The dysbiosis of ovine foot microbiome during the development and treatment of contagious ovine digital dermatitis

Background Contagious Ovine Digital Dermatitis (CODD) is an emerging and common infectious foot disease of sheep which causes severe welfare and economic problems for the sheep industry. The aetiology of the disease is not fully understood and control of the disease is problematic. The aim of this study was to investigate the polybacterial aetiopathogenesis of CODD and the effects of antibiotic treatment, in a longitudinal study of an experimentally induced disease outbreak using a 16S rRNA gene amplicon sequencing approach. Results CODD was induced in 15/30 experimental sheep. During the development of CODD three distinct phenotypic lesion stages were observed. These were an initial interdigital dermatitis (ID) lesion, followed by a footrot (FR) lesion, then finally a CODD lesion. Distinct microbiota were observed for each lesion in terms of microbial diversity, clustering and composition. Porphyromonadaceae, Family XI, Veillonellaceae and Fusobacteriaceae were significantly associated with the diseased feet. Veillonellaceae and Fusobacteriaceae were most associated with the earlier stages of ID and footrot rather than CODD. Following antibiotic treatment of the sheep, the foot microbiota showed a strong tendency to return to the composition of the healthy state. The microbiota composition of CODD lesions collected by swab and biopsy methods were different. In particular, the Spirochaetaceae family were more abundant in samples collected by the biopsy method, suggesting that these bacteria are present in deeper tissues of the diseased foot. Conclusion In this study, CODD presented as part of a spectrum of poly-bacterial foot disease strongly associated with bacterial families Porphyromonadaceae, Family XI (a family in Clostridiales also known as Clostridium cluster XI), Veillonellaceae and Fusobacteriaceae which are predominately Gram-negative anaerobes. Following antibiotic treatment, the microbiome showed a strong tendency to return to the composition of the healthy state. The composition of the healthy foot microbiome does not influence susceptibility to CODD. Based on the data presented here and that CODD appears to be the severest end stage of sheep infectious foot disease lesions, better control of the initial ID and FR lesions would enable better control of CODD and enable better animal welfare. Supplementary Information The online version contains supplementary material available at 10.1186/s42523-021-00078-4.


Introduction
Contagious ovine digital dermatitis (CODD) is an emerging, infectious foot disease of sheep rst reported in the UK in 1997 (1). Epidemiological surveys suggest that CODD now occurs on approximately 50% of UK farms (2,3), exhibiting a within farm prevalence of between 2 and 50% (3) and has now also been reported in Ireland (4), Germany (5) and Sweden (personal communication) with a very similar manifestation has now appeared in UK goats (6) and wild American elk (7). The clinical presentation of CODD is an in ammatory lesion on the dorsal coronary band which develops into progressive underrunning of the hoof horn in a distal direction, eventually resulting in avulsion of the entire hoof capsule (8). In animal welfare terms, CODD is the most severe form of sheep lameness (9). Control relies heavily on antibiotic treatments, in particular the macrolides which are categorized by the World Health Organization as a highest priority -critically important antimicrobials, essential for human health.
Therefore, due to the animal welfare and public health impact, CODD emergence has been identi ed as a priority disease issue for the sheep industry (10). The development of non-antibiotic control strategies such as vaccination and evidenced based biosecurity protocols is crucial and requires a greater understanding of CODD aetiopathogenesis.
Spirochetal bacteria are associated with CODD lesions, speci cally three members of the Treponema genus, namely bacteria of the Treponema medium phylogroup, Treponema phagedenis phylogroup and Treponema pedis (11) (12). These bacteria are also considered causal in bovine digital dermatitis (DD) (13), leading to the suggestion that DD treponemes may have crossed species from cattle to sheep to cause CODD.
However, CODD also shares substantial bacteriological and epidemiological features with footrot, a common sheep foot disease that is endemic in many countries worldwide. Studies of CODD affected feet identi ed the presence of the causal agents of footrot, Dichelobacter nodosus and Fusobacterium necrophorum (11). In addition, key risk factors for both diseases are remarkably similar with congruence acknowledged for wet underfoot conditions, poor biosecurity practices, foot trimming and ock size (14) (15). Footrot presence on farms and failure to apply footrot control measures such as vaccination and/or prompt individual treatment are also strongly associated with CODD occurrence (2,14,15). Therefore, current evidence on CODD aetiology requires clari cation as to whether CODD is an entirely novel infectious foot disease or whether it results from secondary invasion of pre-existing footrot lesions.
Furthermore, dissection of the precise roles of the different bacteria within CODD lesions during lesion development is needed and can help in the development of better treatment and control strategies.
In order to unravel the aetiology of CODD we assessed clinical lesion development and changes in the bacterial communities of sheep's feet during transition from healthy to CODD diseased state. To ensure accurate implication of responsible pathogens this study was novel in investigating a naturally occurring outbreak of disease in a previously CODD naïve ock in a controlled experimental environment.
The objectives of the current study were to:-1) Describe the clinical, phenotypic changes in the ovine foot during the development of CODD, 2) Describe the changes in the microbiome of the ovine foot during the development of CODD, 3) Describe the changes in the microbiome of the ovine foot following antibiotic treatment of CODD, 4) Determine whether differences in the healthy ovine foot microbiome predict susceptibility to CODD, 5) Compare the ovine foot microbiome of CODD lesions obtained by different lesion sampling methods.

