OUP user menu

Characterization of the fecal bacteria communities of forage-fed horses by pyrosequencing of 16S rRNA V4 gene amplicons

Megan L. Shepherd, William S. Swecker Jr, Roderick V. Jensen, Monica A. Ponder
DOI: http://dx.doi.org/10.1111/j.1574-6968.2011.02434.x 62-68 First published online: 1 January 2012


The diversity of the equine fecal bacterial community was evaluated using pyrosequencing of 16S rRNA gene amplicons. Fecal samples were obtained from horses fed cool-season grass hay. Fecal bacteria were characterized by amplifying the V4 region of bacterial 16S rRNA gene. Of 5898 mean unique sequences, a mean of 1510 operational taxonomic units were identified in the four fecal samples. Equine fecal bacterial richness was higher than that reported in humans, but lower than that reported in either cattle feces or soil. Bacterial classified sequences were assigned to 16 phyla, of which 10 were present in all samples. The largest number of reads belonged to Firmicutes (43.7% of total bacterial sequences), Verrucomicrobia (4.1%), Proteobacteria (3.8%), and Bacteroidetes (3.7%). The less abundant Actinobacteria, Cyanobacteria, and TM7 phyla presented here have not been previously described in the gut contents or feces of horses. Unclassified sequences represented 38.1% of total bacterial sequences; therefore, the equine fecal microbiome diversity is likely greater than that described. This is the first study to characterize the fecal bacterial community in horses by the use of 16S rRNA gene amplicon pyrosequencing, expanding our knowledge of the fecal microbiota of forage-fed horses.

  • horse
  • feces
  • microbiome


The horse is a nonruminant herbivore where the hindgut (cecum and colon) is a fermentative chamber for a complex and dynamic microbial population. Gut microorganisms serve the host through energy extraction, immune stimulation, pathogen exclusion, and detoxification of toxic compounds. Despite the horse's reliance on these microorganisms, the scientific literature currently provides limited details on which microorganisms are present and how these microorganisms maintain host health.

The horse's forage-based diet is rich in fiber, a molecule indigestible by host enzymes. Hindgut bacteria, especially those with fibrolytic metabolism, enable herbivores to thrive on a high-fiber forage-based diet by slowly fermenting these fibers in the hindgut. The horse's hindgut serves as an ideal anaerobic environment for fiber fermentation. The cecum and colon make up the majority (∼70%) of the equine gastrointestinal tract, and 75% of the mean transit time (23–48 h) is spent in the hindgut (Argenzio, 1975; Van Weyenberg et al., 2006). Ruminant herbivores obtain up to 80% of total daily calories from microbial fermentation with a mean forage retention time of 57 h (Bergman et al., 1965; Uden et al., 1982). The horse obtains more than 50% of its daily energy requirements from volatile fatty acids that are the microbial products of fiber fermentation (Argenzio et al., 1974; Glinsky et al., 1976; Vermorel & MartinRosset, 1997). In contrast, humans obtain only 10% of total daily calories through fermentation despite having similar mean retention times (Kelsay et al., 1978; Wrick et al., 1983). Species differences could be due to the fact that larger percentages of the gastrointestinal tract of horses and cattle (69% and 76%, respectively) accommodate microbial fermenters in comparison with humans (17%) (Parra, 1978). Furthermore, the differences in the location of microbial fermentation in the horse (hindgut) vs. the ruminant (pregastric/foregut) may also influence members and functions of these communities. Differences in diet between horses and other species likely also influence the members and function of the microbial communities.

Compared to the rumen microbiota, the equine hindgut microbiota has received little attention; furthermore, few studies have characterized the equine hindgut bacterial community using culture-independent methods (Daly et al., 2001; Daly & Shirazi-Beechey, 2003; Hastie et al., 2008; Yamano et al., 2008). No studies to date have evaluated the fecal bacterial community in adult horses on a controlled forage diet by the use of pyrosequencing of 16S rRNA gene amplicons. The objective of this study was to characterize the fecal bacterial community of horses fed grass hay using pyrosequencing of 16S rRNA gene amplicons. We propose that the use of high-throughput sequencing will provide an evaluation of the equine fecal microbiome, which may be used to increase the understanding of the relationship between the microorganisms and the host.

