Composition, spatial distribution, and diversity of the bacterial communities in the rumen of cows fed different forages

Authors


  • Editor: Julian Marchesi

Correspondence: Robert Forster, Lethbridge Research Centre, Agriculture and Agri-Food Canada, 5403 1st Avenue South, Lethbridge, AB, Canada T1J 4B1. Tel.: +1 403 317 2292; fax: +1 403 382 3156; e-mail: robert.forster@agr.gc.ca

Abstract

The species composition, distribution, and biodiversity of the bacterial communities in the rumen of cows fed alfalfa or triticale were investigated using 16S rRNA gene clone library analyses. The rumen bacterial community was fractionated and analyzed as three separate fractions: populations in the planktonic, loosely attached to rumen digesta particles, and tightly attached to rumen digesta particles. Six hundred and thirteen operational taxonomic units (OTUs) belonging to 32 genera, 19 families, and nine phyla of the domain Bacteria were identified from 1014 sequenced clones. Four hundred and fifty one of the 613 OTUs were identified as new species. These bacterial sequences were distributed differently among the three fractions in the rumen digesta of cows fed alfalfa or triticale. Chao 1 estimation revealed that, in both communities, the populations tightly attached to particulates were more diverse than the planktonic and those loosely attached to particulates. S-Libshuff detected significant differences in the composition between any two fractions in the rumen of cows with the same diet and between the communities fed alfalfa and triticale diets. The species richness estimated for the communities fed alfalfa and triticale is 1027 and 662, respectively. The diversity of the rumen bacterial community examined in this study is greater than previous studies have demonstrated and the differences in the community composition between two high-fiber diets have implications for sample selection for downstream metagenomics applications.

Introduction

The rumen is a complex ecosystem, where microorganisms convert feedstuffs into microbial biomass and fermentation end products that can be utilized by animals. In the rumen, macromolecules in complex lignocellulosic plant materials are hydrolyzed under anaerobic conditions to their oligomers and monomers, which are then further fermented to form volatile fatty acids. Three taxonomic groups of microorganisms, bacteria, protozoa, and fungi, carry out this digestion process in the rumen. A renewed interest in the enzymes associated with digestive processes in the rumen has developed because of efforts to increase the efficiency of biomass conversion into ethanol and other higher value products. An improved understanding of rumen microbial ecology can help with these goals, as well as provide knowledge to increase animal feed efficiency, enhance the quality of animal products, and provide insights into other fermentation processes.

More than 200 bacterial species have been isolated from the rumen and many of these have been characterized physiologically (Russell & Hespell, 1981). Culture-independent 16S rRNA gene clone library analyses, as reviewed by Edwards et al. (2004), have demonstrated that the rumen bacterial community is more diverse than the number of cultivated species would suggest and that the rumen may contain 300–400 bacterial species. However, even this could still be an underestimation, because in these studies, only a limited number of clones (<200) were sequenced and the possible effects of diet and other environmental factors on the biodiversity and composition of the rumen microbial community were not considered. Most of the bacteria identified in clone library analyses (Whitford et al., 1998; Koike et al., 2003; Nelson et al., 2003; Edwards et al., 2004; An et al., 2005; Tajima et al., 2007) belong to the phyla Firmicutes and Bacteroidetes.

There is limited information on how diet affects the composition and distribution of the rumen bacterial community. Using quantitative PCR, Tajima et al. (2001) found that the relative abundance of bacterial populations markedly changed when the diet was shifted from hay to grain. When ruminants are changed from one diet to another containing different feedstuffs, such as from a dry cow diet to a transition diet, a period of adaptation is needed for the rumen microbial community to adjust to allow the newly offered substrate(s) to be used in the most efficient manner. However, it is unknown whether there would be any major differences in the rumen microbial community composition when the animal is fed with two different high-fiber diets. Bacteria are distributed in different phases in the rumen content and on the wall of the rumen. The epimural populations are different from the liquid-borne populations as well as the particulate-attached populations (Sadet et al., 2007; Cho et al., 2006). Larue et al. (2005) found that plant particle-associated bacterial populations were more diverse than fluid-phase populations. However, no statistical evidence was available in these studies to support the differences in the microbial composition. The possibility that the difference observed was due to a random variation could not be ruled out.

In this study, we developed a method to separate rumen microbial communities into planktonic (fluid-borne), loosely attached to particulates (biofilm), and tightly attached to particulates (embedded, or fixed) fractions. 16S rRNA gene clone libraries were constructed for each of these fractions to investigate the identity, spatial distribution, and diversity of bacteria in the rumen of cattle fed alfalfa or triticale straw. These feedstuffs were chosen because alfalfa and triticale straw have been proposed as alternative lignocellulosic feedstocks for biomass conversion (e.g. Sheaffer et al., 2000; Davis & Weightman, 2008). The 16S rRNA gene sequences obtained were analyzed phylogenetically and statistically. The similarity of microbial populations among individual fractions from the same diet and the similarity between the two microbial communities with different diets were estimated statistically.