Experimental Design (Longitudinal Experimental Study)
The project was carried out under UK Animal Scienti c Procedure Act (ASPA) 1986; Home O ce Project License PPL 708756 and University of Liverpool Ethics VREC417. The experimental study was supervised at all times by a Named Animal Care and Welfare O cer and a team of three veterinary surgeons. The reporting of the experiment is in accordance with the ARRIVE guidelines (16) (supplementary le 1).
The study design was an observational study of an experimentally induced outbreak of CODD in housed sheep whereby 30 healthy, 18-month-old, Texel cross ewes were housed with 10, mixed age and breed sheep affected by CODD. Inclusion criteria for the "healthy" ewes were that they should be sourced from a single ock with no known history of CODD and they should be of the same sex, breed and age. The inclusion criteria for infected sheep were that they were sourced from farms with a history of CODD in the ock and they should have a con rmed veterinary diagnosis of an active, untreated CODD lesion in one foot. At the start of the study all the infected sheep were PCR positive (13) for at least one of the hypothesised causal pathogens of CODD (Treponema medium phylogroup, Treponema phagedenis phylogroup, Treponema pedis). Sample size power calculations were not made for this study due to lack of data on the expected variation in the microbiological consortium; however, the sample sizes were consistent with other similar studies (17,18). The observational design of the experiment meant that it did not require blinding or randomizing.
Sheep were housed in a Home O ce Designated Building (according to UK Animal Scienti c Procedures Act, Code of Practice for Care and Accommodation of Animals) on deep litter straw bedding at a stocking rate of 1.9 m 2 /sheep. Sheep were fed a maintenance ration of ad libertum hay. A footbath was placed under the feed racks which contained damp straw, water, and contaminated hoof clippings from a CODD infected farm to simulate naturally occurring risk factors for CODD. Sheep welfare was monitored by daily inspection of demeanour and feed intake, twice weekly mobility and body condition score and weekly veterinary clinical examination. Humane endpoints were set and if an animal reached these predetermined points (inappetence, recumbency or non-weight bearing lameness on any limb) the animal was withdrawn from study. When half of the sheep in the ock had developed CODD lesions, all sheep with any foot lesion was treated with 2 doses, 48 hours apart, of a long acting amoxicillin (Betamox LA 150 mg/ml, Norbrook, Northern Ireland, UK) at dose rate of 10 mg/kg by intramuscular injection.

