The prevalence of EGUS in domestic horses is high.1–3 Multiple factors are associated with or have been proposed to contribute to formation of gastric ulcers; these include changes in the production of stomach acid, alterations in the exposure of stomach mucosa to stomach acid, impairment of gastric mucosal blood flow, and bacterial colonization.3,4 The PPI omeprazole is available in a proprietary oral paste formulation, which is the current treatment and preventative of choice owing to its efficacy in reducing acid secretion and ulceration scores, convenient daily dosing, and high margin of safety.5–7 In addition to reducing physical damage to the mucosa, reduction of acidity in the stomach has been proposed to alter gastrointestinal microbiota.8–11 Alterations in gastric microbiota, including decreases in Helicobacter spp, are additional positive effects of PPIs in people and dogs.9,12 However, changes in gastrointestinal microbiota may also contribute to the association between PPI use and increased incidence of gastrointestinal, respiratory, and hematogenous infectious complications in human patients.8,13,14 One study15 of hospitalized foals found an increased incidence of diarrhea and sepsis in association with the use of antiulcer medication, suggesting the potential for infectious complications or dysbiosis as a result of PPI use in horses as well.
The microbiome of the gastrointestinal tract in healthy horses16,17 and microbiota changes associated with management, antimicrobial treatment, and anesthesia of horses18–20 have been previously described. However, to the authors’ knowledge, the potential effects of omeprazole on equine gastrointestinal microbiota have not been investigated. Given the microbiota changes reported following PPI administration in other species and the potential therapeutic and adverse effects associated with these changes, it is important to address this gap in knowledge regarding equine health. The objective of the study reported here was to determine the effects of administration of omeprazole on the fecal and gastric microbiota of healthy adult horses. We hypothesized that oral administration of commercially available omeprazole paste at a dosage of 4 mg/kg, PO, for 28 days would result in a decrease in diversity and shift in composition of the gastric and fecal microbiota of horses during the administration period and that the changes in gastric and fecal microbiota diversity and composition would be reversed 28 days after omeprazole paste administration was discontinued.
Materials and Methods
Animals
Twelve horses owned by the University of Georgia Veterinary Teaching Hospital were enrolled in the study. The study population consisted of 8 mares, 3 geldings, and 1 stallion (median age, 13 years; range, 2 to 21 years); 6 were American Quarter Horses, and 4 were Thoroughbreds. The remaining 2 animals were a warmblood-type horse and a mixed-breed pony. All enrolled horses were deemed healthy on the basis of physical examination results, and none had any history of gastrointestinal disease within the previous 3 months or had received any antimicrobial drugs or NSAIDs within the 3 months prior to the beginning of the study. The study protocol outlining animal housing, handling, and procedures was reviewed and approved in accordance with ethics guidelines of the Institutional Animal Care and Use Committee of the University of Georgia.
Experimental design and sample collection
Horses were randomly allocated into 2 groups (omeprazole and control) of 6 horses each through use of a statistical software program.a An a priori power calculationb determined that a sample size of 6 would be needed for power (1 - β) of 0.80 with an α level of 0.05 for detecting a 25% change in OTU counts, assuming a normal distribution with a mean ± SD OTU count of 2,886 ± 391/sample.21
Horses were managed in 1 of 3 housing environments throughout the study. The type and site of housing was routine for all horses and was determined according to herd dynamics and the roles outside of the study that each horse served as part of the teaching hospital and academic environment. Housing environments of the study horses included grass pasture (n = 8 [4 each of the control and omeprazole groups]), dirt paddock (2 [1/group]), and a combination of stall confinement and dirt paddock access with high-quality grass hay ad libitum from the same source (2 [1/group]). The housing sites were located at 3 university-owned facilities (2 veterinary teaching hospital facilities and 1 research farm) < 20 km apart. Horses housed on grass pasture or in dirt paddocks were at the same facility, and the pastures and paddocks were on similar soil. Each horse had ad libitum access to tap water. Signalment, housing environment, and study group information for each horse used in the study were summarized (Supplementary Table SI, available at avmajournals.avma.org/doi/suppl/10.2460/ajvr.80.1.79).
From day 0 through day 27, omeprazole group horses received commercially available omeprazole pastec (4 mg/kg, PO, q 24 h). Dosages were determined on the basis of approximate body weight for each horse. Control group horses received a sham treatment (tap water [20 mL, PO, q 24 h]) delivered by syringe during the same period. All horses were visually examined for any changes in attitude, appetite, feces, or signs of colic daily at the time of treatment administration.