Materials and methods

Sample collection and fractionation

Four mature rumen-cannulated dry Holstein cows were used in this crossover study, with two sampling periods. In the first 1-month period, two cows were fed alfalfa hay plus 2 kg of a dry cow protein/mineral supplement and two cows were fed triticale straw plus 2 kg of the supplement. The cows were cared for in accordance with the guidelines set by the Canadian Council on Animal Care (CCAC, 1993). Samples were obtained in the last week of the period and then the diets were switched to start the second 1-month period. The second set of samples was obtained in the last week of the second period. Rumen digesta samples were collected in the morning, before the cattle were fed, in 200-mL air-tight centrifuge tubes through the rumen cannula and transferred to the laboratory within 20 min.

The fractionation of the rumen samples was carried out in an anaerobic bag (Sigma-Aldrich, MD, cat. no. Z106089) filled with 99.999% carbon dioxide. Firstly, 100 mL digesta was transferred into a heavy-walled 250-mL beaker and squeezed with a Bodum coffee maker plunger (Bodum Inc., Triengen, Switzerland). The liquid (the planktonic fraction) was poured off into a container. The squeezed digesta left in the beaker was washed twice with 100 mL O2-free phosphate rinse buffer (K2HPO4 30 mmol, KH2PO4 20 mmol, NaHCO3 35 mmol, resazurin 0.5 × 10−9 g mL−1) by stirring gently with a spatula, followed by squeezing and liquid decanting into a waste vessel. After washing, 100 mL O2-free methyl cellulose release buffer (phosphate rinse buffer plus 0.2% methyl cellulose) was added and stirred with a spatula. The mixture was transferred into the homogenizing cup of a Braun hand blender, where it was processed in three 20-s bursts, with a 10-s rest between each burst. The blended digesta was then transferred into a clean heavy-walled beaker and once again the liquid and particles were separated with the Bodum filter. The liquid obtained [the loosely attached particle-associated (biofilm) fraction] was poured into a separate container. The solid residue after squeezing [the tightly attached particle-associated (fixed or embedded) fraction (c. 15 g)] was suspended in 30 mL cold (4 °C) grinding buffer (100 mM Tris-HCl, 500 mM EDTA, 1.5 M NaCl, 1 mg mL−1 proteinase K, pH 8.0) and the mixed sample was spread on a shallow aluminum foil dish and flash frozen on liquid nitrogen. The samples were stored at −80 °C before DNA extraction.

The sample for the planktonic fraction (1 mL) and the particulate biofilm (5 mL) were centrifuged at 10 000 g for 10 min in a desktop centrifuge (Eppendorf 5415D) or at 5000 g for 10 min in a Du Pont RC5 super speed centrifuge, respectively, to collect microorganisms. After decanting the supernatant, 1.4 mL ASL stool lysis buffer (QIAamp DNA Stool Mini Kit) was added to each pellet and resuspended with a combination of vortexing and sonicating (Aquasonic 75HT, 10 min at maximum power). Finally, the suspended samples were transferred into 2-mL screw-cap Eppendorf tubes and stored at –80 °C for later DNA extraction.

DNA extraction and PCR amplification

To extract DNA from the planktonic and biofilm fractions, the frozen ASL-preserved samples (in 2-mL Eppendorf tubes) were heated in a water bath at 95 °C for 5 min and then centrifuged at 10 000 g for 5 min in a desktop centrifuge. The supernatant in each tube was then separated from the pellet, transferred into a clean 2-mL screw-cap Eppendorf tube, and kept on ice for DNA isolation. Potassium phosphate buffer (1.4 mL 0.4 M) was added to the pellet and the pellet was resuspended using vortexing and sonication as described earlier. Lysozyme (100 μL, 100 mg mL−1) and mutanolysin (10 μL, 2.5 U μL−1) were added and mixed, and the mixture was incubated at 37 °C for 30 min. After that, proteinase K (20 μL, 20 mg mL−1) was added and mixed, and the mixture was incubated at 37 °C for another 1 h. After incubation, a brief (3 s) centrifugation in the desktop centrifuge was carried out to spin down moisture from the tube caps. Five hundred milligrams of mixed glass beads (200 mg with 0.5 mm diameter and 300 mg with 1.0 mm diameter) were added to each tube and the samples were processed in a bead-beater homogenizer (B. Braun, Melsungen AG, Germany) for 3 min. After beating, each tube was centrifuged at 10 000 g for 1 min and the supernatant was transferred to a new tube. The supernatant obtained from the pellets after enzyme treatment/bead-beating and the supernatant collected before enzyme treatment and bead-beating were processed using the DNA extraction protocol provided in the QIAamp DNA Stool Mini Kit. Finally, the DNA obtained from the original supernatant and the DNA isolated from the pellet were combined into a single tube.