Animal Sampling (Longitudinal Experimental Study)
At the start of the project and during every week of study the following data/samples were collected from each sheep, mobility score (9), body condition score (19), foot lesion score of each foot (8) and foot skin swab from each foot, regardless of disease status. If no active lesion was present the swab (Copan, Brescia, Italy) was rubbed rmly across the interdigital skin and along the dorsal border of the coronary band. When a foot lesion was present the swab was applied to the entire surface of the visible lesion.
Collected swabs were immediately stored at -80 °C until DNA extraction. Animal metadata was stored on an Access Database (Microsoft; USA).
A subset of these samples was then selected for 16S rRNA gene amplicon sequencing. To study CODD lesion development foot swab samples were selected on the basis that the foot had undergone phenotypic lesion progression from healthy state to CODD lesion. For each sheep/ foot sample set, foot swabs were then further selected in order to represent the type of clinical lesion stages observed during the development of CODD. To study microbiome changes following treatment, an additional set of samples was included for each foot which were collected from sheep's feet two weeks post treatment and classed as healed (grade 5 CODD lesion (20)). In order to examine the differences in the microbiome of sheep's feet in animals that did and those that did not go onto develop CODD, foot swab samples were selected for comparison from sheep at the start of the study (on entry to the experimental unit) and from those who remained healthy throughout the experiment and those sheep that went onto develop CODD during the experiment.
Animal Sampling (Cross-sectional Farm Study) To compare CODD lesions swabs with CODD biopsy material, lesion biopsy samples were collected from 31 sheep diagnosed as having CODD by a veterinary surgeon (8). Prior to sampling, all feet were anaesthetized by in ltrating the local area with procaine hydrochloride with adrenaline (Adrenacaine, Norbrook). Prior to biopsy dry cotton swabs were drawn across the lesion. Surgical biopsies from the lesion were then obtained using a sterile 6 mm punch biopsy. After biopsy sampling all sheep were treated with antibiotics as prescribed by the farmer's own veterinary surgeon.
Extraction of Genomic DNA Given that we have previously established associations for DD treponemes with CODD (11) and that there are known issues surrounding detection of spirochaete DNA in microbiome studies, (21) (22) we used a genomic DNA extraction methodology with a known ability to enable treponeme detection from cotton swabs. Consequently, microbial DNA was extracted from swabs using the DNeasy Blood and Tissue Kits (QIAGEN, Manchester, UK) as previously described in other studies of ruminant DD 16S rRNA gene targeted microbiome studies (23) (24) (25). All steps were performed under sterile conditions. A negative control, not containing any sample material, was included with each extraction run. Extracted DNA samples were stored at -80 °C until use.

16S rRNA Gene Ampli cation and Illumina MiSeq Sequencing
The Qubit™ dsDNA HS Assay Kit (Thermo Fisher Scienti c, Fair Lawn, NJ, USA) was used to measure DNA concentrations before PCR. In addition to the swab samples, ZymoBIOMICS™ Microbial Community DNA Standard was used as a mock microbial community and included as a PCR positive control. Primers (26) were used to amplify the V4 region of the 16S rRNA gene with forward primer:

Sequencing Reads Quality Control and Filtering
Raw FASTQ les were trimmed to remove Illumina adapter sequences using Cutadapt version 1.2.1 (28).
Reads were further trimmed to remove low quality bases using Sickle version 1.200 (29) with a minimum window quality score of 20. After trimming, reads shorter than 20 bp were removed.

Amplicon Sequence Variant (ASV) Identi cation and Taxonomy Assignment
The dada2 plugin (30) tool was used to discriminate between bases that originate from errors (either from PCR or sequencing steps) with bases that correctly correspond with the original template sequences in the denoising step and was used to create a feature table using the QIIME 2 (31) feature table plug-in. Following alignment of sequences using MAFFT (32) the phylogenetic tree was built and converted to a rooted format using the Fast Tree tool (33). Taxonomy was assigned using the q2-feature-classi er plugin (34) with a pretrained NaiveBayes classi er based on the Greengenes database available for download at (https://greengenes.secondgenome.com/). Taxonomy pro le plots were produced using the QIIME2 taxa plug-in.

Bacterial Diversity Analysis
Alpha and beta diversity analyses were performed at a sampling depth of 40,000 sequences using the diversity core metrics-phylogenetic QIIME2 plug-in. This value was obtained using the diversity alpha rarefaction QIIME2 plugin and was chosen to ensure maximum sampling depth, whilst ensuring minimum sample loss (98.2% retained). Inspection of the resulting alpha rarefaction curve was used to ensure adequate sequencing depth. An observed ASV metric was used to assess species richness and compared between categories of samples using a Kruskal Wallis non-parametric test with a false discovery rate (FDR) correction. To study differences in beta diversity between groups, the rare ed abundance table was used to build pairwise sample distance matrices, using a weighted UniFrac (35) dissimilarity measure. The weighted UniFrac distance metric measures the distance between two samples considering the presence, phylogenetic distances and relative abundance of ASVs. Principal Coordinate Analysis (PCoA) was then used to visualize each distance matrix. Differences in microbiota beta diversity between groups in terms of location, dispersion or correlation structure were assessed for strength and signi cance using pairwise ANOSIM tests with a FDR correction (36).

