proceedings of the

National Academy of Sciences of the United States of America 82: 6955-6959.

Larkin MA, Blackshields G, Brown NP, et al. (2007) Clustal W and clustal X version

2.0. Bioinformatics 23: 2947-2948.

Larue R, Yu ZT, Parisi VA, Egan AR & Morrison M (2005) Novel microbial diversity

adherent to plant biomass in the herbivore gastrointestinal tract, as revealed by

181

ribosomal intergenic spacer analysis and rrs gene sequencing. Environmental

Microbiology 7: 530-543.

Leedle JAZ & Hespell RB (1980) Differential carbohydrate media and anaerobic replica

plating techniques in delineating carbohydrate-utilizing subgroups in rumen

bacterial populations. Applied and Environmental Microbiology 39: 709-719.

Leser TD, Amenuvor JZ, Jensen TK, Lindecrona RH, Boye M & Moller K (2002)

Culture-independent analysis of gut bacteria: the pig gastrointestinal tract

microbiota revisited. Applied and Environmental Microbiology 68: 673-690.

Liles MR, Turkmen O, Manske BF, et al. (2010) A phylogenetic microarray targeting

16S rRNA genes from the bacterial division Acidobacteria reveals a lineage-

specific distribution in a soil clay fraction. Soil Biology & Biochemistry 42: 739-

747.

Liu WT, Marsh TL, Cheng H & Forney LJ (1997) Characterization of microbial diversity

by determining terminal restriction fragment length polymorphisms of genes

encoding 16S rRNA. Applied and Environmental Microbiology 63: 4516-4522.

Liu ZZ, DeSantis TZ, Andersen GL & Knight R (2008) Accurate taxonomy assignments

from 16S rRNA sequences produced by highly parallel pyrosequencers. Nucleic

Acids Research 36:e120.

Liu ZZ, Lozupone C, Hamady M, Bushman FD & Knight R (2007) Short pyrosequencing

reads suffice for accurate microbial community analysis. Nucleic Acids Research

35:e120.

182

Lodge-Ivey SL, Rappe MS, Johnston WH, Bohlken RE & Craig AM (2005) Molecular

analysis of a consortium of ruminal microbes that detoxify pyrrolizidine alkaloids.

Canadian Journal of Microbiology 51: 455-465.

Lourenco M, Cardozo PW, Calsamiglia S & Fievez V (2008) Effects of saponins,

quercetin, eugenol, and cinnamaldehyde on fatty acid biohydrogenation of forage

polyunsaturated fatty acids in dual-flow continuous culture fermenters. Journal of

Animal Science 86: 3045-3053.

Lozupone C & Knight R (2005) UniFrac: a new phylogenetic method for comparing

microbial communities. Applied and Environmental Microbiology 71: 8228-8235.

Ludwig W, Strunk O, Westram R, et al. (2004) ARB: a software environment for

sequence data. Nucleic Acids Research 32: 1363-1371.

Lukas F, Simunek J, Mrazek J & Kopecny J (2010) PCR-DGGE analysis of bacterial

population attached to the bovine rumen wall. Folia Microbiologica 55: 345-348.

Mackie RI, Aminov RI, Hu WP, Klieve AV, Ouwerkerk D, Sundset MA & Kamagata Y

(2003) Ecology of uncultivated Oscillospira species in the rumen of cattle, sheep,

and reindeer as assessed by microscopy and molecular approaches. Applied and

Environmental Microbiology 69: 6808-6815.

Maia MRG, Chaudhary LC, Figueres L & Wallace RJ (2007) Metabolism of

polyunsaturated fatty acids and their toxicity to the microflora of the rumen.

Antonie Van Leeuwenhoek International Journal of General and Molecular

Microbiology 91: 303-314.

183

Maluszynska GM & Janota-Bassalik L (1974) A cellulolytic rumen bacterium,

Micromonospora ruminantium sp.nov. Journal of General Microbiology 82: 57-

65.

Margulies M, Egholm M, Altman WE, et al. (2005) Genome sequencing in

microfabricated high-density picolitre reactors. Nature 437: 376-380.

Martin C, Morgavi DP & Doreau M (2010) Methane mitigation in ruminants: from

microbe to the farm scale. Animal 4: 351-365.

Martin C, Millet L, Fonty G & Michalet-Doreau B (2001) Cereal supplementation

modified the fibrolytic activity but not the structure of the cellulolytic bacterial

community associated with rumen solid digesta. Reproduction Nutrition

Development 41: 413-424.

