An analysis of the ruminal bacterial microbiota in West African Dwarf sheep fed grass- and tree-based diets




To measure the impact of supplementing a forage diet with tree-based browse on the ruminal bacterial communities of Nigerian West African Dwarf (WAD) sheep.

Methods and Results

Fifteen WAD sheep were fed a control diet of forage (Panicum maximum), with 12 animals shifted in groups of three to one of four browse-supplemented diets (Albizia saman, Bridelia micrantha, Ficus sur, or Gmelina arborea). These browse plants were shown in a concurrent but separate study to be reasonably nutritious (based on chemical composition and fibre constituents) and nontoxic (based on tannin, phytate, saponin, alkaloid and oxalate levels). Rumen liquids and solids for DNA extraction were collected via intubation from two animals in each group before and after dietary shift. Bacterial 16S rRNA gene regions V6-V8 were sequenced by 454 pyrosequencing. All communities were highly diverse and dominated by the phyla Firmicutes, Bacteroidetes, Tenericutes, Actinobacteria and Proteobacteria. All communities shared members of the genera Butryivibrio, Prevotella and Ruminococcus. Our analysis defined a core sets of bacteria shared by all animals, forage-fed animals and browse-fed animals. Community structure shifted dramatically in animals fed A. saman or G. arborea.


The impact of tree-based browse on the ruminal bacterial community of Nigerian WAD sheep varies by browse species, likely due to differences in browse composition.

Significance and Impact of the Study

Our study describes the first neotropical small ruminant bacterial microbiome and supports diet supplementation with specific tree-based browse for WAD sheep.


There is an ever-increasing need to understand agricultural practices, especially for subsistence farmers in equatorial Africa. The West African Dwarf (WAD) sheep (Ovis aries, Djallonké) is an important meat animal reared on family farms (Adesehinwa et al. 2004) from Senegal to Botswana and number in the millions of animals per country (Shaw et al. 2006). WAD sheep are highly adaptable to a broad range of environments, can live on crop by-products without grain supplementation (Anele et al. 2010), are trypanotolerant (Geerts et al. 2009), have rapid growth (Sowande and Sobola 2007) and are rich in genetic variation (Akinyemi and Salako 2010). Many African farmers use tree-based browse as feed supplements, as these feed sources require no cultivation. Moreover, during the dry season, tree leaves and branches are potentially more nutritious than grasses (D'Mello and Fraser 1981; Aletor and Omodara 1994) and with measurably higher crude protein (Odedire and Babayemi 2008; Omoniyi et al. 2013). Previous work suggests that certain trees are acceptable feed substitutes with no detrimental effects on overall animal production (Bamikole et al. 2004; Osakwe and Drochner 2006; Ouédraogo-Koné et al. 2008). However, some tree-based browse can result in feed refusal and loss of host nitrogen (Reed et al. 1990; Osakwe and Drochner 2006), possibly by adversely changing the ruminal microbial community through increased phenolic and tannin concentrations (Waterman et al. 1980). Thus, it is important to investigate the effect of these diet alterations on animal digestion, particularly with respect to the host ruminal microbial community.

Ruminant digestion relies upon a ruminal microbial community composed of protozoa, fungi, bacteria and archaea (Dehority 2003), with the bulk of cellulose hydrolysis thought to be performed by bacteria (for a recent review, see (Wilson 2011)). Recent diet-based studies characterizing rumen communities in cattle (Callaway et al. 2010; Fernando et al. 2010; Kong et al. 2010; de Menezes et al. 2011; Li et al. 2012b; Petri et al. 2012) and sheep (Perumbakkam et al. 2011; Stiverson et al. 2011; Saro et al. 2012) have highlighted the dynamic responses of ruminal microbes to changes in diet composition. To date, no study has been reported utilizing a sequence-based approach to characterize the ruminal bacterial microbiome for any neo-tropical small ruminant.

