Diversity of the Gastric Microbiota in Thoroughbred Racehorses Having Gastric Ulcer

Abstract

Equine gastric ulcer syndrome is one of the most frequently reported diseases in thoroughbred racehorses. Although several risk factors for the development of gastric ulcers have been widely studied, investigation of microbiological factors has been limited. In this study, the presence of Helicobacter spp. and the gastric microbial communities of thoroughbred racehorses having mild to severe gastric ulcers were investigated. Although Helicobacter spp. were not detected using culture and PCR techniques from 52 gastric biopsies and 52 fecal samples, the genomic sequences of H. pylori and H. ganmani were detected using nextgeneration sequencing techniques from 2 out of 10 representative gastric samples. The gastric microbiota of horses was mainly composed of Firmicutes (50.0%), Proteobacteria (18.7%), Bacteroidetes (14.4%), and Actinobacteria (9.7%), but the proportion of each phylum varied among samples. There was no major difference in microbial composition among samples having mild to severe gastric ulcers. Using phylogenetic analysis, three distinct clusters were observed, and one cluster differed from the other two clusters in the frequency of feeding, amount of water consumption, and type of bedding. To the best of our knowledge, this is the first study to investigate the gastric microbiota of thoroughbred racehorses having gastric ulcer and to evaluate the microbial diversity in relation to the severity of gastric ulcer and management factors. This study is important for further exploration of the gastric microbiota in racehorses and is ultimately applicable to improving animal and human health.

Introduction

Equine gastric ulcer syndrome (EGUS) is one of the most commonly reported diseases in horses and is characterized by ulceration of the mucosa of the esophagus, stomach, or duodenum [3]. Several factors, including dietary pattern, stress, and excessive exercise, are associated with the high prevalence of EGUS in racehorses [12,23]. As the horse is a monogastric animal, its gastrointestinal tract is suited to continual grazing of high-fiber and low-starch material. However, racehorses are usually fed a low-fiber and highstarch diet intermittently to improve racing performance, resulting in a more acidic environment in the gastrointestinal tract, and thereby leading to metabolic diseases such as gastric ulcer, colic, or laminitis [1,15]. Clear differences in the fecal microbial communities between healthy and unhealthy horses were observed in several studies, suggesting that microbiome imbalance may contribute to disease [10,31]. However, information regarding equine gastric microbial communities is limited [9,28]. Thus, the microbiological impact on EGUS remains unclear.

Since Helicobacter pylori was first cultured from the human stomach in 1982, a strong association between H. pylori and gastric diseases, including gastritis, peptic ulcer disease, and gastric adenocarcinoma, has been well characterized in humans [17,22]. After recognition of H. pylori as a pathogenic bacterium, there was increased interest in determining whether Helicobacter spp. could colonize the gastric tissue or were associated with the development of gastric disease. To date, more than 20 Helicobacter spp. have been identified in the stomach and intestines of humans and animals (http://www.bacterio.net/helicobacter.html), and some have been found to be pathogenic [13]. Although there have been several attempts to culture Helicobacter spp. from the horse stomach, Helicobacter has not been cultured to date, yet the possibility of the presence of Helicobacter in the horse stomach remains [5,6,29]. In 2007, however, H. equorum was first isolated from horse feces [24]. Although the likelihood that H. equorum is involved in the development of disease appears to be low [26], the potential pathogenicity of H. equorum in the gastrointestinal tract cannot be dismissed.

The use of high-throughput sequencing-based molecular techniques, namely, next-generation sequencing (NGS), provides more extensive information regarding the microbial ecosystem and is not limited to culturable bacteria, enabling multilateral approaches for understanding the microbial community in diseased and healthy states of the host. Several studies have utilized NGS techniques to characterize the microbiota of various organs such as the oral cavity, esophagus, stomach, intestine, or vagina in humans, revealing not only bacterial species that are difficult to culture and the core microbiota, but also the relationship between microbiota and health status [2,7]. However, studies involving the horse microbiota have been limited. Previously, fecal bacterial communities were characterized in horses, demonstrating that there were profound differences in the microbiota between healthy and unhealthy horses [10,31]. The gastric bacterial community has been investigated in healthy non-racehorses, revealing that the horse gastric mucosa has a diverse and abundant microbiota [28]. Because the health status of the stomach in racehorses is closely influenced by the racing performance, it is necessary to investigate the gastric microbiota just after racing, when the stress and gastrointestinal lesions may become most severe [12,27]. The objective of the present study was to investigate the prevalence of Helicobacter spp. in thoroughbred racehorses from a main racing track located in Gyeonggi-do, Korea and to characterize the gastric microbiota of thoroughbred racehorses with mild to severe gastric ulcer using NGS. The microbial diversity of the microbial communities was analyzed in association with the severity of gastric ulcers, and the relationship between the microbial community and possible factors that may influence relative abundance was investigated.