Mathew B, Eastridge ML, Oelker ER, Firkins JL & Karnati SKR (2011) Interactions of

monensin with dietary fat and carbohydrate components on ruminal fermentation

and production responses by dairy cows. Journal of Dairy Science 94: 396-409.

Matthies C, Evers S, Ludwig W & Schink B (2000) Anaerovorax odorimutans gen. nov.,

sp. nov., a putrescine-fermenting, strictly anaerobic bacterium. International

Journal of Systematic and Evolutionary Microbiology 50: 1591-1594.

McSweeney CS, Denman SE, Wright ADG & Yu Z (2007) Application of recent

DNA/RNA-based techniques in rumen ecology. Asian-Australasian Journal of

Animal Sciences 20: 283-294.

Michalet-Doreau B, Fernandez I, Peyron C, Millet L & Fonty G (2001) Fibrolytic

activities and cellulolytic bacterial community structure in the solid and liquid

phases of rumen contents. Reproduction Nutrition Development 41: 187-194.

184

Morita H, Shiratori C, Murakami M, et al. (2008) Sharpea azabuensis gen. nov., sp. nov.,

a Gram-positive, strictly anaerobic bacterium isolated from the faeces of

thoroughbred horses. International Journal of Systematic and Evolutionary

Microbiology 58: 2682-2686.

Morrison M, Mackie RI & Kistner A (1990) 3-Phenylpropanoic acid improves the

affinity of Ruminococcus albus for cellulose in continuous culture. Applied and

Environmental Microbiology 56: 3220-3222.

Mosoni P, Martin C, Forano E & Morgavi DP (2011) Long-term defaunation increases

the abundance of cellulolytic ruminococci and methanogens but does not affect

the bacterial and methanogen diversity in the rumen of sheep. Journal of Animal

Science 89: 783-791.

Nagaraja TG, Newbold CJ, Ven Nevel CJ & Demeyer DI (1997) Manipulation of rumen

fermentation. Pages 523-632 in The Rumen Microbial Ecosystem. Hobson, P.N.

and Stewart, C.S. ed. Chapman and Hall, New York, NY. .

Nelson KE, Zinder SH, Hance I, et al. (2003) Phylogenetic analysis of the microbial

populations in the wild herbivore gastrointestinal tract: insights into an

unexplored niche. Environmental Microbiology 5: 1212-1220.

Nichols D (2007) Cultivation gives context to the microbial ecologist. FEMS

Microbiology Ecology 60: 351-357.

Nicholson MJ, Evans PN & Joblin KN (2007) Analysis of methanogen diversity in the

rumen using temporal temperature gradient gel electrophoresis: Identification of

uncultured methanogens. Microbial Ecology 54: 141-150.

185

Odongo NE, Or-Rashid MM, Bagg R, et al. (2007) Long-term effects of feeding

monensin on milk fatty acid composition in lactating dairy cows. Journal of Dairy

Science 90: 5126-5133.

Oelker ER, Reveneau C & Firkins JL (2009) Interaction of molasses and monensin in

alfalfa hay- or corn silage-based diets on rumen fermentation, total tract

digestibility, and milk production by Holstein cows. Journal of Dairy Science 92:

270-285.

Ohene-Adjei S, Teather RM, Ivanj M & Forster RJ (2007) Postinoculation protozoan

establishment and association patterns of methanogenic archaea in the ovine

rumen. Applied and Environmental Microbiology 73: 4609-4618.

Or-Rashid MM, McBride BW & Wright TC (2009) Microbial fatty acid conversion

within the rumen and the subsequent utilization of these fatty acids to improve the

healthfulness of ruminant food products. Applied Microbiology and

Biotechnology 84: 1033-1043.

Ouwerkerk D, Klieve AV & Forster RJ (2002) Enumeration of Megasphaera elsdenii in

rumen contents by real-time Taq nuclease assay. Journal of Applied Microbiology

92: 753-758.

Ozutsumi Y, Tajima K, Takenaka A & Itabashi H (2005) The effect of protozoa on the

composition of rumen bacteria in cattle using 16S rRNA gene clone libraries.

Bioscience Biotechnology and Biochemistry 69: 499-506.

Palmer C, Bik EM, Eisen MB, et al. (2006) Rapid quantitative profiling of complex

microbial populations. Nucleic Acids Research 34:e5.

