Nasal bacterial microbiota during an outbreak of equine herpesvirus 1 at a farm in southern Ontario

Abstract

The objective of this study was to investigate the nasal bacterial microbiota of healthy horses and horses infected with equine herpesvirus 1 (EHV-1). The nasal bacterial microbiota of 10 horses infected with EHV-1 and 11 control horses from a farm experiencing an outbreak was characterized using the Illumina MiSeq platform targeting the V4 region of the 16S ribosomal RNA gene. The nasal bacterial microbiota of healthy horses and EHV-1 horses was significantly different in community membership and structure. Horses shedding EHV-1 had lower bacterial richness (P = 0.002), evenness (P = 0.008), and diversity (P = 0.026) than healthy horses. Healthy horses had a higher relative abundance of Firmicutes and Bacteroidetes, but lower Proteobacteria than horses with EHV-1 (P < 0.05). This study provides the basis for generating hypotheses and investigations on the role of bacterial-viral interactions in the health and diseases of adult horses.

Introduction

Equine herpesvirus (EHV) infection in horses can result in respiratory disease, myeloencephalopathy, abortions, and perinatal death. Infection of the upper respiratory tract (URT) with EHV results in colonization of the nasal epithelial cells (1). These processes likely involve a complex synergistic and competitive interaction between resident microbes, particularly between bacteria and viruses. Viral infections of the URT can modify the bacterial adherence and colonization and immune response of the host (2,3). Infection with EHV-1 can affect the mechanisms of defense of the URT, leading to respiratory diseases caused by opportunistic bacteria species that are part of the URT microbiota, e.g., Streptococcus equi subsp. zooepidemicus, Pasteurella spp., Actinobacillus spp., Klebsiella spp., Escherichia coli, and Bacteroides spp.

In children, virus-specific nasopharyngeal bacterial shifts have been identified during viral infections (3–5). The influence of viral colonization of the nasal epithelium on the whole bacterial composition of the URT in horses is unknown. Understanding the interaction between viruses and bacteria in the URT during infection may provide insights into the pathogenesis of respiratory infection. The structure, as well as the functional roles, of the nasal microbiota of horses undergoing viral infection should therefore be examined in more detail. The objective of this study was to profile the nasal bacterial microbiota of healthy horses and horses infected with EHV-1 from a single farm that was experiencing an EHV-1 outbreak.

Materials and methods

The farm, animals, and outbreak

During the winter of 2017, a Quarter Horse farm in southern Ontario, Canada experienced an outbreak of a neuropathogenic strain of EHV-1. Thirty adult horses and 1 adult donkey resided at the farm. Nineteen horses were kept in 1 barn, in individual stalls that allowed nose-to-nose contact, with access to outdoor paddocks during the day. Eleven horses and 1 donkey remained outside 24 h a day in 3 different outdoor paddocks, with nose-to-nose contact. Food and water sources were similar for all horses on the farm. There was no direct contact between horses that were kept in the barn and those that were housed outside at any time during the outbreak. There were 2 pregnant mares in their 9th month of gestation in the indoor group.

The farm was immediately quarantined when the first case was confirmed EHV-1 positive and all horses remaining on the premises were monitored twice a day for the presence of fever, limb edema, and nasal discharge. Fever was defined as a rectal temperature above 38.5°C.

Sample collection

Two nasal swabs were taken from all horses (healthy and those developing clinical signs compatible with EHV-1 infection) during the outbreak, 1 for diagnostic and another for microbiota analyses. A polymerase chain reaction (PCR) assay was used to screen all nasal swabs for the presence of EHV-1 genomic DNA. Nasal swabs of the horses that developed clinical signs were collected within the first 48 h of the onset of fever using a 15-mm long sterile nylon swab without transport media (FLOQSwabs; Copan Flock Technologies, Bescia, Italy). All nasal swabs were taken after wiping the nostrils with a sterile 0.9% saline solution, as described in a previous study (6). Nasal swabs were transported in a container at 4°C within 2 h of collection and stored at −80°C until analysis. Negative controls (swab only, without sample or sterile saline) were not obtained during sampling on the farm or introduced throughout the extraction, amplification, and sequencing process.