 

Materials and Methods

Horse Recruitment for Endoscopic Examination

A gastroscopy was performed on 52 thoroughbred racehorses within 2 days after racing at the Korean Racing Authority during August 2013 to October 2013. The horses were voluntarily recruited through an announcement at the Seoul Racecourse Trainer’s Association, and the horses were selected by order of arrival. The horses were registered as patients of a veterinary medical teaching hospital located in the racing track and written consent was obtained from the owners and/or managers of the horses before gastroscopy. The horses were fasted 12-18 h for feed and 4 h for drinking and sedated with intravenous detomidine HCl (20 μg/kg; Equadin, Dong Bang Co., Ltd., Korea) before conducting gastroscopy; they underwent gastroscopy with a Karl Storz endoscope (Karl Storz, Germany). Macroscopic evaluation of gastric ulcers was performed by three individual veterinarians, and the severity was graded by scoring from 0 to 4 according to the standard grading system [32]. Each grade from the three veterinarians was then summed to categorize the severity of gastric ulcer into four groups; healthy (sum grade 0-1), mild (sum grade 2-4), moderate (sum grade 5-8), and severe (sum grade 9-12). Meanwhile, horse information was obtained from the racehorse trainer and/or manager regarding the animal’s age, gender, and management factors, including dietary patterns (type, frequency, and amount of feed and water), type of bedding, and past records of drug administration within 2 weeks before sampling.

Sample Collection

Two biopsy samples were obtained from the pyloric antrum of the glandular region of each horse to identify the presence of Helicobacter spp. by the culture method and to analyze the bacterial community in the equine gastric mucosa. One biopsy sample was placed in 100 μl of Brucella Broth (Oxoid, UK) for culture, and the other biopsy sample was placed in a sterile tube for DNA extraction. Fresh fecal samples were also obtained from the horses by rectal palpation on the day gastroscopy was performed. The samples were maintained at 4℃ and were directly transported to the laboratory. DNA extraction and the Helicobacter isolation were performed immediately upon arrival of the samples to the laboratory.

Isolation of Helicobacter spp.

The biopsied tissue (n = 52) was mechanically homogenized with 100 μl of Brucella broth and the homogenate was spread on Columbia agar (Columbia agar base (Oxoid, UK) containing 7% horse serum (Gibco, CA, USA), and antibiotic supplement (SR0147E, Oxoid, UK)). For the fecal sample (n = 52), approximately 1 g of each sample was homogenized with 1 ml of Brucella broth, and subsequently 50 μl of homogenate was then streaked onto Columbia agar. The plates were incubated at 37℃ for 5 days under microaerobic conditions, and if no colonies appeared, the plates were incubated for an additional 5 days.

Colonies that appeared on the Columbia agar were screened for the presence of Helicobacter spp. using PCR targeting the Helicobacter-genus specific 16S rRNA, H. pylori-specific ureC, and H. equorum-specific 23S rRNA gene, as described in the PCR section below. Any positive samples on the screening test were further examined using biochemical tests (urease test, oxidase test, and Gram staining) for confirmation.

PCR-Based Screening Assay for the Identification of Helicobacter spp.