As such, understanding the ruminal bacterial community and its response to diet supplementation is of particular interest in the neo-tropics, where tree-based browse usage can decrease reliance on the production of grass-based forage. Moreover, recent work has suggested that for ruminants, a mixed diet including browse plants and forage can not only increase feed efficiency, but also promote higher biodiversity (Broom et al. 2013). Here, we used Nigerian WAD sheep and 454 pyrosequencing to assess the impact of specific tree-based browse on neo-tropical ruminal bacterial communities, with the hypothesis that tree-based browse diet supplementation would shift the bacterial communities to be distinct from grass-fed animals. We also investigated whether or not a core ruminal bacterial community is present in WAD sheep fed grass-based forage or tree-based browse. In this study, we fed WAD sheep a basal diet of the grass Panicum maximum, divided them into groups of three by weight and kept one group on the basal diet while supplementing the diet of the other groups with one of the browse trees Albizia saman, Bridelia micrantha, Ficus sur, or Gmelina arborea. These tree species were chosen for their ubiquity and ease of growth in West Africa, current underutilization as feedstocks, and known lack of antinutritional toxicity (Omoniyi et al. 2013). We expect our results to be informative both for ruminant microbiologists and for future applications of tree-based browse in ruminant agriculture.

Materials and methods

Experimental feed materials and sources

Four indigenous or naturalized browse trees (A. saman (family Fabaceae), G. arborea (Verbenaceae), F. sur (Moraceae) and B. micrantha (Phyllanthaceae)) were harvested from an arboretum and the grass P. maximum from a cultivated pasture, established by the Department of Pasture and Range management of the Federal University of Agriculture, Abeokuta, Ogun State, Nigeria. The study location (latitude: 7°N, longitude 3·5°) is in the savannah agro-ecological zone of southwest Nigeria in Abeokuta that receives an average annual rainfall of 1037 mm (Anele et al. 2010). A total of five diets were used in this study: Diet I (Control) – P. maximum (100%); Diet II – A. saman (60%) + P. maximum (40%); Diet III – G. arborea (60%) + Pmaximum (40%); Diet IV – F. sur (60%) + P. maximum (40%); and Diet V – B. micrantha (60%) + P. maximum (40%). Each diet was measured using dry matter, with total feed volume adjusted to prevent selective refusal in the mixed diets.

Feeding and Management of Animals

All animals were reared in the sheep unit at the Federal University of Agriculture, Nigeria, following institutional guidelines. Fifteen indigenous female West African Dwarf (WAD) sheep aged 5–7 months and weighing between 11·20–14·50 kg were grouped by weight into five cohorts of three animals. All animals were placed in disinfected individual pens with ad libitum water access, de-wormed with Albendazole® 2·5% oral suspension at 1 ml/10 kg body weight and treated against ectoparasites with Cypermethrin® Pour-on at 1 ml/10 kg body weight. The weights of all animals were taken before the commencement of the experiment and weekly thereafter. Each cohort was fed exclusively P. maximum prior to the start of the experiment. Cohorts were then randomly assigned to a dietary treatment as described above, with all diets being offered in equal total quantities and refusals collected daily to track consumption of browse vs Pmaximum consumption. The entire feeding period lasted 84 days, with feed refusals collected and weighed daily after an initial 2-week adaptation period. A one-way anova was performed on all animal metrics using the statistical software SAS, ver. 6 (SAS Inst. Inc., Cary, NC).

Collection of rumen fluids and solids

Rumen contents were collected from two animals on each diet at the start (basal P. maximum forage diet) and completion (after the 2-week adjustment period following diet supplementation with tree-based browse) of the experiment. All samples were collected prior to morning feeding. Each animal was assigned an arbitrary number (1–10) for sample identification purposes. An aliquot of 50 ml of rumen fluid containing solids was taken from the oesophagus via suction tube. The rumen samples were filtered through four layers of cheesecloth to obtain both liquid and solid (fibre-adherent) portions of approximately 20 ml each. Each sample was preserved at −20°C and shipped on dry ice to the University of Wisconsin-Madison, Madison, WI, USA, for analysis of the ruminal microbiota. All samples were imported under USDA import permit #120106.