Analysis of Differentially Abundant Amplicon Sequence Variants (ASV) Among Samples Groups (Gneiss Analysis).
Gneiss analysis (37) was run using QIIME2 (31) to identify alteration of microbial communities with disease progression. In Gneiss analysis, shifts in the balance (ratio of abundance) between subsets of the community rather than the absolute or relative abundance of community members are calculated. The ASV table was ltered to include the most abundant 500 ASVs then the ASV abundance data log transformed and the balance calculated as the log Principal balances for use in Gneiss analysis were obtained via Ward's hierarchical clustering using the correlation-clustering command producing a dendogram with 499 balances (y0-y498) created from the internal nodes of the dendogram tree. Isometric log ratios for each balance were calculated using the ilrtransform command. A multivariate response linear regression model of log ratios balances was constructed with disease status as the only covariate using the ols-regression command. Results were visualised through a regression summary and dendogram heatmaps. Balances signi cantly affected by the disease status were identi ed as those with an FDR corrected p-value less than 0.05. A Student's t-test was used to make pairwise comparisons in balance log ratios between disease states.
Signi cant balances were used to classify ASVs as associated with Healthy, Intermediate or Diseased sheep. ASVs not differentially abundant (NDA) between disease states formed a fourth group. The taxonomy of ASVs identi ed by Gneiss analysis as more abundant in Healthy, Intermediate or Diseased sheep was examined at family level. A chi-square test was used to assess the distribution of taxa across Gneiss analysis groups. The proportion of total ASVs classi ed to each group (Healthy, Intermediate, Diseased, NDA) was calculated. This proportion was applied to the total number of ASVs assigned to each taxon to generate an expected number of each taxon in each group. P-values were corrected for multiple tests using an FDR correction. The dataset was searched for pathogens previously associated with ID, Footrot and CODD such as Dichelobacter nodosus, Fusobacterium necrophorum and Treponema spp. to identify whether they were associated with the disease state.

Quality Control and Sequencing results
The number of reads per sample are summarised in Supplementary le 2. One DNA extraction negative control was ampli ed and sequenced, producing less than100 reads. As this was considerably fewer reads than the proccessed samples, the degree of contamination during DNA extraction was considered negligble and these sequences were not removed prior to analysis. Negative controls included during PCR steps indicated no contamination had occurred at this stage. 156 samples were included in the data set and the number of reads per sample was variable within and between sample types. The median number of reads per sample was 217,628 (Interquartile range,IQR: 103,638). A total of 16,177 different ASVs were identi ed and taxonomically assigned. The median number of ASVs per sample was 120,736 (IQR 48,811). Taxonomic analysis was primarily carried out at family level to minimise information lost due to unclassi ed samples at genera and species level.

Clinical Description of CODD Lesion Development
Only CODD lesions that progressed past CODD stage 2 (8) during the experimental study were included so as to avoid confusion with non-speci c injuries to the coronary band of the foot that can be confused with CODD grade 1 lesions. Applying this case de nition, 15 of the 30 sheep (50%) and 19 of the 120 feet (15.83%) in the study developed CODD lesions.
All these CODD lesions were observed to follow a speci c clinical pattern of lesion development consistent with the descriptions of footrot lesions as described by Egerton et al (38) and CODD lesions as described by Angel et al (8). The disease process began as initial interdigital dermatitis (ID), whereby the interdigital skin was in amed (Fig. 1A). This was the followed by the footrot (FR) stage with progressive underrunning of horn of the hoof beginning at the axial margin and extending across the sole (Fig. 1B).
Finally, at the CODD stage, an in ammatory lesion would be present at the dorsal coronary band which would progress to separate the hoof horn capsule from the underlying dermis in a ventral direction (Fig. 1C). The median survival time for a sheep to develop an ID lesion was 88 days, FR lesion 103 days and CODD lesion was 116 days from the start of the experiment.