EHV-1 droplet-digital PCR assay

The DNA was extracted from nasal swabs using the OMEGA E.Z.N.A Tissue DNA Kit (Omega Bio-Tek, Norcross, Georgia, USA) as per manufacturer’s instructions, with some modifications. The first 3 buffers were used at volumes 1.5× the amount listed in the instructions so that the fluid level completely covered the swab [300 μL of TL Buffer, 37.5 μL of OB Protease Solution, and 330 μL of BL Buffer (all from Omega Bio-Tek)]. Following the BL Buffer step, all volumes were as per the manufacturer’s instructions. The optional RNA removal step was not carried out.

Droplet digital PCR (ddPCR) was conducted by the Guelph Molecular Supercentre, Lab Services Division at the University of Guelph according to the following conditions. The ddPCR reaction mixture consisted of 1× ddPCR Supermix for Probe (Bio-Rad, Mississauga, Ontario), 96 nM each of the primers EHV1F (5′-CATGTCAACGCACTCCCA-3′), EHV1R (5′-GGG TCGGGCGTTTCTGT-3′), 64 nM of EHV1 probe (5′-FAM-CCC TACGCTGCTCC-MGB-NFQ-3′) (7), and 4 μL of sample DNA in a final volume of 25 μL.

From each PCR reaction mixture, 20 μL were mixed with 70 μL of Droplet Generation Oil for Probes (Bio-Rad) in a DG8 Cartridge (Bio-Rad). The PCR droplets were then generated using a QX200 Droplet Generator (Applied Biosystems, Foster City, California, USA). From each droplet mix, 20 μL were transferred to a 96-well PCR plate. The plate was sealed with a foil heat seal using a PX1 TM PCR Plate Sealer (Bio-Rad Laboratories, Hercules, California, USA). Each plate included up to 94 samples with 1 positive control (synthetic gene fragment) and 1 negative control (water). The PCR amplification was carried out on a GeneAmp PCR System 9700 at the following settings: 95°C for 10 min, followed by 48 cycles of 95°C for 20 s, and 62°C for 40 s, and 1 cycle of 98°C for 10 min.

After amplification, droplets from each well were read automatically on a QX200 Droplet Reader (Bio-Rad). The ddPCR data were acquired and analyzed with QuantaSoft software (Bio-Rad) and recorded as copies/microliter (μL). The final viral load [copies/milliliter (mL)] of each sample was determined by multiplying the average copy number per microliter (μL) of PCR mixture in each well and the sample dilution factor. A subset of samples was sent to the Animal Health Laboratory, University of Guelph, Ontario to determine the D752 variant of the DNA polymerase [open reading frame 30 (ORF30)] in the EHV-1.

DNA extraction, amplification, and sequencing of bacterial 16S rRNA gene

Total DNA was extracted from the nasal samples using a commercial kit (EZNA Stool DNA Kit, Omega Bio-Tek). The whole tip of the swab was processed through the lysis stage of extraction. The V4 region of the 16S ribosomal RNA (rRNA) gene was amplified using the forward (5′-AYTGGGYDTAAAGNG-3′) and reverse (5′-TACNVGGGTATCTAATCC-3′) primers. The primers included overhanging adapters (forward: 95′-TCGTCGGCA GCGTCAGATGTGTATAAGAGACAG-3′, reverse: 5′-GTCTCGT GGGCTCGGAGATGTGTATAAGAGACAG-3′) for annealing to Illumina universal index sequencing adaptors added in a later PCR. The reaction mixture and amplification conditions were described in a previous study (8).