On days 0, 7, 28, 35, and 56, fecal and gastric fluid samples were collected. On days 0 and 7, sample collection was performed prior to administration of omeprazole or the sham treatment. Each horse was sedated with xylazine (0.3 to 0.9 mg/kg, IV) and restrained with a lip twitch for sample collection. A nasogastric tube was passed into the stomach and a sample of gastric fluid was collected into a 50-mL conical tube by creating a siphon and lavaging the stomach with approximately 2 L of tap water. A fecal sample was collected from the rectum or a sample was collected from a fecal pile immediately following observed natural passage of manure. Each manure sample (approx half of a fecal ball) was secured in a small plastic bag. Fecal and gastric fluid samples were stored in an ice cooler at 4°C following collection and during transit to the clinical research laboratory at the veterinary teaching hospital. Samples were stored at −80°C (within 4 hours after collection) until batched DNA extraction.
DNA extraction, PCR assay, and sequencing
A commercially available kitd was used to extract DNA from each fecal and gastric fluid sample according to the manufacturer's protocol with slight modifications as previously described.22 Briefly, the contents of each sample were thawed at room temperature ≤ 1 hour until 200 mg of feces or gastric content pellet could be separated from the frozen sample. This aliquot was then placed in a 2-mL microcentrifuge tube that contained 1 mL of a buffer solution designed to remove PCR inhibitors from DNA samplese and 50 mg of 0.1- and 0.5-mm silica zirconium beads. Each sample was homogenizedf at 3,500 oscillations/min for 3 spins of 45 seconds’ duration. Samples were then heated to 70°C for 10 minutes before following the manufacturer's protocol for the remainder of the extraction. Following extraction, all samples were stored at −80°C until analysis.
The PCR assay and sequencing of hypervariable region 4 of the 16S rRNA gene were performed as previously described.23,24 A DNA library was prepared according to a DNA library preparation protocol.25 Sequencing was performed with a next-generation sequencing systemg and consisted of 250-bp paired-end reads. The resulting output files were demultiplexed, merged, and used for downstream analysis.
Sequence and statistical analyses
Data processing and bioinformatics analysis were performed by use of an open-source bioinformatics pipeline.24,h The raw sequence data were demultiplexed and low-quality reads were filtered according to the default settings. Chimeric sequences were detected with a computational algorithm26,i and removed prior to further analysis. Sequences were then assigned to OTUs through an open-reference protocolh that selected OTUs against a reference database27,j filtered at 97% identity. All reads were retained for OTU assignment other than those that were of poor quality, contained chimeras, or were multimapping. To adjust for uneven sequencing depth among the samples, the results for each sample were rarefied to an even sequencing depth (selected as the number of reads present in the sample with the second smallest number of reads) prior to further analysis. These data were filtered to remove low-abundance reads, eliminating all reads present in < 10 horses and those identified < 50 times.
The α rarefaction, β diversity measures, Good coverage estimates, taxonomic summaries, and tests for significance were calculated and plotted with the previously described bioinformatics software.h The β diversity differences were determined by ANOSIM,28,h with an R statistic close to 1.0 interpreted as dissimilarity in β diversity and a value closer to 0 interpreted as similar β diversity among groups. Differences in α diversity were determined by exporting the data from the bioinformatics softwareh for analysis in an open-source statistical computing environmentk by linear and nonlinear mixed-effectsl modeling. Diversity outcomes were modeled as a function of the fixed effects of day, treatment, and sample type (ie, gastric contents or feces), and horse was treated as a random effect with an exchangeable correlation structure; all possible bivariate interaction terms were examined for significance. Model fit was assessed on the basis of Akaike information criterion values and by inspection of residual plots. For all statistical comparisons, values of P < 0.05 were considered significant.
The OTU data table was exported for further analysis and comparisons between treatment groups and over time,29,30,m–r including generation of PCoA plots on the basis of the Unifrac distance metric for visual assessment of β diversitym and testing for differential expression of OTUs.n For analysis of differentially expressed OTUs, raw data were used (ie, results were not rarefied to even sampling depth) as recommended,31 and the data were normalized with a function to scale raw library sizes with software for analysis of differential gene expression data.n The microbiome census data output was converted into a matrix of gene counts, with taxonomic designation reclassified as genes for purposes of analysis.
Results
Horses and sample collection
All 12 horses completed the study. No horses developed any detectable disease, including signs of colic or diarrhea, during the study period. No horses were treated with medications other than that required for the study protocol. All samples were collected as previously described, and all collected fecal samples were formed. Gastric fluid samples varied in gross composition, which included grass-containing, dirt-containing, and predominantly translucent fluid with small floating particles. To obtain samples that would yield sufficient bacterial DNA for testing, as many siphoning attempts as necessary were used to collect a sample that included solid material; however, the number of attempts was limited to the minimum needed to collect an appropriate sample, and the amount of water used was minimized to the extent possible.