Changes in the Sheep Foot Microbiome During the Development of CODD
Foot swab samples from 6 sheep in the study whose feet developed typical, progressive CODD lesions were selected for the analysis of changes in foot microbiome composition during the development of CODD lesions. These sheep were selected on the basis that the foot had undergone the previously described clinical progression, namely from a healthy foot to interdigital dermatitis to footrot and then to CODD. Samples included were those from: when compared with the healthy state (up to 2 weeks before disease onset), as CODD developed there was a reduction in the number of observed ASVs, consistent with dominance of disease associated bacterial species in the lesions across all foot disease states (p < 0.05) (Fig. 2). Pairwise comparisons were then used to identify differences in diversity between the different disease states observed as the clinical lesions progressed. Signi cant reductions in ASV numbers (p < 0.05) were observed when the foot moved from the B_ Healthy state (2 weeks before disease onset) to the ID stage and then to the FR stage. However, there was no difference in number of ASVs between the FR and CODD stages (Table 1).
Distinct clustering patterns of microbiota for each phenotypic stage observed in the development of CODD lesions were also observed in the beta diversity analysis (Fig. 3). When measured with a weighted Uni Frac metric, beta diversity in terms of location and spread was signi cantly affected by disease state overall (ANOSIM R statistic = 0.64329, p = 0.001). Furthermore, pairwise ANOSIM tests demonstrated signi cant differences (p = 0.001) between microbiota at each phenotypic lesion stage of CODD development, in terms of microbiota location, dispersion and correlation structure ( Table 2). Inspection of the dendogram heat map and examination of log ratio of balances clearly show y0 was signi cantly lower in A_Healthy than ID (β = 29.9, p < 0.001), Footrot (β = 74.0, p < 0.001) and CODD (β = 63.8, p < 0.001) samples (Fig. 5A) but was not signi cantly different from B_Healthy samples (β = 1.98, p = 0.9). The dendogram heatmap shows that broadly, y0 denominator ASVs were more abundant in healthy samples while y0 numerator were more abundant in diseased samples with ID samples acting as an intermediate.

Changes in Composition of
Subdivisions of y0 denominator ASVs reveal further differences between A_Healthy, B_Healthy and ID samples; these stages represent the earlier stages of the CODD disease development process and therefore examination of microbial compositional changes here give an indication of earlier changes in the bacterial community. The log ratio of balance y1 was signi cantly lower in A_Healthy samples compared to B_Healthy (β = 16.5, p < 0.001) and ID (β = 18.6, p < 0.001) samples (Fig. 5B). The dendogram heatmap shows that this was due to a higher abundance of some y1 denominator ASVs in A_Healthy samples, speci cally those further identi ed by balance y3 denominator (Fig. 5D). In addition, there was a higher abundance of some y1 numerator ASVs in B_Healthy and ID samples, the most abundant taxa identi ed by y4 denominator (Fig. 5E).
Balance y2 (Fig. 5C) is useful to describe differences in microbial composition as the foot lesions develop from ID stage as this balance was signi cantly lower in ID samples compared to Footrot (t = 10.25, p < 0.001) and CODD (t = 6.4, p < 0.001) samples. The dendogram heatmap shows that this was due to a higher abundance of some y2 denominator ASVs in ID samples, speci cally y5 numerator ASVs as demonstrated by a signi cantly higher log ratio in ID samples when compared to other groups (A_Healthy: t = 8.34, p < 0.001; B_Healthy: t = 3.25, p = 0.002; Footrot: t = 6.12, p < 0.001; CODD: t = 8.40, p < 0.001; Fig. 5F).
Differences in microbiome composition between footrot and CODD are observed in balance y14 whose log ratio was signi cantly higher in ID and footrot samples compared to CODD samples (t = 7.96, p < 0.001 and t = 13.3, p < 0.001 respectively) due to an increased abundance of y14 numerator ASVs in these samples (Fig. 5H).

