The influence of clinical severity and topical antimicrobial treatment on bacteriological culture and the microbiota of equine pastern dermatitis

Abstract

Background

Equine pastern dermatitis (EPD) is a common dermatological problem in horses, yet its aetiology and pathogenesis are poorly understood.

Objectives

This study aimed to investigate the effects of lesion severity and topical antimicrobial treatment on bacterial flora of EPD‐affected skin.

Animals

Sixteen horses with EPD were investigated.

Methods and materials

An observational study was conducted by assigning a clinical severity score ranging from 0 (macroscopically nonlesional) to 21 (severe), and sampling the most and least severely affected limbs of 16 horses (32 limbs) for bacteriological culture and 16S rRNA sequencing. Topical antimicrobial treatment in the month before sampling was recorded. The limbs were allocated to a nonlesional or mildly affected group (Group A, score 0–3) and a moderate to severely affected group (Group B, score 4–21).

Results

The most commonly cultured bacterial species was Staphylococcus aureus (one of 15 Group A versus nine of 17 Group B). Within Group B, S. aureus was found in three of six limbs treated with topical antimicrobials and in six of 11 untreated limbs. β‐haemolytic streptococci (three of 32) and Trueperella pyogenes (two of 32) also were cultured exclusively in the untreated limbs of Group B. Staphylococci and streptococci were found more often by 16S rRNA sequencing than in culture. Limbs with higher lesion severity and topical antimicrobial treatment appeared to have a lower alpha diversity and different beta diversity compared to milder and untreated lesions.

Conclusions and clinical importance

Observed differences in microbiota of equine skin are likely to be linked to the presence and severity of EPD and topical antimicrobial treatment. Further research is needed to establish causal bacteria.

Background –Equine pastern dermatitis (EPD) is a common dermatological problem in horses, yet its aetiology and pathogenesis are poorly understood. Objectives –This study aimed to investigate the effects of lesion severity and topical antimicrobial treatment on bacterial flora of EPD‐affected skin. Conclusions and clinical importance – Observed differences in microbiota of equine skin are likely to be linked to the presence and severity of EPD and topical antimicrobial treatment. Further research is needed to establish causal bacteria.

Introduction

Equine pastern dermatitis (EPD) is a cutaneous reaction pattern ¹ marked histologically by epidermal hyperplasia and orthokeratotic hyperkeratosis. ²

The development of EPD is complex and influenced by a variety of different factors and pathogens. ³ , ⁴ , ⁵ , ⁶ Bacterial colonization is thought to play a substantial role in initiating the pathogenesis or perpetuating the syndrome of EPD. ³ , ⁴ , ⁵ , ⁶

The current method to analyse bacterial flora associated with EPD in clinical practice is bacteriological culture. An alternative method that has become increasingly popular in human and veterinary medicine is next‐generation sequencing (NGS) of the bacterial 16S rRNA gene. In human medicine, the time to healing was shortened by 22.9% with the help of NGS diagnostics compared to culture. ⁷

Macroscopically normal equine skin has been analysed by NGS in two studies. ⁸ , ⁹ To our knowledge, the skin microbiota of horses affected by EPD has not been analysed with NGS to date. The aims of this study were to evaluate the microbiota of EPD lesions using bacteriological culture and NGS, and to describe the microbiota of lesions with different severity and with or without topical antimicrobial treatment.

Methods and materials

A detailed description of the materials and methods used to conduct this study can be found in File S1 and Table S1.

Study design and horses

Horses with EPD were recruited for this observational study between September 2017 and April 2018. The study protocol was approved by the veterinary ethics committees of all 26 cantons of Switzerland (approval no. VD3297). Horses that had received systemic antimicrobials were excluded; there was no withdrawl period for topical treatment. All limbs were scored individually using a standardized clinical scoring system (Table S1) with a severity score ranging from 0 (macroscopically nonlesional) to 21 (severe). The most and least affected limb of each horse was allocated to one of two clinical groups (Group A, score 0–3, or Group B, score 4–21) and sampled for bacteriological culture and NGS.