To extract DNA from the residual particulate sample, each frozen sample was ground coarsely under liquid nitrogen in a precooled porcelain mortar and then transferred into a precooled Retsch RM 100 Mortar Grinder equipped with a stainless-steel mortar bowl and pestle (F. Kurt Retsch GmbH and Co. KG, Haan, Germany) and ground for a further 5 min under liquid nitrogen. Liquid nitrogen was added to the mortar bowl during grinding as needed to maintain the grinding mixture in a semi-fluid state. The ground samples were transferred to a 200-mL wide-mouth centrifuge bottle and incubated in a water bath at 50 °C for 40 min to thaw the samples. The sample (∼30 mL) was then slowly poured into a screw-top Sorval SS34 centrifuge tube containing 3 mL 20% sodium dodecyl sulfate and mixed by gentle inversion of the capped tube. The mixture was incubated in a water bath at 65 °C for 45 min before centrifugation at 19 200 g for 10 min. The supernatant was mixed by gentle rocking with an equal volume of molten (held at 65 °C) 2% agarose (Sigma A9539, Sigma-Aldrich), poured into a 90 mm square Petri dish, and allowed to solidify on a level surface. The DNA-containing agarose was cut into 8 mm strips, placed in sterile plastic trays, and equilibrated at room temperature with 0.5 × TBE/50% glycerol buffer [1 : 20 agar : buffer (v/v), 16–24 h, 3 × , on a gentle platform rocker] and stored at 4 °C. DNA was isolated from the agar using a QIAquick Gel Extraction Kit (Qiagen Inc., Mississauga, ON, Canada) and used as the template for PCR amplification.

Clone library construction and sequencing

Clone libraries were constructed for each diet for the planktonic bacteria, bacteria attached to fiber particles, and tightly attached to fiber particles fractions to investigate the distribution of bacterial populations in rumen digesta. Animal bias was controlled by the crossover design of the experiment. For each time period, the fractions from the two animals on the same diet were pooled before library construction. Each fraction, therefore, was covered by two libraries, and the library results presented represent the summed clones of the two libraries.

The universal bacterial primer pair 63F (5′-CAG GCC TAA CAC ATG CAA GTC-3′) and 1387R (5′-GGG CGG WGT GTA CAA GGC-3′), which has been proved to be able to cover more bacteria than other universal primer sets (Marchesi et al., 1998), was used in this study to amplify bacterial 16S rRNA genes using Amplitaq Gold (Applied Biosystems) polymerase. The PCR cycle used consisted of a 5-min denaturation period at 95 °C, followed by 25 cycles of 95 °C for 1 min, 55 °C for 1 min, and then 72 °C for 1.5 min. After 25 cycles, an 8-min extension period at 72 °C was added. Clone libraries were built using a TOPO TA Cloning Kit (Invitrogen Canada Inc., Burlington, ON, Canada) with electrocompetent cells following the protocol provided. Individual clones were picked with a Qpix II (Genetix USA Inc., MA) into 96-well plates (Genetix USA Inc.). Sequencing of the clone libraries was carried out by Polymorphic (Polymorphic DNA Technologies Inc., CA) using an ABI3730XL Genetic Analyzer with a 50 cm capillary array.

Phylogenetic and biodiversity analyses

16S rRNA gene sequences were edited using sequencher 4.5 (Gene Code Cooperation) and evaluated using the chimera check program provided in RDP-II (Cole et al., 2005) and mallard (version 1.02, http://www.bioinformatics-toolkit.org/Mallard/). For phylogenetic analysis, the 16S rRNA gene sequences contained in the two libraries (period 1 and period 2) for each fraction of rumen digesta with the same diet were pooled and submitted to the classifier tool provided in RDP. The sequences were aligned using the Silva aligner (Pruesse et al., 2007). Trees in Figs 1–3 shown in the text were built with arb (Ludwig et al., 2004) using the neighbor-joining method. Trees in Figs 1 and 2 were built using the filters built with sequences sharing >70% similarity as the threshold. No filter was used for the tree in Fig. 3. Bootstrap trees in Supporting Information, Figs S1–S3, were calculated and built with mega4 (Tamura et al., 2007) using the model Kimura 2-parameter. Biodiversity index calculations were carried out using dotor 1.53 (Schloss & Handelsman, 2005), libshuff 1.0 (Schloss et al., 2004), and sons (Schloss & Handelsman, 2006). These same programs were also used as integrated into mothur version 1.10.2 (http://www.mothur.org) for deeper phylogenetic analysis and comparisons presented in Table 1. The distance files used in all these calculations were prepared using phillip 4.0 (http://evolution.genetics.washington.edu/phylip/phylipweb.html), and also using the model Kimura 2-parameter.

Figure 1.