Materials and methods

Fecal samples for this study were taken from two adult Arabian geldings during a companion study (Shepherd et al., 2011). The protocol was approved by the Virginia Tech Institutional Animal Care and Use Committee (#08-217-CVM). Briefly, the geldings were fed orchardgrass hay to meet daily energy requirements for two 28-day periods; fresh fecal samples were obtained immediately after defecation on day 28 of each period. Fresh manure was placed immediately on ice and stored at −80 °C until analysis. The four fecal samples were individually homogenized with 1% (weight in volume) peptone (Sigma-Aldrich Co., St Louis, MO) (10 g feces: 90 mL of peptone) in a stomacher for 6 min to distribute bacteria throughout the sample (Price et al., 2010). Homogenized samples were centrifuged at 4000× g for 10 min at 4 °C, and pellets were retrieved for microbial DNA extraction. Microbial DNA was extracted from the homogenized fecal pellets using a manual disruption method using the ZR Soil Microbe DNA MiniPrep kit (Zymo Research, Irvine, CA) as per manufacturer's instructions (Khafipour et al., 2009; Cuiv et al., 2011). A 270- to 300-bp nucleotide sequence of the V4 region of the 16S rRNA gene was amplified with primers used by Lopez-Velasco et al. (2011) and Jesus et al. (2010). Amplicons were generated as described by Lopez-Velasco et al. (2011). Libraries were prepared, enrichments titrated, and pyrosequencing performed using a LR70 sequencing kit and 70 × 75 PicoTiterPlates (two samples per plate) performed with a Genome Sequencer FLX System (Roche, Branford, CT) by the core laboratory facility at the Virginia Bioinformatics Institute (Blacksburg, VA). The reads obtained from GS-FLX were preprocessed to identify sequencing errors and trimmed of linker sequences. Unique sequence taxonomic classification and operational taxonomic unit (OTU) assignment were performed using the Pyrosequencing pipeline of the Ribosomal Database Project (http://pyro.cme.msu.edu/) (Cole et al., 2009) software tools. Rarefaction indexes were calculated with 3% dissimilarity (http://pyro.cme.msu.edu/). OTU assignments, estimates of richness (Chao1), and diversity (Shannon index [H′]) were calculated at 3% dissimilarity. Evenness was calculated as E =H′/Hmax; Hmax = ln(Chao1) where being S the total number of species in the sample, estimated with Chao1. Relative bacterial phylum abundance was calculated based on the total number of classified reads for each sample using the rdp classifier tool (Fig. ). Matches with a rdp confidence estimate below 60% were designated as unclassified bacteria. All sequences have been deposited in the GenBank Sequence Read Archive (accession number SRA039855.1).

Figure 1

Rarefaction curves, calculated at 3%, 5%, and 10% dissimilarity, representing the observed number of OTUs within the 16S rRNA gene–based equine fecal bacterial communities in the four samples combined.

Results and discussion

Pyrosequencing of the 16S rRNA gene amplicons was used to characterize the fecal bacteria of healthy adult horses fed a controlled forage diet. Mean length of the pyrosequencing reads and the number of reads per sample were 250 bp and 28458 (range 24802–31164), respectively. Reads meeting the quality parameters (100% match over 25 bases; minimum of two reads) were trimmed. On average, 5898 unique sequences were identified from the four fecal samples. The unique sequences were classified into 1510 (range 1414–1778) different bacterial class to order-level OTUs based on 3% dissimilarity, which is greater than that reported in the rumen (Hess et al., 2011). The bacterial richness of the horse fecal microbiome presented in this study (Chao1 = 2359) is comparable to human feces (2363) (Larsen et al., 2010) but less than that reported for beef cattle feces (5725) (Shanks et al., 2011), or soil (3500) (Acosta-Martinez et al., 2008). In contrast, the bacterial richness was greater than that reported in fecal samples of pigs (114) (Lamendella et al., 2011) or the rumen of cattle (1000) (Hess et al., 2011). Rarefaction curves did not reach an asymptote at 3% dissimilarity (Fig. ); therefore, the richness of equine fecal bacteria is likely greater than that described in the present study. Fecal bacterial diversity of the horses in the present study is higher (Shannon Index = 6.7) than that found in swine (3.2) (Lamendella et al., 2011), humans (4.0) (Andersson et al., 2008; Dethlefsen et al., 2008), and cattle (4.9) (Durso et al., 2010) feces. The high-fiber nature of the horse's diet and location of the fermentation chamber likely influence this difference in bacterial diversity across species. Bacterial evenness, a measurement of how equally abundant species are in a community, indicates that the species within the horse fecal bacterial community (E = 0.9) are more evenly distributed, and not as dominated by individual taxonomic groups as in humans (E = 0.6) (Dethlefsen et al., 2008).