186

Patel GB, Khan AW, Agnew BJ & Colvin JR (1980) Isolation and characterization of an

anaerobic, cellulolytic , Acetivibrio cellulolyticus gen. nov., sp.

nov. International Journal of Systematic Bacteriology 30: 179-185.

Patra AK & Saxena J (2009) Dietary phytochemicals as rumen modifiers: a review of the

effects on microbial populations. Antonie Van Leeuwenhoek International Journal

of General and Molecular Microbiology 96: 363-375.

Pitta DW, Pinchak WE, Dowd SE, et al. (2010) Rumen bacterial diversity dynamics

associated with changing from bermudagrass hay to grazed winter wheat diets.

Microbial Ecology 59: 511-522.

Pope PB, Smith W, Denman SE, et al. (2011) Isolation of Succinivibrionaceae implicated

in low methane emissions from Tammar wallabies. Science 333: 646-648.

Pope PB, Denman SE, Jones M, et al. (2010) Adaptation to herbivory by the Tammar

wallaby includes bacterial and glycoside hydrolase profiles different from other

herbivores. Proceedings of the National Academy of Sciences of the United States

of America 107: 14793-14798.

Pruesse E, Quast C, Knittel K, Fuchs BM, Ludwig WG, Peplies J & Glockner FO (2007)

SILVA: a comprehensive online resource for quality checked and aligned

ribosomal RNA sequence data compatible with ARB. Nucleic Acids Research 35:

7188-7196.

Qu A, Brulc JM, Wilson MK, et al. (2008) Comparative metagenomics reveals host

specific metavirulomes and horizontal gene transfer elements in the chicken

cecum microbiome. PLoS One 3: e2945.

187

Quince C, Lanzen A, Curtis TP, et al. (2009) Accurate determination of microbial

diversity from 454 pyrosequencing data. Nature Methods 6: 639-641.

Rainey FA & Janssen PH (1995) Phylogenetic analysis by 16S ribosomal DNA-sequence

comparison reveals 2 unrelated groups of species within the genus Ruminococcus.

FEMS Microbiology Letters 129: 69-73.

Rajilic-Stojanovic M, Heilig HGHJ, Molenaar D, Kajander K, Surakka A, Smidt H & de

Vos WM (2009) Development and application of the human intestinal tract chip,

a phylogenetic microarray: analysis of universally conserved phylotypes in the

abundant microbiota of young and elderly adults. Environmental Microbiology

11: 1736-1751.

Rea S, Bowman JP, Popovski S, Pimm C & Wright ADG (2007) Methanobrevibacter

millerae sp. nov. and Methanobrevibacter olleyae sp. nov., methanogens from the

ovine and bovine rumen that can utilize formate for growth. International Journal

of Systematic and Evolutionary Microbiology 57: 450-456.

Reveneau C, Adams SE, Cotta MA & Morrison M (2003) Phenylacetic and

phenylpropionic acids do not affect xylan degradation by Ruminococcus albus.

Applied and Environmental Microbiology 69: 6954-6958.

Rimour S, Hill D, Militon C & Peyret P (2005) GoArrays: highly dynamic and efficient

microarray probe design. Bioinformatics 21: 1094-1103.

Roesch LF, Fulthorpe RR, Riva A, et al. (2007) Pyrosequencing enumerates and

contrasts soil microbial diversity. ISME Journal 1: 283-290.

Russell JB (2002) Rumen microbiology and its role in ruminant nutrition. 1st ed. Ithaca,

NY.

188

Russell JB & Strobel HJ (1989) Effect of Ionophores on Ruminal Fermentation. Applied

and Environmental Microbiology 55: 1-6.

Russell JB & Houlihan AJ (2003) Ionophore resistance of ruminal bacteria and its

potential impact on human health. FEMS Microbiology Reviews 27: 65-74.

Saeed AI, Hagabati NK, Braisted JC, et al. (2006) TM4 microarray software suite. DNA

Microarrays, Part B: Databases and Statistics 411: 134-193.

Schauer R, Bienhold C, Ramette A & Harder J (2010) Bacterial diversity and

biogeography in deep-sea surface sediments of the South Atlantic Ocean. ISME

Journal 4: 159-170.

Schloss PD (2010) The effects of alignment quality, distance calculation method,

sequence filtering, and region on the analysis of 16S rRNA gene-based studies.

