Appl Microbiol Biotechnol (2007) 75:165174
DOI 10.1007/s00253-006-0802-y
`
APPLIED MICROBIAL AND CELL PHYSIOLOGY
Dominance of Prevotella and low abundance of classical
ruminal bacterial species in the bovine rumen revealed
by relative quantification real-time PCR
David M. Stevenson & Paul J. Weimer
Received: 26 September 2006 / Revised: 28 November 2006 / Accepted: 10 December 2006 / Published online: 18 January 2007
# Springer-Verlag 2007
Abstract Relative quantification real-time PCR was used
to quantify several bacterial species in ruminal samples
from two lactating cows, each sampled 3 h after feeding on
two successive days. Abundance of each target taxon was
calculated as a fraction of the total 16S rRNA gene copies
in the samples, using taxon-specific and eubacterial
domain-level primers. Bacterial populations showed a clear
predominance of members of the genus Prevotella, which
comprised 42% to 60% of the bacterial rRNA gene copies
in the samples. However, only 2% to 4% of the bacterial
rRNA gene copies were represented by the classical
ruminal Prevotella species Prevotella bryantii, Prevotella
ruminicola and Prevotella brevis. The proportion of rRNA
gene copies attributable to Fibrobacter succinogenes,
Ruminococcus flavefaciens, Selenomonas ruminantium
and Succinivibrio dextrinosolvens were each generally in
the 0.5% to 1% range. Proportions for Ruminobacter
amylophilus and Eubacterium ruminantium were lower
(0.1% to 0.2%), while Butyrivibrio fibrisolvens, Strepto-
coccus bovis, Ruminococcus albus and Megasphaera
elsdenii were even less abundant, each comprising
<0.03% of the bacterial rRNA gene copies. The data
Disclaimer: Mention of products is for informational purposes only
and does not imply a recommendation or warranty by the USDA over
other products that may also be suitable.
D. M. Stevenson : P. J. Weimer (*)
United States Dairy Forage Research Center, Agricultural
Research Service, U.S. Department of Agriculture,
Madison, WI 53706, USA
e-mail: pjweimer@wisc.edu
P. J. Weimer
Department of Bacteriology, University of Wisconsin-Madison,
Madison, WI 53706, USA
suggest that the aggregate abundance of the most inten-
sively studied ruminal bacterial species is relatively low and
that a large fraction of the uncultured population represents
a single bacterial genus.
Keywords Bovine . PCR . Prevotella . Primers .
Real-time PCR . Rumen
Introduction
The rumen is a complex habitat in which feedstuffs are
fermented primarily to a mixture of volatile fatty acids
(VFA, predominantly acetic, propionic, and butyric acids)
that serve as the major nutrient source for the ruminant
animal. Much of our knowledge of the ruminal metabolism
of these feedstuffs was gained through in vitro study of
approximately two dozen bacterial species, most of which
were isolated in the early decades of rumen microbiology
(Hungate 1966; Krause and Russell 1996). Techniques of
molecular microbial ecology provide an opportunity to
quantify these ruminal species with great sensitivity and
precision, and several recent reports have provided a sense
of which species are most abundant in the rumen under
particular feeding conditions (Tajima et al. 2001; Klieve et
al. 2003). However, the abundance of individual species as
a fraction of the total bacterial population in the rumen was
not clarified. In natural environments examined thus far,
known species typically represent less than 10%, and often
on the order of 1%, of the estimated species (Janssen 2006).
The purpose of this study was to determine the relative
contribution of the classical species of ruminal bacteria to
the total bacterial population of the rumen. To do this, we
have quantified three Prevotella species and ten non-
Prevotella species in the bovine rumen using quantitative
166 Appl Microbiol Biotechnol (2007) 75:165174