The majority of sequences were classified to the Bacteria domain (99%). The remainder sequences (1%) were classified to the Archaea domain; members of Archaea are commonly identified when targeting the 16S rRNA gene V4 region (Yu et al., 2008). The Methanomicrobia class, of the Euryarchaeota phylum, represented Archaea in all samples (mean 47 reads per sample). From all classified bacterial sequences, 10 phyla and 27 genera each represented at least 0.01% of total sequences (Table 2). Sequences from an additional six phyla including Acidobacteria (0–1 read per sample), Deinococcus–Thermus (0–10 reads per sample), Chloroflexi (0–6 reads per sample), Lentisphaerae (0–3 reads per sample), Planctomycetes (0–1 read per sample), and SR1 (0–1 read per sample) were not identified in all samples, suggesting that these are rare, possibly transient members of the horse fecal bacterial community. These infrequently occurring phyla, not previously described in the horse, were detected by the use of pyrosequencing owing to the ability of pyrosequencing to sequence thousands of nucleotide sequences simultaneously. It is unclear whether these bacteria are functionally important in the degradation and metabolism of grass forage in horses.

The dominant phyla in each of the four samples were Firmicutes, Proteobacteria, Verrucomicrobia, and Bacteroidetes (Table ), with the majority of bacterial sequences (43.7%) belonging to the Firmicutes phylum. Firmicutes and Bacteroidetes are the dominant phyla in equine hindgut clone library reports (Daly et al., 2001; Yamano et al., 2008); however, the abundance of Firmicutes in the present study is lower than that reported by Daly et al. (2001) (72%). The Firmicutes phylum dominates the bacterial community in pig (55%), human (56%), and beef cattle (70%) feces suggesting an ecological and functional importance of this group within the gut across species (Larsen et al., 2010; Lamendella et al., 2011; Shanks et al., 2011). The abundance of Bacteroidetes (3.7%) in the present study is much less than that reported in human (35.4%) and pig (35%) using high-throughput sequencing technologies (Larsen et al., 2010; Lamendella et al., 2011). The total percentage of Bacteroides in this report is also lower than that previously reported in horses; however, the percentage of Bacteroides has been shown to range between 12% and 49% of the total number of clones sequenced (Daly et al., 2001, 2011; Daly & Shirazi-Beechey, 2003; Yamano et al., 2008; Willing et al., 2009). These differences may be associated with source of sample, differences in diet and the sensitivity and numbers of clones examined. Daly et al. (2001, 2011) collected colonic samples from euthanized horses that grazed pasture, and some received supplemental grain. Yamano et al. (2008) collected fecal samples from horses on bamboo grass pastures. Willing et al. (2009) described a higher abundance of Bacteroidetes in horses that were fed an early cut timothy/meadow fescue haylage as compared to horses fed late cut timothy/meadow fescue and concentrate (27%). Unfortunately, a thorough nutrient analysis that documents carbohydrate content is lacking in the previous citations and thus comparisons related to the role of dietary composition are speculative. While it is likely that forage vs. concentrate feeding influences the equine gut microbial community to a greater degree than forage alone, the influence of different types of forages on this community has not been determined. In this study, the low relative abundance of Bacteroidetes may be in part due to the differences in diet; however, it is also possible that the primers used are not inclusive of all members of the phyla. Aquatic members of the Bacteroidetes phylum have been previously underrepresented by PCR primer-based methodologies (Cottrell & Kirchman, 2000).

View this table:
Table 1

Mean relative abundance of the 10 phyla present in the equine fecal samples, in the order of abundance

Unclassified bacteria38.143.84
  • The percentage of sequenced reads was calculated based on the mean total number of reads from the Bacteria domain of the four fecal bacteria libraries using the rdp classifier tool.