Template preparation and 454 pyrosequencing

Total genomic DNA was extracted separately from solid and liquid fractions using mechanical disruption with hot/cold phenol as previously described (Stevenson and Weimer 2007). DNA quantification and integrity were measured using a Nanodrop (Thermo Scientific, Wilmington, DE) and by gel visualization (1% agarose in TAE). PCR was carried out on each liquid and solid sample to amplify the V6-V8 variable region of the 16S rRNA gene using primers constructed from the universal 16S rRNA sequences 926F and 1392R coupled to the Roche 454 A or B Titanium sequencing adapters, respectively. Specifically, the forward primer used was 926F-5′-CCTATCCCCTGTGTGCCTTGGCAGTCTCAGAAACTYAAAKGAATTGACGG-3′ and the reverse primer included one of 20 barcodes, 5 bp in length, as indicated by XXXXX: 1392R-5′-CCATCTCATCCCTGCGTGTCTCCGACTCAG-XXXXX-ACGGGCGGTGTGTRC-3′. Each sample was amplified in 20 μl reactions containing 40 ng of DNA and 0·125 μmol l−1 final concentration of each primer with the high-fidelity DNA polymerase Platinum Blue (Invitrogen Life Technologies, Grand Island, NY). The following PCR cycling conditions were used: initial denaturation of 94°C for 2 min followed by 30 cycles of 94°C for 30 s, 50°C for 45 s and 68°C for 1 min 45 s, with the final extension at 68°C for 10 min. Amplicon creation without secondary products was determined by gel electrophoresis (1% agarose in TAE).

Total PCR products for liquid and solid DNA samples were combined to make a single equimolar pool for each animal. Each pool was cleaned twice with the Agencourt AMPure XP system (Beckman Coulter, Inc., San Diego, CA) to remove primers and short DNA fragments and then quantified using a Qubit® Fluorometer (Invitrogen, San Diego, CA). Amplicon quality in each pool was verified using an Agilent Bioanalyzer with the DNA 1000™ chip (Agilent Technologies, Santa Clara, CA, USA), and a final pool containing equimolar portions of all samples was made at 1 × 109 molecules μl−1. An emPCR reaction was performed using an approximate ratio of 0·8:1 (amplicon:emPCR beads). Amplicon sequencing was performed following the manufacture's protocols (Roche Applied Science, Indianapolis, IN) for Titanium sequencing on a Roche 454 GS Junior Titanium sequencer using a Lib-L kit.

Sequence analysis

All sequences have been deposited with sample IDs and barcodes at the National Center for Biotechnological Information's Short Read Archive projects under accession SRP027328. Data analysis was performed using the bioinformatics program mothur v.1.28.0 (Schloss et al. 2009). In brief, sequences were allowed to have a maximum of two differences in the primer and none in the barcode and de-noised using an implementation of the Amplicon Noise algorithm (Quince et al. 2011). Sequences were trimmed to a minimum length of 250 bp, aligned against the SILVA 16S rRNA gene reference alignment database (Pruesse et al. 2007) and checked for putative chimeric sequences (chimera.uchime). All sequences were classified (classify.seqs) to operational taxonomic units (OTUs) at a 95% identity level (classify.otu) using the Greengenes database (DeSantis et al. 2006) at a confidence level of at least 60% with Cyanobacteria, Eukaryota and Archaea lineages removed as our primers were not designed to amplify these groups. Sequence coverage was determined using rarefaction and Good's coverage (Good 1953), with diversity measured using Simpson's diversity index (Simpson 1949). The bacterial communities were analysed using AMOVA (iters = 1 000 000; Yue and Clayton Theta (Exchoffier et al. 1992; Yue and Clayton 2005)), Good's coverage (Good 1953), principal coordinates analysis (PCA) (Gower 1966), weighted UniFrac (Lozupone et al. 2006) and unweighted pair group method with arithmetic mean (UPGMA) clustering (Sokal and Michener 1958) based on the Morisita–Horn index (Horn 1966) as implemented in mothur.


All tree-based browses were consumed without negative host impact

Weight and feed consumption data for all study animals are reported in Table 1. None of the browse-supplemented diets (Diets II–V) had a negative impact on weight gain as compared to the control P. maximum diet (Diet I). In all cases, the addition of browse reduced the daily consumption of P. maximum. All dietary treatments resulted in positive body mass changes, with an average daily gain ranging from of 20·64 g (Diet I) to 41·72 g (Diet V). No animals showed signs of disease or ill health during the course of the trial.