real-time PCR referenced to the total bacterial domain to was immediately placed on ice in a sealed tube while the
calculate the relative abundance of 16S rRNA. solids were processed. As chilling was shown to maximize
the release of particle-associated bacteria from ruminal
contents (Dehority and Grubb 1980), squeezed solids were
Materials and methods placed in a chilled Waring blender with 75 ml ice-cold
extraction buffer (EB; 100 mM Tris/HCl, 10 mM ethyl-
Rumen samples Two lactating, fistulated Holstein cows enediaminetetraacetic acid [EDTA], 0.15 M NaCl pH 8.0)
(Table 1) were maintained in adjacent indoor tie stalls and blended for 2 min. The material was then transferred to
according to a protocol approved by the University of a 250 ml centrifuge bottle, along with 25 ml of cold EB
Wisconsin-Madison Animal Care and Use Committee, and used to rinse the blender. This suspension was then
were milked twice daily. Cows were allowed ad libitum centrifuged gently at 500g for 15 min at 4C to remove
access to feed and water, and feed was supplied once daily the considerable amount of broken plant particles while
(3 h after morning milking) as a total mixed ration (TMR) keeping bacterial cells in suspension. The supernatant was
that contained lucerne silage, corn silage, corn grain, then filtered through a small quantity of glass wool into a
soybean meal, roasted soybeans, blood meal, whole fresh 250 ml centrifuge bottle. The squeezed solids were
cottonseed and supplemental vitamins and minerals. TMR then shaken with an additional 25 ml cold EB and similarly
component composition was 27.5% neutral detergent fiber centrifuged and filtered into the same fresh bottle. The
and 18% crude protein (N6.25), and forage represented 25 ml of filtered ruminal liquid was then added and the
37% of TMR dry matter. Ruminal samples (approximately mixture centrifuged (10,000g, 25 min, 4C). The bacterial
equal volumes of solids and liquid totaling 1.8 l) were cell pellet was resuspended in 4 ml cold extraction buffer
collected 3 h after feeding on two consecutive days (30 and and 700 L aliquots were transferred to each of three or
31 days after adaptation to the TMR) and transported to the four screw cap microcentrifuge tubes containing 0.5 g
laboratory in sealed, pre-warmed thermos bottles. The 3 h zirconium/silica beads (0.1 mm diameter), 50 l 20%
time point was selected as near the midpoint of the upper sodium dodecyl sulfate [SDS] and 700 l equilibrated
and lower pH values of ruminal contents within the feeding phenol (pH 8.0). Cell lysis was achieved by bead disruption
cycle (Weimer et al. 1999). (Weimer et al. 1999). The aqueous and phenol phases were
separated by centrifugation (12,000g, 5 min) and the
DNA preparation Ruminal samples collected each day aqueous phase was extracted twice more with 500 l
were processed separately for each cow. Within 10 min of phenol (pH 8.0), then twice with 500 l phenol/chloroform,
collection, a quantity of the rumen contents was tightly pH 8.0 and finally twice with 500 l chloroform. Small
squeezed through four layers of cheesecloth into a vessel amounts of EB were occasionally added to keep the
continually sparged with CO2 to yield ruminal fluid and aqueous volume above 450 l. The final supernatant was
squeezed solids. For each sample, 25 ml of fluid and 25 g combined with 0.1 vol of 3.0 M Na acetate and precipitated
of solids were used for DNA extraction. The fluid portion with 0.6 vol of isopropanol. The precipitated DNA pellet
was then washed twice by adding 1,000 l of 70% ethanol
Table 1 Production and ruminal pH data of cows used in this study
without disturbing the pellet, centrifuging (12,000g,
5 min) and decanting off the supernatant. After a final
Parameter Cow 4884 Cow 4991 Pooled SE additional centrifugation, the remaining liquid was removed
by aspiration and the pellet was dried at room temperature
Days in milk at start of trial 74 84
Body weight (kg)a 6511 7172 12
for 20 min, then resuspended overnight in 50100 l TE
Dry matter intake (kg/day)b 29.21 25.62 6.0 (10 mM TrisHCl, 1 mM EDTA, pH 8.0) depending on the
Milk production pellet size. DNA concentrations were estimated by spectro-
Milk yield (kg/day)b 47.11 36.12 4.0 photometry and the DNA preparations were stored at 4C
Fat content (%)c 3.421 3.672 0.63 for short-term use, or were archived at 80C.
Protein content (%)c 2.441 3.002 0.06
Ruminal pH Primer design and validation Primer pairs for quantitative
Day 30 6.07 5.71
analysis (Table 2) were designed to 16S rRNA gene
Day 31 6.44 5.88
sequences using Applied Biosystems (Foster City, CA)
Different superscript numbers indicate that the mean values between Primer Express software. Sequences for as many examples
cows differ (P<0.05). of the target species as possible were aligned using the
a
Mean value from days 30 and 31.
b
Mean values from days 29, 30, and 31.
PILEUP program within the GCG Wisconsin software
c
Mean values from six samples (two milkings each on days 29, 30, package (Accelrys, San Diego, CA). From this alignment,
and 31). primer pairs were selected using the Primer Express
Appl Microbiol Biotechnol (2007) 75:165174 167

Table 2 PCR primers used in this study

Target taxon and specific strain Primer set Primer sequences Tma Eb Amplicon
tested Tmc

Butyrivibrio fibrisolvens H17c ButFib2F ACCGCATAAGCGCACGGA 62 1.98 74

ButFib2R CGGGTCCATCTTGTACCGATAAAT 61
Eubacterium EubRum2F CTCCCGAGACTGAGGAAGCTTG 61 1.97 80
ruminantium GA195 EubRum2R GTCCATCTCACACCACCGGA 60
Fibrobacter FibSuc3F GCGGGTAGCAAACAGGATTAGA 59 1.93 77
succinogenes S85 FibSuc3R CCCCCGGACACCCAGTAT 59
Megasphaera elsdenil T81 MegEls2F AGATGGGGACAACAGCTGGA 59 1.97 79
MegEls2R CGAAAGCTCCGAAGAGCCT 58
Prevotella brevis B14 PreBre1F GGTTTCCTTGAGTGTATTCGACGTC 61 1.98 82
PreBre1R CTTTCGCTTGGCCGCTG 60
Prevotella bryantii PreBry2F AGCGCAGGCCGTTTGG 61 1.95 83
GA33 PreBry2R GCTTCCTGTGCACTCAAGTCTGAC 61
Prevotella ruminicola 23 PreRum1F GAAAGTCGGATTAATGCTCTATGTTG 58 1.93 71
PreRum1R CATCCTATAGCGGTAAACCTTTGG 59
Ruminobacter RmbAmy2F CTGGGGAGCTGCCTGAATG 60 1.96 82
amylophilus H18 RmbAmy2R GCATCTGAATGCGACTGGTTG 60
Ruminococcus albus 7 RumAlb3F TGTTAACAGAGGGAAGCAAAGCA 60 1.85 75
RumAlb3R TGCAGCCTACAATCCGAACTAA 59
Ruminococcus RumFla3F TGGCGGACGGGTGAGTAA 60 1.79 78
flavefaciens FD-1 RumFla3R TTACCATCCGTTTCCAGAAGCT 60
Selenomonas SelRum2F CAATAAGCATTCCGCCTGGG 61 1.95 82
ruminantium D SelRum2R TTCACTCAATGTCAAGCCCTGG 61
Streptococcus bovis JB1 StrBov2F TTCCTAGAGATAGGAAGTTTCTTCGG 59 1.95 82
StrBov2R ATGATGGCAACTAACAATAGGGGT 59
Succinivibrio dextrinosolvens 22b SucDex1F CGTCAGCTCGTGTCGTGAGA 60 1.95 80
SucDex1R CCCGCTGGCAACAAAGG 60
Domain bacteriad BAC338F ACTCCTACGGGAGGCAG 52 1.922.01e f