The PCR products were purified using magnetic beads (Agencourt AMPure XP; Beckman Coulter, Mississauga, Ontario). Illumina universal adapters (forward: 5′-AATGATACGG CGACCACCG AGATCTACAC-index-TCGTCGGCAGCGTC-3′, reverse: 5′-CAA GCAGAAGACGGCATACGAGAT-index-GTCTCGTGGGCTC GG-3′) were then added to the purified 16S rRNA gene product by PCR. The PCR products were assessed by electrophoresis in 1.5% agarose gel and purified with magnetic beads (Agencourt AMPure XP, Beckman Coulter). Following purification, the PCR products were quantified using a NanoDrop Spectrophotometer (NanoDrop Technologies, Wilmington, Delaware, USA). Samples were normalized to a final concentration of 2 nM. The library pool was submitted to the Genomics Facility of the University of Guelph and sequenced with an Illumina MiSeq (Illumina, San Diego, California, USA) for 250 cycles from each end. Negative controls were not included during the sampling, extraction, or sequencing processes.

Bioinformatic analysis

The Mothur v1.38 software package (Michigan State University, East Lansing, Michigan, USA) was used for sequence processing and analysis. Paired end reads were merged to fully overlapping reads and then aligned to the SILVA 16S rRNA reference database (9). Sequences inconsistent with the target amplicon size (240 bp) or containing ambiguous base calls or those with runs of homopolymers > 8 base pairs were removed. Sequences misaligned with the V4 region were also removed. Chimeras were identified with UCHIME and then removed, as were sequences belonging to nonbacterial domains. The remaining sequences were assigned to operational taxonomic units (OTUs) using an open OTU-picking approach, with a distance limit of 97% similarity. The Ribosomal Database Project Classifier was used for taxonomy assignment (Michigan State University).

Statistical analysis

Subsampling was completed to normalize sequence number by random selection of a number of sequences that corresponded to the lowest number of reads for any sample. The relative abundances of the main phyla, classes, orders, and families (median relative abundance > 0.1%) and the main genera (median relative abundance > 0.05%) were then calculated and compared using the non-parametric Mann-Whitney U-test. P-values were adjusted for multiple comparisons with Benjamini-Hochberg False Discovery Rate using statistical software (R Core Team, 2013, R Foundation for Statistical Computing, Vienna, Austria) to generate q-values. A q of < 0.05 was considered statistically significant.

Good’s coverage value was used to assess sampling coverage (10). Richness, evenness, and diversity indexes were calculated using the Chao1, Shannon’s evenness, and inverse Simpson’s indexes, respectively. Comparisons among groups were made using Wilcoxon rank-sum test. The core microbiota was investigated by identifying genera with relative abundances of at least 1% in all samples from a group.

Similarities of the bacterial microbiota membership and structure were calculated using a metric based on OTU richness (Jaccard index) (11) or abundance (Yue and Clayton index) (12), respectively. Differences in community membership and structure among groups were assessed using the Unweighted-Unique Fraction Metric (Unweighted-UNIFRAC) analysis, parsimony test, the analysis of molecular variance (AMOVA), and the homogeneity of molecular variance analysis (HOMOVA). Dendrograms were created to visualize similarities among groups (FigTree v1.4.0.1. Institute of Evolutionary Biology, University of Edinburgh, Edinburgh, Scotland). Clustering of the groups was represented by principal coordinate analysis plotted using statistical software (JMP 12, SAS Institute, Cary, North Carolina, USA).

Linear discriminant analysis effect size (LEfSe) (13) was used to identify bacterial taxa that were enriched in nasal samples from control and EHV-1 horses, based on P < 0.05 and linear discriminant analysis (LDA) score of > 3.0. Dirichlet multinomial mixture model (DMM) was used to assess the number of different meta-communities into which the data could be clustered (14). The number of different meta-communities was determined by selecting the number of components that gave the minimum Laplace approximation to the negative log model. Samples were assigned to their community type based on the maximum posterior probability.

