Effects of Administration of Live or Inactivated Virulent Rhodococccus equi and Age on the Fecal Microbiome of Neonatal Foals

Background Rhodococcus equi is an important pathogen of foals. Enteral administration of live, virulent R. equi during early life has been documented to protect against subsequent intrabronchial challenge with R. equi, indicating that enteral mucosal immunization may be protective. Evidence exists that mucosal immune responses develop against both live and inactivated micro-organisms. The extent to which live or inactivated R. equi might alter the intestinal microbiome of foals is unknown. This is an important question because the intestinal microbiome of neonates of other species is known to change over time and to influence host development. To our knowledge, changes in the intestinal microbiome of foals during early life have not been reported. Thus, the purpose of this study was to determine whether age (during the first month of life) or administration of either live virulent R. equi (at a dose reported to protect foals against subsequent intrabronchial challenge, viz., 1×1010 colony forming units [CFU]) or inactivated virulent R. equi (at higher doses, viz., 2×1010 and 1×1011 [CFU]) altered the fecal microbiome of foals. Methodology/Principal Findings Fecal swab samples from 42 healthy foals after vaccination with low-dose inactivated R. equi (n = 9), high-dose inactivated R. equi (n = 10), live R. equi (n = 6), control with cholera toxin B (CTB, n = 9), and control without CTB (n = 8) were evaluated by 454-pyrosequencing of the 16S rRNA gene and by qPCR. No impact of treatment was observed among vaccinated foals; however, marked and significant differences in microbial communities and diversity were observed between foals at 30 days of age relative to 2 days of age. Conclusions The results suggest age-related changes in the fecal microbial population of healthy foals do occur, however, mucosal vaccination does not result in major changes of the fecal microbiome in foals.


Introduction
Rhodococcus equi is a facultative intracellular pathogen that primarily infects macrophages [1]. Although human beings may be infected (primarily those who are immunocompromised by HIV infection or immunosuppressive treatments), R. equi is most commonly recognized clinically as a leading cause of severe pneumonia in foals [1][2][3][4]. The disease occurs among foals worldwide [1][2][3][4]. Isolates that are virulent in foals bear a plasmid that encodes for a pathogenicity island, which includes the gene for the virulence-associated protein A (vapA); vapA is necessary but not sufficient to cause disease [5,6].
Despite the global importance of the disease, an effective vaccine is lacking for control and prevention of R. equi pneumonia in foals. The lack of an effective vaccine is likely attributable to the complexity of immunity to R. equi [7][8][9], and the finding that foals appear to be infected very early in life [10,11], when immune responses are naïve or deficient. It is generally accepted that a vaccine must be able to provide foals with protection against infection with R. equi during early life [5].
To date, the only vaccination strategy that has been demonstrated repeatedly to be effective for protecting against experimental intrabronchial challenge with virulent R. equi has been oral administration of live, virulent R. equi [12][13][14]. Protection against respiratory pathogens induced by oral vaccination also has been documented in mice [15][16][17], and evidence exists that bacillus Calmette-Guerin (BCG) administered orally is protective against tuberculosis in people and animals [18][19][20]. Moreover, inactivated bacteria and viruses also can elicit protective immune responses against systemic infections, including those of the respiratory tract [21][22][23][24]. Despite the success of oral administration of live organisms to protect foals against experimental challenge, very limited information is available regarding immune and other biological responses to the enteral route of vaccination.
One issue of importance with regard to enteral vaccination with live organisms is the impact of enteral administration of bacteria on the intestinal microbiome. This question might be particularly important for neonates. Although the microbiome of foals has not been systematically evaluated, evidence exists in other species, including humans, that the intestinal microbiome of neonates develops with age [25][26][27], and is linked to the functional development of the gut and gut immunity [25][26][27][28][29]. Thus, the purpose of the study reported here was to determine whether agerelated changes in the microbiome occur in foals and whether ageassociated changes are impacted by administration of either live virulent R. equi at a dose documented to protect foals against experimental challenge or 2 doses of inactivated virulent R. equi higher than the dose of live R. equi.

Ethics statement
All procedures for this study, including collection of rectal swab samples and enteral treatments/vaccinations, were reviewed and approved by the Texas A&M University Institutional Animal Care and Use Committee (protocol number AUP # 2011-124) and the Texas A&M University Institutional Biosafety Committee (permit number 20110183-Cohen). The foals used in this study are owned by Texas A&M University, and permission for their use was provided in compliance with the Institutional Animal Care and Use Committee procedures.