BAC805R GACTACCAGGGTATCTAATCC 48
Genus Prevotellag PreGen4F GGTTCTGAGAGGAAGGTCCCC 60 1.92 83
PreGen4R TCCTGCACGCTACTTGGCTG 61

All primers were designed for this study except for BAC338F and BAC805R, which were designed by Yu et al. (2005).
a
Tm values calculated using Primer Express software (Applied Biosystems, Foster City, CA). Tm varied among species.
b
E represents the efficiency calculated from an amplification of a dilution series of target DNA as described in the text. For the domain-level
eubacterial primer, the indicated range spans the efficiencies determined against all of the indicated genus (Prevotella) or species-level taxa
listed. An efficiency of 2.0 is equivalent to 100% of theoretical amplification per PCR cycle.
c
Amplicon Tm calculated from the dissociation protocol run on the resultant amplicon generated with the selected primer against the pure species
indicated.
d
Domain-level primers.
e
Efficiency values varied among species within this range.
f
Tm values varied among species. Five strains were tested, which included B. fibrisolvens (Tm=82C), E. ruminantium (Tm=83C), F.
succinogenes (Tm=85C), R. flavefaciens (Tm=83C), and S. bovis (Tm=82C).
g
Genus-level primers. P. bryantii GA33 was used as test strain.

software and screened or were designed manually and
analyzed with the Primer Express software. Primer pairs
were selected for regions conserved among all examples of
a particular species or genus and for as much sequence
divergence as possible from those of other closely related
species. In addition, pairs were selected to have Tm values
near 60C and within 2C of each other, to be as close to
20 bp in length as possible, to amplify a region of between
60 and 200 bp in total, to avoid runs of greater than 4
identical nucleotides, to avoid stable secondary structures
and to mimimize the number of G+C residues within the
last 5 nucleotides near the 3 end. An exception was the
domain-level eubacterial pair, which was designed by Yu et
al. (2005) to amplify a total of 468 bp and to have lower
Tm values (see Table 2). Inevitably, a few primers designed
to include all strains of a target species displayed identity to
a few strains of closely related species as indicated in
Table 3. Primer pairs specific for all members of the target
species and not identical to any non-target, named species
will be referred to here by the species name (e.g.,
Fibrobacter succinogenes), while primer pairs that display
identity for all members of the target species plus one or
168 Appl Microbiol Biotechnol (2007) 75:165174

Table 3 Results of BLAST searches of primers

Primer Number of matches to named species Number of matches to

unclassified sequences

ButFib2 Butyrivibrio fibrisolvens (4) Pseudobutyrivibrio ruminis (1) 0

EubRum2 Eubacterium ruminantium (1) 0
FibSuc3 Fibrobacter succinogenes (9) 2
MegEls2 Megasphaera elsdenii (13) Megasphaera sp. (2) 15
PreBre1 Prevotella brevis (1) 0
PreBry2 Prevotella bryantii (2) 5
PreRum1 Prevotella ruminicola (5) Prevotella sp. (2) 14
RumAlb3 Ruminococcus albus (2) 0
RmbAmy2 Ruminobacter amylophilus (2) 0
RumFla3 Ruminococcus flavefaciens (30) Ruminococcus callidus (4) Ruminococcus sp. (8) 48
Lachnospiraceae sp. (3)
SelRum2 Selenomonas ruminantium (11) Selenomonas sputigena (2) Selenomonas sp. (5) Mitsoukella 90
jalalundinii (1) Mitsoukella multiacidus (1) Mitsoukella sp. (1)
StrBov2 Streptococcus bovis (21) Streptococcus infantarius (3) Streptococcus luteciae (3) Streptococcus 17
lutetiensis (1) Streptococcus sp. (10)
SucDex1 Succinivibrio dextrinosolvens (1) Anaerobiospirillum thomasii (1) Anaerobiospirillum sp. (1) 9
Mycobacterium sp. (1)

Values indicate the number of sequences for named species and for unclassified sequences that show 100% identity to both forward and reverse
primers.

