Effect of Environmental Factors and Influence of Rumen and Hindgut Biogeography on Bacterial Communities in Steers

Abstract

Feces from cattle production are considered important sources of bacterial contamination of food and the environment. Little is known about the combined effects of arctic temperatures and fodder tannins on rumen and hindgut bacterial populations. Individual rumen liquor and rectal fecal samples from donor steers fed either alfalfa silage or sainfoin (Onobrychis viciifolia Scop.) silage and water ad libitum were collected weekly on the first three sampling days and fortnightly afterwards. The daily ambient temperatures were registered and averaged to weekly mean temperatures. Steers fed sainfoin silage had lower (P < 0.05) concentrations of branched-chain volatile fatty acids (VFA) than those fed alfalfa silage. All VFA concentrations were higher (P < 0.001) in rumen liquor samples than in fecal samples. The interaction of sample type and diet showed a significant effect (P < 0.05) on the proportions of the bacterial community that were from the phyla Proteobacteria and Verrucomicrobia. Ambient temperature had an indirect effect (P < 0.05) on the phylum Firmicutes, as it affected its proportional balance. The bacterial population diversity in samples appeared to decrease concurrently with the ambient temperature. The phylum Firmicutes explained the first principal component at 64.83 and 42.58% of the total variance in rumen liquor and fecal samples, respectively. The sample type had a larger effect on bacterial communities than diet and temperature. Certain bacterial populations seemed to be better adapted than others to environmentally adverse conditions, such as less access time to nutrients due to higher motility and rate of passage of digesta caused by extreme temperatures, or antimicrobials such as tannins, possibly due to an influence of their biogeographical location within the gut.

Cattle production has been identified as one important source of microbial contamination, including potentially pathogenic bacteria, of food, soil, and water reservoirs (7, 26, 43). Fecal shedding and the spreading of manure on fields are believed to be some of the agents of contamination (24, 45). Better knowledge of the bacterial communities in the rumen and those found in fecal matter may provide insights into the mitigation of fecal bacterial contamination. To date, however, our understanding of the bacterial communities in both the rumen and the hindgut ecosystem remains limited (33).

The bacterial communities residing in the rumen and those found in feces are believed to be similar, as it is generally accepted that there are far fewer differences within individual animals than between individual animals (50). The principles of competitive exclusion have established that microorganisms with similar physiological needs should not coexist in ecosystems where the environment and flow of nutrients fluctuate greatly (24). The rumen and hindgut can be included in this category. Nonetheless, recent work has indicated that bacteria which are phylogenetically closely related can be found in the same environment (18). Indeed, differences which may exist in the genome size and the number of chromosomes may allow that coexistence to take place (18). Nevertheless, to date, very little work has been conducted to estimate similarities in the bacterial populations in the rumen and hindgut ecosystems.

The activity of microbial populations in the rumen and hindgut can be affected by environmental factors, such as antimicrobial compounds in plants, e.g., tannins (19, 32, 46), and temperature, e.g., cold conditions (48). However, previous data have shown that bacteria can develop tolerance to tannins (5, 36, 37) and even degrade them (13). Likewise, previous research has shown that bacteria remain active even when the environmental temperature is sharply reduced (14, 20), as would occur to fecal bacteria during excretion in a cold environment. Past work has indicated that a combination of very low temperature (−20 and 25°C and −20 to 25°C, respectively) with high pressure (50 to 450 MPa and 100 to 350 MPa, respectively) may cause a synergistic inhibition of bacterial growth (34, 42). No information has been reported on the possible combined effect of arctic temperatures and fodder tannins on the bacterial populations in the rumen and hindgut of cattle.

Recent molecular techniques, such as terminal restriction fragment length polymorphism (T-RFLP), have provided microbiologists with tools to estimate the community composition of microbial ecosystems (31). Several studies have analyzed the diversity of bacterial communities using T-RFLP (9, 27, 28). Analysis by T-RFLP may help elucidate the differences between bacterial populations in the rumen and hindgut.

The objective of the present work was to estimate the nature and diversity of the bacterial communities in the rumen liquor and feces of beef steers subjected to different environmental factors, such as a tanniferous diet and the ambient temperature, by means of T-RFLP analyses.

MATERIALS AND METHODS

Animals and sampling.

Forty beef cattle steers (initial weight: 301.3 ± 3.77 kg [mean ± standard deviation]) were divided into groups of 10 according to live weight and allocated into four pens. Two groups received a diet of hay silage, and the remaining two groups a diet of sainfoin (Onobrychis viciifolia Scop.) silage. The nutrient composition of the diets is shown in Table 1. All groups were given the diets and water ad libitum. The treatments were imposed on the steers for 10 weeks (10 December 2007 to 20 February 2008). Rumen liquor was aspirated using an oral probe (10). Samples of rumen liquor (ca. 50 ml) were removed from individual animals on the 10th and 17th of December 2007, the 7th of January 2008, and fortnightly afterwards. The first 200 ml of aspirated rumen liquor was discarded, and the subsequent 50 ml of rumen liquor was collected. Fecal grab samples (250 to 500 g) were taken directly from the rectum of individual animals on the same days as the rumen liquor samples. Individual fecal and rumen liquor samples were divided by sample type and separately pooled per pen, and an aliquot was taken (rumen liquor, ca. 20 ml, and feces, ca. 100 g). Fecal and rumen liquor samples were kept in a −20°C freezer until further use. On the day of analysis, individual fecal and rumen liquor samples were subsampled, yielding four duplicates per pen and eight per sampling day.

TABLE 1.

Analyzed nutrient composition, gross energy, and condensed tannin content of experimental diets

Dietary componenta	Amt (g/kg of DM, unless otherwise indicated) in dietb
	SainSil1	AlfSil2	SainSil3	AlfSil4
Condensed tannin	11.4	0.0	11.8	0.0
DM	551.1	625.6	554.7	624.9
CP	157.1	190.9	156.0	191.9
ADF	345.5	322.7	343.0	323.1
NDF	427.4	412.3	422.4	418.8
Ash	82.9	93.6	82.2	94.9
GE (Kcal/g)	4.4	4.4	4.4	4.4

^aDM, dry matter; CP, crude protein; ADF, acid detergent fiber; NDF, neutral detergent fiber; GE, gross energy.