Animals and housing
Forty-two healthy Quarter Horse foals were used for this study. All foals were born healthy and had age-appropriate results of complete blood count (CBC) on day 2 of life, and adequate transfer of passive immunity as assessed by a commerciallyavailable qualitative immunoassay for serum concentration of total IgG (SNAP Foal IgG test; IDEXX, Inc., Westbrook, ME). The foals were assigned into 1 of 5 experimental groups prior to birth (please see section on Vaccine Preparation and Treatment Groups below). All foals were monitored daily by Texas A&M University Horse Center staff for clinical signs of disease, and inspected at least twice weekly by a veterinarian for clinical signs of disease. All foals remained free of clinical signs of disease and in good health throughout the study.

Mare Diet
The respective dams were fed 6.4 kg per horse per day of a 13% horse pellet (crude protein: 13.5%; crude fat: 4.5%; crude fiber: 10%). Also, the foals and their mares were allowed free access to coastal Bermuda grass hay, plus grazing of pastures at the Texas A&M University Horse Center where the mares were maintained.

Vaccine Preparation and Treatment Groups
Rhodococcus equi strain EIDL 5-331, obtained from a Texas foal confirmed to have R. equi pneumonia, was used to prepare live and inactivated vaccines used for this project. Physiological saline (NaCl 0.9%) was used as a diluent to achieve the specified concentration of all vaccine preparations, as well as for the negative control. The vaccine was produced by inoculating blood agar plates with 1 colony forming unit (CFU) of R. equi strain 5-331 and incubating at 37uC for 48 hours. One colony from this pure culture was selected and used to inoculate 1,000 ml of brain heart infusion (BHI, Bacto TM Brain Heart Infusion, BD Diagnostic Systems, Sparks, MD) broth. The flask with inoculated broth was placed on an orbital shaker (VWR OS-500, VWR, Radnor, PA) at 200 rpm for 24h at 37uC to allow bacterial growth. Isolates were repeatedly tested by PCR for the vapA gene to confirm that the isolates were virulent [30]. The bacterial culture was inactivated by electron-beam irradiation (irradiation dose between 3.5 and 5 kGy). After inactivation, the irradiated bacterial cells were plated out on BHI agar plates and incubated for 2 weeks at 37uC to confirm inactivation.
The number of foals in each group was determined a priori, and foals were assigned randomly to each of the groups. The study groups were as follows: 1) low-dose inactivated virulent R. equi group (n = 9), receiving 2610 10 CFUs of inactivated R. equi combined with 100 mg of cholera toxin subunit B (CTB, List Biological Laboratories, Campbell, CA) as a mucosal adjuvant, diluted in 100 ml of saline administered via nasogastric intubation; 2) high-dose inactivated virulent R. equi group (n = 10), receiving

Fecal swabbing
Rectal swabs were collected by inserting a 16-inch, cottontipped swab that was pre-moistened with 3 ml of sterile saline Figure 1. Rarefaction analysis of 16 S rRNA gene sequences obtained from fecal swabs from foals. Lines represent the average of each vaccination group at all ages (panel A) or at 30 days only (panel B), while the error bars represent the standard deviations. The analysis was performed on a randomly selected subset of 1,300 sequences per sample and included samples from 42 foals. Note that both the greatest and least number of species observed occurred among foals that received no enteral bacteria (live or inactivated), indicating an absence of evidence of treatment effect. Control = control plus CTB group; Control_no_CTO = control without CTB group; High = high-dose inactivated R. equi group; Live = live R. equi group; Low = low-dose inactivated R. equi group. doi:10.1371/journal.pone.0066640.g001 Figure 2. Rarefaction analysis of 16 S rRNA gene sequences obtained from fecal swabs from foals. Lines represent the average numbers obtained at each age (legend numbers refer to the age in days), while the error bars represent the standard deviations. The analysis was performed on a randomly selected subset of 1,300 sequences per sample and included samples from 42 foals. Note the progressive increase in observed species (representing microbial diversity) with sequential age. The numbers for the legend represent age (in days). doi:10.1371/journal.pone.0066640.g002   Other class 0% (0 to 0.5%) 0.3% (0 to 6.5%) ,0.0001 Other order 0% (0 to 0.5%) 0.3% (0 to 6.5%) ,0.0001 Other family 0% (0 to 0.5%) 0.3% (0 to 6.5%) ,0.0001
Raw sequence data were screened, trimmed, filtered, de-noised, and chimera-depleted with default settings using the QIIME pipeline version 1.6.0 (http://qiime.sourceforge.net) and with USEARCH using the OTU pipeline (www.drive5.com). Operational taxonomic units (OTUs) were defined as sequences with at least 97% similarity using QIIME. For classification of sequences on a genus level the naïve Bayesian classifier within the Ribosomal Database Project (RDP, v10.28) was used [31].
The obtained data were compiled to determine the relative proportions of bacteria for each individual sample. The subsequent analysis was performed on a randomly selected subset of 1,300 sequences per sample to account for unequal sequencing depth across samples. Alpha diversity and beta diversity measures were calculated and plotted using QIIME. To determine differences in microbiota composition between the animal groups, the analysis of similarities (ANOSIM) function in the statistical software package PRIMER 6 (PRIMER-E Ltd., Lutton, UK) was used on the unweighted Unifrac distances matrices. This analysis measures the phylogenetic distance among bacterial communities in a phylogenetic tree, and thereby provides a measure of similarity among microbial communities present in different biological samples. The linear discriminant analysis (LDA) effect size (LEfSe) method was used to represent taxonomic relevant agerelated differences in foal fecal swabs [32].