more strains of another named species will be referred to as
representing an only slightly larger group composed
primarily of the target species (e.g., the Ruminococcus
flavefaciens group). In addition, most primer pairs dis-
played sequence identity to some taxonomically unidenti-
fied sequences. Such matches reflect the very large number
of sequences in the database derived from environmental
samples relative to sequences from cultured species and are
likely to reflect a high degree of relatedness between the
target species and these uncultured bacterial strains from
which the environmental sequences were obtained.
Real-time PCR Each microtiter plate well contained the
following: 2X SYBR Green Master Mix (Applied Bio-
systems), which contained all the nucleotides, polymerase,
reaction buffer and SYBR green dye; forward and reverse
primers at concentrations of 300, 450 or 900 nM (depend-
ing on previously determined optimal concentration) and
nuclease-free water to a total of 23 l per well. To this
solution, 2.0 l of each sample or standard was added and
the plate was briefly centrifuged and placed in the
thermocycler for analysis.
Real-time PCR was performed using the Applied
Biosystems Prism 7300 sequence detection system with
fluorescence detection of SYBR green dye. For most of the
primer pairs, amplification consisted of an initial hold for
10 min followed by 40 to 50 cycles of 95C for 1525 s
and 5861C for 6090 s depending on the Tm of the
specific primer pair. For the domain-level primer pair,
which had lower Tm values (Table 2), amplification
consisted of an initial hold for 10 min followed by 40 to
50 cycles of 95C for 15 s and 50C for 60 s and finally
72C for 90 s. Standards used for PCR efficiency
calculations were run on the same plate with quadruplicate
samples of 20 ng (estimated spectrophotometrically) of
DNA purified from individual ruminal samples. For the
standards, genomic DNA purified from a pure culture of the
target species was subjected to seven sequential fivefold
dilutions, each in quadruplicate. A linear relationship was
observed between Ct and log of DNA concentration when
each primer pair was tested against purified DNA from its
target taxon (r2 =0.997 to 0.999) and when the domain-
level primer pairs were tested against purified DNA from
each of the test species (r2 =995 to 0.999).
To verify the accuracy of the PCR methods, amplifica-
tion efficiencies for DNA isolated from pure cultures were
compared to amplification efficiencies for the same species
in DNA isolated from ruminal contents and spiked with a
known amount of target DNA (Fig. 1). The efficiencies of
amplification were essentially identical, indicating that the
same target DNA was amplified at similar efficiency
regardless of the presence of non-target DNA.
Relative quantification Ruminal contents contain high
concentrations of organic matter, much of it in the form
of plant materials and various byproduct feeds. This makes
it difficult to purify bacterial DNA from whole ruminal
contents. To minimize the problem of DNA purification and
absolute quantification, the relative quantification method
of Pfaffl (2001) was used in this study. In relative
Appl Microbiol Biotechnol (2007) 75:165174
Fig. 1 Efficiency of PCR amplification of DNA derived from pure
cultures and ruminal fluid. Real-time PCR was carried out on DNA
purified from B. fibrisolvens H17c, to generate a standard curve
displaying a loglinear effect of target concentration (r2 =0.999) and
an amplification efficiency (antilog of negative reciprocal of line
slope) of 2.00. Known amounts (0.00025 to 1 ng) of this same
B. fibrisolvens DNA were added to DNA purified from a ruminal
sample (calculated from the above standard curve to contain
0.00064 ng of B. fibrisolvens DNA), resulting in a nearly super-
imposed regression line, having a loglinear relationship (r2 =0.999)
and an amplification efficiency of 1.99, not significantly different
(P>0.10) from the pure culture standard. Data points are the mean
values of four analytical replicates. Coefficients of variation among
the Ct values ranged from 0.11% to 0.68% (mean CV=0.40%), too
small to visualize as error bars on the graph. No-template (primer-
only) controls yielded Ct values of 41
quantification (most often used as a means of comparing
gene expression within a single species without the
necessity of synthesizing an in vitro RNA standard),
amplification is expressed relative to the amplification of
reference primers utilizing an experimentally derived
amplification efficiency (see Appendix). In this study, a
non-degenerate, domain-level primer set that amplified all
eubacterial species was used as the reference primer set.
Efficiencies of PCR reactions were determined for each
primer set by running a set of seven serial dilutions (1/5) of
the target species for each primer set along with the
unknowns, all in quadruplicate. A plot of Ct vs log DNA
concentration for the dilution series allowed the efficiency
to be calculated, as the negative reciprocal of the slope of
this line is equal to the log of the efficiency. In addition,
purified DNA from each target species and each ruminal
sample was run on a separate plate using the domain-level
eubacterial primers, BAC338F and BAC805R (Yu et al.
2005). This allowed the efficiency of this primer set for
each specific target species to also be determined. Finally, it
was assumed that amplification using the domain-level
169
primers would be equal to that of the target primer set, as
both were designed to amplify regions of the same 16S
rRNA gene. Thus, the lowest dilution of the target species
could be set to be theoretically equal on both plates
(domain-level and genus-level or species-level), allowing
this Ct value to be used as the control Ct value in
equation A1 in the Appendix. A sample calculation for
the relative proportion of 16S rRNA genes is provided in
the Appendix.
Statistical analysis Results from quadruplicate PCR reac-
tions for each combination of target taxon, cow and sample
day were averaged and the means were compared using the
general linear model of the SAS Statistical Software
package, version 7.0 (SAS, Cary, NC) with taxon, cow
and sample day as class variables.
Results
Each species-specific primer set was tested against all the
other target species used in this study and no primer pair
was found to significantly cross-react with any other than
the desired target species. The genus-level Prevotella
primers amplified DNA purified from all three of the
individual Prevotella species, but did not amplify any of the
10 non-Prevotella species tested. A significant reaction was
defined for this study as a Ct difference of at least 12 cycles
equivalent to a difference in amplification efficiency of 34
orders of magnitude depending on the efficiency of the
primer pair.
Table 2 shows the efficiencies of the real-time PCR
reactions, as calculated from equation A1 in the Appendix.
Efficiencies of the genus-specific or species-specific primers
ranged from 1.79 to 1.98, as would be expected from
different primers, though most were in the 1.92 to 1.98
range, close to the theoretical value of 2.0 that represents a
doubling of the target sequence during each PCR cycle. The
efficiencies of the domain-level eubacterial primer set
(BAC338F and BAC805R) when tested against all the
bacterial species utilized in the study ranged from 1.92 to
2.01. This much smaller range was expected as this universal
primer set was not degenerate and each species should have
the same target sequence, although the total amplicon length
may have varied very slightly among species. The slight
differences in PCR efficiency values among the primers
were incorporated into the relative quantification determi-
nations for comparison of the proportions of 16S rRNA
genes attributable to an individual taxon (see Appendix).
Table 2 also shows the Tm values of the final amplicons,
determined empirically by the generation of a dissociation
curve at the conclusion of each PCR run. These dissociation
170 Appl Microbiol Biotechnol (2007) 75:165174