^bSainSil1, sainfoin silage in pen 1; AlfSil2, alfalfa silage in pen 2; SainSil3, sainfoin silage in pen 3; AlfSil4, alfalfa silage in pen 4.

The experimental protocol was reviewed and approved by the University of Manitoba Animal Care Committee. Beef steers were cared for according to the guidelines of the Canadian Council on Animal Care (CCAC) of 1993.

Ambient temperature measurement.

Ambient temperature (range, −4.6 to −30.0°C) was registered daily by the weather station located at the Glenlea Research Station in Glenlea, Manitoba, 20 km south of Winnipeg City (49°39′0″N, 97°7′0″W). The annual precipitation in the area was ca. 535 mm, and of this, approximately 30% was as snow. The average sampling day temperature was obtained by using the daily temperatures to calculate the weekly mean temperatures for those weeks during which samples were collected.

Analyses of condensed tannin content, volatile fatty acids, ammonia, and lactic acid.

The condensed tannin content of alfalfa and sainfoin silages was analyzed as previously described (51). For analysis of volatile fatty acids (VFA), frozen rumen liquor samples were thawed at room temperature and 1 ml of 25% metaphosphoric acid solution was added to 5 ml of rumen liquor. The tubes were vortexed and placed in a −20°C freezer for 17 h. The thawed samples were centrifuged for 10 min at 1,900 × g. The concentrations of VFA and lactic acid were determined by using a gas chromatograph (Clarus 500; PerkinElmer Life and Analytical Sciences, Inc.) fitted with a flame ionization detector (FID). A Silane-treated empty glass column (TightSpec model 2-1179-U with a length of 2.0 m, outer diameter of 6.35 mm, and inner diameter of 2 mm; Supelco, Oakville, Ontario, Canada) packed with Carbopack B-DA (60 to 200°C range, 80/120 mesh, 4% Carbowax 20 M phase, Supelco catalogue no. 11889) was used for the analysis. Approximately 2 ml of supernatant was decanted, transferred into gas chromatograph vials, and placed in the built-in autosampler device for analysis. The injector and detector temperatures were set at 200°C, with an initial column temperature of 175°C, held for 20 min, and final column temperature set at 200°C, held for 0 min. The total run time was 21.25 min with no thermal equilibration period.

The ammonia nitrogen concentration of rumen liquor samples was determined using the method described by Novozamsky et al. (40). Absorbance was read at 630 nm on a Pharmacia Biotech Ultraspec 2000 UV/visible light spectrophotometer (Biochrom, Cambridge, United Kingdom).

DNA analysis.

Two aliquots (ca. 150 mg) were taken from each pool of rumen liquor and fecal subsamples. The DNA was extracted using a ZR fecal DNA kit (Zymo Research Corp., Orange, CA) according to the manufacturer's recommended protocol. On a few subsamples, extraction of DNA with this kit failed for unknown reasons. The DNA extraction from these subsamples was repeated using a QIAamp DNA stool mini kit (Qiagen, Inc., Mississauga, Ontario, Canada). Extraction was conducted using 200-mg aliquots from the subsamples following the manufacturer's guidelines. Sequences of 16S rRNA genes were amplified (25) using primer blue fluorescent forward 27, 5′-AGAGTTTGATCMTGGCTCAG (WellRED D4dye; Sigma-Proligo Co., St. Louis, MO), to allow detection of the PCR fragments by capillary electrophoresis and reverse primer 1100, 5′-GGGTTGCGCTCGTTG, in a mixture (25). Amplicon production was carried out using 35 cycles of 94°C for 1 min followed by 61°C for 1 min and 72°C for 5 min.

PCR amplicons were digested with the restriction enzyme HhaI (T-RF) according to the manufacturer's instructions, with some modifications as previously described (3), in order to produce terminal restriction fragments. The digested products (20 μl) were precipitated with 0.25 μl of 2 mg/ml glycogen and 2 μl each of 3 M sodium acetate (NaOAc, pH 5.2) and cold 95% ethyl alcohol (EtOH). The pellets were washed with cold 70% EtOH twice and allowed to air dry on Kimwipes (SPI Supplies/Canada, Toronto, Ontario, Canada) for 30 to 45 min, resuspended in 20 μl of sample loading solution (SLS) (Beckman Coulter Canada, Inc., Mississauga, Ontario, Canada), and stored in a −20°C freezer until further use.

Terminal restriction fragment length polymorphism (T-RFLP) data analysis.

The lengths and peaks of blue-fluorescence-labeled T-RF were estimated with CEQ 8800 genetic analysis system software (Beckman Coulter Canada, Inc., Mississauga, Ontario, Canada). A protocol recommended by the manufacturer was used for the analysis, with modifications (3). The T-RF with peak heights smaller than 50 fluorescence units were excluded from the analysis. The T-RFLP analyses of individual rumen liquor and fecal samples produced electropherograms with peaks of different sizes. Each peak represented an operational taxonomic unit and was identified by its fragment size.

Bioinformatic analysis of T-RFLP data.

The fragment size and peak area data obtained from the T-RFLP analyses by the CEQ 8800 genetic analysis system software were analyzed as previously described (3).

Statistical analysis.

The statistical analyses of the concentrations of volatile fatty acids, lactic acid, and ammonia, as well as the proportions of the sequences in relation to the total populations of the bacterial communities, were conducted with GenStat release 10.1, 2007 (Lawes Agricultural Trust, Harpenden, United Kingdom), using linear mixed models (REML) in which the fixed effects were sample type, diet, and ambient temperature (with each value treated as a distinct factor level) and the random effects were week and pen. Simple richness (S) and Margalef richness indices, as well as Shannon-Wiener and Simpson diversity indices, were calculated using the diversity analyses in GenStat release 10.1.