Table 1. Weight and feed intake metrics for all study animals (= 3 per diet)
 Control Diet I (kg)Supplementation
Albizia saman Diet II (kg)Gmelina arborea Diet III (kg)Ficus sur Diet IV (kg)Bridelia micrantha Diet V (kg)SEM
  1. Significant within row (different superscript letters) (< 0·05, anova).

  2. a

    Dry matter weight.

Average animal weight
Average daily feed intakea
Tree-based browse0·29b0·30b0·37a0·35a0·04
Panicum maximum 0·40a0·19c0·20c0·24b0·23b0·02

Sequencing the WAD sheep ruminal bacterial community

To assess each animal's ruminal bacterial community, we performed a 454 pyrosequencing-based analysis of 16S rRNA genes on pooled (solids and liquids) rumen samples from 19 of our 20 samples. We used the variable regions V6-V8 to maximize sequence length and minimize the impact of any individual region on the final diversity and identity metrics. Because our primary interest was in the whole ruminal bacterial population and not on phase-specific shifts, and to maximize per-sample sequence coverage, we pooled the solid- and liquid-based PCR products. One sample (Animal 6 fed F. sur) did not survive transport and processing. From an initial set of 82 366 sequences, a total of 44 262 sequences were retained through all clean-up and filtering steps. Of these, 6533 were unique, with an average length of 429 bp. Sequence distribution, Good's coverage and inverse Simpson's diversity index values are given in Table S1. An average of 2330 ± 301 SD sequences per sample was obtained. Importantly, sufficient coverage for each sample was achieved, as measured by a Good's value of at least 92% for each sample (Table S1), a levelling of the associated rarefaction curves (Fig. S1), and the closeness of our final sequence counts to the theoretical maxima calculated from second order equations (all R2 > 0·97) fitted to each rarefaction curve (Table S1). The inverse Simpson's index for all 10 animals (= 12 samples) on the control P. maximum diet ranged from 3·51–45·45, with an average of 19·8 ± 16·4 SD.

WAD sheep ruminal bacterial communities are small but rich in diversity

Among all animals, there were a total of 1272 unique OTUs at 95% sequence similarity; 794 of these were classifiable to at least the family level with a minimum confidence of 60%. The mean number of OTUs across all 10 animals fed P. maximum prior to browse supplementation was 260 ± 70 SD. These were dominated by sequences belonging to the phyla Firmicutes (57·0%) and Bacteroidetes (17·9%) (Fig. S2). The remaining sequences were distributed among the Tenericutes (7·6%), unclassified Bacteria (8·8%), Actinobacteria (2·7%), Proteobacteria (1·7%) and a number of low-abundance phyla. These P. maximum-fed animals were the only animals containing the phyla Fibrobacteres (0·3%), Lentisphaera (0·3%) and Fusobacteria (0·1%). Other low-abundance phyla, such as TM7 and Fusobacteria, were found at similar levels in all diets, although there was individual variation by animal. The general pattern and order of relative OTU dominance were similar among all P. maximum-fed animals (Fig. S3a), but this pattern was less clear when examined by relative sequence abundance (Fig. S3b). Both analysis methods indicated that the dominant phyla are the Firmicutes and Bacteroidetes for all P. maximum-fed animals.

When analysed at 95% OTUs (approximately genus level), the individual ruminal bacterial communities diverged between both treatments and animal pairs. There was a high degree of variation in the total community structure among all Diet I samples (Fig. 1a,b), with each animal's total bacterial community significantly different from all others as determined by a weighted UNIFRAC (< 0·001). When the total community structure was analysed by UPGMA (Fig. 1a) or PCA (Fig. 1b), there was a general pattern of P. maximum-fed animals separating from the tree browse-fed animals, but this included a high degree of mixing. Although the large total loading values of our PCA plot (73·18% of total variance) indicate that the community pattern could be explained by two major variables, there is a lack of clustering by our chosen variable of diet treatment.

Figure 1.

Comparison of the West African Dwarf sheep ruminal community structure by UPGMA, PCA, and vector analysis. (a) UPGMA tree and (b) PCA plot showing total community structure relationships based on 95% similarity (0·05) operational taxonomic units. (c) The degree of change for each sample pair (from before to after diet supplementation) was calculated as a vector from their respective PCA coordinates. Individual animals are identified by number (1–10). Diets: (○) I control, (□) II, (Δ)III, (×) IV, (♢) V.