 A simplified distance tree (neighbor-joining) built using the 16S rRNA gene sequences of Firmicutes and their close relatives obtained from this study. The numbers in the quadrilateral polygons represent the number of sequences in that taxonomic group obtained in this study. The individual sequences with names set in a bold font were also obtained from this study. The 16S rRNA gene sequence of Aquifex pyrophilus (M83548) was selected as the outgroup. The scale bar=1 substitution per 10 nucleotides.

Figure 2.

 A simplified distance tree (neighbor-joining) built using the 16S rRNA gene sequences of Bacteroidetes and their close relatives obtained from this study. The numbers in the quadrilateral polygons represent the number of sequences in that taxonomic group obtained in this study. The 16S rRNA gene sequence of Aquifex pyrophilus (M83548) was selected as the outgroup. The scale bar=1 substitution per 10 nucleotides.

Figure 3.

 A simplified distance tree (neighbor-joining) built using all the 16S rRNA gene sequences obtained from this study other than those of Firmicutes and Bacteroidetes and their close relatives. The numbers in the quadrilateral polygons represent the number of sequences in that taxonomic group obtained in this study. The individual sequences with names set in a bold font were also obtained from this study. The 16S rRNA gene sequence of Aquifex pyrophilus (M83548) was selected as the outgroup. The scale bar=1 substitution per 10 nucleotides.

Table 1.   Biodiversity index, distribution, and similarity of bacterial populations in different rumen digesta fractions in animals fed with alfalfa or triticale
Libraries*Number of sequencesNumber of OTUsSpecies richnessOTUs shared (%)P value (XY, YX)
  • *

    A_ and T_ represent the rumen microbial community fed with alfalfa and triticale, respectively. _L, _PT, and _RS represent the fluid-borne, particle-attached biofilm, and tightly particle-attached embedded bacterial fractions, respectively. _PTRS represents the combined sequence libraries for PT and RS.

  • The numbers before the parentheses represent the total number of sequences, OTUs, or estimated species obtained from the compared sequences. The two numbers in the parenthesis represent the number of sequences, OTUs, or estimated species obtained from the individual groups that were compared.

  • XY represents the P values obtained when the first library, fraction, or community (as indicated in column 1) is compared with the second library, fraction, or community. YX represents the P values obtained when the second library, fraction, or community is compared with the first library, fraction, or community. If the P value of either of these comparisons has a P<0.005, the libraries are considered significantly different at an experiment-wide rate of P<0.01.

A_L vs. T_L354 (166, 188)208 (108, 117)368 (270, 175)80.0001, 0.0001
A_PT vs. T_PT330 (111, 219)217 (82, 144)385 (175, 220)40.0664, 0.0001
A_RS vs. T_RS312 (160, 152)241 (128, 131)1089 (533, 599)70.0023, 0.3289
Prevotellaceae_PTRS
 A vs. T152 (63, 89)95 (45, 58)182 (80, 121)8<0.0362,<0.0001
Bacteroidales_PTRS
 A vs. T126 (65, 61)92 (43, 52)285 (155, 134)3<0.0027,<0.0007
Lachnospiraceae_PTRS
 A vs. T141 (53, 88)101 (44, 61)254 (155, 109)4<0.4102,<0.0001
Ruminococcaceae_PTRS
 A vs. T99 (44, 55)71 (33, 41)160 (83, 81)4<0.0265,<0.0001

Nucleotide sequence accession number

The 16S rRNA gene sequences obtained in this study have been deposited in the GenBank database under accession numbers GQ326914GQ327927.

Results

16S rRNA gene clone library analysis

In total, 1048 clones of almost full length (c. 1300 bp) were successfully sequenced. Thirty four sequences were putatively identified as chimeras and were removed. The remaining 1014 sequences were classified into 613  operational taxonomic units (OTUs) at a distance of 0.03. Of these, 451, comprised of 592 sequences, are putative new bacterial species sharing <97% similarity to any other available 16S rRNA gene sequence (data not shown). The number of clones and OTUs obtained from each fraction and fiber source are summarized in Table 2. Each of the 12 individual libraries contains 57–74 OTUs and 75–96 clones. The bacteria identified in the six fractions belong to nine phyla (two of them are putative phyla represented by only 16S rRNA gene sequences), 17 families, and 32 genera (Table 2). Figures 1–3 show simplified distance trees showing the phylogenetic position (as individual groups) of all the Firmicutes, Bacteroidetes, other bacteria, and their close relatives, respectively. Figures S1–S3 show bootstrap trees showing all the Firmicutes, the Bacteroidetes, other bacteria, and their close relatives identified in this study.

Table 2.   Identification and distribution of 16S rRNA gene clones in different rumen digesta fractions
PhylumPhylogenetic classification*Clone distribution in rumen digesta
Class, order, or familyGenusA_LA_PTA_RST_LT_PTT_RSA_T_
  • *

    U, represents unclassified.