To screen for the presence of Helicobacter spp., DNA was extracted from randomly swiped areas of heavy bacterial growth by the boiling method. Briefly, the cells were suspended in 100 μl of distilled water, boiled for 10 min, chilled on ice for 3 min, and centrifuged at 16,000 ×g for 3 min. The supernatant was used as a template for the PCR assay. The PCR was performed as described previously, targeting 16S rRNA for Helicobacter spp. detection (420 bp) [14], the urease gene for H. pylori detection (294 bp) [20], and 23S rRNA for H. equorum detection (1,074 bp) [25] (Table 1).

Table 1.Primer sets used for the PCR assay.

Pyrosequencing

Metagenomic DNA from gastric tissue was extracted from 52 biopsy samples using the FastDNA Spin extraction kit (MP Biomedicals, CA, USA) according to the manufacturer’s instructions, with a minor modification. Vigorous vortexing for 15 min was used to replace a bead beater for cell lysis. From the 52 metagenomic DNA samples, two or three qualified DNA samples were randomly selected from each severity group (healthy, mild, moderate, and severe) of gastric ulcer to select representative samples. The DNA samples showing a final concentration of less than 40 ng/μl were excluded from the analysis. For amplification of a fragment spanning the V1-V3 regions of the 16S rRNA gene, PCR was performed in a total volume of 50 μl, containing 1× buffer, 2.0 mM MgCl2, 2.5 mM deoxynucleoside triphosphate (dNTP), 20 pmol/μl of each forward and reverse barcoded fusion primer [18] (http://www.ezbiocloud.net/resource/M1001), 5 U/μl of Amplitaq Gold, and 1 μl of DNA polymerase. The mixture was then amplified with the following conditions: 5 min at 94℃ for initial denaturation; 30 cycles of 30 sec at 94℃ for denaturation, 30 sec at 55℃ for annealing, and 30 sec at 72℃ for extension; with a final extension for 7 min at 72℃ followed by a hold at 4℃. The 16S rRNA V1-V3 amplicons were then purified using a PCR purification kit (Qiagen, CA, USA) and quantified using the PicoGreen dsDNA Assay kit (Invitrogen, CA, USA). Equal concentrations of the purified amplicons were pooled and sequenced using a 454 Junior system (Roche, CT, USA) according to the manufacturer’s instructions. The sequencing data were deposited in the NCBI Short Read Archive (SRA) database under Accession No. SRP066932.

Data Analysis

Raw sequence files were filtered by demultiplexing, removing short and low-quality reads (read length <300 bp or average quality score <25), and separating barcoded primers and non-16S reads. The representative sequences were then taxonomically assigned by pairwise sequence alignment and BLAST searched using the Eztaxon-e database (http://eztaxon-e.ezbiocloud.net/). After removing chimeric sequences using the UCHIME algorithm, microbial community analysis was performed using the CLcommunity software (ChunLab, Inc., Korea).

The species richness and diversity index were calculated using the Chao 1 estimation and the non-parametric (NP) Shannon diversity index, respectively. Operational taxonomic units (OTUs) were calculated using three different methods: cluster database at high identity with tolerance (CD-HIT), taxonomy-based clustering (TBC), and taxonomy-dependent clustering-taxonomy-based clustering (TDC-TBC). To compare multiple microbial communities, the taxonomic composition (identified taxa and their relative abundance) of each horse was determined at each level of biological classification (phylum to species). The core microbiota was investigated by examining the taxa that were present and shared in all samples. A heatmap was drawn to illustrate the proportion of each taxon, with a cutoff value of 5%, using the CLcommunity software.

To analyze the divergence of OTUs among the microbial communities, Fast UniFrac and canonical correlation analysis (CCA) was performed using the CLcommunity software. An unweighted pair group method with arithmetic mean analysis (UPGMA) dendrogram was generated via hierarchical clustering, and principal coordinate (PCo) analysis was performed to generate an ordination diagram. To investigate the relationship between management factors and the microbial community, CCA was performed on the factors that were found to be significantly associated with the microbial community using the Kruskal-Wallis test.

Statistical Analysis

The relative abundance of each taxon was compared between two groups of horses: group 1 (healthy horses and horses with mild gastric ulcer) and group 2 (horses with moderate to severe gastric ulcer). The difference in relative abundance of each taxon was considered significant when the p-value was less than 0.05 (two-tailed Mann-Whitney-U test) using the SPSS statistical software ver. 22.0 (SPSS IBM, NY, USA). The associations between the microbial community and each variable were analyzed by the Kruskal-Wallis test using the SAS 9.3 system (SAS Institute Inc., NC, USA). Small sample sizes in each group were adjusted by using the exact option provided in the SAS system. The factors were considered to be significantly associated when the p-value was less than 0.05.