2) Bacterial Taxa Associations with Changes in Disease State
The Gneiss analysis results were then used to classify the ASVs into four groups as follows: 1. Healthy ASVs -ASVs with higher abundance in A_Healthy samples (y3 denominator ) and ASVs with higher abundance in A_Healthy, B_Healthy and ID samples (y3 numerator and y4 numerator ); 2. Intermediate ASVs -ASVs with higher abundance in B_Healthy and ID samples (y4 denominator ) and ASVs with higher abundance in ID samples (y5 numerator ); 3. Diseased ASVs -ASVs with higher abundance in ID and Footrot samples (y14 numerator ), ASVs with higher abundance in ID, Footrot and CODD samples (y14 denominator ) and ASVs with higher abundance in Footrot and CODD samples (y6 denominator and y11 denominator ); 4. ASVs not differentially abundant between sample groups (y11 numerator ).
The families of ASVs identi ed by Gneiss as more abundant between disease states were plotted in a taxa bar plot to compare the relative abundance of the most abundant 10 families from both the Healthy and Diseased ASV groups (Fig. 6). The most abundant families in the healthy group were Moraxellaceae, Corynebacteriaceae Pseudomonaceae Saccharimonadaceae Acholeplasmataceae, Flavobacteriaceae, Ruminococcaceae Carnobacteriaceae Aerococcaceae, Family XI. Whilst in the diseased sheep the most abundant 10 families were Porphyromonadaceae, Family XI, Bacterioida Peptostreptococcaceae Fusobacteriaceae Lachnospiraceae, Wohlfahrtiimonadaceae, Ruminococcaceae Veillonellaceae, Acidaminococcaceae Examination of the distribution of ASVs in these families between Healthy, Intermediate and Diseased ASV groups demonstrated signi cant associations (p < 0.05) between speci c taxa, namely Porphyromonadaceae, Family XI, Veillonellaceae and Fusobacteriaceae and the Diseased ASV group (Table 3). Furthermore, the taxa plot shows the relative abundance of these four families increasing from Healthy samples to ID, footrot and CODD samples (Fig. 6), although an increased abundance of Veillonellaceae and Fusobacteriaceae was most associated with ID and footrot rather than CODD.
No other families were signi cantly differently distributed between Healthy, Intermediate and Diseased groups. This may be an artefact of small numbers of ASVs assigned to these families. For example, Bacteroidia was not identi ed as signi cantly differently distributed between groups even though 4 of 5 ASVs were assigned to the Diseased ASV group. The taxa plot (Fig. 6) shows that the relative abundance of Bacteroidia was especially increased in footrot samples. Equally, the majority of ASVs assigned to Corynebacterium, Pseudomonadaceae, Saccharimonadaceae, Acholeplasmataceae, Flavobacteriaceae, Carnobacteriaceae and Aerococcaceae were assigned to the Healthy ASV group with these families featuring prominently in the taxa plots of Healthy samples. Ruminococcaceae and Lachnospiraceae, families commonly associated with the intestinal microbiome, were more divided between Healthy and Diseased ASV groups. The taxa plot shows that a different pro le of Ruminococcaceae and Lachnospiraceae was found in Healthy samples compared to footrot and CODD samples.
Pathogens previously identi ed as important in the pathogenesis of ID, footrot and CODD were found within the dataset. One ASV assigned to Dichelobacter nodosus was classi ed as an Intermediate ASV and was found mainly in ID samples. Three ASVs assigned to Fusobacterium necrophorum were classi ed as Diseased ASVs and were found mainly in ID and footrot samples. Six ASVs were identi ed as Treponema with all six classi ed as Diseased ASVs which were present in ID and footrot samples. When these six treponeme sequences were compared with a wide range of relevant, previously isolated treponeme 16S rRNA gene sequences (39), three of the sequences had 100% sequence identity to the key DD associated treponemes, speci cally Treponema medium-phylogroup strain T19, Treponema phagedenis phylogroup strain T320A and Treponema pedis strain T3552B T . The further three treponeme sequences identi ed all shared 98.4-99.6% nucleotide sequence identity with the Treponema medium phylogroup strain T19.
Changes in the Sheep Foot Microbiome Following Antibiotic Treatment of CODD Foot swab samples from ve sheep in the study whose feet developed progressive CODD lesions and were subsequently treated (2 doses of long acting amoxicillin every 48 hours), were selected for the analysis of changes in foot microbiome composition following treatment. Antibiotic treatment of sheep was effective in achieving a clinical cure (CODD lesion stage grade 5 (20)) in all sheep within 7 days of initial treatment. Based on clinical presentation foot lesions were classi ed as A_Healthy (samples collected from sheep feet classed as healthy upon entry to study (n = 5)); B_CODD (samples collected from sheep's feet clinically assessed as having active CODD lesions (n = 21)); C_Treat (samples collected from sheep's feet 2 weeks post treatment and classed as healed, grade 5 CODD lesions (n = 5)).

Changes in Bacterial Diversity of Sheep Foot Microbiome (Alpha and Beta) Following Treatment of CODD
Changes in microbial community diversity for A_Healthy, B_CODD and C_Treated feet were measured by examining the number of observed ASVs (Fig. 7). Pairwise comparisons found signi cant reductions in observed ASV numbers between B_CODD feet and both A_Healthy feet (p < 0.01) and C_Treated feet (p < 0.001). However, no differences were observed in diversity of samples from A_Healthy feet and C_Treated feet (p = 0.62) indicating similar bacterial diversity in the microbiomes of the healthy and treated feet.
Similarly, when measured with a weighted UniFrac metric, the ANOSIM tests show beta diversity was signi cantly different between all three groups B_CODD and both A_Healthy, C_Treated feet (R test statistic 0.825846, p = 0.001)( Table 5). However, the PCoA plot shows that the A_Healthy and C_Treated feet tend to cluster together suggesting that following treatment the community clustering tends to return towards the healthy state (Fig. 8).