PLoS Computational Biology 6:e1000844.

Schloss PD & Handelsman J (2004) Status of the microbial census. Microbiology and

Molecular Biology Reviews 68: 686-691.

Schloss PD & Handelsman J (2005) Introducing DOTUR, a computer program for

defining operational taxonomic units and estimating species richness. Applied and

Environmental Microbiology 71: 1501-1506.

Schloss PD & Handelsman J (2006) Introducing SONS, a tool for operational taxonomic

unit-based comparisons of microbial community memberships and structures.

Applied and Environmental Microbiology 72: 6773-6779.

Schloss PD, Westcott SL, Ryabin T, et al. (2009) Introducing mothur: open-source,

platform-independent, community-supported software for describing and

189

comparing microbial communities. Applied and Environmental Microbiology 75:

7537-7541.

Shinkai T, Ueki T & Kobayashi Y (2010) Detection and identification of rumen bacteria

constituting a fibrolytic consortium dominated by Fibrobacter succinogenes.

Animal Science Journal 81: 72-79.

Skillman LC, Evans PN, Strompl C & Joblin KN (2006) 16S rDNA directed PCR

primers and detection of methanogens in the bovine rumen. Letters in Applied

Microbiology 42: 222-228.

Small J, Call DR, Brockman FJ, Straub TM & Chandler DP (2001) Direct detection of

16S rRNA in soil extracts by using oligonucleotide microarrays. Applied and

Environmental Microbiology 67: 4708-4716.

Smith PH & Hungate RE (1958) Isolation and characterization of Methanobacterium

ruminantium n. sp. Journal of Bacteriology 75: 713-718.

Sogin ML, Morrison HG, Huber JA, et al. (2006) Microbial diversity in the deep sea and

the underexplored "rare biosphere". Proceedings of the National Academy of

Sciences of the United States of America 103: 12115-12120.

Stack RJ & Hungate RE (1984) Effect of 3-phenylpropanoic acid on capsule and

cellulases of Ruminococcus albus 8. Applied and Environmental Microbiology 48:

218-223.

Stack RJ & Cotta MA (1986) Effect of 3-phenylpropanoic acid on growth of and

cellulose utilization by cellulolytic ruminal bacteria. Applied and Environmental

Microbiology 52: 209-210.

190

Stackebrandt E & Ebers J (2006) Taxonomic parameters revisited: tarnished gold

standards. Microbiology Today 33: 152-155.

Stackebrandt E & Goebel BM (1994) Taxonomic Note: A place for DNA-DNA

reassociation and 16S rRNA sequence analysis in the present species definition in

bacteriology. International Journal of Systematic Bacteriology 44: 846-849.

Stahl DA, Flesher B, Mansfield HR & Montgomery L (1988) Use of phylogenetically

based hybridization probes for studies of ruminal microbial ecology. Applied and

Environmental Microbiology 54: 1079-1084.

Stevenson DM & Weimer PJ (2007) Dominance of Prevotella and low abundance of

classical ruminal bacterial species in the bovine rumen revealed by relative

quantification real-time PCR. Applied Microbiology and Biotechnology 75: 165-

174.

Stiverson J, Morrison M & Yu Z (2011) Populations of select cultured and uncultured

bacteria in the rumen of sheep and the effect of diets and ruminal fractions.

International Journal of Microbiology doi:10.1155/2011/750613

Sundset MA, Kohn A, Mathiesen SD & Praesteng KE (2008) Eubacterium rangiferina, a

novel usnic acid-resistant bacterium from the reindeer rumen.

Naturwissenschaften 95: 741-749.

Sundset MA, Praesteng KE, Cann IKO, Mathiesen SD & Mackie RI (2007) Novel rumen

bacterial diversity in two geographically separated sub-species of reindeer.

Microbial Ecology 54: 424-438.

191

Sundset MA, Edwards JE, Cheng YF, et al. (2009) Rumen microbial diversity in

Svalbard reindeer, with particular emphasis on methanogenic archaea. FEMS

Microbiology Ecology 70: 553-562.

Sylvester JT, Karnati SKR, Yu ZT, Morrison M & Firkins JL (2004) Development of an

assay to quantify rumen ciliate protozoal biomass in cows using real-time PCR.

Journal of Nutrition 134: 3378-3384.