Feed supplementation can alter the individual ruminal communities

To determine the impact of tree-based browse on the total ruminal community structure in each animal, we performed a vector analysis for each animal using our PCA coordinates (pre and postdiet supplementation) (Fig. 1c) and an AMOVA for each diet. By PCA vector comparison, each animal pair's vector magnitudes showed two major groups of communities: those with magnitudes less than 0·103 (Diets I, IV and Diet V) and those with magnitudes greater than 0·270 (Diets II and III). Given that larger magnitudes correspond to a greater degree of total community structure change, our results indicate that Diets II and III resulted in dramatic shifts within the WAD sheep rumen community, regardless of the community's starting structure (Fig. 1b). Diets IV and V resulted in a degree of community shift, as measured by vector analysis, comparable to the animal pair maintained on the control Diet I. By AMOVA, only Diet V was not significantly different from the control animals (= 0·333 for Diet V and < 1 × 10−6 for all other diets). The community structure changed significantly (< 1 × 10−6, AMOVA) for each animal over the course of the experiment except for the single F. sur-supplemented animal.

To measure specific changes occurring within these communities, we analysed the distribution of OTUs among all diets. To reduce interanimal variation, we required each OTU to be present in half or more of the samples for Diet I. For Diets II–V, we required the OTU be present in all animals on that diet. Unclassified OTUs were not considered in our analysis. All OTUs were pooled at the genus level, and the relative abundance of OTUs within each genus was calculated relative to the total number of OTUs in each diet (Table 2). Using an OTU-based abundance metric, we were able to approximate a measure of the diversity present within each genus. In all diets, the highest percentage of OTUs was within Prevotella (16·67–33·33%). Decreases in OTU diversity, relative to Diet I, were seen for Eubacterium (5·4-fold) and Ruminococcus (2·9-fold) on Diet IV, while Ruminococcus and Coprococcus decreased 2·3-fold in Diet II. Changes greater than twofold, relative to Diet I, included Oribacterium (4·4-fold) in Diet III and Shuttleworthia (4·1-fold) and Oscillospira (3·1-fold) in Diet V.

Table 2. Relative sequence abundance distribution of 0·05 operational taxonomic units (OTUs) pooled at the genus level
Shared OTUsaControl Diet I = 12 samplesSupplementation (= 2 samples per diet)
Albizia saman Diet IIGmelina arborea Diet IIIBridelia micrantha Diet V
  1. Data are expressed as percentage.

  2. a

    Shared by at least six samples for Panicum maximum-fed animals and by both animals for other diets (Diet IV, Ficus sur, is not included due to being a singlet).

Prevotella 27·1717·1416·6733·33
Eubacterium 11·9614·2911·902·22
Butyrivibrio 7·6111·437·146·67
Clostridium 6·524·766·67
Coprococcus 6·522·867·144·44
Ruminococcus 6·522·869·522·22
Acetivibrio 4·356·67
Selenomonas 3·262·862·22
Oscillospira 2·172·386·67
Lactococcus 2·172·862·38
Bulleidia 2·175·714·44
Desulfovibrio 1·092·862·382·22
Shuttleworthia 1·092·862·384·44
Carnobacterium 1·092·862·38
Enterococcus 1·092·862·38
Psychrobacter 1·092·862·38
Streptococcus 1·092·862·38
Pseudobutyrivibrio 1·092·38
Staphylococcus 1·092·38
Oribacterium 1·092·864·762·22
Dehalobacterium 1·092·22
Treponema 1·092·22
Atopobium 1·092·86
Aerococcus 1·09
Arthrobacter 2·38
Paludibacter 2·38
Planomicrobium 2·38
Pseudomonas 2·86
Succinivibrio 2·86
Synergistes 2·86

To determine changes in relative abundance of highly represented genera, we compared the ten most abundant genera in each diet (Fig. S4). In many cases, both animals on each diet showed either an increase or decrease relative to Diet I, but with only one of the two animals, it was outside of the expected level of variation (calculated as the standard deviation of the mean in all 12 Diet I control samples). In the following cases, both animals on a given diet had the same trend and were outside of the expected variation: Carnobacterium increased in Diets II, III and IV; Psychrobacter increased in Diet IV; Staphyloccocus increased in Diet II; and SHD-231 (family Anaerolinaceae) decreased in Diet V.