Fecal microbiota
There was a mean of 28,803 sequencing reads/sample (median, 27,581; range, 8,913 to 42,954); of these, 16,981 were unique reads representing a mean of 1,814 OTUs/sample. To account for uneven sampling depth, results for each sample were rarefied to a depth of 17,000 reads. This sampling depth appeared adequate, as the Good coverage index estimates indicated that a mean of 94.9% of the OTUs were represented among all samples with no differences between treatment groups or among time points (Figure 1). Filtering of low-abundance reads removed < 0.05% of total reads.
Scatterplot of the Good coverage estimates for microbial sequencing of fecal samples from 12 healthy horses that received omeprazole (4 mg/kg, PO, q 24 h) or a sham (control) treatment (tap water [20 mL, PO, q 24 h]) for 28 consecutive days (omeprazole [black circles] and control [gray circles] groups, respectively; n = 6/group). Horizontal lines represent the mean, and error bars indicate the 95% CIs for each group and time point. The day of the first treatment was considered day 0; samples on days 0 and 7 were collected prior to treatment.
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
Scatterplot of the Good coverage estimates for microbial sequencing of fecal samples from 12 healthy horses that received omeprazole (4 mg/kg, PO, q 24 h) or a sham (control) treatment (tap water [20 mL, PO, q 24 h]) for 28 consecutive days (omeprazole [black circles] and control [gray circles] groups, respectively; n = 6/group). Horizontal lines represent the mean, and error bars indicate the 95% CIs for each group and time point. The day of the first treatment was considered day 0; samples on days 0 and 7 were collected prior to treatment.
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
Scatterplot of the Good coverage estimates for microbial sequencing of fecal samples from 12 healthy horses that received omeprazole (4 mg/kg, PO, q 24 h) or a sham (control) treatment (tap water [20 mL, PO, q 24 h]) for 28 consecutive days (omeprazole [black circles] and control [gray circles] groups, respectively; n = 6/group). Horizontal lines represent the mean, and error bars indicate the 95% CIs for each group and time point. The day of the first treatment was considered day 0; samples on days 0 and 7 were collected prior to treatment.
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
α Diversity—Several α diversity indices were calculated for fecal microbiota. The α rarefaction curves determined on the basis of observed OTUs (an uncorrected measure of richness) and α diversity as determined by use of the chao1 diversity index revealed no visual difference between treatment groups or among time points (Figure 2). This lack of visual difference was confirmed by mixed-effects modeling, with no significant effects of day (P = 0.139), treatment (P = 0.558), or their interaction (P = 0.052) on the number of observed OTUs in the samples. Similarly, there were no significant effects of day (P = 0.117), treatment (P = 0.621), or their interaction (P = 0.096) on chao1 diversity estimates for fecal samples. The composition of the fecal microbiota for both groups was plotted (Figure 3).
Plots of α rarefaction curves depicting α diversity of observed OTUs (A) and chaol diversity estimates (B) for fecal microbiota in samples from the horses in Figure 1. Data points indicate the mean per sample for each group, and error bars indicate 95% CIs. No visual difference is detectable between treatment groups or over time.
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
Plots of α rarefaction curves depicting α diversity of observed OTUs (A) and chaol diversity estimates (B) for fecal microbiota in samples from the horses in Figure 1. Data points indicate the mean per sample for each group, and error bars indicate 95% CIs. No visual difference is detectable between treatment groups or over time.
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
Plots of α rarefaction curves depicting α diversity of observed OTUs (A) and chaol diversity estimates (B) for fecal microbiota in samples from the horses in Figure 1. Data points indicate the mean per sample for each group, and error bars indicate 95% CIs. No visual difference is detectable between treatment groups or over time.
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
Stacked bar charts depicting composition of the microbiota in fecal samples from the horses in Figure l on each sample collection day. A—Major bacterial phyla. B—Phyla on the basis of class (c) and order (o).
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
Stacked bar charts depicting composition of the microbiota in fecal samples from the horses in Figure l on each sample collection day. A—Major bacterial phyla. B—Phyla on the basis of class (c) and order (o).
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
Stacked bar charts depicting composition of the microbiota in fecal samples from the horses in Figure l on each sample collection day. A—Major bacterial phyla. B—Phyla on the basis of class (c) and order (o).
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
β Diversity—Examination of fecal microbiota PCoA plots created on the basis of UniFrac distances revealed no apparent visual clustering of results by treatment group or by time (Figure 4). This visual lack of association with treatment was confirmed by ANOSIM of weighted UniFrac distances (R value, −0.0234; P = 0.715). The data did cluster visually by horse, and this was confirmed by ANOSIM (R value, 0.804; P = 0.001). Evaluation of the UniFrac distance networks also confirmed clustering by horse (Figure 5).