As some data were not normally distributed, they were transformed using a natural logarithm. For these data, the levels of significance and standard errors of the means were calculated using the transformed data. The transformed data have been described where necessary, including identification of the type of transformation and statistical significance calculated using the transformed data. The significance of differences between individual samples was estimated using Student's t test based on the means and standard errors estimated using REML. The levels of significance were tested using the Wald test (41).

Principal component analysis (PCA) based on a correlation matrix was carried out using the proportions of the phyla detected in the total bacterial communities in the rumen liquor and fecal samples to reveal patterns within the data set. The PCA was carried out using GenStat release 10.1. Eigenvalues, percentage of variation, latent vectors, PCA scores, residuals, and significance tests were calculated. The first and second principal components were plotted against the ambient temperature to determine sample type differences and the effect of the temperature on the bacterial communities found in the rumen liquor and fecal samples recovered from steers fed on different diets, namely, sainfoin silage versus alfalfa silage.

A three-dimensional scatter plot was created using Jump 7.01 (SAS Institute, Inc., Cary, NC). Colored contours in the scatter plot were created, and colors were enhanced to highlight the data grouping. Multivariate methods in Jump 7.01 were used to analyze clustering and generate a phylogenetic tree. The clustering analysis was calculated using Ward's hierarchical method. The data were standardized and ordered by sampling week. The scale of the dendrogram is by even spacing.

RESULTS

The sainfoin silage contained a condensed tannin concentration of ca. 11 g/kg of dry matter (Table 1), and as expected, no condensed tannins were detected in alfalfa silage. Other than condensed tannin content, only minor differences were observed between the measured components in sainfoin and alfalfa silage diets (Table 1).

Volatile fatty acids and other rumen fermentation products: effects of diet and sample type on the production of volatile fatty acids.

There were lower concentrations (P < 0.05) of isobutyric acid in both the rumen liquor and feces and of isovaleric acid in the feces of steers fed on sainfoin silage than in samples from steers consuming alfalfa silage (Table 2). The concentrations of ammonia in rumen liquor and feces were lower in steers fed sainfoin silage (P = 0.052) than in steers fed alfalfa silage (Table 2).

TABLE 2.

Effects of sample type, diet, and ambient temperature on the concentrations of volatile fatty acids, total volatile fatty acids, lactic acid, and ammonia in the fecal and rumen liquor samples of steers fed sainfoin silage or alfalfa silage

Rumen fermentation product	Concn (mmol/liter) in samples from steers fed indicated dieta					Level of significanceb of:
	Rumen liquor (n = 12)		Feces (n = 12)		SEb
	Alfalfa silage	Sainfoin silage	Alfalfa silage	Sainfoin silage		Sample type	Diet	Ambient temp (°C)	Sample type × diet
Acetic acid	21.76	19.89	4.53	3.78	0.119	***	NS	NS	NS
Propionic acid	4.53	4.10	0.83	0.68	0.117	***	NS	NS	NS
Isobutyric acid	0.39	0.25	0.12	0.10	0.120	***	*	NS	NS
Butyric acid	2.59	2.14	0.24	0.19	0.120	***	NS	NS	NS
Isovaleric acid	0.21	0.12	0.04	0.04	0.152	***	*	NS	NS
Valeric acid	0.28	0.20	0.06	0.05	0.253	***	NS	NS	NS
Total VFA	29.67	26.58	5.81	4.81	0.119	***	NS	NS	NS
Lactic acid	0.02	0.03	0.04	0.06	0.585	NS	NS	NS	NS
Ammonia	1.01	0.64	2.86	2.44	0.131	***	0.052	NS	NS

^aData presented are mean values for fecal and rumen liquor samples from two pens at six sampling times. As data were not normally distributed, they were transformed using natural logarithms.

^bLevels of significance and standard errors of the means were calculated using the transformed data. Statistical significance: *, P < 0.05; **, P < 0.01; ***, P < 0.001; NS, not significant.

The estimated concentrations of all volatile fatty acids were very significantly (P < 0.001) different in fecal and rumen liquor samples, with the concentrations in rumen liquor being substantially higher than those in feces (Table 2). The only exception was lactic acid, whose concentrations were similar in both the feces and the rumen liquor. The concentrations of ammonia were significantly higher in rumen liquor than in feces (Table 2).

Bacterial communities. (i) Effects of diet, ambient temperature, and sample type on bacterial communities.

Although there was no evidence of a direct effect of diet on the proportions of the taxons in the bacterial communities estimated in both the feces and the rumen liquor (Table 3), the interaction of sample type versus diet did alter (P < 0.05) the bacterial community proportions found for the phyla Deinococcus-Thermus, Proteobacteria, and Verrucomicrobia (Table 3). These observed changes in proportion may be due to the difference in tannin content between the diets (Table 1). Similarly, there were differences found in the interaction of the sample type and diet on the bacterial community proportions of the phyla Bacteriodetes, Fusobacteria, and Tenericutes (P = 0.061, P = 0.065, and P = 0.063, respectively) (Table 3). Visual inspection of the phylogenetic tree generated by the cluster analysis showed that diet affected the clustering of bacterial population data (Fig. 1). Indeed, with a clustering distance of more than 80%, the bacterial population data aggregated mainly into one large cluster for steers fed sainfoin silage and one smaller cluster for those fed alfalfa silage (Fig. 1). However, other clusters were unaffected by diet, with the exception of a small cluster associated with fecal samples from steers fed alfalfa silage (Fig. 1).

TABLE 3.