Tajima K, Nagamine T, Matsui H, Nakamura M & Aminov RI (2001b) Phylogenetic

analysis of archaeal 16S rRNA libraries from the rumen suggests the existence of

a novel group of archaea not associated with known methanogens. FEMS

Microbiology Letters 200: 67-72.

Tajima K, Aminov RI, Nagamine T, Matsui H, Nakamura M & Benno Y (2001a) Diet-

dependent shifts in the bacterial population of the rumen revealed with real-time

PCR. Applied and Environmental Microbiology 67: 2766-2774.

Tajima K, Aminov RI, Nagamine T, Ogata K, Nakamura M, Matsui H & Benno Y (1999)

Rumen bacterial diversity as determined by sequence analysis of 16S rDNA

libraries. FEMS Microbiology Ecology 29: 159-169.

Tajima K, Arai S, Ogata K, et al. (2000) Rumen bacterial community transition during

adaptation to high-grain diet. Anaerobe 6: 273-284.

Tajima K, Nonaka I, Higuchi K, et al. (2007) Influence of high temperature and humidity

on rumen bacterial diversity in Holstein heifers. Anaerobe 13: 57-64.

Turnbaugh PJ, Ley RE, Mahowald MA, Magrini V, Mardis ER & Gordon JI (2006) An

obesity-associated gut microbiome with increased capacity for energy harvest.

Nature 444: 1027-1031.

192 van Gylswyk NO (1995) Succiniclasticum ruminis gen. nov., sp. nov., a ruminal

bacterium converting succinate to propionate as the sole energy-yielding

mechanism. International Journal of Systematic Bacteriology 45: 297-300.

Van Gylswyk NO (1980) Fusobacterium polysaccharolyticum sp. nov., gram-negative

rod from the rumen that produces butyrate and ferments cellulose and starch.

Journal of General Microbiology 116: 157-163. van Nevel CJ & Demeyer DI (1996) Control of rumen methanogenesis. Environmental

Monitoring and Assessment 42: 73-97.

Wang Q, Garrity GM, Tiedje JM & Cole JR (2007) Naive Bayesian classifier for rapid

assignment of rRNA sequences into the new bacterial taxonomy. Applied and

Environmental Microbiology 73: 5261-5267.

Weimer PJ, Stevenson DM, Mertens DR & Thomas EE (2008) Effect of monensin

feeding and withdrawal on populations of individual bacterial species in the

rumen of lactating dairy cows fed high-starch rations. Applied Microbiology and

Biotechnology 80: 135-145.

Whitford MF, Teather RM & Forster RJ (2001a) Phylogenetic analysis of methanogens

from the bovine rumen. BMC Microbiology 1:5.

Whitford MF, Yanke LJ, Forster RJ & Teather RM (2001b) Lachnobacterium bovis gen.

nov., sp. nov., a novel bacterium isolated from the rumen and faeces of cattle.

International Journal of Systematic and Evolutionary Microbiology 51: 1977-

1981.

193

Whitford MF, Forster RJ, Beard CE, Gong JH & Teather RM (1998) Phylogenetic

analysis of rumen bacteria by comparative sequence analysis of cloned 16S rRNA

genes. Anaerobe 4: 153-163.

Williams WL, Tedeschi LO, Kononoff PJ, Callaway TR, Dowd SE, Karges K & Gibson

ML (2010) Evaluation of in vitro gas production and rumen bacterial populations

fermenting corn milling (co)products. Journal of Dairy Science 93: 4735-4743.

Woese CR, Gutell R, Gupta R & Noller HF (1983) Detailed analysis of the higher-order

structure of 16S-like ribosomal ribonucleic acids. Microbiological Reviews 47:

621-669.

Wolf M, Muller T, Dandekar T & Pollack JD (2004) Phylogeny of Firmicutes with

special reference to Mycoplasma () as inferred from phosphoglycerate

kinase amino acid sequence data. International Journal of Systematic and

Evolutionary Microbiology 54: 871-875.

Wood TM, Wilson CA & Stewart CS (1982) Preparation of the cellulase from the

cellulolytic anaerobic rumen bacterium Ruminococcus albus and its release from

the bacterial cell wall. Biochemical Journal 205: 129-137.

Wright ADG, Toovey AF & Pimm CL (2006) Molecular identification of methanogenic

archaea from sheep in Queensland, Australia reveal more uncultured novel

archaea. Anaerobe 12: 134-139.