Plot depicting results of PCoA of UniFrac (phylogenetic tree) distances for the OTUs of fecal microbiota of the horses in Figure 1. Values on the x- and y-axes represent arbitrary units and should be assigned no specific biological meaning. Percentages in the axis titles indicate the amount of variation explained by the components that make up the axes. Each PCoA axis is made of the components (OTUs) that explain the greatest amount of variability among the samples based on the UniFrac phylogenetic distance metric.
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
Plot depicting results of PCoA of UniFrac (phylogenetic tree) distances for the OTUs of fecal microbiota of the horses in Figure 1. Values on the x- and y-axes represent arbitrary units and should be assigned no specific biological meaning. Percentages in the axis titles indicate the amount of variation explained by the components that make up the axes. Each PCoA axis is made of the components (OTUs) that explain the greatest amount of variability among the samples based on the UniFrac phylogenetic distance metric.
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
Plot depicting results of PCoA of UniFrac (phylogenetic tree) distances for the OTUs of fecal microbiota of the horses in Figure 1. Values on the x- and y-axes represent arbitrary units and should be assigned no specific biological meaning. Percentages in the axis titles indicate the amount of variation explained by the components that make up the axes. Each PCoA axis is made of the components (OTUs) that explain the greatest amount of variability among the samples based on the UniFrac phylogenetic distance metric.
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
UniFrac distance network depicting phylogenetic distances of the microbiota in fecal samples from the horses in Figure 1. There is clustering by horse with no evidence of clustering by treatment or day. Numbers associated with values on the x- and y-axes represent arbitrary units. Numbers within the graphic represent individual horses in the omeprazole (treatment) and control groups.
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
UniFrac distance network depicting phylogenetic distances of the microbiota in fecal samples from the horses in Figure 1. There is clustering by horse with no evidence of clustering by treatment or day. Numbers associated with values on the x- and y-axes represent arbitrary units. Numbers within the graphic represent individual horses in the omeprazole (treatment) and control groups.
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
UniFrac distance network depicting phylogenetic distances of the microbiota in fecal samples from the horses in Figure 1. There is clustering by horse with no evidence of clustering by treatment or day. Numbers associated with values on the x- and y-axes represent arbitrary units. Numbers within the graphic represent individual horses in the omeprazole (treatment) and control groups.
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
Differentially expressed OTUs—Raw data sets were examined for differentially expressed OTUs between the omeprazole and control groups at each time point. A total of 1,376 OTUs were evaluated in the analysis. Each comparison resulted in a small number of OTUs (2 to 18) that were differentially abundant, with a false discovery rate P value < 0.05 (data not shown). However, the OTUs were unclassified to any taxonomic level, which prevented determination of any biological meaning for these findings. Similar differences were observed between horses of the omeprazole and control groups at all time points, including day 0.
Gastric microbiota
There was a mean of 20,644 sequencing reads/sample (median, 20,547; range, 54 to 26,311), representing a mean of 300 OTUs/sample. Of these, 4,565 were unique reads. To account for uneven sampling depth, results for each sample were rarefied to a depth of 10,000 reads. The Good coverage estimates revealed that this sampling depth covered a mean of 98.6% of the OTUs among all gastric fluid samples (data not shown). Raw data were specifically examined for presence of members of the family Helicobacteraceae, which includes Helicobacter pylori, and none were found. Data were then filtered to eliminate low-abundance reads, and this removed < 0.05% of total reads.
α Diversity—The α rarefaction curves for gastric microbiota plateaued at 10,000 reads, additionally indicating adequate sampling depth. The α diversity measures revealed substantial variation among samples, with no apparent effects of treatment or day (Figure 6). This lack of visual difference was confirmed by mixed-effects modeling, with no significant effects of day (P = 0.549), treatment (P = 0.724), or their interaction (P = 0.329) on the number of observed OTUs in the samples. There were significantly (P < 0.001) more OTUs in the fecal samples than in the gastric fluid samples. Adjusting for the effects of day and treatment, there were approximately 1,512 more observed OTUs (95% CI, 1,442 to 1,581 more OTUs) in fecal samples. Similarly, chao1 diversity estimates were significantly (P < 0.001) greater for fecal samples than for gastric fluid samples. Adjusting for effects of day and treatment, the chao1 values were 2,606.6 units (95% CI, 2,485.7 to 2,727.5 units) higher for fecal samples.