 

Results

Isolation of Helicobacter spp.

Helicobacter spp. was not detected in 52 tissue samples from gastric biopsies and 52 fecal samples using a combination of selective culture, biochemical, and PCR techniques.

Species Richness and Diversity of Gastric Microbiota

The microbial communities of the gastric tissue (glandular region) of 10 thoroughbred racehorses were analyzed to determine the microbial composition of horses with mild to severe gastric ulcers (Table 2). A total of 121,200 sequences (mean: 12,120 per horse; range: 6,360-18,941) were obtained from pyrosequencing of the V1-V3 regions of the 16S rRNA gene. After discarding the unmatched, low-quality, and chimeric sequences, 4,127 to 18,807 reads per horse (mean: 11,256) were taxonomically assigned (97% similarity) (Table 3). Rarefaction curves were generated by plotting the number of reads against the number of OTUs after normalization with 4,127 reads (the minimum number of reads across the samples) (Fig. 1).

Table 2.All horses were thoroughbred racing horses. aThree individual veterinarians evaluated the severity of gastric ulcer by scoring from 0 to 4 according to the standard grading system. The sum grade was calculated by adding each grading score from the three veterinarians, and the severity of gastric ulcer was categorized as follows; healthy (sum grade 0-1), mild (sum grade 2-4), moderate (sum grade 5-8), and severe (sum grade 9-12). bDietary pattern is described as the following; type of feed (C: concentrated feed; R: roughage; W: water), no. of feeding per day, total amount of feeding. cThe type and number of NSAID administration within 2 weeks before sampling was described. The type, route, dose, time, and duration (consecutive days) of antibiotic administration were collected within 30 days before sampling (G: gentamicin; IV: intravenous; sid: once per day).

Table 3.Data summary of diversity estimates for each horse using three different calculation methods.

Fig. 1.Rarefaction curves for gastric microbial communities of 10 racehorses. The curves were generated after normalization with 4,127 reads, which was the minimal number across the 10 samples.

The average composition of the microbial communities is shown in Fig. 2A. A total of 27 phyla, 63 classes, 147 orders, 329 families, 898 genera, and 1,853 species were identified. Among the 27 phyla identified, four phyla, namely, Firmicutes (mean: 50.0%), Proteobacteria (mean: 18.7%), Bacteroidetes (mean: 14.4%), and Actinobacteria (mean: 9.7%), were present in all horses, which ranged from 78.0% to 99.3% (mean: 92.7%), but the proportion of each phylum varied among the samples (Fig. 2B). Whereas Proteobacteria was the predominant phylum in samples 20130029 and 20130052, Firmicutes was the predominant phylum in the rest of the samples. Twelve classes, 19 orders, 25 families, 20 genera, and 15 species composed a core microbiota that was present in the gastric tissues of all horses (Table S1).

Fig. 2.The compositions of microbial communities from the gastric tissues of racehorses. (A) Double pie chart showing the average composition of gastric microbial communities of 10 racehorses. The inner circle indicates the phylum composition and the outer circle indicates the genus composition of the bacterial communities. (B) Bar chart showing the microbial distribution (at the phylum level) of 10 horses. The four most dominant phyla were Firmicutes, Bacteroidetes, Proteobacteria, and Actinobacteria, which accounted for 78.0-99.3% of all phyla.

Microbial Communities of Horse Stomachs with Mild to Severe Gastric Ulcers

The microbial compositions were compared according to the severity of gastric ulcers, which were organized into two groups; group 1 consisted of healthy and mild (n = 2 and n = 3, respectively) and group 2 consisted of moderate and severe (n = 3 and n = 2, respectively). In horses in group 1, the top four most prevalent phyla were Firmicutes (43.6%), Proteobacteria (28.1%), Bacteroidetes (11.8%), and Actinobacteria (10.8%). The microbial composition of horse stomachs in group 2 was dominated with Firmicutes (56.3%), Proteobacteria (9.2%), Bacteroidetes (17.0%), and Actinobacteria (8.6%). The average composition of the gastric microbial communities at the phylum and genus levels is presented in Fig. S1. Two genera, Clostridium_g19 and Staphylococcaceae_uc, were only present in group 2, whereas one OTU, Dietzia cinnamea, was only present in group 1.