Results

The horses and the outbreak

During the outbreak, 18/19 horses kept in the barn developed clinical signs, including fever (14/18), limb edema (10/18), and nasal discharge (1/18). Two pregnant mares aborted during the outbreak. As 10/18 horses that developed clinical signs tested positive for EHV-1 in the ddPCR of nasal samples at the time of sampling, only 10 EHV-1 positive horses were enrolled in this study.

The anamnesis of the 10 EHV-1 positive horses was similar: clinical signs started with fever (n = 10) and then the horses developed limb edema (n = 6). Both mares that aborted did so after 2 d of fever, which generally lasted for 3 to 5 d, but was controlled after NSAIDs were administered. None of the 10 EHV-1 positive horses developed neurological signs. All samples positive for EHV-1 on ddPCR were also positive on quantitative PCR (qPCR) at the Animal Health Laboratory, University of Guelph. In addition, all horses with EHV-1 carried the D752 variant of the DNA polymerase (ORF30). None of the 11 healthy control horses had detectable levels of EHV-1 DNA in nasal samples.

Analysis of 16S rRNA gene sequencing

A total of 858 263 reads were obtained with a mean of 40 860 reads per horse (standard deviation: 21 132, median: 38 263, and range: 11 977 to 107 141). A random subsample of 11 977 reads per sample was used to normalize data. Subsampling was considered adequate, as evidenced by the coverage of 99.9% obtained for all samples.

Alpha diversity

Horses infected with EHV-1 had significantly lower bacterial richness, evenness, and diversity than control horses (Figure 1).

Figure 1.

A — Inverse-Simpson (diversity), B — Shannon-evenness (evenness), and C — Chao-1 (richness) indexes observed in healthy horses and horses infected with equine herpesvirus type-1 (EHV-1).

Relative abundance

A total of 33 different phyla were identified, but Proteobacteria, Firmicutes, Actinobacteria, Verrucomicrobia, Spirochaetes, Fibrobacteroides, and Bacteroidetes accounted for 90% of sequences (Figure 2). These phyla were identified in control and EHV-1 horses at > 1% of the total number of sequences. Comparison between control and EHV-1 horses identified a higher relative abundance of Firmicutes (P = 0.04), Bacteroidetes (P = 0.01), and Fusobacteria (P = 0.01) in control horses. However, high inter-individual variation was identified within each group (Figure 2A).

Figure 2.

A — Median relative abundance of the main bacterial phyla in nasal samples of healthy horses and horses infected with equine herpesvirus type-1 (EHV-1). B — Median relative abundance of the main bacterial genera in nasal samples of healthy horses and horses infected with EHV-1.

While 25 different classes, 43 orders, and 135 families were identified, only 11, 27, and 50 accounted for ≥ 0.1% of sequences overall, respectively. The relative abundance of bacterial taxa that were significantly different between control and EHV-1 horses is presented in Table I. The relative abundance of the most abundant bacterial genera identified in control and EHV-1 horses is presented in Table II.

Table I.

Relative abundance (median in percentage and ranges) of selected^* bacterial taxa identified in nasal samples of healthy horses and horses infected with equine herpesvirus 1 (EHV-1).