Plot of α diversity as determined by α rarefaction curves of observed OTUs of gastric microbiota for the horses in Figure I. There is marked variation among samples within each group and over time.
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
Plot of α diversity as determined by α rarefaction curves of observed OTUs of gastric microbiota for the horses in Figure I. There is marked variation among samples within each group and over time.
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
Plot of α diversity as determined by α rarefaction curves of observed OTUs of gastric microbiota for the horses in Figure I. There is marked variation among samples within each group and over time.
Citation: American Journal of Veterinary Research 80, 1; 10.2460/ajvr.80.1.79
β Diversity—Assessment of gastric microbiota PCoA plots created on the basis of weighted UniFrac distances revealed no visual evidence of clustering of results by treatment group, time, or horse (data not shown). Additionally, ANOSIM confirmed the lack of clustering by group (R = 0.0136; P = 0.346). The degree of clustering by horse was very small (R = 0.145) but significant (P = 0.01). Given the clear lack of treatment effect and the large amount of variation within groups, no attempt was made to identify differentially expressed OTUs for gastric microbiota.
Discussion
To our knowledge, this study was the first one conducted to investigate the effects of omeprazole administration on gastric and fecal microbiota of healthy adult horses. Contrary to our main hypothesis, we did not find any effect of omeprazole treatment on diversity of the fecal microbiota. Because there was no significant change in response to treatment, the second hypothesis, concerning a shift back to the original microbiota composition, was not testable. The clustering of data by horse in β diversity analysis suggested that each horse, regardless of treatment group, had a unique fecal microbiota population that changed minimally throughout the study.
As expected, we found that gastric fluid samples contained significantly fewer OTUs, compared with fecal samples overall; however, the gastric sample results were too variable within groups over time and among horses to allow meaningful interpretation of the data, and therefore, our hypotheses regarding changes in gastric microbiota could not be tested. The wide variability in diversity of the gastric microbiota samples, combined with the lack of visual clustering by treatment group or horse, suggested that the samples were more representative of other gastric contents than of host microbiota. The weak but significant clustering effect by horse may have been reflective of the consistent environments in which horses were housed throughout the study or may have been a spurious result. Although we used the fewest attempts and smallest volume of water needed to collect a usable sample, the comprehensive microbiota inhabiting the stomach could have been diluted by lavage. Previous evaluation of microbiota of the stomach of horses has been performed for mucosal samples.32 In a study by Sung et al33 that compared the microbiota of gastric fluid and mucosal samples from 4 human patients, gastric fluid samples were found to have greater OTU diversity and to contain subjectively fewer reads relative to those for gastric mucosa. Only a small proportion of gastric fluid samples contained H pylori, compared with the findings for mucosal samples in that study,33 consistent with a review34 of other investigations indicating that H pylori colonizes the mucous layer and adheres to gastric epithelial cells in people. This is consistent with our study findings that no members of the family Helicobacteraceae were identified in gastric fluid samples. In the aforementioned study,33 there was presence of bacteria thought to be transiently passing from the oral cavity and esophagus rather than inhabiting the stomach. In the present study, we chose to collect gastric fluid instead of gastric mucosa samples for several reasons, including maintenance of the normal diet and routine for the horses to minimize confounding factors that might alter microbiota of the gastrointestinal tract, in addition to financial constraints.
In our study, no significant change was observed in the fecal microbiota of healthy adult horses following omeprazole administration. The few OTUs that were differentially expressed between groups were unclassified, and significance of these findings was likely related to the fact that a large number of OTUs was included in the comparisons. Further, similar differences were seen between treated and untreated horses at all time points, including day 0, confirming that these small differences were attributable to factors other than omeprazole treatment. Administration of a PPI to healthy human subjects has been shown to decrease OTU counts in fecal samples to values similar to those in samples from patients with Clostridium difficile infection within 1 week; this effect was partially reversed 1 month after PPI administration was discontinued.10 A recent study15 showed an association between treatment with antiulcer medications and development of diarrhea (OR, 2.0; 95% CI, 1.4 to 2.9) and sepsis (OR, 1.5; 95% CI, 1.1 to 2.1) in neonatal foals. Diarrhea in young foals has been presumptively associated with gastrointestinal dysbiosis.35 Changes in the fecal and hindgut microbiota of adult horses have been previously associated with antimicrobial use,20 management changes,18,19 and anesthesia,18 and microbiota alterations have been identified in broodmares prior to development of postpartum colic.36 However, because horses are hindgut fermenters, it is plausible that the effects of acid reduction at the level of the stomach are minimized at the level of the hindgut because of the relatively large numbers of bacteria present and the area involved. The fecal microbiota of foals is known to change significantly with age as they become hindgut fermenters.37,38 We speculate that the fecal microbiota would be more likely affected in response to omeprazole administration in adult horses with concurrent risk factors such as stress or medications and in neonatal foals with monogastric digestion.