Bacterial culture

Direct culture conditions were targeted to known pyogenic bacteria. After incubation, bacterial species were confirmed by matrix assisted laser desorption ionisation time‐of‐flight mass spectrometry (MALDI‐TOF MS). The bacterial growth was assessed semiquantitatively: one plus (+) equalled <30 colonies per plate, two plus (++) 30–100 colonies per plate, and three plus (+++) >100 colonies per plate.

Sequencing of the 16S rRNA gene

Skin swabs for NGS were processed as described previously. ¹⁰ The amplification was performed on the V4 region of the bacterial 16S rRNA gene. The threshold for sequencing was 1 ng/µL. Negative control samples were processed with the other samples. They produced <1 ng/µL DNA and were thus classified as negative and not sent for sequencing. The samples were then indexed and 2 x 250 bp paired‐end sequenced on the Illumina MiSeq platform (Illumina Inc.; San Diego, CA, USA) by the genetics department of the University of Berne, Switzerland. Various procedures were performed for quality filtering. Sequences are available at the NCBI Sequence Read Archive (SRA) under the accession number PRJNA647995. The amplicon sequence variants (ASV) were further analysed in the programs excel 2010 (Microsoft Corporation; Redmond, WA, USA) and R ¹¹ (v3.4.3, base, vegan and microbiome packages).

Descriptive statistics

Descriptive statistics were performed using NCSS 12 (NCSS Statistical Software; Kaysville, UT, USA, ncss.com/software/ncss) and R. Total sequences in all samples, mean sequences per sample, standard deviation (SD) and range were calculated in excel. Alpha diversity indices of groups A and B were investigated; within Group B the lesions treated with antimicrobials were compared to the untreated lesions. A permutational multivariate ANOVA (PERMANOVA) using the distance matrices was performed in R to detect significant differences in group centroids for groups A and B, and for lesions treated or not treated with antimicrobials.

Results

Horses, samples and characteristics

A total of 16 horses with EPD were included in this study (Tables S2 and S5).

All limbs of each horse were clinically evaluated except for one right forelimb of one horse that was in a cast (individual J). The characteristics of the lesions are documented in Table S2.

Thirty‐two limbs were sampled in total, of which 17 were allocated to group B and 15 to group A based on their clinical severity score. Six of 32 samples collected from limbs of four horses had been treated with topical antimicrobials, all of which were in Group B. The median score of untreated lesions in Group B was 9 (range 6–12) and the median score of the treated lesions was 8 (range 5–17).

Bacterial culture

Staphylococcus aureus, β‐haemolytic streptococci and Trueperella pyogenes were the opportunistic pathogens detected (Table 1). The prevalence of S. aureus was more than seven‐fold higher in Group B (nine of 17) than in Group A (one of 15). Within Group B, there was no difference in the prevalence of S. aureus between treated and untreated lesions (three of six and six of 11, respectively). β‐haemolytic streptococci and T. pyogenes were found only in untreated lesions of Group B.

Table 1.

Bacterial culture results from 32 samples taken from the pastern skin of 16 horses with equine pastern dermatitis (EPD).

AT+	Group A	Group B
	n = 0	n = 6
Mixed flora	0	4
S. aureus	0	3
AT−	n = 15	n = 11
Mixed flora	14	10
Staphylococcus aureus	1	6
β‐haemolytic streptococci		3
Trueperella pyogenes		2

Samples from limbs that were mildly affected (severity score <4 of 21) were assigned to Group A and those from limbs with more severe lesions (score of ≥4) were assigned to Group B.

AT+, treated with topical antimicrobials in the month before sampling; AT−, not treated with topical antimicrobials in the month before sampling.