Effects of sample type, diet, and ambient temperature on the bacterial communities in the fecal and rumen liquor samples of steers fed sainfoin silage or alfalfa silage

Taxon	Proportion of sequences in total bacterial community (×100)a					Level of significanceb of:
	Rumen liquor (n = 12)		Feces (n = 12)		Avg SE
	Alfalfa silage	Sainfoin silage	Alfalfa silage	Sainfoin silage		Sample type	Diet	Ambient temp (°C)	Sample type × diet
Phylum Actinobacteria	1.50	1.44	1.08	1.40	0.118	NS	NS	NS	NS
Phylum Bacteroidetes	0.37A	0.14B	0.40A	0.48A	0.286	0.053	NS	NS	0.061
Phylum Deferribacteres	0.01	0.01	0.01	0.01	0.042	NS	NS	NS	NS
Phylum Deinococcus-Thermus	0.02B	0.01A	0.01A	0.01A	0.095	*	NS	NS	*
Phylum Firmicutes	86.93	87.80	93.23	90.47	0.038	NS	NS	*	NS
Class Bacilli
Order Bacillales	45.84	44.04	26.98	30.42	0.133	**	NS	NS	NS
Order Lactobacillales	46.76	43.60	67.70	61.26	0.158	0.057	NS	0.088	NS
Class Clostridia
Order Clostridiales	98.01	98.01	98.01	98.01	0.004	NS	NS	NS	NS
Phylum Fusobacteria	0.06	0.05	0.04	0.07	0.190	NS	NS	NS	0.065
Phylum Proteobacteria	4.64B	3.30A	3.11A	4.37B	0.172	NS	NS	NS	*
Phylum Tenericutes	0.15	0.18A	0.15	0.12B	0.092	*	NS	NS	0.063
Phylum Verrucomicrobia	0.83B	0.10	0.08A	0.46B	0.756	NS	NS	NS	*

^aData presented are the mean values for fecal and rumen liquor samples from two pens at six sampling periods. As data were not normally distributed, they were transformed using natural logarithms. Values with different superscripts are statistically different (P < 0.05).

^bLevels of significance and standard errors of the means were calculated using the transformed data. Statistical significance: *, P < 0.05; **, P < 0.01; ***, P < 0.001; NS, not significant.

FIG. 1.

Phylogenetic tree based on the proportions of the phyla detected in the bacterial communities in the rumen liquor (RumLiq) and fecal (Faec) samples of steers fed either alfalfa silage or sainfoin silage, by T-RFLP analysis. Relationships were based on data from 16S rRNA gene sequences. The paired fractions on the y axis indicate rumen liquor sampling observations (upper fractions) and fecal sampling observations (lower fractions). The denominators in these values are the proportions of the total number of observations in the cluster; numerators are the total numbers of the predominant sampling observations.

The proportions of bacterial communities belonging to the phyla Deinococcus-Thermus and Tenericutes were larger in the rumen liquor samples than in the fecal samples (Table 3). Although the proportion of bacterial communities of phylum Firmicutes remained comparatively unaltered in both the feces and the rumen liquor (Table 3), closer observation of the order Bacillales, under class Bacilli, indicated that there was a significant difference (P < 0.01) between sample types. The proportion of the Bacillales was larger in the rumen liquor than in feces (Table 3). Conversely, there was a greater proportion (P = 0.053) of the bacterial community belonging to the phylum Bacteriodetes in feces than was estimated in the rumen liquor (Table 3).

Arctic conditions prevailed throughout the experiment (Fig. 2), and thus, sampling was conducted indoors. To avoid further handling stress on the animals, it was decided not to measure the temperature in the rumen and intestine of steers, and only the ambient temperature was registered and analyzed. In general, there was little direct effect of the ambient temperature on the bacterial community proportions estimated in either the feces or the rumen liquor. Nonetheless, the ambient temperature did affect (P < 0.05) the proportional balance of phylum Firmicutes in the bacterial communities (Fig. 3) in both types of samples, although there was no evidence for a similar effect for most orders under this phylum (Table 3). There were, however, temperature-dependent differences in the proportions of the order Lactobacillales, under class Bacilli, in the bacterial communities (P = 0.088) (Table 3 and Fig. 3).

FIG. 2.

Fluctuation of the ambient temperature (°C) that beef steers experienced during the present experiment.

FIG. 3.

Effect of ambient temperature on the proportions of phylum Firmicutes and its order Lactobacillales in bacterial communities in fecal and rumen liquor samples from steers fed either alfalfa silage or sainfoin silage. Histograms show results for fecal samples from steers fed alfalfa silage (A), rumen liquor samples from steers fed alfalfa silage (B), fecal samples from steers fed sainfoin silage (C), and rumen liquor samples from steers fed sainfoin silage (D). Firmicutes is a bacterial phylum; Lactobacillales is a bacterial order under class Bacilli belonging to phylum Firmicutes.

The three-dimensional scatter plot of the first and second principal components showed that the ambient temperature had either a positive or inverse relation with the bacterial populations detected in both the fecal and rumen liquor samples, a relationship that was more obvious for feces (Fig. 4).

FIG. 4.

Three-dimensional scatter plot of the first and second principal components calculated on the natural log-transformed proportions of the taxons in the bacterial communities detected in the rumen liquor and fecal samples of steers fed on either sainfoin silage or alfalfa silage. Principal component data were plotted against ambient temperature. PC1, principal component 1 (x axis); PC2, principal component 2 (y axis); z axis, ambient temperature. Symbols denote results for groups in pens as follows: square, pen 1; circle, pen 2; plus sign, pen 3; diamond, pen 4. Colors of symbols represent weeks during the sampling: yellow, week 1; blue, week 2; brown, week 3; pink, week 4; black, week 5; orange, week 6. Steers in pens 1 and 3 were fed sainfoin silage, while steers in pens 2 and 4 were fed alfalfa silage. Blue contour color denotes rumen liquor samples (A), and red contour color denotes fecal samples (B).

Cluster analysis revealed an effect of the biogeography of the gastrointestinal tract of the experimental animals, as it showed some difference in the aggregation of bacterial population data around the two sample types. Indeed, while the clusters for fecal samples moved toward the top end of the phylogenetic tree, those for rumen liquor moved toward the bottom end of the tree (Fig. 1). This aggregation was also evident in the three-dimensional scatter plot of the first and second components of the bacterial population, as fecal data and rumen liquor data formed clusters (Fig. 4; fecal sample data are contoured in red, and rumen liquor sample data are contoured in blue).

(ii) Richness and diversity of the bacterial populations in the samples.