Limitations of the present study should be acknowledged when considering these results. The sample size (n = 6) was small but was based on an a priori calculation. The mean OTU count was 1,814, which was slightly lower than our a priori assumption. Additionally, there is empirical precedent for small sample size in equine microbiota studies.17,21,39–41 The small sample size limited statistical power and the ability to detect differences; however, the results of qualitative, graphic clustering of microbiota composition of individual horses over time supported the lack of difference in composition and diversity of the fecal microbiota between horses that did and did not receive omeprazole.
The study design did not control for slight differences in housing and environment among horses and associated minor differences in feed and water composition. However, horses were primarily pastured in close proximity to each other, horses of each group were housed in equal numbers in each environment, and, when stall housed, feed sources were the same for all horses (ad libitum feeding of grass hay). A previous study39 that identified differences in fecal microbiota associated with differences in diet included much more distinct differences in feed than existed for the horses in our study. Finally, we looked for differences in the microbiota over time beginning with day 0 (prior to the start of treatment), which allowed each horse to serve as its own control and revealed that fecal microbial populations in each horse changed minimally over time, thus confirming no change in response to treatment, irrespective of minor differences in environment. Gastric fluid microbiota β diversity measures were only minimally clustered by horse and not by time or treatment. Thus, the effects of environment on the study results were likely minimal. Ideally, each horse would have been housed in the same environment and offered the same feed and water from 1 source over the course of the study, but because of the facilities available and herd dynamics, this was not feasible, and such management changes would likely have confounded study results.
Our study included healthy horses that were part of a university teaching program and essentially retired from all physical performance, and these may not have been representative of the population of horses most commonly administered omeprazole for prophylaxis or for treatment of EGUS. Furthermore, although the horses were deemed healthy, the presence of EGUS was not ruled out before, during, or after treatment. Notably, these results cannot be extrapolated to neonates, horses with colic or other diseases, or horses receiving other medical treatments. The effect of oral omeprazole administration on the fecal microbiota should be evaluated in such horses.
Although the small sample size might have limited the power to detect a difference in the gastric and fecal microbiota of horses attributable to omeprazole treatment if one truly existed, qualitative examination of graphic results suggested that any difference would likely have been small and of limited clinical importance. The study findings supported that administration of omeprazole to healthy adult horses at the dosage used is unlikely to result in clinically relevant gastrointestinal complications owing to effects on these microbial populations. Further understanding of the role of gastric microbiota in the effects of omeprazole treatment of gastric ulcers would require a modified gastric fluid collection technique or endoscopically guided gastric mucosal biopsy.
Acknowledgments
Funded by the For the Love of the Horse Research Endowment Fund through University of Georgia Equine Programs. The product used in this study was provided without cost by Merial Ltd. The provider was not involved in study procedures, data analysis, or writing of the manuscript.
The authors thank Londa Berghaus, Lorelai Branch, and Zac Turner for project support.
ABBREVIATIONS
| ANOSIM | Analysis of similarity |
| CI | Confidence interval |
| EGUS | Equine gastric ulcer syndrome |
| OTU | Operational taxonomic unit |
| PCoA | Principle coordinates analysis |
| PPI | Proton pump inhibitor |
Footnotes
MedCalc, version 18.5, MedCalc SoftwareBVBA, Ostend, Belgium.
STATA, version 12.1, STATA Corp, College Station, Tex.
GastroGard, Merial Ltd, Duluth, Ga.
QIAmp Fast DNA Stool Mini Kit, Qiagen, Hilden, Germany.
InhibitEX buffer, Qiagen, Hilden, Germany.
Mini-Beadbeater-24, BioSpec Products Inc, Bartlesville, Okla.
MiSeq, Illumina, San Diego, Calif.
QIIME (Quantitative Insights Into Microbial Ecology), version 1.9, version 3.0. Available at: qiime.org. Accessed Jan 7, 2017.
UCHIME, version 4.2, Edgar RC, Haas BJ. Available at: drive5.com/uchime/uchime_download.html. Accessed Jan 7, 2017.
Greengenes database, version gg_13_8, The Greengenes Database Consortium. Available at: greengenes.secondgenome.com. Accessed Jan 7, 2017.
R: A language and environment for statistical computing, version 3.4.0, R Foundation for Statistical Computing, Vienna, Austria. Available at: cran.r-project.org/. Accessed Jan 7, 2017.
nlme, Linear and nonlinear mixed effects models, R package, version 3.1-122, Pinheiro J, Bates D, DebRoy S, et al. Available at: CRAN.R-project.org/package=nlme. Accessed Jan 7, 2017.