curves confirmed the presence of a single highly dominant
peak. This was true for all the species-specific primers,
while multiple peaks were observed for the bacterial
domain-level primers, as would be expected with a primer
amplifying targets from many divergent species. The
Prevotella genus-level primer displayed an intermediate
profile with two or three peaks observed with ruminal
templates. However, both domain-level and genus-level
primers showed only a single dissociation peak when tested
against pure species templates.
Despite substantial differences in the productivity of the
two cows and in ruminal pH (Table 1), the proportion of
16S rRNA genes in the sample that were attributable to
each taxon were remarkably similar (P>0.05) between
cows and between sample days (Table 4). While there was a
trend toward higher relative population size (based on
proportion of 16S rRNA gene copies) of the genus
Prevotella in the samples from cow 4991, this trend
likewise was not statistically significant (P>0.05). Conse-
quently, the data from both cows on both sampling days
were combined within a target taxon for comparison of
population sizes across taxa.
The genus-level primer for Prevotella sp. revealed a
clear dominance (42% to 60% of the ruminal bacterial 16S
rRNA copies) of this genus in the ruminal samples from
both cows (Table 4). However, quantification with species-
specific primers for the three Prevotella species considered
most common in the rumen, together accounted for only
2% to 4% of the bacterial 16S rRNA gene copies. After
Prevotella, the most common species of those tested were
found to be F. succinogenes, the R. flavefaciens group, the
Selenomonas ruminantium group and the Succinivibrio
dextrinosolvens group, each of whose 16S rRNA gene
copies generally accounted for 0.5% to 1% of the total
bacteria. Proportions of 16S rRNA genes attributable to
Ruminobacter amylophilus and Eubacterium ruminantium
were each generally in the 0.1 to 0.2% range. The
Butyrivibrio fibrisolvens group, the Streptococcus bovis
group, Ruminococcus albus and Megasphaera elsdenii
were even less abundant, each comprising <0.03% of the
bacterial 16S rRNA gene copies. In aggregate the 13
individual species accounted for only about 5% to 7% of
the bacterial 16S rRNA gene copies in these ruminal
samples.
Discussion
Molecular-based analysis of several environments have
revealed that cultured bacterial species represent only a
small fraction of the total bacterial diversity (Janssen 2006)
and the rumen is no exception. Several studies showed that

Table 4 Percentages of target species in ruminal samples relative to total Eubacterial content