(a) Richness. The Margalef richness index indicated a steady loss in richness of the bacterial population of the phyla concurrently with the temperature reaching lower values, although there was a clear struggle by the bacterial population to recover from damage at two different temperature levels (Table 4). The richness did not recover from the initial loss, independently of the temperature status, although further decreases in richness were too small to be significant (Table 4). The Margalef index consistently showed a significant decrease in the richness of the bacterial orders of phyla Firmicutes detected in all animal samples as the experiment progressed in time. This occurred in a manner that was independent of temperature (Table 5). There was little effect of sample type and diet on the Margelef richness scores (Tables 4 and 5).

TABLE 4.

Effects of sample type, ambient temperature, and diet on richness and diversity indices of bacterial phyla in the fecal and rumen liquor samples of steers fed either alfalfa silage or sainfoin silage

Diet and ambient temp (°C)	Index value for sample typea
	Rumen liquor				Feces
	Richness		Diversity		Richness		Diversity
	S	Margalef	Shannon	Simpson	S	Margalef	Shannon	Simpson
Alfalfa silage
−11.1	9	5.78	0.69	0.52C	9	5.78	0.65	0.51C
−16.5	8	6.38BY	0.87	0.72B	8	6.37BY	0.71	0.68B
−18.7	9	4.97	0.58	0.41	7	3.73	0.53	0.40
−19.1	9	4.47	0.53	0.34D	9	4.47	0.52	0.34D
−19.3	9	11.62AX	0.85	1.07AX	9	11.58AX	0.82	1.05AX
−20.2	9	4.11	0.48	0.29D	8	3.60	0.48	0.29D
Total	9	2.43C	0.64	0.37	10	2.73C	0.60	0.37D
SE	0.000	0.142	0.069	0.051	0.833	0.330	0.052	0.048
Sainfoin silage
−11.1	7	4.34C	0.79	0.54	8	5.05C	0.67	0.51
−16.5	9	7.29B	0.72	0.68B	9	7.29B	0.78	0.70BY
−18.7	8	4.35C	0.53	0.40	9	4.97C	0.59	0.41C
−19.1	9	4.47C	0.49	0.34CY	9	4.47C	0.52	0.34CZ
−19.3	8	10.14A	0.82	1.05A	8	10.13A	0.91	1.09A
−20.2	9	4.11C	0.47	0.29CY	9	4.11C	0.46	0.29CZ
Total	9	2.43DZ	0.63	0.37C	9	2.43DZ	0.62	0.37C
SE	0.000	0.142	0.075	0.051	0.000	0.142	0.065	0.050

^aS, number of phyla present in the community. The richness and diversity indices shown were calculated from the proportions of the phyla Actinobacteria, Bacteroidetes, Deferribacteres, Deinococcus-Thermus, Firmicutes, Fusobacteria, Proteobacteria, Tenericutes, and Verrucomicrobia in the total bacterial populations of the samples. Values with different superscripts are statistically different (P < 0.05). ABC, the lack of a common letter for values in the same column denotes significant differences within the diet (P < 0.05); XYZ, the lack of a common letter for values in the same column denotes significant differences between the diets (P < 0.05). Index values do not represent numbers of operational taxonomic units; they are abstract values. Data presented are mean values for fecal and rumen liquor samples from two pens at six sampling times.

TABLE 5.

Effects of sample type, ambient temperature, and diet on richness and diversity indices of the orders of the Firmicutes phylum in the fecal and rumen liquor samples of steers fed either alfalfa silage or sainfoin silage

Ambient temp (°C)	Index value for sample typea
	Rumen liquor				Feces
	Richness		Diversity		Richness		Diversity
	S	Margalef	Shannon	Simpson	S	Margalef	Shannon	Simpson
Alfalfa silage
−11.1	5	2.24C	1.36	0.82	5	2.24C	1.34	0.81
−16.5	5	2.50BX	1.46AX	0.92AX	5	2.49BX	1.45	0.92
−18.7	5	2.06	1.24	0.72	5	2.06D	1.25	0.73
−19.1	5	1.93	1.15B	0.65	5	1.93	1.15	0.65
−19.3	5	2.89AX	1.54AX	1.04AX	5	2.89AX	1.50	1.03
−20.2	5	1.82D	1.08B	0.59B	5	1.82	1.10	0.59
Total	5	1.09E	1.30	0.67	5	1.09E	1.29	0.67
SE	5	0.039	0.081	0.049		0.039	0.078	0.048
Sainfoin silage
−11.1	5	2.23CY	1.33	0.81	5	2.24CY	1.34	0.81
−16.5	5	2.49B	1.45A	0.92A	5	2.49B	1.47	0.93
−18.7	5	2.06Y	1.24	0.72	5	2.06DY	1.24	0.72
−19.1	5	1.93Y	1.16	0.65Y	5	1.93Y	1.15	0.65
−19.3	5	2.89A	1.56A	1.04A	5	2.89A	1.53	1.03
−20.2	5	1.82EY	1.08BY	0.59BY	5	1.82Y	1.10	0.59
Total	5	1.09FY	1.30	0.67	5	1.09EY	1.29	0.67
SE		0.039	0.081	0.049		0.039	0.082	0.049

^aS, number of orders present in the community. The richness and diversity indices shown above were calculated from the proportions of orders Bacillales and Lactobacillales in class Bacilli and orders Clostridiales and Erysipelotrichales in class Clostridia in the total bacterial populations in the samples. Values with different superscripts are statistically different (P < 0.05). ABC, the lack of a common letter for values in the same column denotes significance within the diet (P < 0.05); XYZ, the lack of a common letter for values in the same column denotes significance between the diets (P < 0.05). Index values do not represent numbers of operational taxonomic units; they are abstract values. Data presented are mean values for fecal and rumen liquor samples from two pens at six sampling times.

(b) Diversity in phyla.