qPCR
To validate the pyrosequencing results, quantitative PCR (qPCR) assays were performed as described previously [33]. Briefly, EvaGreen-based reaction mixtures (total 10 mL) contained 5 mL of SsoFast TM EvaGreenH supermix (Biorad Laboratories), 2.2 mL of water, 0.4 mL of each primer (final concentration: 400 nM), and 2 mL of DNA (normalized to 5 ng/ul)). PCR conditions were 98uC for 2 min, and 40 cycles at 98uC 5 s, and 5 s at the optimized annealing temperature (Table 1). A melt curve analysis was performed for under the following conditions: beginning at 65uC, gradually increasing 0.5uC/5 s to 95uC with acquisition data every 5 s. The qPCR data was expressed as log amount of DNA (fg) for each particular bacterial group per 10 ng of isolated total DNA [34].

Data analysis
Pairwise comparisons between ages 2 days and 30 days were made at the levels of phylum, class, order, and family of bacteria for 2 outcomes: the observed percentage of sequences of bacteria at a given level, and the proportion of foals in which any amount of a given sequence for a given level was observed (i.e., the dichotomous outcome of whether or not a specific phylum [or class or order or family] was represented). The paired differences in percentages were compared using a Wilcoxon sign-rank test, and the paired proportions were compared using McNemar's test. Because of the multiplicity of comparisons, P values at a given level (e.g., order) were adjusted using the method of Hochberg [35]. An adjusted P value ,0.05 was considered significant for these analyses. Analyses were conducted using S-PLUS (Version 8.0; Insightful, Inc.) and R (Version 2.12.1; R Statistical Project). To assess the diversity of the GI microbiota, the Shannon-Weaver [36] and Chao 1 [37] diversity indices were calculated in QIIME.

Sequence analysis
The 454-pyrosequencing pipeline yielded 499,419 quality sequences for the 42 samples analyzed. For technical reasons attributed to random error, 5 foals (2 foals from the control group without CTB, and 3 foals from the live R. equi group) did not generate sufficient sequences (cut-off value of 1,300 sequences) in at least 1 sample from 1 sampling time-point (either 2 or 30 days) by 454-pyrosequencing. Those foals were included in the descriptive analysis (Figures PCoA and rarefaction). For comparing age-related changes of the microbiome, however, the analysis was restricted to 37 foals with samples available from both collection time-points (2 and 30 days).
Across all vaccination groups and ages, sequences were classified into 18 phyla (Table 2 and 3). For the rarefaction curves of all vaccination groups ( Figure 1A and 1B) and age groups (Figure 2), 1,300 sequences per sample yielded stable estimates of sample diversity.

Microbial communities in control and vaccinated foals
No differences in microbial composition were observed among animals from control, live and inactivated treatment/vaccination groups ( Figures 1A, 1B, and 3). The rarefaction curves for the treatment groups revealed no clear pattern of greater number of observed species (i.e., diversity) among foals receiving either live or inactivated R. equi, or those foals in the 3 control groups that did not receive R. equi ( Figure 1A). Because the samples at age 2 days were not affected by treatment (because treatment was administered after sample collection on day 2), we also performed analysis restricting data to samples collected at age 30 days ( Figure 1B). Once again, there was no pattern of differences in the rarefaction curves among treatment groups receiving either live or inactivated R. equi or the control groups. Using PCoA (Figures 3 and 4), there was no qualitative evidence of differences among groups; the clustering observed in Figure 3 panel A was attributable to effects of age (please see next section). When considering only the data from foals at 30 days of age (because samples on day 2 were collected prior to treatment administration), the PCoA plots revealed no clustering by group and the ANOSIM test statistic for differences among groups was not significant (P = 0.494).