Target taxon Cow 4884 Cow 4991 MeanSEM P>Ff P>Ff

Day 30 Day 31 Day 30 Day 31 Cow Day

Butyrivibrio fibrisolvens groupa 0.022 0.027 0.022 0.024 0.0240.003 0.584 0.256
Eubacterium ruminantium 0.170 0.158 0.163 0.213 0.1760.025 0.584 0.658
Fibrobacter succinogenes 0.838 0.889 0.615 0.995 0.8350.160 0.783 0.416
Megasphaera elsdenii 0.0011 0.0001 0.0003 0.0004 0.00050.0004 0.728 0.564
Prevotella brevis 0.162 0.099 0.152 0.128 0.1350.028 0.693 0.266
Prevotella bryantii 1.226 0.730 1.942 1.830 1.4320.56 0.133 0.359
Prevotellaruminicola 1.600 1.582 1.756 2.032 1.7430.208 0.288 0.541
Ruminobacter amylophilus 0.170 0.141 0.392 0.189 0.2230.115 0.364 0.410
Ruminococcus albus 0.003 0.001 0.004 0.008 0.0040.003 0.361 0.811
Ruminococcus flavefaciens groupb 0.757 0.336 0.558 0.799 0.6130.212 0.759 0.831
Selenomonas ruminantium groupc 0.706 0.341 0.468 0.688 0.5510.176 0.883 0.345
Streptococcus bovis groupd 0.008 0.002 0.003 0.002 0.0040.003 0.525 0.477
Succinivibrio dextrinosolvens 0.715 0.656 1.071 0.799 0.8100.184 0.257 0.365
groupe
Sum of individual species 6.19 4.92 6.90 7.21 6.301.02 0.308 0.654
Genus Prevotella 49.60 42.44 58.12 59.93 52.528.09 0.211 0.658
a
BLAST search results for primer sequences included one strain of Pseudobutyrivibrio ruminis.
b
BLAST search results for primer sequences included four strains of Ruminococcus callidus and three strains of Lachnospiraceae.
c
BLAST search results for primer sequences included two strains of S. sputigena and three strains of Mitsuokella.
d
BLAST search results for primer sequences included a total of seven strains of three other Streptococcus species.
e
BLAST search results for primer sequences included two strains of Anaerobiospirillum species and one strain of Mycobacterium species.
f
Probability of a greater value in an F-test, from reduced statistical models in which cow or sampling day was analyzed as the sole independent
variable. The reduced models were applied after an initial statistical model in which cow and sampling day were treated as independent variables,
did not yield statistical significance for either variable at P<0.10.
Appl Microbiol Biotechnol (2007) 75:165174
the readily cultured bacterial population of the rumen (as
estimated by viable cell count in rich media) ranges from
5% to 15% of the total direct cell count determined
microscopically (Bryant and Burkey 1953; Varel and
Dehority 1989; Krause et al. 1999).
Real-time PCR was used in several studies to quantify
the abundance of particular bacterial species in the rumen
under a variety of conditions (e.g., see Klieve et al. 2003
and Tajima et al. 2001). However, there are few reports of
the abundance of individual taxa as a fraction of the total
ruminal bacterial population. In this study, we used real-time
PCR with genus-specific and species-specific primers,
coupled with a relative quantification method that employed
a non-degenerate, domain-level bacterial primer set and
corrected for slight variations in PCR efficiency among
primers to show that 42% to 60% of the bacterial 16S rRNA
gene copies in the rumen of two cows 3 h after feeding were
attributable to a single genus, Prevotella, although only 2%
to 4% of the bacterial 16S rRNA gene copies were
represented by the three most commonly isolated species
of Prevotella (Prevotella bryantii, Prevotella ruminicola and
Prevotella brevis). While more complete studies with larger
numbers of cows and different feeds are clearly warranted,
the numerical dominance of Prevotella observed in this
study is in accord the work of Wood et al. (1998) who used a
less specific ribotyping method to estimate that (1) the
genera Bacteroides and Prevotella together can account for
up to 62% of the rumen bacterial population and (2) most
Prevotella strains in the rumen represent species other than
the classical ruminal Prevotella species. Our data also find
analogy with the real-time PCR study of Tajima et al. (2001)
who reported that 16S rRNA gene copies of Prevotella
species far exceeded those of 11 other bacterial species
(including 8 species examined in our study). The abundance
of Prevotella is also indicated by their frequency in 16S
rDNA clone library analysis and by their very high isolation
frequencies (up to 50% of the culturable population) using
classical enumeration methods (see Avgustin et al. 1997 for
numerous references).
The ruminal Prevotella species represent a re-classifica-
tion of a disparate group of strains formerly classified as
Bacteroides ruminicola. The four recognized ruminal
species (the three tested here plus the relatively uncommon
Prevotella albensis) vary in mol% G+C and in the ability to
utilize certain substrates (Avgustin et al. 1997). A hallmark
of all four species is their nutritional versatility with many
individual sugars, amino acids and small peptides able to
support growth. While this nutritional versatility has
obvious selective advantage in the rumen where many
carbohydrate and protein feed components are encountered,
it also renders problematic interpretation of the ecological
roles of individual Prevotella species, both cultured and
yet-uncultured.
171
An additional 5% of the bacterial 16S rRNA gene copies
were distributed among 10 non-Prevotella species classically
regarded as important agents of the rumen fermentation
(Hungate 1966; Krause and Russell 1996). This value may
overestimate the abundance of some species because the use
of primers that amplify all known strains of a target species
in the NCBI database sometimes amplified one or more
strains of a different species; similarly, a species would be
underestimated if a currently unclassified strain representing
that species had one or more nucleotide mismatches that
would prevent amplification by the primer sets used here.
Assuming that these overestimations or underestimations
were minor, the data indicate that less than 10% of the 16S
rRNA gene copies are represented by 13 of the most well-
characterized ruminal species. The relatively modest numer-
ical contribution of classical ruminal species to the ruminal
bacterial population parallels similar observations in another
complex ecosystem, namely, soil. Janssen (2006) has
reported that based on clone library construction, the 9
genera once thought to comprise the bulk of the bacterial
population in soil actually represent <5% of the total
bacterial population and that even the most abundant
subphylum (Alphaproteobacteria) typically represents less
than 19% of the total clones.
The three cellulolytic species examined here (F. succino-
genes, R. flavefaciens and R. albus) represent 2% of the
bacterial 16S rRNA, an additive proportion generally within
the range reported for these three species determined by use
of oligonucleotide probes to 16S rRNA (Krause et al. 1999;
Weimer et al. 1999) and by the fraction of cellulolytic
isolates determined by culture-based enumeration techniques
(Slyter et al. 1970; Varel and Dehority 1989). Such a small
cellulolytic population suggests that a significant fraction of
ruminal cellulose degradation may be due to contributions
from some combination of cellulolytic eucaryotes (fungi or
protozoa) and novel, currently uncultured cellulolytic bacte-
rial species. By the same token, classical starch-fermenting
bacteria other than Prevotella species (viz., S. ruminantium,
S. dextrinosolvens, S. bovis and R. amylophilus) are unlikely
to support a level of starch degradation observed in modern
high-grain dairy rations, again suggesting the presence of
alternative starch-fermenting species that await isolation in
pure culture (including novel prevotellas) and/or the involve-
ment of eucaryotic species.
Real-time PCR is one of several molecular techniques
for quantifying individual taxa in complex environmental
samples. This technique has the advantage of very high
sensitivity and precision relative to membrane hybridization
methods (White et al. 1999). An alternative technique,
fluoroscence in situ hybridization (FISH) permits both
quantitation and direct visualization of specific taxa and
was used to quantify Prevotella and other taxa in
environmental samples lacking large amounts of particulate
172 Appl Microbiol Biotechnol (2007) 75:165174