To investigate the influence of Helicobacter species on gastric ulcer development, OTUs that matched with known Helicobacter species sequences were investigated. Two OTUs were taxonomically assigned to a Helicobacter spp., including H. pylori and H. ganmani, from horses 20130018 and 20130035, respectively.

Phylogenetic Analysis

Three distinct clusters of bacterial communities of the EGUS horse stomach were generated using the UPGMA method based on their taxonomic composition: Cluster 1 (20130013, 20130018, 20130037, 20130042, and 20130049), Cluster 2 (20130035, 20130043, and 20130048), and Cluster 3 (20130029 and 20130052) (Fig. 3A). These patterns were similar to the results of PCo analysis, which produced three distinct groups: Group A (20130013, 20130018, 20130037, 20130042, and 20130049), Group B (20130035, 20130043, and 20130048), and Group C (20130029 and 20130052) (Fig. 3B). In Clusters 1 and 2, Firmicutes (61.7%, 54.8%), Bacteroidetes (8.5%, 21.7%), Actinobacteria (14.0%, 4.1%), and Proteobacteria (12.3%, 5.6%) were the predominant phyla, respectively, whereas in Cluster 3, Proteobacteria (53.9%) was predominant, followed by Bacteroidetes (18.2%), Firmicutes (13.3%), and Actinobacteria (7.6%) (Fig. 4A). Each cluster featured distinct patterns: Cluster 1 had a higher level of the class Bacilli, order Lactobacillales, and family Streptococcaceae; Cluster 2 had a higher level of the classes Clostridia and Bacteroidia, orders Clostridiales and Bacteroidales, and families Ruminococcaceae and Lachnospiraceae; and Cluster 3 had a higher level of the class γ-Proteobacteria and order Pseudomonadales (Fig. 4B).

Fig. 3.Similarity of gastric microbial communities. (A) Three distinct clusters were generated with the UPGMA method. (B) PCo analysis for comparing microbial communities. The distance between samples represents the dissimilarity between microbial communities of samples. Cluster 1 (Group A): 20130013, 20130018, 20130037, 20130042, and 20130049; Cluster 2 (Group B): 20130035, 20130043, and 20130048; and Cluster 3 (Group C): 20130029 and 20130052.

Fig. 4.Gastric microbial composition of each cluster. (A) The bar chart shows the average microbial composition of each cluster at the phylum level. (B) The heat map shows the proportion of taxa in class, order, and family levels (displayed with a cutoff value of 5% of each taxon).

Relationships between OTU Divergence and Management Factors

Among the management factors, the frequency of concentrated feed (p = 0.031) and roughage (p = 0.022), amount of water consumption (p = 0.022), and type of bedding (p = 0.022) were found to be significantly associated with gastric microbial communities, whereas the severity of gastric ulcer and antibiotic administration were not significantly related to the variation in microbial communities. Regarding dietary patterns, the horses belonging to Cluster 3 (Group C) were fed a concentrated feed and roughage four times a day with 20-30 L of water, whereas other horses were fed two or three times a day with more than 40 L of water. In addition, in contrast to the use of sawdust for the bedding in horses of Cluster 3 (Group C), pellets or a mixture of pellets and sawdust were used for the rest of the horses (Table 2). The CCA results showed that the frequency of concentrated feed per day and the amount of water consumption had the greatest influence on the gastric microbial community (Fig. S2).

 

Discussion

In this study, to elucidate the impact of the microbiological community on EGUS, the presence of Helicobacter spp., including H. pylori, a well-known microbiological risk factor for gastric ulcers in humans, was investigated. Furthermore, the microbial communities of thoroughbred racehorses were characterized using NGS and evaluated for the association between the diversity of the microbial communities and the severity of gastric ulcers. Finally, the relationship between the diversity of the microbial communities and horse management factors was evaluated.