WAD sheep have a core microbiome

We then identified the set of core OTUs present in WAD sheep across all animals on all diets (Diets I–V); those fed tree-based browse (Diets II-IV); and those fed only P. maximum grass forage (Diet I and initial samples from all animals). Many of the OTUs share genus- or family-level taxonomy, and included both classical ruminal genera such as Ruminococcus and unclassified genera. As shown in Table 3, there were 11 OTUs present in all samples, the most abundant of which belonged to an unclassified member of the Catabacteriaceae. Other members of the shared core microbiome include OTUs in the Chloroflexi, Clostridiales, Ruminococcaceae, Butyrivibrio, Prevotella and Ruminococcus. The core set of OTUs found in all animals on all diets was not significantly different in relative abundance when split by forage vs browse diets (1-way anova,> 0·05).

Table 3. Core ruminal bacterial communities found in West African Dwarf sheep
Control Diet I (= 12)aAlbizia saman Diet II (= 2)Gmelina arborea Diet III (= 2)Ficus sur Diet IV (= 1)Bridelia micrantha Diet V (= 2)OTU (confidence)b
  1. Data are expressed as percentage.

  2. a

    Includes 10 samples from all animals prior to diet supplementation and two samples from the control diet animals post-diet supplementation.

  3. b

    Taxonomic identifications for each OTU are given at the phylum level followed by the most specific classification that could be defined, with the confidence for each classification given in parentheses.

Present in all animals, times and diets (Diets I–V)
14·6114·0313·3416·5810·55Firmicutes: Catabacteriaceae (96)
4·772·246·402·666·02Firmicutes: Butyrivibrio (100)
2·852·121·831·316·23Firmicutes: Ruminococcaceae (100)
2·080·811·801·314·51Firmicutes: Clostridiales (94)
1·080·620·310·511·70Firmicutes: Clostridiales (100)
0·930·380·172·030·22Chloroflexi: SHD-231 (100)
0·640·741·270·130·79Firmicutes: Clostridiales (92)
0·800·190·261·180·46Bacteroidetes: Prevotella (100)
0·500·330·790·300·77Firmicutes: Butyrivibrio (100)
0·570·260·630·130·38Firmicutes: Clostridiales (100)
0·580·480·220·340·30Firmicutes: Ruminococcus (100)
Present only in all browse-supplemented animals (Diets II–V)
0·430·480·340·77Firmicutes: Lachnospiraceae (100)
0·450·670·590·12Bacteroidetes (100)
0·710·140·130·40Firmicutes: Veillonellaceae (95)
0·050·070·130·04Bacteroidetes: Prevotella (100)
0·210·050·040·14Proteobacteria: Desulfovibrio (100)
0·100·120·080·12Firmicutes: Lachnospiraceae (100)
Present only in Panicum maximum-fed animals both pre and postsupplementation
2·24Firmicutes: Staphylococcus (100)
1·72Firmicutes: Lachnospiraceae (100)
1·73Firmicutes: Catabacteriaceae (100)
0·62Chloroflexi: SHD-231 (100
0·50Actinobacteria: Atopodium (73)
0·46Firmicutes: Clostridiales Family XIII Incertae Sedis (100)
0·26Tenericutes: p-75-a5 (100)
0·20Firmicutes: Lachnospiraceae (88)
0·17Firmicutes: Lachnospiraceae (100)
0·16Firmicutes: Clostridiales (100)
0·19Firmicutes: Selenomonas (99)

OTUs present only in all browse-supplemented animals included six OTUs, of which the most abundant was an unclassified member of the Lachnospiraceae. The other members of this browse-fed core OTU set belong to the Bacteroidetes, Veillonellaceae, Prevotella and Desulfovibrio. The animals fed only P. maximum had 11 OTUs present in all samples that did not appear in any animals fed Diets II–V, with the highest relative sequence abundance OTU being a member of Staphylococcus at 2·24%. The remaining OTUs included members of the Chloroflexi, Tenericutes, Catabacteriaceae, Clostridiales, Lachnospiraceae, Atopodium and Selenomonas.