Wright ADG, Auckland CH & Lynn DH (2007) Molecular diversity of methanogens in

feedlot cattle from Ontario and Prince Edward Island, Canada. Applied and

Environmental Microbiology 73: 4206-4210.

194

Wright ADG, Ma XL & Obispo NE (2008) Methanobrevibacter phylotypes are the

dominant methanogens in sheep from Venezuela. Microbial Ecology 56: 390-394.

Wright ADG, Williams AJ, Winder B, Christophersen CT, Rodgers SL & Smith KD

(2004) Molecular diversity of rumen methanogens from sheep in western

Australia. Applied and Environmental Microbiology 70: 1263-1270.

Yanagita K, Kamagata Y, Kawaharasaki M, Suzuki T, Nakamura Y & Minato H (2000)

Phylogenetic analysis of methanogens in sheep rumen ecosystem and detection of

Methanomicrobium mobile by fluorescence in situ hybridization. Bioscience

Biotechnology and Biochemistry 64: 1737-1742.

Yang S, Ma S, Chen J, et al. (2010a) Bacterial diversity in the rumen of Gayals (Bos

frontalis), Swamp buffaloes (Bubalus bubalis) and Holstein cow as revealed by

cloned 16S rRNA gene sequences. Molecular Biology Reports 37: 2063-2073.

Yang LY, Chen J, Cheng XL, Xi DM, Yang SL, Deng WD & Mao HM (2010b)

Phylogenetic analysis of 16S rRNA gene sequences reveals rumen bacterial

diversity in Yaks (Bos grunniens). Molecular Biology Reports 37: 553-562.

Youssef N, Sheik CS, Krumholz LR, Najar FZ, Roe BA & Elshahed MS (2009)

Comparison of species richness estimates obtained using nearly complete

fragments and simulated pyrosequencing-generated fragments in 16S rRNA gene-

based environmental surveys. Applied and Environmental Microbiology 75: 5227-

5236.

Yu Z & Mohn WW (2001) Bacterial diversity and community structure in an aerated

lagoon revealed by ribosomal intergenic spacer analyses and 16S ribosomal DNA

sequencing. Applied and Environmental Microbiology 67: 1565-1574.

195

Yu Z & Morrison M (2004a) Comparisons of different hypervariable regions of rrs genes

for use in fingerprinting of microbial communities by PCR-denaturing gradient

gel electrophoresis. Applied and Environmental Microbiology 70: 4800-4806.

Yu Z & Morrison M (2004b) Improved extraction of PCR-quality community DNA from

digesta and fecal samples. Biotechniques 36: 808-812.

Yu Z, Yu M & Morrison M (2006) Improved serial analysis of V1 ribosomal sequence

tags (SARST-V1) provides a rapid, comprehensive, sequence-based

characterization of bacterial diversity and community composition.

Environmental Microbiology 8: 603-611.

Yu Z, Garcia-Gonzalez R, Schanbacher FL & Morrison M (2008) Evaluations of

different hypervariable regions of archaeal 16S rRNA genes in profiling of

methanogens denaturing by Archaea-specific PCR and gradient gel

electrophoresis. Applied and Environmental Microbiology 74: 889-893.

Zhang K & Dong X (2009) Selenomonas bovis sp. nov., isolated from yak rumen

contents. International Journal of Systematic and Evolutionary Microbiology 59:

2080-2083.

Zhou M, Hernandez-Sanabria E & Guan LL (2009) Assessment of the microbial ecology

of ruminal methanogens in cattle with different feed efficiencies. Applied and

Environmental Microbiology 75: 6524-6533.

196

APPENDIX A:

ADDITIONAL PYROSEQUENCING METHODS

Supplementary Materials and Methods for Chapter 8:

Processing of the sequence dataset:

The Dataset A was obtained from the OSU sequencing center, while the Dataset B and the Dataset C were generated by Research and Testing Laboratory (Lubbock, TX).

Each Dataset was processed using the QIIME bioinformatics software package, v1.3.0.

(Caporaso et al., 2010) or the Mothur program (Schloss et al., 2009). The split_libraries.py command as implemented in the QIIME program was used to separate sample libraries with the following parameters: -q qual.file -l 200 -L 650 -H 8 –M 2 –b 8.