  • Number of sequences classified into each group. A_L, A_PT, and A_RS represent the fluid-borne, weakly particle-attached, and tightly particle-attached fractions from the rumen with alfalfa as the diet; correspondingly. T_L, T_PT, and T_RS represent the fluid-borne, weakly, and tightly attached fractions from the rumen with triticale as the diet. Values listed for A_ and T_ are the sum of their respective fractions.

  • NA, not applicable.

BacteroidetesPorphyromonadaceaeTannerella20001021
U-PorphyromonadaceaeNA4373461413
PrevotellaceaePrevotella3729253025239178
Xylanibacter01001011
Hallella741011122
U-PrevotellaceaeNA81111815103033
RikenellaceaeRikenella00010001
U-BacteroidalesNA1423282122206563
U-BacteroidetesNA 3625631114
FirmicutesLachnospiraceaeButyrivibrio10743372113
Lachnospiraceae INS134654815
Moryella04110051
Lachnobacterium01002012
Pseudobutyrivibrio41135169
Oribacterium10000010
U-LachnospiraceaeNA1712241727245368
RuminococcaceaeAcetivibrio00010001
Ruminococcus3447761120
Papillibacter00301132
Sporobacter00007108
U-RuminococcaceaeNA161717291895056
ClostridiaceaeClostridium10000010
VeillonellaceaeSucciniclasticum1160511177
Mitsuokella30100040
U-VeillonellaceaeNA00001001
ErysipelotrichaceaeBulleidia20200141
Incertae Sedis XIIIAnaerovorax60320294
U-ClostridialesNA8612158152638
U-ClostridiaNA46912671925
U-FirmicutesNA 00022004
ProteobacteriaSuccinivibrionaceaeSuccinivibrio20110132
Succinimonas00001001
Ruminibacter00013004
U-SuccinivibrionaceaeNA02000020
SphingomonadaceaeSphingomonas01000010
U-AlphaproteobacteriaNA00021205
AlcaligenaceaeSutterella30000030
U-AlcaligenaceaeNA01000010
U-DesulfovibrionaceaeNA10000010
FibrobacterFibrobacteraceaeFibrobacter02001021
TenericutesAnaeroplasmataceaeAnaeroplasma12010031
Asteroleplasma01000010
ActinobacteriaCoriobacteriaceaeOlsenella00100010
ChloroflexiAnaerolinsaceaeLevilinea06021164
TM7NATM7_genera_incertae_sedis00040105
SR1NASR1_genera_incertae_sedis00100010
(U-domain Bacteria)NANA231444612

Composition of the bacterial populations in different rumen digesta fractions

Bacteroidetes and Firmicutes constituted a large proportion of the bacteria identified in each rumen fraction of cows fed alfalfa or triticale. Others were Proteobacteria, Fibrobacter, Tenericutes, Actinobacteria, Chloroflexi, TM7, and SR1 (Table 2). Many clones could only be identified at the family (e.g. Prevotellaceae, Lachnospiraceae, and Ruminococcaceae), order (Bacteroidales and Clostridiales), class (e.g. Clostridia), or phylum (e.g. Bacteroidetes) level, and a few only at the domain level. Members of the genera Prevotella, Butyrivibrio, Lachnospiraceae INS, Pseudobutyrivibrio, and Ruminococcus were found in bacterial populations in the planktonic (_L), loosely attached on particulates (_PT), and tightly attached on particulates (_RS) fractions in the rumen of cows fed either alfalfa or triticale.

In addition to the genera universally present in all fractions, members of Succiniclasticum, Anaerovorax, Succinivibrio, and Anaeroplasma were identified in the _L fraction of cows fed either alfalfa or triticale. Members of Hallella, Oribacterium, Clostridium, Mitsuokella, Bulleidia, and Sutterella were only identified in the _L fraction of cows fed alfalfa. Members of Rikenella, Moryella, Acetivibrio, Levilinea, and TM7_genera_incerta_sedis were only identified in the _L fraction of cows fed triticale. Similarly, members of the genera Xylanibacter, Hallella, Lachnobacterium, Succiniclasticum, Fibrobacter, and Levilinea were found in the _PT fraction of cows fed either alfalfa or triticale, while members of Moryella, Sphingomonas, Anaeroplasma, and Asteroleplasma were only identified in the _PT fraction of cows fed alfalfa and members of Tannerella, Papillibacter, Sporobacter, Succinimonas, and Ruminobacter were only identified in the _PT fraction of cows fed triticale. For the bacterial populations in the _RS fraction, the cows fed alfalfa shared members of the genera Hallella, Papillibacter, Bulleidia, Anaerovorax, and Succinivibrio with the cows fed triticale, while members of Moryella, Mitsuokella, Olsenella, and SR1_genera_incertea_sedis were only identified in cows fed alfalfa and members of Sporobacter, Succiniclasticum, Levilinea, and TM7_genera_incertae_sedis were only identified in cows fed triticale.