Classical ruminal bacteria are also present in the WAD sheep rumen

Given the known functional importance of classical ruminal bacteria in other ruminants, we determined whether there existed a correlation between these bacteria and community shifts on the various tree-based browse-supplemented diets. Specifically, we examined sequence abundances for Diets II–V, relative to Diet I, that were classified as belonging to the following important ruminal genera: Bacteroides (Wallace and Brammall 1985), Butyrivibrio (Wallace and Brammall 1985), Fibrobacter (Osborne and Dehority 1989), Megasphaera (Counotte et al. 1981), Prevotella (Osborne and Dehority 1989), Ruminococcus (Sijpesteijn 1951), Selenomonas (Wallace and Brammall 1985) and Streptococcus (Hudson et al. 2000). All results are given in Table S2, with fold changes considered significant when more than twofold. The genera Fibrobacter, Lachnospira and Megasphaera were in extremely low abundance in all diets and were not detectable in most samples. In Diet II, Bacteroides, Butyrivibrio, Prevotella and Streptococcus decreased, while Succinivibrio increased. In Diet III, Bacteroides, Prevotella and Selenomonas were decreased. In Diet IV, Lachnospira, Selenomonas, Streptococcus and Succinivibrio all increased, with Succinivibrio being nearly 20-fold higher. In Diet V, which was the diet that by other metrics resulted in a ruminal community most closely resembling that of the control diet, there were decreases in Bacteroidetes, Selenomonas and Succinivibrio.


The ability to shift livestock feeding practices to local browse plants has the potential to increase the availability of arable land for other agricultural uses such as the cultivation of cash crops. Importantly, such a shift should not negatively impact livestock health or production. Of the diets compared in this study, only the control diet of P. maximum is a grass (guineagrass, Diet I), while all browse plants are trees indigenous or naturalized to tropical Africa that grow without deliberate cultivation on suboptimal land. We found that all tree-based browse diets tested were accepted as feed by WAD sheep, and none negatively impacted weight gain or health over the course of our study.

Our WAD sheep rumen community analysis revealed that the total bacterial community for each animal was small, relative to other ruminants like cattle (Shanks et al. 2011; Li et al. 2012a; de Oliveira et al. 2013), but remained highly diverse. This may reflect a tightly knit ruminal community enabling WAD sheep to flourish on suboptimal feeds. In particular, we found that the WAD sheep rumen is dominated by the Firmicutes, Bacteroidetes and Tenericutes, with the minor presence of Actinobacteria and Proteobacteria (Fig. S1). Previous work in sheep using clone libraries found similar trends for the Firmicutes and Bacteroidetes (Larue et al. 2005; Perumbakkam et al. 2011; Stiverson et al. 2011), and a recent multispecies (cows, sheep and red deer) pyrosequencing-based study found that among all samples, the phyla with the most abundant families were the Firmicutes, Bacteroidetes and Fibrobacteres (Kittelmann et al. 2013). In our data set, the high abundance of sequences and OTUs in the Tenericutes suggest that this phylum may play an important role in WAD sheep.

Our analysis also revealed 11 OTUs shared across all animals and diets, of which many are well-known ruminal bacteria including Clostridiales, Ruminococcaceae, Butyrivibrio and Prevotella (Table 3). Bacteria in the hemicellulolytic genera Prevotella and Butyrivibrio accounted for at least 25% of the sequences recovered from all of our samples, whereas bacteria in the cellulolytic genera Ruminococcus and Fibrobacter contributed to at most 9·5% of the sequences in any given diet (Table 2). Comparisons between grass and browse-supplemented diets showed only modest increases and decreases in the abundances of these bacteria (Table S2), suggesting that their populations were stable throughout the experiment. We also found a member of the Catabacteriaceae and the Chloroflexi bacterium SHD-231 conserved across all diets; members of these groups have likewise been found in other ruminants (Kong et al. 2010; Li et al. 2011; Samsudin et al. 2011). Given that all of these OTUs persisted after diet shifts, it is likely that these bacteria play important roles within the WAD sheep ruminal community, possibly by participating in the fermentation of short-chain fatty acids from plant polysaccharides.