Age-related changes in microbial communities in foals
There were strong and significant differences in the fecal microbiome of foals associated with age. The rarefaction curves demonstrated a pattern of increasing number of species (diversity) with increasing age (Figure 2). These results should be interpreted with caution because there were only 2 foals for which data for ages other than 2 days and 30 days were available. The PCoA plots by age revealed an obvious separation of samples by age, attributable to differences between the time-points of days 2 and 30 ( Figure 5); the ANOSIM test statistic for differences between day 2 and day 30 was significant (P = 0.0010).
Significant differences in the number of OTUs, the Shannon index, and the Chao1 metric were observed between the age groups ( Because of the apparent differences of the microbiota between age groups, we also compared the distribution of bacteria by phylum, class, order, and family between foals aged 2 days and 30 days. In total, 18 phyla were detected in fecal samples from foals ( Table 2). Of those, Bacteroidetes (40.6%, day 30), Firmicutes (40.4%, day 2), and Proteobacteria (36.6%, day 2) had the highest percentages of sequences reported. Proteobacteria and Firmicutes were detected in all samples from 2-day-old foals, followed by Bacteroidetes (92%) and Actinobacteria (73%) ( Table 3). Among 30-day-old foals, Bacteroidetes and Firmicutes were detected in all fecal samples, followed by Actinobacteria (97%) and Proteobacteria (97%), Verrucomicrobia (89%), and Fusobacteria (84%) ( Table 3). The following phyla increased significantly with age (i.e,    Figure 6). Within the phylum Proteobacteria, the class Gammaproteobacteria (P,0.0001) and the family Enterobacteriaceae (P,0.0001) decreased significantly with age. Other classes of Proteobacteria, such as Deltaproteobacteria (P = 0.0084) and Epsilonproteobacteria (P,0.0001) significantly increased with age ( Table 2).
To confirm results of pyrosequencing, we also performed realtime quantitative PCR. Significant differences were observed in specific microbial communities between the 2 age groups based on qPCR analysis, with age-related decreases for Escherichia coli (P,0.0001) and for Enterococcus (P,0.0001). These data were consistent with genus-level results observed by pyrosequencing (Table 5) for Enterococcus (P = 0.0009) and for Escherichia (P,0.0001). We also found agreement for a lack of evidence of a significant difference between the pyrosequencing and the qPCR results for Bacteroidetes (P = 0.9519 by qPCR and P = 0.5376 by pyrosequencing) and Fusobacteria (P = 0.1051 on qPCR and P = 0.1000 on pyrosequencing).