The Shannon-Weiner diversity index did not produce any evident differences between the bacterial populations in the rumen liquor of steers fed either diet (Table 4). Differences were observed when the Shannon-Weiner index was calculated across time, with the diversity decreasing in the second part of the experiment (Table 4), but these differences were not significant. This trend was similar to that of the ambient temperature effect (Fig. 2). The Simpson diversity index showed little difference between the bacterial populations detected in the rumen liquor of steers fed either diet. However, diversity suffered sharp drops through time in the rumen liquor from steers fed both diets. Again, diversity followed a trend similar to that exhibited by the ambient temperature effect (Table 4 and Fig. 2).

The Shannon-Weiner diversity index seemed to consistently decrease across time, apparently independently of temperature (Table 4). For both diets, the lowest nonsignificant Shannon-Weiner diversity index values were observed when the temperature was the lowest (Table 4). The Simpson diversity index reported significant differences within sampling periods, where diversity was higher in the first stages of the experiment before showing a sharp drop (Table 4). This followed the same trend as temperature, which dropped consistently in the late stages of the experiment (Table 4).

(c) Diversity in Firmicutes orders.

Diversity calculated by the Shannon index detected a decrease in the bacterial population diversity as time passed and the temperature dropped further (Table 5). The Simpson diversity index was lower on the last sampling day for the samples from steers across both diets, apparently coincidentally with a drop in ambient temperature (Table 5).

The Shannon-Wiener index detected decreases in the diversity of bacterial populations through time and under different temperatures in the feces of steers fed both diets, but they were not significant (Table 5). The Simpson diversity index estimated that the diversity of bacterial populations in feces from all steers decreased as the experiment neared the end and following the trend of the ambient temperature, which decreased further, but these decreases were not significant (Table 5).

(iii) Principal component analysis.

(a) Principal component analysis of phyla. Table 6 presents the results of the principal component analysis of the data for phyla in the bacterial populations detected in the rumen liquor of steers. The first principal component explained 64.83% of the total variance. The principal component is explained by the phylum Firmicutes (Table 6). Deinococcus-Thermus represented the second principal component, accounting for 16.71% of the total variance. The results of the principal component analysis showed that the first three principal components accounted for 91.78% of the total variance in phyla in the bacterial population in the rumen liquor (Table 6). The results of principal component analysis of the data for phyla detected in the bacterial populations in the feces of the steers are shown in Table 7. Similar to the results of the principal component analysis of rumen liquor data, the first principal component was accounted for by the phylum Firmicutes, accounting for 42.58% of the total variance (Table 7). This time, Tenericutes was the second principal component, which explained 14.66% of the total variance. All three principal components accounted for 69.44% of the total variance (Table 7).

TABLE 6.

Eigenvector values for principal components^a using the proportions of the bacterial phyla in rumen liquor samples

Phylum	PC1	PC2	PC3
Actinobacteria	−0.42153	−0.00356	0.12775
Bacteroidetes	−0.37059	0.28497	−0.27177
Deferribacteres	0.00000	0.00000	0.00000
Deinococcus-Thermus	0.06491	0.76571	−0.30451
Firmicutes	0.43487	−0.02698	0.06000
Fusobacteria	0.20860	0.51060	0.63318
Proteobacteria	−0.40522	0.23902	0.17577
Tenericutes	−0.32342	−0.10539	0.59050
Verrucomicrobia	−0.42348	−0.05288	−0.18166
Eigenvalue	5.186	1.336	0.819
% of total variation	64.83	16.71	10.24

^aPC1, PC2, and PC3, principal components 1, 2, and 3.

TABLE 7.

Eigenvector values for principal components using the proportions of the phyla in the bacterial communities in fecal samples

Phylum	PC1	PC2	PC3
Actinobacteria	−0.37830	0.11522	0.35551
Bacteroidetes	−0.30331	−0.01776	0.39181
Deferribacteres	0.13041	0.18839	0.35568
Deinococcus-Thermus	−0.25444	−0.57015	0.41173
Firmicutes	0.47462	0.06485	0.27637
Fusobacteria	−0.31499	0.44209	0.00911
Proteobacteria	−0.46881	0.03569	0.05083
Tenericutes	−0.09286	0.63526	0.03667
Verrucomicrobia	−0.36090	−0.14619	−0.58005
Eigenvalue	3.832	1.319	1.098
% of total variation	42.58	14.66	12.20

(b) Principal component analysis of the Firmicutes orders.

The results for the principal component analyses of the total bacterial populations detected in both the rumen liquor and feces for orders of the phylum Firmicutes are shown in Table 8. For both sample types, the order Clostridiales represented the first principal component, which explained 82.66 and 71.63% of the total variance in rumen liquor and feces, respectively (Table 8). In contrast, the order Bacillales accounted for the second principal component, which explained 17.34 and 28.37% of the variance for the rumen liquor and feces, respectively (Table 8). The first two principal components explained 100% of the total variance in the rumen liquor and feces.

TABLE 8.

Eigenvector values for principal components using the proportions of the orders of the phylum Firmicutes in fecal and rumen liquor samples of steers fed either sainfoin silage or alfalfa silage

Sample type and Firmicutes class and order	PC1	PC2	PC3
Rumen liquor
Class Bacilli
Bacillales	0.66531	0.23949	0.24125
Lactobacillales	−0.66531	−0.23949	0.24125
Class Clostridia
Clostridiales	0.33869	−0.94090	0.00000
Erysipelotrichales	0.00000	0.00000	0.94000
Eigenvalue	2.1489	0.8511	0.0000
% of the total variation	71.63	28.37	0.00
Feces
Class Bacilli
Bacillales	0.614447	0.34989	0.70711
Lactobacillales	−0.61447	−0.34989	0.70711
Class Clostridia
Clostridiales	0.49482	−0.86900	−0.00001
Erysipelotrichales	0.00000	0.00000	0.00000
Eigenvalue	2.4798	0.5202	0.0000
% of the total variation	82.66	17.34	0.00

Clustering of the data at different sampling dates for both rumen liquor and feces from steers fed either alfalfa silage or sainfoin silage suggested that the proportions of phyla and orders in the bacterial populations were similar across all samples (Fig. 4).