Taxonomy of the microbiota

All samples had DNA concentrations >1 ng/µL. Of the 32 samples pooled, a total of 14,236 different ASVs were found [total sequences: 2,526,500, mean ± SD sequences/sample: 78,953 ± 54,344 (range 1,009–301,094)], which were classified into 41 phyla, 102 classes, 220 orders and 357 families. As all rarefaction curves reached their plateau, sequencing depth was found to be sufficient and further rarefaction of samples was therefore dismissed (Figure S1). ¹² Main phyla across all samples consisted of Proteobacteria (34.6%), Actinobacteria (19.4%), Firmicutes (16.6%), Bacteroidetes (10.4%), Kiritimatiellaeota (5.3%) and others with lower abundances. The distribution of the phyla in the 32 samples can be seen in Figure 1 and Table S3, and in greater detail in Figure S2; the distribution of the bacterial families is shown in Table S4.

Figure 1.

Overview of the 16S rRNA sequencing results of 32 samples taken from the pastern skin of 16 horses with equine pastern dermatitis (EPD). Bar graphs show the taxonomic composition on the level of phyla for all samples, Group A, Group B, and within Group B, the samples from skin treated with topical antimicrobials (AT+) and without treatment (AT‐) in the month before sampling.

Samples from limbs that were mildly affected (severity score <4 of 21) were assigned to Group A and those from limbs with more severe lesions (score ≥4) were assigned to Group B. All amplicon sequence variants (ASV) with a relative abundance <1% in all bars are pooled in “Others”. Unclassified bacterial phyla, ASVs that could not be allotted to a phylum.

The 10 bacterial genera with the highest relative abundance across all samples consisted of Acinetobacter (6.9%), Sphingomonas (5.1%), Corynebacterium (2.7%), Pantoea (2.6%), Staphylococcus (2.6%), Moraxella (2.1%), Rothia (2.0%), Psychrobacter (2.0%), Actinobacillus (2.0%) and Fusobacterium (2.0%). The 10 main genera in Group A were Sphingomonas (8.5%), Pseudomonas (2.5%), Rothia (2.4%), Pantoea (2.3%), Acinetobacter (2.1%), Chryseobacterium (2.0%), Pedobacter (1.6%), Psychrobacter (1.6%), Massilia (1.4%) and Methylobacterium (1.3%). In Group B, the genera consisted of Acinetobacter [11.1% (within Group B: 4.7% no antimicrobial treatment, 22.9% with antimicrobial treatment)], Corynebacterium [4.5% (within Group B: 3.4% no antimicrobials treatment, 6.4% with antimicrobial treatment)], Staphylococcus [4.4% (6.2%, 1.1%)], Fusobacterium [3.7% (5.6%, 0.1%)], Moraxella [3.6% (4.6%, 1.8%)], Actinobacillus [3.4% (3.1%, 3.9%)], Pantoea [3.0% (0.1%, 8.1%)], Streptococcus [2.5% (2.2%, 2.9%)], Psychrobacter [2.3% (3.4%, 0.4%)] and Sphingomonas [2.0% (1.5%, 3.0%)].

A total 180 of 14,236 (1.3%) ASVs could not be allocated to a phylum and constituted 0.1% of the relative abundance of the pooled samples. Concerning bacterial families, 3,400 ASVs (23.9%) remained unclassified (12.3% of the relative abundance). At the genus level, 7,436 (52.2%) ASVs and 22.0% of the relative abundance remained unclassified.

Alpha and beta diversity

Figure 2 shows the dot plots of indices for alpha diversity for groups A and B and antimicrobial treatment within Group B. Mean values for alpha diversity appear lower in Group B (richness = 867.8, Pielou’s evenness = 0.7, Shannon = 4.2, Simpson = 0.9) compared to Group A (richness = 1,034.1, Pielou’s evenness = 0.8, Shannon = 5.4, Simpson = 1.0). Treated lesions within Group B (richness 685.8, Pielou’s evenness 0.6, Shannon 3.8, Simpson 0.8) also seemed lower on visual inspection compared to untreated lesions in the same group (richness 978.0, Pielou’s evenness 0.7, Shannon 4.5, Simpson 0.9).

Figure 2.