matter yielding results similar to parallel measurements Appendix

using real-time PCR (Fredricks et al. 2005). FISH was not
applied quantitatively in ruminal samples largely because of Sample calculation of the relative population size
the intense autofluorescence of feed particles and the using real-time PCR
difficulty in counting bacteria adherent to these particles
(Tepsic and Avgustin 2001). The ratio of the amounts of one target sequence to another,
Despite its advantages, real-time PCR, like all molecular first described by Pfaffl (2001), is given by equation A1.
techniques, is subject to certain artifacts. In particular, real-
DCttarget controlsample
time PCR can be biased by differences in amplification Etarget
ratio A1
efficiency of different primers and can also amplify free Ereference DCtreference controlsample
(non-cellular) DNA. The relative quantification method
where E is the efficiency of the PCR reaction and Ct is
used here corrects for differences in amplification efficien-
the difference in the number of cycles to crossing
cy, and a contribution by free DNA is unlikely, owing to the
threshold. In the example here, the target is the sequence
unusually high nuclease activity of ruminal contents (Ruiz
amplified by EubRum2, designed to be species-specific
et al. 2000). Real-time PCR is also subject to bias due to
for E. ruminantium. The reference is the sequence
differences among taxa with respect to target gene copy
amplified by BACset1, specific to the entire domain
number. Thus, translating the proportions of 16S rRNA
bacteria. Calculations are provided here for a single rumen
gene copies for an individual taxon to a true relative
sample designated R1.
population size is only possible if the rRNA gene copy
Two PCR plates were run, one using the primer set
number for the individual taxon and the average rRNA
EubRum1 and the other using BACset1. The actual data for
gene copy number for the entire population are known in
these two plates are shown in Table A1 (EubRum1) and
the respective genomes. However, because the 16S rRNA
Table A2 (BACset1). Note that in both plates, the same
proportions attributable to a single taxon are calculated
DNA sample was used at the same series of concentrations.
based on the total bacterial 16S rRNA of the entire
The varying concentrations for the E. ruminantium GA195
population (whose average copy number is unlikely to vary
DNA was produced by fivefold serial dilutions. The amount
substantially over time), the relative quantification method
of DNA loaded was estimated spectrophotometrically, but
should be useful for comparing changes in specific taxa in
relative quantification assures that any inaccuracies in this
response to different experimental treatments (e.g., dietary
estimate of the DNA amount do not affect the accuracy of
manipulations). Among the three cellulolytic species tested
the final ratio results. Only the mass relationships between
here, 16S rRNA gene copies from both F. succinogenes and
the serial dilutions are critical and then only for the
R. flavefaciens were substantially more abundant than those
construction of the line used for calculation of PCR
of R. albus. Because each of these species appears to have
efficiency.
three rRNA gene copies per genome (M. Morrison,
Next, Ct results for the E. ruminantium GA195 standard
personal communication), it is likely that the 16S rRNA
were plotted against the log of the ng DNA loaded for both
abundance data obtained by real-time PCR can be used for
plates (Tables A1 and A2). The two resultant lines are
direct comparisons of the relative population sizes of these
plotted together in Fig. A1.
three particular species.
The efficiency values for the EubRum2 and BACT1
An important final consideration regards the relationship
primers sets, calculated as the antilog of the negative
between the population of an individual taxon in a given
reciprocal of the line slope, were 1.968 and 1.932,
habitat and its metabolic activity in that habitat. This
consideration is as old as the field of microbial ecology
Table A1 Results from PCR plate 1 using the EubRum2 primer set
itself. While a high population of an individual species does
specific for E. ruminantium
not necessarily indicate a high metabolic activity at the
moment of sampling, an abundant species is likely abundant Template ng DNA Ct (quadruplicate) Ct mean
because its high metabolic activity has provided sufficient
Eubacterium 20 10.06 10.08 10.09 10.14 10.09
energy for reproduction necessary to maintain this numerical
ruminantium 4 12.05 12.09 12.18 12.05 12.09
abundance. It remains the challenge to the microbiologist to GA195 std 0.8 14.61 14.79 14.74 14.80 14.74
quantitatively relate populations of specific taxa to metabolic 0.16 17.07 17.04 17.03 16.96 17.03
activities within particular ecological niches. 0.032 19.34 19.21 19.33 19.31 19.30
0.0064 21.81 21.88 21.92 22.01 21.91
Acknowledgments This work was supported by the Agricultural 0.00128 24.25 24.11 24.26 24.22 24.21
Research Service through the CRIS project 3655-21000-033-00D. We Unknown rumen 20 19.35 19.48 19.33 19.52 19.42
thank R. Zeltwanger for supplying pure cultures of the bacterial sample R1
strains, D.R. Mertens and M.B. Hall for the valuable discussions.
Appl Microbiol Biotechnol (2007) 75:165174 173