Taxa	Control	EHV-1	q-value
Phylum
Bacteroidetes	1 (0.5 to 2.9)	0.3 (0.1 to 3.2)	0.019
Fusobacteria	0.2 (0.02 to 1.8)	0.07 (0 to 0.2)	0.019
Firmicutes	36 (20 to 53)	20 (2.8 to 50)	0.047
Class
Fusobacteriales	0.2 (0.02 to 1.8)	0.08 (0 to 0.2)	0.017
Sphingobacteriales	0.2 (0.04 to 7.4)	0.2 (0.03 to 1.6)	0.017
Aeromonadales	0.2 (0.1 to 1)	0.06 (0 to 0.3)	0.021
Selenomonadales	2 (0.6 to 5.8)	0.6 (0.1 to 2.2)	0.049
Family
Dermabacteraceae	0.01 (0 to 0.08)	0.1 (0.03 to 0.77)	0.04
Fusobacteriaceae	0.2 (0.02 to 2)	0.08 (0 to 0.2)	0.04
Succinivibrionaceae	0.2 (0.08 to 1)	0.06 (0 to 0.3)	0.04
Alteromonadaceae	0.04 (0.01 to 0.4)	0.01 (0 to 0.3)	0.04
Flavobacteriaceae	0.07 (0.03 to 1.4)	0.03 (0 to 0.3)	0.04
Sphingobacteriales_unclass.	0.06 (0 to 0.15)	0.01 (0 to 0.06)	0.04
Lactobacillaceae	2.5 (0.4 to 8)	0.6 (0.07 to 2.2)	0.048
Enterococcaceae	0.6 (0.3 to 6)	0.2 (0.02 to 1.4)	0.048
Clostridiaceae_2	0.08 (0.03 to 0.2)	0.01 (0 to 0.1)	0.048
Genus
Lactobacillus	2.3 (0.2 to 6)	0.5 (0.07 to 2.1)	0.03
Gallibacterium	0.5 (1.8 to 3.9)	0.1 (0 to 0.4)	0.028
Fusobacterium	0.1 (0.01 to 1.7)	0.06 (0 to 0.1)	0.028
Succinivibrio	0.2 (0.08 to 0.1)	0.04 (0 to 0.2)	0.028
Butyricicoccus	0.2 (0.1 to 0.5)	0.05 (0 to 0.1)	0.028
Megamonas	0.08 (0.03 to 1)	0.03 (0 to 0.4)	0.028
Lachnospiracea_incertae_sedis	0.1 (0.03 to 0.2)	0.01 (0 to 0.1)	0.028
Psychrobacter	0.3 (0.1 to 0.9)	1.4 (0.2 to 20)	0.032
Enterococcus	0.5 (0.2 to 5)	0.2 (0.02 to 1)	0.032
Ruminococcus	0.3 (0.05 to 1)	0.09 (0.01 to 0.2)	0.032
Viridibacillus	0.2 (0.1 to 1.2)	0.05 (0.02 to 0.5)	0.032
Clostridiaceae_1_unclass.	0.3 (0.1 to 0.7)	0.1 (0 to 0.3)	0.032
Selenomonas	0.1 (0.03 to 0.7)	0.04 (0 to 0.1)	0.032
Roseburia	0.1 (0.02 to 0.3)	0.04 (0 to 0.1)	0.032
Alkaliphilus	0.1 (0.02 to 0.2)	0.01 (0 to 0.1)	0.036
Peptostreptococcus	0.1 (0.01 to 0.5)	0.05 (0 to 0.3)	0.044
Lysinibacillus	0.05 (0.02 to 0.6)	0.02 (0 to 0.7)	0.044
Desulfovibrio	0.03 (0 to 0.02)	0 (0 to 0.02)	0.044
Succiniclasticum	0.05 (0 to 0.4)	0.01 (0 to 0.09)	0.044

^*Only significantly different taxa are presented.

q-value — P-value adjusted based on the Benjamini-Hochberg False Discovery Rate; unclass — an unclassified taxonomy within the respective taxonomic group.

Table II.

Relative abundance (median in percentage and ranges) of the 10 more abundant bacterial genera identified in nasal samples of healthy horses and horses infected with equine herpesvirus 1 (EHV-1).