DISCUSSION

Together with a few other studies (for further information, see reference 50), the current work is one of the first serious attempts to estimate and compare the bacterial communities in the rumen liquor and feces of steers. This is also the first work that includes as combined environmental factors a diet with phytochemical components, such as tannins, and extreme arctic conditions. Previous work has reported that external factors such as diet, especially diets consisting of plants rich in tannins (19, 32, 46), can have a deleterious effect on the bacterial population in the rumen of animals. Similarly, it has previously been reported that cold temperatures can negatively affect the activity of microbial populations of the rumen (48), possibly due to fluctuations in feed intake or digesta passage rate. In the present study, the effect of a tanniferous diet with sainfoin silage when the animals remained under arctic conditions was investigated. In addition, the analyses were carried out with T-RFLP of 16S rRNA as the fingerprinting technique because it has been reported in the literature to be a high-throughput, robust, reproducible method for the characterization of microbial populations in intestinal samples (27).

Use of the T-RFLP technique to estimate the diversity of bacterial populations.

In the present work, the use of the T-RFLP technique was a satisfactory means to estimate the diversity of the bacterial population in the samples. The T-RFLP technique is a powerful and yet rapid method for the analysis of bacterial communities (28, 31). Unfortunately, the resolution of this technique and, therefore, the quality of the data obtained, becomes substantially lower as the taxonomical classification approaches the species rank. For the present work, it was decided that a compromise between quality and resolution had to be made. The lowest taxonomical rank for our analyses was order. Therefore, it was not possible to investigate species shifts that may have occurred in the bacterial populations as a result of changes in dietary or environmental conditions.

Effects of diet.

The T-RFLP analysis showed that the concentrations of VFA, except for the branched-chain VFA in the feces and rumen liquor of steers fed sainfoin silage, were no different than those found in the samples from steers fed alfalfa silage. VFA are produced mainly during the digestion of carbohydrates and fermentation of sugars but also during the digestion of lipids and the breakdown of amino acids (2). Although not well established, it is believed that only a few “generalist” bacterial species are capable of producing more than one VFA (47). Instead, it seems that a great number of “specialist” bacterial species are capable of producing a single VFA (47). As the rumen is a very complex and dynamic ecosystem (16), it seems that a symbiotic association through cross-feeding strongly contributes to its maintenance and survivability (44, 49). The REML results showed no major changes in the proportions of bacterial communities (Table 3), but the observed decrease in branched-chain fatty acids (Table 2) in samples from steers fed sainfoin silage compared to the amounts in samples from steers fed alfalfa silage may suggest a shift of bacterial species within communities. This may imply that certain bacterial species were affected by the inclusion of tanniferous sainfoin in the diet, while other bacterial species, more tolerant to the tannins in the sainfoin, took the place of the diminishing ones. Jones et al. (19) determined that the condensed tannins present in the leaves of sainfoin (Onobrychis viciifolia Scop.) negatively affected the growth and protease activity of Butyrivibrio fibrisolvens A38 and Streptococcus bovis 45S but not of Prevotella ruminicola B₁4 or Ruminobacter amylophilus WP225.

Conversely, cluster analyses of the present sample data seemed to at least partially confirm a recovery of tannin-tolerant bacteria (Fig. 4). Indeed, the two major clusters aggregated to the greatest extent around the results for the samples from the steers that consumed sainfoin silage. This may suggest that the surviving bacterial populations not only recovered from the initial tannin effect but may well have thrived in the tanniferous environment (Table 1). Previous work has reported that certain bacteria can partially degrade tannins (for further information, see reference 13). Another possible explanation may be that, although the bacterial phyla could be similar in the rumen and the hindgut, the bacterial species may differ between these two locations in the gastrointestinal tract. This may explain why the concentrations of branched-chain fatty acids differed between these two sites, as different species may fulfill different tasks and, thus, occupy separate ecological niches.

Effects of ambient temperature.

It has long been considered that temperature is the most influential factor within any ecosystem (15). The microbial ecosystem of the rumen is not an exception. However, when ruminants are subject to cold temperatures, the rumen environment remains thermally stable (13). Thermoregulation of the animal maintains the temperature of the internal organ systems (e.g., the brain, heart, lungs, and digestive tract). The extremities (feet, tail, ears, etc.) are usually affected by ambient temperature first, and only in extreme situations are the organ systems affected (12). In the present work, the REML analysis found little effect of the ambient temperature on the proportions of the taxons in the bacterial communities in either rumen liquor or fecal samples. One exception was the phylum Firmicutes, which showed a significant effect of the environmental temperature on the bacterial communities, with emphasis on the order Lactobacillales.

Although no direct measurements of the effects of ambient temperature on either the host animals or the endogenous temperature were carried out, it can be hypothesized that it was unlikely that the effects were direct. As mentioned previously, homeostasis maintains a formidable control on temperature changes; nonetheless, biogeographically, the rumen can potentially be exposed to greater fluctuations of temperatures. For example, external factors, such as water and food intake and even air inhalation during rumination under arctic conditions, could result in modest temperature fluctuations within the rumen. Degen and Young (8) and Nicol and Young (39) estimated that the consumption of cold food and water decreases the temperature inside the rumen by lowering the overall endogenous temperature. Nonetheless, the temperature fluctuation is only about 3 to 4°C and this temperature decrease occurs for only about 0 min. (T. A. McAllister, personal communication). Experimental exposure of bacteria to a suboptimal temperature has indicated that a low temperature limited growth by inhibiting protein synthesis (4). Furthermore, it has been reported that closely shorn sheep fed alfalfa and brome grass diets and exposed to low temperatures (1 to 5°C) exhibited a depressed digestibility of the apparent organic matter (8 to 5%) and nitrogen (4%) in the stomach and intestines (21). These effects may be due to increased gut motility, rate of passage of digesta through the gut, and levels of circulating thyroid hormone caused by a higher food intake as the appetite of the animals is stimulated by the ambient cold (52). On the other hand, it is well documented that some bacterial species can exhibit more low-temperature tolerance than others (14, 17, 20) by limiting their metabolic activity and protein synthesis.