Table A2 Results from PCR plate 2 using the domain bacteria- provides mean Ct values for E. ruminantium of 10.09
specific primers BAC338F and BAC805R (Table A1) and 11.40 (Table A2). The mean Ct values for
Template ng DNA Ct (quadruplicate) Ct mean the unknown rumen sample R1 were 19.42 (Table A1) and
11.31 (Table A2). Applying these values to equation A1, the
Eubacterium 20 11.51 11.45 11.36 11.27 11.40 mean Ct was subtracted from the control Ct:
ruminantium 4 13.43 13.59 13.76 13.79 13.64
GA195 std 0.8 16.23 16.29 16.30 16.30 16.28 for plate1 : 10:09  19:42 9:33; and
0.16 18.48 18.65 18.65 18.58 18.59 for plate2 : 11:40  11:31 0:09:
0.032 21.09 21.10 21.10 21.08 21.10
0.0064 23.44 23.63 23.62 23.62 23.58 The efficiency of each reaction, from Fig. A1, was then
0.00128 25.96 25.93 26.04 25.98 25.98 raised to the power of the calculated value, (control Ct)
Unknown rumen 20 11.30 11.25 11.30 11.40 11.31
(unknown sample Ct). Thus, the numerator of the ratio (for
sample R1
the E. ruminantium-specific primer, plate 1) was
(1.968)9.33 =0.001806 and the denominator of the ratio,
(for the domain bacteria primer, plate 2) was (1.932)0.09 =
respectively, near the theoretical maximum of 2.0 (i.e., a
1.0611. The ratio in equation A1 can thus be calculated as
doubling of the target DNA in each PCR cycle).
(0.001806)/(1.0611)=0.00170, or 0.170%. This represents
An essential assumption is made at this point: that the
the ratio of the total E. ruminantium 16S rRNA gene copies
number of species-specific targets per cell is equal to the
to the total bacterial 16S rRNA gene copies for the
number of domain-level primer sites. This allows data from
unknown rumen sample R1.
both plates to be compared to one another. An additional
To verify that the relative quantification method is robust
assumption is that the efficiency of the primers is the same for
across a wide range of target concentrations, repeating the
DNA isolated from pure cultures and for DNA isolated from
above calculation with the most dilute of the standards
rumen samples. Separate experiments have confirmed this.
(nominal 0.00128 ng DNA) yields a ratio of 0.00162 (vs
As per equation A1, relative quantification requires both
0.00170 for the most concentrated standard).
the amplification efficiency and a control Ct for each sample.
Selection of the control Ct can be made at any nominal DNA
concentration, as long as the same concentration is selected
for both plates. In general, there is less variation among References
analytical replicates at higher DNA concentrations. Calcu-
lations using the most concentrated standard (viz., 20 ng) Avgustin G, Wallace RJ, Flint HJ (1997) Phenotypic diversity among
ruminal isolates of Prevotella ruminicola: proposal of Prevotella
brevis, sp. nov., Prevotella bryantii sp. nov., and Prevotella
albensis, sp. nov. and redefinition of Prevotella rumincola. Int J
Syst Bacteriol 47:284288
Bryant MP, Burkey LA (1953) Numbers and some predominant
groups of bacteria in the rumen of cows fed different rations.
J Dairy Sci 36:218224
Dehority BA, Grubb JA (1980) Effect of short-term chilling of rumen
contents on viable bacterial numbers. Appl Environ Microbiol
39:376381
Ferme D, Banjac M, Calsamiglia S, Busquet M, Kamel C, Avgustin G
(2004) The effects of plant extracts on microbial community
structure in a rumen-simulating continuous-culture system as
revealed by molecular profiling. Folia Microbiol (Praha) 49:151
155
Fredricks DN, Fielder TL, Marrazzo JM (2005) Molecular identifica-
tion of bacteria associated with bacterial vaginosis. New Engl J
Med 353:18991911
Hungate RE (1966) The rumen and its microbes. Academic, New
York, NY
Janssen PH (2006) Identifying the dominant soil bacterial taxa in
libraries of 16S rRNA and 16S rRNA genes. Appl Environ
Microbiol 72:17191728
Klieve AV, Hennessy D, Ouwerkerk D, Forster RJ, Mackie RI,
Attwood GT (2003) Establishing populations of Megasphaera
Fig. A1 E. ruminantium standard curves. Coefficients of variation for elsdenii YE34 and Butyrivibrio fibrisolvens YE44 in the rumen
the data points for the bacterial domain primer set BAC338F and of cattle fed high grain diets. J Appl Microbiol 95:621630
BAC805R and the EubRum2 primer sets averaged 0.49% and 0.39%, Krause DO, Russell JB (1996) How many ruminal bacteria are there?
respectively, too small to visualize as error bars on the graph J Dairy Sci 79:14671475
174
Krause DO, Dalrymple BP, Smith WJ, Mackie RI, McSweeney CS
(1999) 16S rDNA sequencing of Ruminococcus albus and
Ruminococcus flavefaciens: design of a signature probe and its
application in adult sheep. Microbiology 145:17971807
Pfaffl MW (2001) A new mathematical model for relative quantifi-
cation in real-time RT-PCR. Nucleic Acids Res 29:20022007
Ruiz TR, Andrews S, Smith GB (2000) Identification and character-
ization of nuclease activities in anaerobic environmental samples.
Can J Microbiol 46:736740
Slyter LL, Oltjen RR, Kern DL, Blank FC (1970) Influence of type
and level of grain and diethylstilbestrol on rumen microbial
populations of steers fed all-concentrate diets. J Anim Sci
31:9961002
Tajima K, Aminov RI, Nagamine T, Matsui H, Nakamura M, Benno Y
(2001) Diet-dependent shifts in the bacterial population of the
rumen revealed with real-time PCR. Appl Environ Microbiol
67:27662774
Tepsic K, Avgustin G (2001) Molecular monitoring of ruminal
prevotellas. Folia Microbiol (Praha) 46:8790
Appl Microbiol Biotechnol (2007) 75:165174
Varel VH, Dehority BA (1989) Ruminal cellulolytic bacteria and
protozoa from bison, cattlebison hybrids, and cattle fed three
alfalfacorn diets. Appl Environ Microbiol 55:148153
Weimer PJ, Waghorn GC, Odt CL, Mertens DR (1999) Effect of diet
on populations of three species of ruminal cellulolytic bacteria in
lactating dairy cows. J Dairy Sci 82:122134
White BA, Cann IKO, Kocherginskaya SA, Aminov RI, Thill LA,
Mackie RI, Onodera R (1999) Molecular analysis of archaea,
bacteria and eucarya communities in the rumen: review. Asian
Australas J Anim Sci 12:129138
Wood J, Scott KP, Avgustin G, Newbold CJ, Flint HJ (1998)
Estimation of the relative abundance of different Bacteroides
and Prevotella ribotypes in gut samples by restriction enzyme
profiling of PCR-amplified 16S rRNA gene sequences. Appl
Environ Microbiol 64:36833689
Yu Y, Lee C, Kim J, Hwang S (2005) Group-specific primer and
probe sets to detect methanogenic communities using quanti-
tative real-time polymerase chain reaction. Biotechnol Bioeng
89:670679