Genus	Healthy	EHV-1	q-value
Nicoletella	2.4 (0 to 51)	41 (0.6 to 93)	0.09
Corynebacterium	0.8 (0.1 to 42)	1 (0.3 to 58)	0.34
Unclassified Clostridiales	4.4 (0.6 to 9)	1.3 (0.2 to 9)	0.07
5 genus incertae sedis	5 (0.2 to 7)	1.5 (0.1 to 11)	0.2
Moraxella	0.1 (0 to 13)	2.7 (0 to 20)	0.09
Unclassified Lachnospiraceae	3.1 (0.5 to 8)	0.7 (0.1 to 5)	0.05
Unclassified Ruminococcaceae	3.2 (0.5 to 7)	1 (0.1 to 4)	0.07
Treponema	2.3 (0.2 to 6)	0.6 (0 to 3.6)	0.09
Lactobacillus	2.3 (0.2 to 6)	0.5 (0.07 to 2.1)	0.03
Staphylococcus	0.3 (0.1 to 10)	0.8 (0.1 to 14)	0.13

q-value — P-value adjusted based on the Benjamini-Hochberg False Discovery Rate.

Overall, 850 genera were detected. A total of 126 of those were present at a relative abundance of > 0.05%. The relative abundance of the main bacterial genera found in control and EHV-1 horses is presented in Figure 2B, which shows the high inter-individual variation that was identified within each group.

LEfSe analysis: Enriched phylotypes

When comparing control and EHV-1 horses, 19 bacterial taxa enriched in control horses and 5 enriched in EHV-1 horses were identified. Enriched phylotypes in control horses were predominantly from the phylum Firmicutes, whereas most from EHV-1 horses were Proteobacteria and Actinobacteria (Figure 3). The genera Brachybacterium, Dietzia, Arthrobacter, Psychrobacter, and Moraxella were significantly associated with EHV-1 infection, whereas Lactobacillus, and unclassified genera of the family Clostridiales and Lachnospiraceae among others were enriched in control horses (Figure 3).

Figure 3.

Plot from linear discriminant analysis effect size (LEfSe) indicating enriched bacterial genera in nasal samples of healthy horses [right panel; values of linear discriminant analysis (LDA) score from 0 to 4] and horses infected with equine herpesvirus type 1 (EHV-1) (left panel; values of LDA score from 0 to -4). All taxa displayed in graph had a relative abundance (median) of > 0.05% in either or both groups of horses. Different colors represent phylum to which each genus belongs. All taxa had a relative abundance of ≥ 0.05% in either or both groups.

Bacterial community analysis

Nasal bacterial microbiota of healthy horses and horses with EHV-1 were significantly different in community membership (Jaccard index) and structure (Yue and Clayton index) (Parsimony and AMOVA; P < 0.05). These differences in community membership and structure were visualized by principal coordinates analysis (PCoA) plots (Figure 4). Nasal bacterial microbiota of healthy horses from different outdoor paddocks was similar in community membership (Jaccard index) and structure (Yue and Clayton index) (Parsimony and AMOVA; P > 0.05) (data not presented).

Figure 4.

Three-dimensional principal coordinates analyses (PCoA) of the community membership. (Jaccard index, A) and structure (Yue and Clayton index, B) of nasal samples of healthy horses (red) and horses infected with equine herpesvirus type 1 (EHV-1) (blue).

Discussion

It was found that the nasal bacterial microbiota of healthy horses is richer and more diverse than that previously reported using culture-based methodology (15,16). Proteobacteria, Firmicutes, and Actinobacteria phyla dominated the nasal microbiota of healthy horses, similar to what had been reported in humans (17), pigs (18), dogs (19), and cattle (20). The same taxa were identified as the more abundant phyla in nasopharyngeal samples of a small group of healthy horses (n = 6) in a previous study (6).

The most abundant genera in the nasal cavity of healthy horses in this study were Nicoletella, Treponema, Streptococcus, Staphylococcus, Lactobacillus, and genera from the order Clostridiales and the families Ruminococcaceae, Lachnospiraceae, and Clostridiales. Nicoletella spp. have been isolated from the respiratory tract of healthy horses of different ages and breeds and those with signs of bronchopneumonia and chronic airway inflammation (21–23). Most members of the family Pasteurellaceae occur as opportunistic organisms. In fact, Nicoletella has been associated with pulmonary disease in combination with other bacterium, e.g. Streptococcus equi var. zooepidemicus (22,23). Our study failed to detect differences in the relative abundance of Nicoletella in healthy and EHV-1 infected horses. Similar to our study, Streptococcus and Staphylococcus genera have been identified as major components of the nasopharynx microbiota of healthy horses (6,16), while Lactobacillus and genera from the order Clostridiales have been detected in nasal samples of healthy pigs (18).