Dot plot showing the richness, Pielou’s evenness, Shannon index and Simpson index of 32 samples taken from the pastern skin of 16 horses with equine pastern dermatitis (EPD).

AT+, samples in Group B with topical antimicrobial treatment in the month before sampling; AT−, samples in Group B without topical antimicrobial treatment in the month before sampling.

Figure 3 shows the beta diversity of groups A and B and of topical treatment or no treatment. The weighted and unweighted Bray–Curtis index showed a difference in centroids between groups (PERMANOVA; P = 0.007 and P = 0.005, respectively; Figure 3). Antimicrobial treatment showed a difference (PERMANOVA) in the unweighted Bray–Curtis index (P = 0.032), and not in the weighted Bray–Curtis index (P = 0.112).

Figure 3.

Spider plots showing the weighted (quantitative dissilimarity) and unweighted (qualitative dissimilarity) Bray–Curtis indices (beta diversity indices) of 32 samples taken from 16 horses affected by equine pastern dermatitis grouped by their clinical scores [Group A (macroscopically nonlesional or mild disease, score 0–3) and Group B (moderate to severe disease, score ≥4)]. This figure shows dissimilarities between samples.

(a) Qualitative dissimilarity between groups A and B. (b) Quantitative dissimilarity between groups A and B. (c) Qualitative dissimilarity between samples taken from horses treated and not treated with antimicrobials within Group B. (d) Quantitative dissimilarity between samples taken from horses treated and not treated with antimicrobials within Group B. AT+, samples in Group B with topical antimicrobial treatment in the month before sampling; AT–, samples in Group B without topical antimicrobial treatment in the month before sampling; CAP1, constrained analysis of principal coordinates, one eigenvalue for constrained axes; MDS, multidimensional scaling, one eigenvalue for unconstrained axes.

Comparison of the bacterial culture and NGS

The term “mixed flora”, which was found in 28 of 32 cultures in addition to specific bacteria, indicates that more than three bacterial phenotypes were grown and were not further specified. Therefore, only the specific bacteria found in culture were compared to the microbiota (Table S5).

Staphylococci sequences could be found in 12 of 15 Group A samples and 14 of 17 Group B samples. Within Group B, sequences were found in six of six treated and eight of 11 untreated samples. The similar prevalence of staphylococci sequences in groups A and B does not agree with the higher prevalence in Group B observed in the culture. Streptococci sequences were detected in nine of 15 Group A samples and all (17 of 17) Group B samples, while the culture detected streptococci in only three untreated samples of Group B. Trueperellae sequences were found only in two of 15 Group A samples and four of 17 Group B samples. Within Group B, sequences were found in one of six treated and three of 11 untreated samples.

Staphylococci accounted for 2.6% of the mean relative abundance of the microbiota of all samples [0.6% Group A, 4.4% Group B (within Group B: 6.2% no antimicrobial treatment, 1.1% with antimicrobial treatment)], streptococci for 1.5% [0.5% Group A, 2.5% Group B (2.2%, 2.9%)] and trueperellae for 0.1% [0.2% Group A, 0.1% Group B (0.1%, 0.1%)].

Discussion

Equine pastern dermatitis affects the bacterial composition of the skin. Specific opportunistic pathogens (S. aureus, β‐haemolytic streptococci and T. pyogenes) were cultured more often from more severe lesions (Group B). Staphylococcus aureus was cultured equally from lesions that were treated or untreated with antimicrobials, while streptococci and T. pyogenes were found exclusively in samples from untreated lesions. In the NGS results, although staphylococci and streptococci had a higher relative abundance in the microbiota of Group B, interestingly within Group B, staphylococci were found more often in the untreated lesions. The alpha diversity seemed to be lower in more severe lesions (Group B), and within this group even lower if they were treated with antimicrobials. However, these observations were not evaluated with statistical analysis.