Garner et al. (1975) concluded that the equine bacterial community is dominated by fibrolytic bacteria by the use of culture-based techniques. Fibrobacter spp. represented 0.75% of total bacteria in the present study, which is similar (1.2%) to cecal contents as reported by Julliand et al. (1999). However, other authors who also quantified Fibrobacter spp. by the use of oligonucleotide probes reported Fibrobacter spp. abundance to be 12% in the cecum and around 4% in the colon (Lin & Stahl, 1995; Daly & Shirazi-Beechey, 2003). When quantified by the use of clone library generation, a report of Fibrobacter spp. abundance was lower (0.01%) (Daly et al., 2001). Ruminococcus spp. and Eubacterium spp., additional fibrolytic bacteria described in the horse (Julliand et al., 1999; Daly et al., 2001), represented a sum of 0.6% of total bacterial sequences (Table ). The abundance of Ruminococcus spp. in the present study is lower than that reported in the hindgut by Daly et al. (2001) and Julliand et al. (1999) (4.4%). The Ruminococcus abundance in equine cecal samples from Julliand et al. (1999) is similar to the reports in cattle feces (Dowd et al., 2008; Durso et al., 2010). Hydrogen-utilizing microorganisms work with fibrolytic bacteria to produce the volatile fatty acids, like acetate, that the host uses (Robert et al., 2001). Treponema spp., a hydrogen-utilizing acetogen, represented 1.9% of total fecal bacteria in the present study, which is similar to equine hindgut reports from Daly et al. (2001) (3%) and higher than that reported in cattle feces (0.93%) (Dowd et al., 2008). Acetogenic Treponema spp. compete with methanogens for H+, and the abundance of these two groups is inversely related in the termite gut and human oral cavity (Leadbetter et al., 1999; Lepp et al., 2004). Methane production in the horse is less than that of ruminants (Vermorel, 1997), which may be due to the higher abundance of Treponema spp.

View this table:
Table 2

Taxonomic identification of bacterial genera present at ≥ 0.01% of fecal bacterial sequences, listed in alphabetical order by phylum

ActinobacteriaAsaccharobacter spp.0.03
Denitrobacterium spp.0.05
BacteroidetesPrevotella spp.0.23
FibrobacteresFibrobacter spp.0.75
FirmicutesAcetivibrio spp.0.19
Acidaminococcus spp.0.31
Anaerosporobacter spp.0.09
Blautia spp.0.92
Coprococcus spp.0.23
Eubacterium spp.0.09
Faecalibacterium spp.0.08
Lactobacillus spp.0.36
Mogibacterium spp.0.16
Oscillibacter spp.0.43
Papillibacter spp.0.07
Pseudobutyrivibrio spp.0.23
Roseburia spp.0.27
Ruminococcus spp.0.50
Schwartzia spp.0.12
Sporobacter spp.0.22
Streptococcus spp.0.17
ProteobacteriaActinobacillus spp.0.03
Succinivibrio spp.0.16
SpirochaetesTreponema spp.1.93
TenericutesMollicutes spp.0.13
TM7TM7 Incertae sedis spp.1.82
VerrucomicrobiaSubdivision 5 Incertae sedis spp.2.81
  • The percentage of sequenced reads was calculated based on the mean total number of reads from the Bacteria domain of the four fecal bacteria libraries using the rdp classifier tool.

Thirteen genera, Actinobacillus, Asaccharobacter, Denitrobacterium, Acetivibrio, Acidaminococcus, Anaerosporobacter, Blauta, Mogibacterium, Oscillibacter, Papillibacter, Roseburia, Schwartzia, and Sporobacter (Table ), and three phyla (in addition to the infrequent phyla described above), Actinobacteria, TM7, and Cyanobacteria (Table ), that were identified in the present study have not been previously reported in the horse (Daly et al., 2001; Milinovich et al., 2008; Yamano et al., 2008). The function of the uncultivated bacterial group TM7 (Table ) in the equine gut is unknown; however, this phylum has been identified in the soil and gut of humans, mice, ruminants, and termites (Hugenholtz et al., 2001). Members of the Cyanobacteria phylum likely correspond to chlorophyll sequences from the forage diet; however, Cyanobacteria have been reported in man and mice, but their role in the equine gut is unknown (Ley et al., 2005).