Limitations of the results reported here include the lack of inclusion of negative controls during the microbiota assessment and the descriptive nature of the study with sampling at a single time point. The lack of inclusion of negative controls during the sample collection and processing limited the interpretation of our results. Several studies have reported the presence of contaminant DNA during microbiota analysis of low biomass samples and demonstrated how this can skew the results and confound their interpretation (24–26).

Currently, there are no systematic recommendations or requirements for examining or reporting contaminants within microbiota studies despite the information collected from different investigations describing amplification of contaminants in microbiota studies (26). Contaminant DNA can originate from the sampling, extraction, or sequencing processes. In order to monitor the background levels of contaminant DNA, the following controls should have been included: sampling blank controls, such as blank swabs and swabs exposed to the sampling environment; DNA extraction blank controls to monitor for contaminant DNA present in the extraction kits and reagents; and no template amplification controls to monitor for DNA contaminant originated from reagents and laboratory environment during library preparation and sequencing (26). The use of negative controls would have provided a semi-quantitative assessment of DNA contaminants to identify taxa that can be removed from the analysis or compared to nasal samples to determine their impact on the results reported here (26).

The descriptive nature of the study prevented us from extrapolating our results to the overall population as our results originate from a group of horses maintained in specific conditions at one particular time. The single time point sampling precluded the analysis of temporal changes in the bacterial communities of horses infected with EHV-1. A longitudinal characterization of the nasal microbiota of EHV-1 horses would have been ideal to determine whether those changes were transient or permanent.

Despite some statistically significant differences identified in this study, there was large variation between horses, which made comparisons between groups difficult. For instance, Proteobacteria was the most abundant phylum in some healthy horses, while Firmicutes or Actinobacteria predominated in others. Similarly, large individual variation was observed with regard to the most abundant genera. For example, while 5 of 11 healthy horses had a relative abundance of Nicoletella (> 25%), 6/11 had lower than 5% and, when the data were summarized, healthy horses had a low relative abundance of Nicoletella (< 5%) (Figure 2B).

This high biological inter-subject variability, especially in healthy subjects, can be a major confounding factor in microbiota studies (4,27). Our data highlight the fact that summaries of bacterial community, as seen in Figures 2A and 2B, must be interpreted cautiously, as they can hide the high degree of individual variability within each group and do not necessarily represent actual patterns seen at the individual level. Therefore, our analyses highlight the need to examine beyond simple comparisons of relative abundances when trying to interpret the microbiota because grouped summaries of relative abundance can result in spurious associations, especially when the sample is small.

There are no explanations for the high inter-individual variability, but differences in age, breed, diet, source of the diet components, pregnancy status, and exercise are associated with changes in the gut microbiota of different species. All horses involved in this outbreak were adults (> 3 y), Quarter Horses (except for 1 donkey), fed the same hay, and did not have access to pasture. The level of exercise was different among the horses reported here, which could potentially explain the variability among individuals. The nature of the results presented, however, prevents us from drawing further conclusions about the reasons for the high inter-individual variability.

The nasal bacterial microbiota of healthy horses and EHV-1 infected horses in the outbreak reported here differed significantly in community membership and structure. The differences detected between groups could have resulted from the viral infection, although housing was different between groups involved in this outbreak, and the environmental variable could therefore have confounded our results (20,28). Interestingly, the changes in alpha (lower richness and diversity) and beta diversity observed in the EHV-1 positive group were similar to those identified in humans after viral infection (4,5,27,29) and cattle with respiratory disease (30). While these findings suggest that the observed differences could have occurred as a consequence of EHV-1 infection, the environmental differences among horses should be considered when interpreting the results reported here (20,28).