Biodiversity analysis

In the rumen of cows fed either alfalfa or triticale, the bacterial populations in the _RS fraction are more diverse than those in the _L and _PT fractions (Table 1). The species richness estimated at a distance of 0.03 is the highest in the _RS fraction (1089), medium in the _PT fraction (385), and the lowest in the _L fraction (368) of cows fed alfalfa and triticale (Table 1). Similarly, the Shannon index calculated for the bacterial populations in different digesta fractions is the highest in the _RS fraction (4.74 and 4.78), medium in the _PT fraction (4.69 and 4.66), and the lowest in the _L fraction (4.50 and 4.62) of cows fed alfalfa and triticale, respectively (Table 3).

Table 3.   Biodiversity index, distribution, and similarity of microbial populations in different rumen digesta fractions in animals fed with alfalfa or triticale
Libraries*Number of sequencesNumber of OTUsSpecies richnessOTUs shared (%)Shannon indexP value§ (XY, YX)
XYYX
  • *

    A_ and T_ represent the rumen microbial community fed with alfalfa and triticale, respectively. _L, _PT, and _RS represent the fluid-borne, weakly particle-attached, and tightly particle-attached fractions, respectively.

  • The numbers before the parentheses represent the total number of sequences, OTUs, or estimated species obtained from the two libraries on the same fraction, the four libraries on two different fractions or the six libraries on different fractions of a community. The two numbers in the parenthesis represent the number of sequences, OTUs, or estimated species obtained from each of the two libraries on the same fraction, each of the two fractions (four libraries) or each of the two communities, respectively.

  • XY represents the number of OTUs in the first library, fraction, or community (as indicated in column 1) shared with the second library, fraction, or community. YX represents the number of OTUs in the second library, fraction, or community shared with the first library, fraction, or community.

  • §

    XY represents the P values obtained when the first library, fraction, or community (as indicated in column 1) is compared with the second library, fraction, or community. YX represents the P values obtained when the second library, fraction, or community is compared with the first library, fraction, or community.

  • NA, not applicable.

A_L, 1 vs. 2172 (89, 83)128 (71, 57)291 (206, 453)25324.500.1338, 0.2027
A_PT, 1 vs. 2162 (82, 80)140 (66, 74)300 (223, 458)27244.690.3073, 0.0693
A_RS, 1 vs. 2163 (78, 85)138 (66, 72)564 (359, 952)18164.740.4745, 0.2146
A_4973271027 (801, 1362)NANA5.58NA
T_L, 1 vs. 2186 (96, 90)134 (72, 62)195 (159, 259)24274.620.0179, 0.0058
T_PT, 1 vs. 2179 (96, 83)134 (66, 68)209 (170, 278)22224.66<0.0001, 0.8495
T_RS, 1 vs. 2152 (77, 75)134 (68, 66)736 (444, 1360)10114.780.7851, 0.7926
T_517329662 (561, 807)NANA4.78NA
A_L vs. A_PT334 (172, 162)268 (128, 140)NA1211NA<0.0001,<0.0001
A_L vs. A_RS335 (172, 163)266 (128, 140)NA1413NA<0.0001, 0.0031
A_PT vs. A_RS325 (162, 163)278 (140, 138)NA99NA<0.0001, 0.0001
T_L vs. T_PT365 (186, 179)268 (134, 134)NA98NA<0.0001,<0.0001
T_L vs. T_RS338 (186, 152)268 (134, 134)NA2119NA<0.0001, 0.9109
T_PT vs. T_RS331 (179, 152)268 (134, 134)NA1211NA<0.0001,<0.0001
A_ vs. T_1014 (497, 517)656 (327, 329)NA1616NA<0.0001,<0.0001

When the bacterial community fed alfalfa is compared with the one fed triticale, the total species richness estimated for the community fed alfalfa (1027) is 1.6 times higher than that estimated for the community fed triticale (662) (Table 3). The Shannon index estimated for the community fed alfalfa is 5.58, also markedly higher than the value of 4.78 estimated for the community fed triticale. For each fraction analyzed, there were significant differences between the bacterial populations when the cows were fed alfalfa as compared with triticale (Table 1). A deeper analysis of phylogenetic groups (Prevotellaceae, Bacteroidales, Lachnospiraceae, and Ruminococcaceae) present on or embedded in the fiber fractions (_PTRS) indicates that these populations also show phylogenetic differences when animals are fed alfalfa or triticale (Table 1).

S-Libshuff estimation detected a significant difference in the bacterial population composition between any two digesta fractions of cows fed alfalfa (Table 3). In cows fed triticale, significant differences were detected in the bacterial composition between any two fractions, except between the _RS and the _L fractions. A significant difference in the bacterial composition was also detected between the rumen bacterial communities of cows fed alfalfa or triticale.