It was not possible to classify a substantial amount of ASVs at the genus level, which is most probably a consequence of the lack of data on equine skin in the databases used rather than sequencing errors. Many of the found ASVs may be unique to equine skin and warrant further studies to add information to these databases. Bearing in mind that 52.2% of the overall relative abundance could not be allocated to a specific genus, some of the 10 genera with the highest relative abundance, especially in Group A, often are found in soil. Ross et al. also found that a large proportion of the main phyla that they sequenced were associated with soil microbes and discussed how although this was likely owing to regular contact with the environment, there was a possibility that these genera belonged to the actual skin flora. ⁸ The geographical location of the animals tested in their study had a greater influence on the skin microbiota than the anatomical location of the sampling site, although the mammalian order of the horse (Perissodactyla) showed a stronger difference between body regions than many other mammalian orders. ⁸ Kamus et al. found that anatomical location had a strong effect on bacterial skin composition. ⁹ They discovered that experimentally induced injured skin wounds were strongly associated with Fusobacterium and Actinobacillus in the initial healing phase. These two genera also were more dominant in Group B than in Group A in our samples. They also found more diverse communities in thoracic wounds compared to limb wounds, and in unbandaged limb wounds compared to their bandaged counterparts. After healing was completed, the skin microbiota had a similar composition to the controls with a higher diversity, providing evidence that healthy skin has a stable microbial composition. ⁹ This conclusion is in accordance with our samples, where the less severely affected samples seemed to have a higher diversity.

The relative abundance of staphylococci in NGS results appears to be higher in more severely affected limbs as well as in untreated lesions. These findings are supported by studies conducted on humans, mice and dogs, which found that the prevalence of S. aureus was higher in more severely inflamed samples ¹³ , ¹⁴ and lower in skin sites treated with antimicrobials. ¹⁴

The main limitation in our study was the small sample size and the inability to pair the samples for each horse, as some individuals did not have a limb meeting the requirements for Group A and another for Group B. The limited number of samples prevented the evaluation of possible confounders such as individual, sex and environment. Also, our study did not include healthy horses as controls. For these reasons, statistical analysis was not performed on our data. In addition, the horses were not sampled during the same time of year and there were more forelimbs in Group A and hind limbs in Group B. This bias may have influenced the results.

Contamination by environmental bacterial DNA in laboratory reagents and extraction kits is a common issue in NGS. ¹⁵ In order to investigate this possibility, the negative control samples would have needed further sequencing. A study describing the microbiota of healthy equine pastern skin would have been valuable in addressing this problem. Further study with more samples, a control group without macroscopic lesions, and more narrowly defined clinical groups will help enable us to assess our findings with better statistical power.

In conclusion, the results of this study suggest that the differences in microbiota of equine skin are likely linked to the presence and severity of EPD and treatment with antimicrobial agents. Bacterial composition and antimicrobial treatment may play a relevant role in the development or perpetuation of EPD, and warrant further study. Further study also is required to investigate the role of causal bacteria and reduction in bacterial diversity in EPD in order to enhance the efficacy of therapy.

Supporting information

Figure S1. Rarefaction curve for 32 samples taken from the pastern skin of 16 horses with EPD.

Figure S2. 16S rRNA sequencing results of 32 samples taken from the pastern skin of 16 horses with EPD.

Table S1. Standardized scoring system ranging from 0 (macroscopically nonlesional) to 21 (severe) used to evaluate the severity of EPD by skin pathologies commonly associated with EPD

Table S2. Characteristics of interest from 16 horses affected by EPD grouped by their clinical scores [Group A (macroscopically nonlesional or mild lesions, severity score <4 of 21) and Group B (moderate to severe lesions, severity score ≥4)] and antimicrobial treatment [treated (AB+) or untreated (AB–)].

Table S3. 16S rRNA sequencing results from 32 samples taken from the pastern skin of 16 horses with EPD.

Table S4. 16S rRNA sequencing results from 32 samples taken from the pastern skin of 16 horses with EPD

Table S5. Comparison of the results from bacteriological culture to NGS from 32 samples taken from the pastern skin of 16 horses with EPD

File S1. Material and Methods describing the study design, clinical NGS of the 16S rRNA gene processing, and descriptive analysis.