Discussion
In this study, our first objective was to evaluate changes in the microbiome of foals following vaccination with both live and inactivated R. equi. Although the number of CFUs administered were as high (for the live R. equi group) or higher than the number of CFU documented to protect foals against intrabronchial challenge with virulent R. equi (viz., 1610 10 CFU), no apparent differences in microbial communities were observed among vaccinated groups ( Figures 1A and 3). Because all but 2 foals had samples collected only on days 2 and 30, and because fecal  samples on day 2 were not influenced by treatment (because they were collected immediately prior to treatment), the effect of group also was examined among only samples collected at 30 days of age. Results restricted to 30 days of age also revealed no pattern distinguishing vaccinated and non-vaccinated foals ( Figures 1B  and 4). Thus, we failed to detect evidence of a significant effect of enteral administration of either live or inactivated R. equi on microbial populations in neonatal foals. These results are consistent with reports in which probiotics (administered at similar or higher numbers of CFUs) have failed to alter the intestinal/fecal microbiome [38][39][40]. Our results should be interpreted with caution because of the relatively small number of foals, particularly in the live R. equi group. For technical reasons attributed to random error, pyrosequencing failed for samples from 3 foals from the live R. equi group and 2 foals from the control group without CTB group; therefore, only 3 foals from the live R. equi group and 6 from the controls without CTB group were included in the analysis. A significant difference between the fecal microbial populations between day 2 and day 30 of age was observed (Table 2 and 3; Figures 2, 5, and 6). For descriptive purposes, we included the results from the 2 foals from which we had data at other ages (these data were not included in the statistical analysis comparing ages). The resident intestinal or fecal microbiota has been described for neonates of other species, such as cats [41,42], dogs [43], and humans [25][26][27]44]. To the authors' knowledge, this is the first report of age-related changes of the fecal microbiome in foals. Significant changes in the number of OTUs, the Shannon index, and the Chao1 metric were observed between the age groups (Table 4), showing clear evidence of strong diversification of bacterial populations between 2 and 30 days of age.
Firmicutes were detected in 100% of foals at both 2 and 30 days of age, with reported median sequences of 40% in 2-day-old foals decreasing (albeit not significantly) to 23% in 30-day-old foals. In 2 previous studies using fecal samples from adult horses, Firmicutes represented 44% [45] and 72% [46] of the bacteria. Within the Firmicutes, the family Enterococcaceae significantly decreased with age (P = 0.0080), which was likely attributable at least in part to decreases in the genus Enterococcus that were observed to decrease significantly by qPCR (P,0.0001) and by pyrosequencing (Table 5). Proteobacteria were detected in the feces of all 2day-old foals and 97% of 30-day-old foals, a difference that was not significant; however, the median percentage of sequences decreased significantly (P,0.0001) between day 2 (median, 36.3%; range, 0.5 to 85.8%) and day 30 (median, 2.7%; range, 0 to 40.9%). In adult horses, Proteobacteria have been reported to represent 6% [45] and 12% [47] of fecal sequences. These results from adult horses are interesting in light of our findings, particularly our observation that the family Enterobacteriaceae decreased with age, a finding substantiated by our qPCR results with a significant decrease in the amount of E. coli (P,0.0001) between ages 2 and 30 days.
The sterile GI tract of newborn puppies and kittens is presumably colonized by bacteria present in the birth canal and from the environment [48], and human neonates appear to become colonized by these sources as well as through the intestinal microbiota of the mother [25,49]. In humans, the initial microbes colonizing infants are facultative anaerobic bacteria, such as E. coli and Streptococcus spp. [49], which was also observed in 2-day old foals by the presence of Enterobacteriaceae (E. coli) and Streptococcaceae families (Streptococcus spp.). We observed a significant decrease in both these families by 30 days of age, suggesting that a similar phenomenon might happen in foals. In human beings, after the initial colonization by facultative anaerobic bacteria, colonization occurs by Staphylococcus-, Enterococcus-, and Lactobacillus-like species, and this change might contribute to generating an anaerobic environment [44]. The development during the first month of life in foals of an anaerobic environment is supported by the age-related increase in the detection of the phylum of Bacteroidetes (P = 0.0066), which is also a common constituent of the gut microbiota of dogs and cats [48]. However, we also observed a significant decrease in the Enterococcaceae family (P = 0.0080) and Enterococcus spp. by qPCR (P,0.0001), as well as the Lactobacillaceae family (P = 0.0281).
Our study has a number of important limitations. One limitation is the use of fecal swab samples for analysis, because feces might not be representative of other compartments of the gut. In humans, the composition of the mucosal-surface microbiota is distinct from that recovered in the feces [50]. The situation is probably similar in the horse, because of the complexity of the equine gastrointestinal tract. For example, the microbial population of adult horse fecal samples is likely to represent that of the right dorsal colon, but not that of the cecum [51].
A second limitation of our study is the small number of foals enrolled. Our sample size was limited both by financial considerations and the number of foals available to us during the study period. Because of the small sample size, we were only able to observe large changes in fecal microbial populations. Nevertheless, our results provide useful data for those exploring enteral vaccination of foals [14,52]. It is worth noting that there were significant differences in immune responses that were detectable among these groups of foals despite the small sample size (data not shown). Also, we were able to detect significant agerelated differences in the microbiome of foals, irrespective of the treatment groups.
Another limitation of our study is that we only characterized age-related changes at 2 ages during the first month of life. Although our data from 2 foals with more frequent sampling appears to demonstrate a progressive diversification of microbial flora with age ( Figure 5), further studies using more foals with more frequent sampling times are needed to better characterize microbial diversification. Our focus on the first month of life was based on current understanding that vaccination of foals against R. equi will have to occur during early life [53].
In conclusion, no differences were observed in the fecal microbiome of foals following enteral vaccination with either live or inactivated R. equi. These results demonstrate that administration of the doses of bacteria used in this study does not likely cause an alteration of the fecal microbiome of foals. More notably, the results indicate significant age-related changes in the microbiome composition of foals during the first month of life.