Phyloseq, Handling and analysis of high-throughput microbiome census data, R package, version 1.23.1, McMurdie PJ, Holmes S. Available at: bioconductor.org/packages/release/bioc/html/phyloseq.html. Accessed Jan 7, 2017.
edgeR, Empirical analysis of digital gene expression data in R, R package, version 3.18.1, Chen Y, Lun A, McCarthy D, et al. Available at: bioconductor.org/packages/release/bioc/html/edgeR.html. Accessed Jan 7, 2017.
DESeq2, Differential gene expression analysis based on the negative binomial distribution, R package, version 3.6, Love M, Anders S, Huber W. Available at: bioconductor.org/packages/release/bioc/html/DESeq2.html. Accessed Jan 7, 2017.
ggplot2, Create elegant data visualisations using the grammar of graphics, R package, version 2.2.1, Wickham H, Chang W, RStudio. Available at: CRAN.R-project.org/package=ggplot2. Accessed Jan 7, 2017.
ape, Analysis of phylogenetics and evolution, R package, version 5.1, Paradis E, Blomberg S, Bolker B, et al. Available at: CRAN.R-project.org/package=ape. Accessed Jan 7, 2017.
plyr, Tools for splitting, applying, and combining data, R package, version 1.8.4, Wickham H. Available at: CRAN.R-project.org/package=plyr. Accessed Jan 7, 2017.
References
1. Vatistas NJ, Snyder JR, Carlson G, et al. Cross-sectional study of gastric ulcers of the squamous mucosa in Thoroughbred racehorses. Equine Vet J Suppl 1999;29:34–39.
2. McClure SR, Glickman LT, Glickman NW. Prevalence of gastric ulcers in show horses. J Am Vet Med Assoc 1999;215:1130–1133.
3. Videla R, Andrews FM. New perspectives in equine gastric ulcer syndrome. Vet Clin North Am Equine Pract 2009;25:283–301.
4. Camacho-Luna P, Andrews FM. Equine gastric ulcer syndrome. In: Sprayberry KA, Robinson NE, eds. Robinson's current therapy in equine medicine. 7th ed. St Louis; Saunders-Elsevier, 2015;280–284.
5. Daurio CP, Holste JE, Andrews FM, et al. Effect of omeprazole paste on gastric acid secretion in horses. Equine Vet J Suppl 1999;29:59–62.
6. Andrews FM, Sifferman RL, Bernard W, et al. Efficacy of omeprazole paste in the treatment and prevention of gastric ulcers in horses. Equine Vet J Suppl 1999;29:81–86.
7. Plue RE, Wall HG, Daurio C, et al. Safety of omeprazole paste in foals and mature horses. Equine Vet J Suppl 1999;29:63–66.
8. Bavishi C, Dupont HL. Systematic review: the use of proton pump inhibitors and increased susceptibility to enteric infection. Aliment Pharmacol Ther 2011;34:1269–1281.
9. Freedberg DE, Lebwohl B, Abrams JA. The impact of proton pump inhibitors on the human gastrointestinal microbiome. Clin Lab Med 2014;34:771–785.
10. Seto CT, Jeraldo P, Orenstein R, et al. Prolonged use of a proton pump inhibitor reduces microbial diversity: implications for Clostridium difficile susceptibility (Erratum published in Microbiome 2016;4:10). Microbiome 2014;2:42.
11. Wu WM, Yang YS, Peng LH. Microbiota in the stomach: new insights. J Dig Dis 2014;15:54–61.
12. Garcia-Mazcorro JF, Suchodolski JS, Jones KR, et al. Effect of the proton pump inhibitor omeprazole on the gastrointestinal bacterial microbiota of healthy dogs. FEMS Microbiol Ecol 2012;80:624–636.
13. Laheij RJ, Sturkenboom MC, Hassing RJ, et al. Risk of community-acquired pneumonia and use of gastric acid-suppressive drugs. JAMA 2004;292:1955–1960.
14. Graham PL III, Begg MD, Larson E, et al. Risk factors for late onset gram-negative sepsis in low birth weight infants hospitalized in the neonatal intensive care unit. Pediatr Infect Dis J 2006;25:113–117.
15. Furr M, Cohen ND, Axon JE, et al. Treatment with histamine-type 2 receptor antagonists and omeprazole increase the risk of diarrhoea in neonatal foals treated in intensive care units. Equine Vet J Suppl 2012;41:80–86.
16. Proudman CJ, Hunter JO, Darby AC, et al. Characterisation of the faecal metabolome and microbiome of Thoroughbred racehorses. Equine Vet J 2015;47:580–586.