Differences between prior studies and the present study may be due to the culture-independent method employed to study the microorganisms, biological effect of gastrointestinal tract region, and/or host diet. There is not a gold standard to studying complex microbial populations, and the studies reviewed here have represented a variety of techniques that produce some degree of bias owing to the preferential cloning of some sequences during 16S rRNA gene clone library generation (Daly et al., 2001; Yamano et al., 2008; Willing et al., 2009) or the use of specific probes for the identification of bacterial groups (Lin & Stahl, 1995; Daly & Shirazi-Beechey, 2003; Hastie et al., 2008). Furthermore, PCR primer-based methodologies have underrepresented equine gut bacterial members, such as fibrolytic bacteria (Daly & Shirazi-Beechey, 2003). Gut region appears to influence the abundance of equine gut microbial population (Willing et al., 2009). Feces provide a noninvasive and more humane means to study the gut bacterial community. De Fombelle et al. (2003) reported that the number of anaerobic bacterial CFUs differed between the equine hindgut and feces; however, the numbers of cellulolytic bacterial CFUs were similar between the hindgut and feces. Furthermore, Milinovich et al. (2007) used nucleic acid hybridization to provide evidence that the relative abundance of targeted groups (i.e. Streptococcus spp.) was similar in cecum and fecal samples of healthy horses. However, owing to the differences described in bacterial community along the equine gut (de Fombelle et al., 2003), future studies should evaluate gut contents to shed light on the etiology and pathogenesis of chronic diseases that plague horses.

Pyrosequencing provides a rapid and robust description of the equine fecal bacterial community; however, the present study has limitations. These limitations include use of a single region (V4) of the 16S rRNA gene for amplicon generation, generation of short sequence read lengths, inability to achieve a rarefaction asymptote at 3% dissimilarity, and presence of a large number of unclassified sequences. The V4 region of the 16S rRNA gene was targeted for the evaluation of equine fecal bacterial communities based on the ability to detect bacterial sequences (Claesson et al., 2009). Kumar et al. (2011) reported that the region of 16S rRNA gene amplification does not appear to impact the numbers of rare or abundant taxa detected; however, the relative abundance of several genera was influenced by targeted 16S rRNA gene region amplified. The abundance of Eubacterium, Prevotella, Streptococcus, and Treponema, as found in human gingiva, varied depending on the 16S rRNA gene amplified (Kumar et al., 2011). Therefore, the abundance of some groups presented here may be biased owing to primer selectivity. In this study, we did detect groups, TM7, using the V4 region primers that were not detected with the use of V4–V6 primers by Kumar et al. (2011). Future studies should use two primer sets spanning different regions of the 16S rRNA genes. The sequence read length was limited by the primers utilized; however, the chosen primers have been used previously in bacterial community pyrosequencing studies (Wang et al., 2007; Lopez-Velasco et al., 2011). Furthermore, increasing the specificity by targeting the 16S rRNA gene V4 region helps to overcome the limitations of read length (Nossa et al., 2010). Another source of bias in the present study is DNA extraction technique used; however, Cuiv et al. (2011) reported that beading-based extraction is superior to Gram-positive (i.e. Firmicutes members) lysis. These limitations along with the presence of a large proportion of previously uncultivated microorganisms in the horse feces inhibit complete exposure of the true richness and diversity of the equine fecal bacterial community. In this study, 38% of bacterial sequences were unclassified (Table ), representing novel bacterial sequences that have not been placed into a recognized taxonomic classification within the rdp database (www.cme.msu.edu) when a 60% similarity cutoff was used. The high abundance of unclassified sequences has been reported in prior human (Eckburg et al., 2005; Gill et al., 2006) and horse (Daly et al., 2001) studies.

This difference between the equine fecal bacterial community and bacterial communities of other environments (i.e. human feces, rumen feces, and soil) may be due to substrate concentration and availability. The horse's diet is markedly different than that of humans (i.e. high fiber and reduced fat, protein, and digestible carbohydrates), and the bacterial environment is different between the hindgut, rumen, and soil. Data presented here provide further insight into the hindgut bacterial community.


This study was funded by the Virginia Bioinformatics Institute/Fralin Life Science Institute Core Resources/Equipment Exploratory Grant. The authors wish to acknowledge the assistance of Dr Gabriela Lopez-Velasco who assisted with the completion of the study.


  • Editor: Rustam Aminov


View Abstract