The exact effects of the ambient temperature and their proximate causes were not fully elucidated in the present study. Based on previous studies, it can cautiously be speculated that shifts in the bacterial population detected in the rumen liquor samples may have been due to poor access of bacteria to nutrients as a result of higher motility and rate of passage of digesta throughout the gut. The reasons for the slightly stronger effect of the ambient temperature on the bacterial population in the fecal samples are less clear. Biogeographically, it can be argued that the hindgut is less likely to experience temperature fluctuations than the rumen. However, it is very probable that, similar to the case of the bacteria in the rumen samples, the colonic bacterial population was affected by a change in the hindgut motility induced by external cold temperatures (52). Indeed, the bacterial communities in the hindgut may have been affected by a lack of access to nutrients and their proportional balance shifted by a higher rate of excreta.

Effect of sample type.

The results of the current work showed that the lactic acid concentrations in the fecal and rumen liquor samples were similar. Since lactic acid accounts for about 10% of the total dry matter of good quality silage and the diets consisted of alfalfa and sainfoin silages, it is not surprising that the analyses detected the presence of this acid. It is generally accepted that lactic acid-producing bacteria, usually residing in the rumen (29), are also found in the colon and feces (23). On the other hand, the concentrations of VFA and ammonia were higher in the rumen liquor than in the feces. This is in agreement with previous data reported in the literature (1, 35). Indeed, unlike monogastrics, in which the major production of VFA takes place in the hindgut, in ruminants, the greater production of VFA is in the rumen (49). As VFA and ammonia are products of bacterial digestion and fermentation, it was expected that higher concentrations of these products would be found in the rumen liquor.

Although it is accepted that diet influences the concentration and type of ruminal bacteria (38), in the current study, this effect seemed to be more pronounced for fecal bacteria. Indeed, the results showed that bacteria of the phyla Deinococcus-Thermus and Tenericutes and order Bacillales were in higher proportions in the rumen liquor than in the feces. Furthermore, fecal populations have been shown elsewhere to have an inverse linear relationship with intake (38). Nonetheless, in the present work, the proportions of the phylum Bacteriodetes and order Lactobacillales in the bacterial communities tended to increase (Table 3). This may be due to the fact that Bacteriodetes are more closely related to hindgut bacteria, as has previously been suggested (47). In addition, fecal Lactobacillales were found to be more prevalent than Bacillales in the hindgut (Table 3). This may be due to the fact that species comprising these orders are nutritional specialists, similar to other species previously reported (6), and had to compete for the same substrate. Species belonging to the order Lactobacillales have shown great flexibility in adapting to different environmental niches (22), which may also explain their higher resilience in the gut environment.

Cluster analysis also showed that the type of sample produced some differences in the aggregative activity of the data. This agreed with previous findings by Krause et al. (23), who found that the clustering data for different biogeographic sites of the gastrointestinal tract of ruminants aggregated when the animals shared a common diet. In the present work, packets of clusters aggregated at the top end of the phylogenetic tree for data intrinsically linked to rumen liquor samples (Fig. 1).

Richness of bacterial communities in the samples.

Simple richness is considered to be the number of species in each community (11). For the current work, this assertion was redefined as the number of phyla/orders, rather than species, due to the intrinsic limitation of resolution of the T-RFLP analysis. Sampling forces an inherent dependence on the sample size for the number of individual communities observed (11). The Margalef index was used to estimate a more representative indication of the true bacterial richness in the samples. In order to analyze density in the structure of bacterial communities and unlike simple richness (S), the Margalef index also takes into account the number of individuals and normalizes the species richness for the sample size without any further complex rarefaction techniques (12, 30). In contrast to diversity indices, the richness index indeed seemed to corroborate the stabilization of the bacterial communities throughout the experiment (Tables 4 and 5).

Diversity of bacterial communities in the samples.

The diversity indices presented herein suggested a positive relationship of decreased diversity with a decrease in temperature. For example, the initial temperature fluctuation in the first two stages of the experiment decreased the relative richness of the bacterial population by altering the putative abundance of individual bacterial species in all samples but especially in the rumen liquor (Tables 4 and 5).

It can be cautiously speculated that certain species found in the bacterial communities in the rumen, which also may have been producers of branched-chain fatty acids, were indeed affected by an increased rate of passage of digesta throughout the rumen. This change in the transit time of digesta in the gut was likely to have been brought about by an increased appetite due to very cold ambient temperatures (52). It can be hypothesized that, as the number of these bacterial species decreased, the vacant ecological niche in the rumen may have been rapidly occupied by other bacterial species within the same bacterial communities and with similar metabolite-producing ability and common substrate needs but that were more resilient and abundant. This way, the proportions of the taxons in the bacterial communities studied never seemed to have diminished (Table 3).

Conclusions.

The data presented here are among the first to characterize the bacterial communities in the rumen and the feces of steers subject to negative factors such as phytochemical tannins and arctic temperatures. Based on the results obtained, it can be cautiously speculated that, although bacterial communities have the ability to overcome adverse conditions and survive, only those better equipped would prevail and occupy the ecological niches left vacant by others unable to cope with adverse factors. Certain bacterial populations seemed to be better able to adapt to changes in the rumen and hindgut environment, although the reasons remain equivocal. The results of the present experiment may suggest an influence of the biogeographical location of the gut site. Another reason may be the greater abundance and resilience of certain bacterial communities than of others, which help them to remain and thrive even under poor access to nutrients and the presence of antimicrobial compounds in the diet, although this may forfeit the diversity in the ecosystem. It may well be that a combination of these two environmental factors affected the proportions of the taxons in the bacterial communities. The analyses of principal components and the Margalef richness index seemed to confirm the prevalence of certain types of bacteria over others. The use of T-RFLP analyses was satisfactory. Nonetheless, due to the inherent limitations of this technique, further analysis of the identities of the species that may have experienced shifts in their population was not possible. Further research on the bacterial populations in the rumen and the hindgut of ruminants using a more sophisticated fingerprinting technique is recommended, in order to further characterize these bacterial communities and the environmental factors which affect them.