The percentage of OTUs shared between two fractions or communities was further estimated using the function provided in sons (Table 3). The bacterial populations in any two fractions shared 9–14% of their OTUs and 8–21% of their OTUs in the cows fed either alfalfa or triticale, respectively. When the rumen bacterial community of cows fed alfalfa was compared with that of cows fed triticale, 16% of the OTUs in the former community were identical to the same percentage of the OTUs in the latter community.

To investigate the effect of using different animals on the statistical data obtained, the similarity between the two libraries built on the same digesta fraction from different animals was also estimated (Table 3) using the S-Libshuff estimation. No significant difference in the bacterial composition was detected between any two libraries constructed for any given fraction of the rumen in cows with alfalfa as the diet. For the libraries constructed for the rumen of cows with triticale as the diet, significant differences in the bacterial composition were detected between the two libraries constructed for both the _L and the _PT fractions. Overall, the bacterial populations identified in the library of period 1 for a specified fraction shared 16–32% and 10–27% of their OTUs with the bacterial populations identified in the corresponding library of period 2 in the rumen of cows fed alfalfa and triticale, respectively. At the community level, 16% of the OTUs in the alfalfa-fed bacterial community were identical to the same percentage of the OTUs in the triticale-fed bacterial community.

Discussion

A high level of biodiversity in the rumen bacterial community has been revealed in this study. In total, 656 OTUs were identified in the rumen of animals fed alfalfa or triticale, with 74% of the OTUs (451) belonging to new bacterial species. The species richness (Chao 1) estimated for the bacterial communities fed alfalfa and triticale was 1027 and 662, respectively. The values are markedly higher than the 300–400 estimated (also using Chao 1 estimation) by Edwards et al. (2004) based on the rumen bacterial clones obtained by Whitford et al. (1998), Tajima et al. (1999), and Tajima et al. (2000). This is likely because in each of these previous studies, only limited (80–90) 16S rRNA gene clones were sequenced for each library and multiple phases of the rumen were not considered. The high species richness estimations obtained in this study are relatively close to those (∼621) estimated by Larue et al. (2005) for different fractions of the rumen of sheep (no data were provided for the entire community). In both studies, fractionation of digesta samples and construction of relatively large clone libraries for each fraction were carried out. The sequences obtained by Larue et al. (2005) were<500 bp in length; therefore, possible inconsistencies in the phylogenetic placement could result in biases to the estimation of diversity. We believe that the species richness estimation arrived at in this study is reflective of the bacterial biodiversity of the rumen of cows fed high-forage diets, indicating that the rumen bacterial community is more diverse than expected.

The biodiversity of the rumen bacterial community defined in this study could also be seen at the genus level. Using the naïve bayesian classifier provided in RDP, the 656 OTUs obtained in our study were classified into 32 genera (27 and 23 genera in the three fractions of rumen fed with alfalfa and triticale, respectively), markedly more than the 19 genera reported by Larue et al. (2005) in the rumen of sheep fed orchard grass hay, with or without corn (c. 150 clones sequenced for each diet), or the eight genera defined by Tajima et al. (1999) in the rumen of cows fed timothy hay and concentrate consisting of corn and barley (84 clones sequenced). The genus number is also substantially greater than the eight genera identified by An et al. (2005) in the rumen of Jinhua cattle fed corn and corn stover (197 clones sequenced) and the eight genera identified by Nelson et al. (2003) in the rumen of African Zebu cows (wild animals, 252 clones sequenced). The higher genus detection ratio in our study than that in others is probably due to the fractionation of the rumen digesta and the larger libraries constructed (c. 500 sequences from the rumen of cattle on each diet). Differences between the animals examined, the diets used to feed the animals, the season of sampling, and the sequence classification method may also play a role. Of the 42 genera identified in all of these previous studies, only eight were common to genera that we identified in this study.

In line with the other studies mentioned above, Bacteroidetes and Gram-positive low G+C Firmicutes identified in this study constituted a major fraction of the total clones in each fraction of the rumen fed alfalfa or triticale. Members of Prevotella, Butyrivibrio, Lachnospiraceae incertae sedis, Pseudobutyrivibrio, and Ruminococcus were universally present in all the fractions. Except for Lachnospiraceae incertae sedis and Pseudobutyrivibrio, the presence of the other genera is reported in other clone libraries constructed from the rumen, confirming that some of these are common species. Moreover, Prevotella are the most abundant (comprising 14–22% of the total clones) of all genera identified in our study, as well as in most other studies, suggesting that they play a fundamental role in the rumen ecosystem.