Several contradictory results have been reported regarding the presence of Helicobacter species in horse stomachs. Several groups have reported the detection of genus-specific [8,29] or H. pylori-specific genes [6,16] in various regions of the horse stomach; yet, no Helicobacter species have been isolated from the horse stomach until now. On the other hand, Helicobacter was not found using FISH [19] or NGS [28] techniques. In this study, the presence of Helicobacter spp. was investigated using culture, PCR, and NGS techniques. Although Helicobacter spp. were not detected using the culture and PCR techniques from 52 gastric biopsies and 52 fecal samples, a few sequences that aligned with H. pylori (3/12,111 reads) and H. ganmani (1/12,683 reads) were detected using the NGS techniques from 2 out of 10 horses having mild and severe gastric ulcers, respectively. A potential reason for not detecting Helicobacter by culture and PCR may be due to the presence of dead or injured Helicobacter species in the sample. Considering that only four reads matched to known sequences of Helicobacter spp., the overall abundance of Helicobacter spp. appears to be very low, which may not be detectable by culture or PCR. Although we were unable to isolate Helicobacter species, detection of sequences of H. pylori and H. ganmani could be evidence of the existence of Helicobacter spp. in the horse stomach. Further study may be needed to investigate the presence of Helicobacter species. In addition, further development of culture or PCR techniques and the use of a high-throughput NGS technique such as Miseq may clarify these results, and could be helpful to elucidate the role of Helicobacter in the horse stomach, where gastric ulcers are very common.

In this study, the microbial communities in the stomachs of 10 racehorses were analyzed using 16S rRNA gene pyrosequencing. In horses, gastric ulceration is known to commonly affect squamous mucosa, especially in margo plicatus, which results from a low pH environment associated with stress, excessive exercise, and consumption of highly concentrated feed [4,33]. In this study, however, microbial communities of the pyloric antrum of the glandular region were investigated in order to focus on microbial factors specifically, including Helicobacter spp., which colonize the glandular region [30]. The α-diversity was calculated using three different methods, but the trends of α-diversity were similar in individual horses regardless of the analytical methods used. The core microbiota (i.e., taxa that were shared by all horses) included 4/27 phyla, 12/63 classes, and 19/147 orders, but these proportions ranged from 75.0% to 99.3% of the total microbial composition. This result indicates that a relatively small number of taxa contributed to more than 75% of the gastric microbiota in the horse stomach.

In the horse stomach, Firmicutes, Proteobacteria, Bacteroidetes, and Actinobacteria were found to be the predominant phyla, accounting for a mean of 92.7% (range: 78.0-99.3%) of all operational taxonomic units. Generally, Firmicutes, Proteobacteria, and Bacteroidetes are known to be the dominant phyla in the mammalian gastrointestinal tract [21,28]. However, their proportions varied among species and even between sampling sites within the same subject [2]. There are limited data regarding the horse gastric microbiota. In a previous study, it was found that the gastric microbial community of the horse stomach was subdivided into two clusters: cluster 1 had a higher proportion of Firmicutes and cluster 2 had a higher proportion of Proteobacteria [28]. Moreover, it was found that the dominant phylum differed according to sampling method (postmortem vs. biopsy) and type of management (pastured vs. non-pastured), but not by anatomical site (glandular, non-glandular, or ulcerated regions) [28]. A previous study also investigated the gastric microbial community from horses kept on various diets [9]. Although Firmicutes was identified as the most predominant phylum in the glandular mucosa from euthanized horses, this study only calculated the average relative abundance of the bacterial community, and did not consider the effect of variations in the feeding material or other factors. We similarly found that the predominant phylum was either Firmicutes or Proteobacteria. However, our samples were collected from the glandular region by biopsy, but the microbial composition varied markedly among horses, indicating that the microbial composition may be more affected by health status, management, or other factors than sampling method.