Our data also show clear shifts in the bacterial community structure. It is unlikely that these shifts are due to changes in the ruminal community due to maturation of the animals over the course of the feeding trial, as the degree of change for the control diet animals (and B. micrantha) was extremely small (Fig. 1). Differences in ruminal flora due to browse supplementation may instead be due to compositional differences between P. maximum and the browse plants used in our study. These feeds are similar (0·6–1·5-fold relative to P. maximum) in levels of cellulose, lignin, neutral detergent fibre and dry matter (Odedire and Babayemi 2008; Omoniyi et al. 2013). B. micrantha was the only feed with an increased level of acid detergent fibre (2·0-fold) (Odedire and Babayemi 2008; Omoniyi et al. 2013). All of the tree-based browses were higher in crude protein (1·8–2·5-fold) and lower in hemicelluloses (2·0–4·9-fold) except for B. micrantha, which has a hemicellulose composition similar to P. maximum (1·1-fold) (Odedire and Babayemi 2008; Omoniyi et al. 2013) and which did not significantly change the ruminal bacterial community as compared to P. maximum during our study. It is also possible that specific antinutritional compounds (such as alkaloids) could have an impact on the ruminal bacterial community. However, we did not see this effect, as the diet most like the P. maximum control, B. micrantha, has increased levels of tannins, phytates, alkaloids and oxalates (Omoniyi et al. 2013). Of the seven OTUs shared between P. maximum and B. micrantha-fed animals, five were in the Clostridiales. The shared Clostridiales OTUs included Eubacterium, two Lachnospiraceae, two Clostridiales and one Catabacteriaceae; many bacteria in these groups are known to be involved in plant matter degradation, including hemicellulose fermentation specialists. It is probable that feed quality, digestibility and freshness all play significant roles in the composition of the ruminal community, and the impact of these factors remains to be investigated.

Our whole-community structure analysis revealed that browse-supplemented diets resulted in increased OTU diversity, although there was a high degree of interanimal variation per diet (Fig. 1). We suspect that this high degree of variation is due to the complex genetic diversity of the WAD sheep, and the free-range flock management strategy used outside of the feeding trial. Of the diets tested, we found that Diets IV and V were comparable to the control diet (Fig. 1), with Diet V having a total community structure indistinguishable from the control diet. This finding suggests that, instead of shifting the ruminal community to better digest the browse plants, these diet supplements were best suited to take advantage of the bacteria already present in the WAD sheep rumen.

In conclusion, we have defined a core set of bacterial OTUs for the WAD sheep based on the variable regions V6-V8 of the 16S rRNA gene and further characterized the major shifts in the ruminal populations due to diet supplementation with multiple tree-based browses. We acknowledge that the number of animals used for individual diets other than P. maximum is relatively small, as we were restricted by available resources related to sampling, storing, and shipping specimens between Nigeria and the USA. One further area of investigation not addressed in our research, but of potential future interest, is that methanogens in sheep can be suppressed by increasing dietary protein (McAllister and Newbold 2008). The browse used in our study, with nearly twice the protein of P. maximum (Odedire and Babayemi 2008; Omoniyi et al. 2013), may have a similar impact on WAD sheep. Similarly, one of the classifiable OTUs conserved among only browse-supplemented diets, Desulfovibrio, is known members of the sheep rumen community that reduce sulphate to sulphide (Howard and Hungate 1976) and can act as competitors for methanogens (Ellis et al. 2008). To fully understand the impact of diet upon the rumen community, future studies should include methanogens, protists and fungi, in addition to bacteria. The analysis presented here can be used as a framework for advancing our knowledge of the general ruminal microbial community structure in neo-tropical small ruminants, in addition to providing a framework useful for the management and understanding of WAD sheep agriculture.


This work was supported by the University of Wisconsin-Madison College of Agriculture and Life Sciences and a USDA NIFA Fellowship Grant (2012-01193) to K. Jewell. All authors wish to thank the members of the sheep unit at the Federal University of Agriculture for their support during the feeding trial. We also wish to thank all members of the Suen Lab for their support, insightful discussions and careful reading of the manuscript.

Conflict of Interest

No conflict of interest declared.