All typical fibrolytic rumen bacteria commonly isolated from the rumen were identified in this study. Ruminococcus flavefaciens is represented by 16 (1.6% of total clones) 16S rRNA gene sequences and is identified in all fractions associated with plant particles, but not the rumen liquid. Ruminococcus albus is represented by five (0.5% of the total clones) sequences and is found in T_PT and T_RS. Fibrobacter succinogenes is represented by only one sequence (0.1% of the total clones) and is found in A_PT. The lack of detection of significant numbers of Fibrobacter sequences in this study is a concern. Larue et al. (2005) suggested that low recoveries of Fibrobacter sequences in clone libraries may be due to biases in amplification and cloning. However, we have been able to recover Fibrobacter sequences as high as 16% of all clones in a single library (data not shown) in our laboratory using the methods described in this paper. It is possible that the previous dietary history of the cattle used in this study, which included very high levels of concentrates, may have impacted the species profile in the rumen.

In this study, the differences in the bacterial populations between different rumen digesta fractions and communities were estimated statistically for the first time. S-Libshuff estimation shows that, with the exception of one in 12 comparisons, bacterial populations between any two of the _L, _PT, and _RS fractions within the same diet are significantly different. In contrast, no significance difference was found, except in two of 12 comparisons, between the bacterial populations of the same fraction of the rumen from different animals. The results generally agree with the percentage of species shared between two different fractions or two different samples of one fraction. The bacterial populations present in the same rumen fraction from different animals share more species (22 ± 6%) than those between two different fractions (12 ± 4%). This indicates that although the difference in bacterial populations caused by animal variation is important, the difference between the planktonic, loosely particle-associated, and tightly particulate-associated fractions, is greater.

The results from the species richness estimation carried out in this study also support this conclusion. The species richness of the _RS fraction is 1.9 and 3.5–3.8 times higher than either of the other two fractions from cows fed alfalfa or triticale, respectively. The result is also in line with Larue et al. (2005), who found that the richness of bacterial populations attached to particles in the rumen of sheep is higher than that of the planktonic fraction. In this study, we further fragmented the bacterial populations attached on fiber particles into biofilm and tightly attached or embedded fractions, and demonstrated that the largest bacterial biodiversity in the rumen fed alfalfa or triticale was present on or in particulates and were not easily dislodged. This result also indicates that the rumen sample fractionation and DNA extraction methods used in this study can be reliably used to investigate the bacterial distribution and biodiversity in different functional compartments of rumen digesta, and further suggests their applicability for the targeted isolation of DNA from bacteria actively involved in plant fiber degradation for metagenomic studies.

We have provided statistical data to support the observation that diet affects the composition, spatial distribution, and biodiversity of the rumen bacterial community. S-Libshuff estimation shows that significant differences exist between the community fed alfalfa and the one fed triticale. Most of the species in the bacterial community with alfalfa fed as the diet (84%) are different from those in the community with triticale fed as the diet. The effect of diet could also be seen from the species richness estimates for these two communities. The species richness of the community with alfalfa as the diet (1027) is 1.6 times of that of the community with triticale as the diet (662). This is likely due to the greater richness and complexity of nutrients within the alfalfa hay as compared with the relatively restricted nutrient content of triticale straw. Alfalfa hay has a much higher crude protein content and a lower level of acid detergent fiber than straw. The greater complexity of the nutrients within the alfalfa hay must demand a greater complexity of bacteria to efficiently utilize all the forage components. Deeper phylogenetic analysis indicates that the major populations of bacteria in the rumen are different when animals are fed different diets. However, because the majority of the discovered OTUs are most similar to the sequences of uncultured bacteria, little can be said of the functional significance of these findings. In the future, high-throughput sequencing and functional metagenomic analysis may provide linkages between phylogeny and phenotype.

The data generated by this study add to the growing literature that suggests that the rumen is one of the most phylogenetically complex ecosystems. The rumen ecosystem is being chosen for an increasing number of gene mining studies that aim to isolate novel enzymes for the conversion of lignocellulosic material to high-value products. This study suggests that the community of rumen bacteria that develops to digest a particular forage is quite specific. It is also possible that the enzymes expressed by the selected community may be more specific to the substrate being fed. It would therefore make sense to present the rumen with the material for which biomass conversion is being proposed, and allow the community to adapt and express the most suitable enzymes before gene mining, rather than selecting genes isolated from rumens fed unrelated feedstuffs. Our understanding of the composition and biodiversity of the rumen bacterial community is still limited. High-throughput sequencing studies that generate a large number of short read sequences will be valuable to economically study the rumen ecosystem of many more animals on different diets. However, long sequence reads obtained from full-length 16S rRNA gene cloning studies will still be needed to generate enough information to design specific PCR assays for quantitative analysis and to reliably place species into phylogenetic trees.

Acknowledgements

Financial support for this study was provided by the Alberta Agricultural Research Institute and the Canadian Triticale Biorefinery Initiative through the Agricultural Bioproducts Initiative Program of Agriculture and Agri-Food Canada. Financial support for Y.H.K. was through a Peer Review Project of Agriculture and Agri-Food Canada. We thank Edith Valle and Lyn Paterson for excellent technical assistance.

Ancillary