Changes in the gastric microbial community may be related to the health status of individual horses. Differences in fecal microbial communities were reported between healthy and diseased horses [10,31]. The gastric microbial compositions were compared based on the severity of gastric ulcers. Although there were no significant differences observed in the microbial community at the phylum level, the genus Actinobacteria, which belongs to the class Gammaproteobacteria, order Pseudomonadales, family Moraxellaceae, was found at a much higher frequency in the horses of group 1 than in group 2 (8.7% vs. 0.61%; Fig. S1). Several taxa were found to show a different relative abundance according to the severity of gastric ulcer, but their overall proportions were relatively small, which may be due to the small sample size tested in this study. Although some OTUs were only present in horses either having mild or severe gastric ulcer, their influence on gastric ulcer severity seemed to be very low owing to the small population size. In addition, although the difference in the predominant phylum due to the presence of H. pylori has been reported for the human stomach [2], there were no significant differences in the bacterial composition of gastric tissues regardless of the existence of OTUs that matched to the sequences of Helicobacter species. Considering that only four reads were assigned to Helicobacter species, the abundance of these OTUs in the samples may be too low to substantially affect the microbial community.

To investigate the individual and management factors that might be related to microbial diversity, management information was analyzed in conjunction with the clusters of microbial communities generated by UPGMA and PCo analysis. Kruskal-Wallis analysis showed that the frequency of feeding, total amount of water consumption per day, and type of bedding appeared to be linked to the composition of the gastric microbiota, but individual factors such as gender or age, severity of gastric ulcer, and antibiotic administration within 30 days did not appear to have an effect. The CCA results indicated that the frequency of concentrated feed and the amount of water consumption had a greater influence on the gastric microbiota compared with the other factors. Several groups have reported that the consumption of highly concentrated feed generates gastric fermentation metabolites, such as volatile fatty acids and lactic acids, which give rise to an increase in lactic acid bacteria (LAB), especially Streptococcus spp. and Lactobacillus spp. [11]. In this study, the proportions of the phylum Firmicutes, order Lactobacillales, and family Streptococcaceae, which includes LABs, were higher in Cluster 1 than the other clusters (Fig. 4B). Cluster 1 tended to include horses fed with concentrated feed two times per day and Cluster 3 included horses fed four times per day, which is consistent with the previous knowledge that consumption of highly concentrated feed increases the proportion of LABs. Clusters 1 and 2 also differed from Cluster 3 by the type of bedding and amount of water consumption, but how these factors affect the microbial community is unclear. Horses may accidentally or purposely eat the bedding material, and thus the bedding material may have a direct impact on the microbiota of the horse stomach. The total content of microbiota in sawdust has been reported to be 2.1 × 105 CFU/m3, which is 7.6-17.6 times lower than straw or peat [3], but no study has reported the microbial content of pellet bedding. It was assumed that pellet bedding had a smaller influence than sawdust on the microbiota of the horse stomach owing to its size and coating. In addition, to the best of our knowledge, no studies have been conducted to examine the change in microbiota with the amount of water consumption. Perhaps there might be other confounding factors, which were not accounted for in this study, which could be associated with the alterations in the composition of the microbial communities. In this study, several variables were analyzed using non-parametric statistical analysis because of the small sample size, which somewhat limits the interpretation for generalizing these results owing to the low statistical power and potential for sampling bias. Therefore, further study may be needed to investigate the factors affecting the microbial community with a larger number of subjects in order to reduce the influence of sampling error.

To the best of our knowledge, this is the first study to investigate the gastric microbiota of thoroughbred racehorses having gastric ulcer and to evaluate the microbial diversity in relation to the severity of gastric ulcers and management factors. Although Helicobacter species were not detected with the culture and PCR methods, a few sequences generated by the NGS technique matched those of H. pylori and H. ganmani, providing preliminary evidence of the presence of Helicobacter spp. in the horse stomach. Furthermore, some variables, including frequency of feeding, amount of water consumption, and type of bedding material, were found to be related to the microbial community; however, owing to the small sample size in each group, the statistical power of this analysis was limited. Therefore, further study may be needed with a larger group of horses to investigate the variables that may affect the gastric microbial community. Nevertheless, this study provides important baseline information and suggests potential candidate factors to motivate further exploration of the gastric microbiota in racehorses, which should ultimately help to develop suitable management strategies for improving animal and human health.