These parameters could remove sequences with homopolymer stretches greater than 8nt

(-H 8), primer mismatches greater than 2nt (-M 2), sequence length less than 200 bp (-l

200), sequence length greater than 650 bp (-L 650), overall quality scores less than 25 (-q qual.file). Forward primers and reverse primers were also trimmed off, and the barcode sequences were detected (-b 8). The resultant sequences were aligned using the align.seqs command as implemented in the Mothur program against the Greengenes core set database. Sequences that are expected to result from pyrosequencing errors were removed from the aligned file using the pre.cluster command, and then chimeric sequences were checked using the chimera.slayer command as implemented in the Mothur program. The high-quality sequences obtained from each Dataset were combined into one composite

197

Dataset. The number of sequences among samples was normalized using the sub.sample command as implemented in the Mothur program.

The dist.seqs, the cluster, the make.shared, and the summary.single commands as implemented in the Mothur program were used to generate OTUs and calculate diversity indices. An OTU list file compatible with the QIIME program was created using the get.otulist command as implemented in the Mothur program. Representative sequences for each OTU were selected using the pick_rep_set.py command as implemented in the

QIIME program. The aligned representative sequences were classified to bacterial taxa using the assign_taxonomy.py command as implemented in the QIIME program. The

OTU table was created from the OTU mapping file and the taxonomy file generated from the pick_otus.py and assign_taxonomy.py commands as implemented in the QIIME program. A phylogenetic tree was constructed using the filter_alignment.py and make_phylogeny.py commands and then used to calculate beta diversity measures using the beta_diversity.py command as implemented in the QIIME program. Alpha diversity measures were calculated using the alpha_rarefaction.py command.

198

GTNTBACCGCDGCTGCTG

GTNTBACCGCDGCTGCTG

GTNTBACCGCDGCTGCTG

GTNTBACCGCDGCTGCTG

GTNTBACCGCDGCTGCTG

GTNTBACCGCDGCTGCTG

GTNTBACCGCDGCTGCTG

GTNTBACCGCDGCTGCTG

GTNTBACCGCDGCTGCTG

GTNTBACCGCDGCTGCTG

GTNTBACCGCDGCTGCTG

GTNTBACCGCDGCTGCTG

GTNTBACCGCDGCTGCTG

GTNTBACCGCDGCTGCTG

CGAGAGATAC

CATAGTAGTG

TACTGAGCTA

TGATACGTCT

TCTCTATGCG

TAGTATCAGC

CTCGCGTGTC

CGTGTCTCTA

ATATCGCGAG

ATCAGACACG

AGCACTGTAG

AGACGCACTC

ACGCTCGACA

ACGAGTGCGT

analysis

CTATGCGCCTTGCCAGCCCGCTCAG

CTATGCGCCTTGCCAGCCCGCTCAG

CTATGCGCCTTGCCAGCCCGCTCAG

CTATGCGCCTTGCCAGCCCGCTCAG

CTATGCGCCTTGCCAGCCCGCTCAG

CTATGCGCCTTGCCAGCCCGCTCAG

CTATGCGCCTTGCCAGCCCGCTCAG

CTATGCGCCTTGCCAGCCCGCTCAG

CTATGCGCCTTGCCAGCCCGCTCAG

CTATGCGCCTTGCCAGCCCGCTCAG

CTATGCGCCTTGCCAGCCCGCTCAG

CTATGCGCCTTGCCAGCCCGCTCAG

CTATGCGCCTTGCCAGCCCGCTCAG

CTATGCGCCTTGCCAGCCCGCTCAG

CGTATCGCCTCCCTCGCGCCATCAGAKRGTTYGATYNTGGCTCAG

Sequence

pyrosequencing

for

primers

CGAGAGATAC

CATAGTAGTG

TACTGAGCTA

TGATACGTCT

TCTCTATGCG

TAGTATCAGC

CTCGCGTGTC

CGTGTCTCTA

ATATCGCGAG

ATCAGACACG

AGCACTGTAG

AGACGCACTC

ACGCTCGACA

ACGAGTGCGT

Barcode

bacterial

of

List

MID14

MID13

MID12

MID11

MID10

MID9

MID8

MID7

MID6

MID5

MID4

MID3

MID2

MID1

.

-

-

-

-

-

-

-

-

-

-

-

-

-

-

B

B

B

B

B

B

B

B

B

B

B

B

B

B

A

A

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

BactR

BactR

BactR

BactR

BactR

BactR

BactR

BactR

BactR

BactR

BactR

BactR

BactR

BactR

BactF

PrimerName

Table

199