The planning of how this outbreak investigation was conducted prevented us from drawing conclusions about temporal changes, but studies in humans have demonstrated that changes occur in bacterial composition of the nasal microbiota during the symptomatic rather than the asymptomatic phase of viral infection with human rhinovirus (5). This finding supports the hypothesis that the community structure of the nasal bacterial microbiota changes during viral infection, which is possibly explained by the overgrowth of potential pathogens triggered by the viral infection (5,27,28).

In horses, the most well-known interaction leading to respiratory disease is the synergism between influenza virus or EHV-4 and Streptococcus equi var. zooepidemicus (31). The postulated mechanisms explaining how viral infections can predispose to bacterial colonization include decreased mucociliary clearance, loss of barrier function, upregulation of adhesion proteins, altered expression of antimicrobial peptides, and impairment of neutrophil and monocyte function (32). It can be hypothesized that infection of the upper respiratory tract with EHV-1 in the horses reported here had an impact on the respiratory epithelium and immune system, which led to a shift in the bacterial communities present in the nasal cavity of affected horses.

Linear discriminant analysis effect size (LFfSe) identified enrichment of Moraxella, Dietzia, Psychrobacter, Brachibacterium, and Arthrobacter in EHV-1 infected horses. Arthrobacter, and Brachibacterium are genera of bacteria that are commonly found in soil, feces, and a wide range of moist cold habitats. Its presence appears to be associated with environmental contamination of the nasal cavity (33). However, the role of this likely transient bacteria microbiota in health remains to be determined.

Psychrobacter and Moraxella are a genera of the family Moraxellaceae that has been detected in the upper respiratory tract of healthy cattle (20) and horses (6). Although our analysis showed both Psychrobacter and Moraxella as different organisms, it is possible that both bacteria are in fact the same organism, as the low resolution of the next generation of sequencing makes it difficult to detect differences at the genera and species level. Similarly, Moraxella can be detected in the nasal and nasopharyngeal samples of healthy children (5), dogs (19), cats (34), and cattle (20). It has also been associated, however, with increased risk of developing asthma and pneumonia in children (35,36) and respiratory disease in cattle (37). A higher abundance of Moraxellaceae family members was found in humans after infection with human rhinovirus (5), acute respiratory infection (17), and viral colonization (38). A higher abundance of Moraxellaceae was also identified in cats with upper respiratory tract diseases (34). These findings support the hypothesis of persistent bacterial outgrowth, especially genera from the Moraxellaceae family, after viral infection (5,38). No studies have associated the presence of Moraxella spp. in horses with respiratory disease, but this bacterium can be isolated from horses with lymphoid hyperplasia (grades III and IV) (16) and in horses with conjunctivitis (39). Given the importance of this genus as a causative agent for disease in other species and in different organ systems in horses, further investigation is warranted in order to clarify the role of Moraxella during viral infection of the respiratory system of horses.

The genus Dietzia was also enriched in nasal samples of horses infected with EHV-1. The Gram morphology and colony appearance of the species of this genus is remarkably similar to Rhodococcus equi (40). While the pathogenic role of this bacterium has not been completely clarified, a recent report in human medicine suggested this genus as a possible opportunistic bacterium (40). Further studies are necessary to establish the role of these taxa in the homeostasis of the respiratory system of horses.

In conclusion, it was observed in this study that the nasal cavity of healthy horses is inhabited by a large variety of bacteria with significant individual variation. Acute infection with EHV-1 appears to affect the nasal microbiota, as changes were seen in the diversity, membership, and structure of the bacterial communities. However, given the limitations of this study, e.g., environmental differences between groups, experimental studies controlling for environmental conditions are needed to confirm our results.