17. Costa MC, Silva G, Ramos RV, et al. Characterization and comparison of the bacterial microbiota in different gastrointestinal tract compartments in horses. Vet J 2015;205:74–80.
18. Schoster A, Mosing M, Jalali M, et al. Effects of transport, fasting and anesthesia on the faecal microbiota of healthy adult horses. Equine Vet J 2016;48:595–602.
19. Daly K, Proudman CJ, Duncan SH, et al. Alterations in microbiota and fermentation products in equine large intestine in response to dietary variation and intestinal disease. Br J Nutr 2012;107:989–995.
20. Harlow BE, Lawrence LM, Flythe MD. Diarrhea-associated pathogens, lactobacilli and cellulolytic bacteria in equine feces: responses to antibiotic challenge. Vet Microbiol 2013;166:225–232.
21. Costa MC, Stampfli HR, Arroyo LG, et al. Changes in the equine fecal microbiota associated with the use of systemic antimicrobial drugs. BMC Vet Res 2015;11:19.
22. Whitfield-Cargile CM, Cohen ND, Suchodolski J, et al. Composition and diversity of the fecal microbiome and inferred fecal metagenome does not predict subsequent pneumonia caused by Rhodococcus equi in foals. PLoS One 2015;10:e0136586.
23. Whitfield-Cargile CM, Cohen ND, Chapkin RS, et al. The microbiota-derived metabolite indole decreases mucosal inflammation and injury in a murine model of NSAID enteropathy. Gut Microbes 2016;7:246–261.
24. Caporaso JG, Kuczynski J, Stombaugh J, et al. QIIME allows analysis of high-throughput community sequencing data. Nat Methods 2010;7:335–336.
25. Illumina. TruSeq DNA sample preparation guide, revision C. San Diego, Calif: Illumina, 2012.
26. Edgar RC, Haas BJ, Clemente JC, et al. UCHIME improves sensitivity and speed of chimera detection. Bioinformatics 2011;27:2194–2200.
27. DeSantis TZ, Hugenholtz P, Larsen N, et al. Greengenes, a chimera-checked 16S rRNA gene database and workbench compatible with ARB. Appl Environ Microbiol 2006;72:5069–5072.
28. Clarke KR. Non-parametric multivariate analyses of changes in community structure. Aust J Ecol 1993;18:117–143.
29. McMurdie PJ, Holmes S. phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data. PLoS One 2013;8:e61217.
30. Robinson MD, McCarthy DJ, Smyth GK. edgeR: a bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 2010;26:139–140.
31. McMurdie PJ, Holmes S. Waste not, want not: why rarefying microbiome data is inadmissible. PLoS Comput Biol 2014;10:e1003531.
32. Perkins GA, den Bakker HC, Burton AJ, et al. Equine stomachs harbor an abundant and diverse mucosal microbiota. Appl Environ Microbiol 2012;78:2522–2532.
33. Sung J, Kim N, Kim J, et al. Comparison of gastric microbiota between gastric juice and mucosa by next generation sequencing method. J Cancer Prev 2016;21:60–65.
34. Amieva MR, El-Omar EM. Host-bacterial interactions in Helicobacter pylori infection. Gastroenterology 2008;134:306–323.
35. Schoster A, Staempfli HR, Guardabassi LG, et al. Comparison of the fecal bacterial microbiota of healthy and diarrheic foals at two and four weeks of life. BMC Vet Res 2017;13:144.
36. Weese JS, Holcombe SJ, Embertson RM, et al. Changes in the faecal microbiota of mares precede the development of post partum colic. Equine Vet J 2015;47:641–649.
37. Costa MC, Stampfli HR, Allen-Vercoe E, et al. Development of the faecal microbiota in foals. Equine Vet J 2016;48:681–688.
38. Kuhl J, Winterhoff N, Wulf M, et al. Changes in faecal bacteria and metabolic parameters in foals during the first six weeks of life. Vet Microbiol 2011;151:321–328.
39. Harlow BE, Lawrence LM, Hayes SH, et al. Effect of dietary starch source and concentration on equine fecal microbiota. PLoS One 2016;11:e0154037.
40. Bordin AI, Suchodolski JS, Market ME, et al. Effects of administration of live or inactivated virulent Rhodococccus equi and age on the fecal microbiome of neonatal foals. PLoS One 2013;8:e66640.
41. Bond SL, Timsit E, Workentine M, et al. Upper and lower respiratory tract microbiota in horses: bacterial communities associated with health and mild asthma (inflammatory airway disease) and effects of dexamethasone. BMC Microbiol 2017;17:184.