1) Identi cation of Balances Signi cantly Associated with Changes in Disease State
The overall linear regression model t was R 2 = 0.468 with covariate "Treatment" as the only variable.
B_CODD accounted for 17.44% of variance with C_Treat samples accounting for 37.83% of variance. The dendogram heat map (Fig. 9), in conjunction with balance analysis demonstrates a clear pattern of microbiome composition change as foot lesions move from healthy to CODD state and then healed state. There was a strong tendency for the microbiome of the healed state to return to that of the healthy microbiome. However, 3 log ratio balances y6 (B=-39.18, p < 0.001), y11(B = 7.74, p = 0.004) and y14 (B = 28.78, P < 0.001) showed signi cant differences in log ratios between the groups (Fig. 10A-C).

2) Bacterial Taxa Associations with Changes in Disease State
The log ratio of balance y6 (Fig. 10A) was signi cantly lower in C_Treated feet compared with the healthy feet due to higher abundance of some y6 denominator ASV in the treated feet compared to healthy feet.
These ASVs were mainly assigned to Ruminococcaceae, Lachnospiraceae, Carnobacteriaceae, Staphylococcaceae, Pseudomonadaceae. The log ratio balance of y11 (Fig. 10B) was signi cantly higher in C_Treated feet compared to the healthy state due to higher abundance of some y11 numerator ASVs which were assigned to Family XI, Corynebacteriaceae, Acholeplasmataceae, Staphylococcaceae, and Rhodobacteraceae. The log ratio balance of y14 (Fig. 10C) was signi cantly higher in C Treated feet compared to healthy state. The heatmap demonstrates here a clear decrease in y14 denominator ASVs in the treated feet which were assigned to Moraxellaceae, Saccharinonadaceae, Micrococcaceae, Microbacteriaceae and Pseudomonadaceae.

Lesions
To investigate associations between the healthy foot microbiome of the sheep and disease outcome, the differences in the microbial community diversity and composition of the feet of animals that did (case n = 8) or did not (control n = 8) go on to develop CODD lesions at the start of the study were compared using alpha and beta diversity metrics and Gneiss analysis.
Kruskall Wallis pairwise comparisons found no signi cant differences in ASV numbers between the foot microbiomes of the two groups (p = 0.207) (Fig. 10) and pairwise ANOSIM tests found no difference in the beta diversity metrics (R test statistic = -803, p = 0.803) (Fig. 11). Neither the dendogram heat map (Fig. 12) or Gneiss analysis of log ratio balances (y0-y100) revealed any signi cant microbiome compositional differences between sheep's feet which did or did not go onto to develop CODD at the start of the study.

Comparison of Microbial Taxa in CODD Lesions Sampled by Swab and Biopsy Methods
The majority of studies on the microbiology of CODD have been carried out using biopsy material from CODD lesions. Whilst recognising that different sampling approaches may introduce sampling bias and thus make comparisons of results between studies di cult, repeated invasive sampling of sheep was not possible in this longitudinal experimental study. Therefore, to investigate the extent of bias caused as a result of sampling method employed, we compared the microbiomes of sheep's feet diagnosed with clinical CODD sampled by swabbing (n = 28) with those sampled by biopsy (n = 31). From the taxa bar plots (Fig. 14), the top 15 taxa present in the two samples types were obtained. 11 bacterial families were common to both sample types with Porphyromonadaceae and Family XI the most abundant in both sample types.
The most notable difference between the two sample typesthat in the CODD swab samples there was the absence of the Spirochaetaceae family in the top 15 taxa; the family to which the putative causal organisms of CODD belong.

Clinical Description of CODD Lesion Development
This is the rst study to describe the changes in microbial communities of the feet of sheep occurring during the development of CODD and subsequently, following treatment, to the healed state All previous work describing CODD has been cross sectional in nature and has described the epidemiology of the disease (14) as well as allowing development of a ve point lesion scoring system to classify the clinical observations (8). However, the cross-sectional nature of these observations means that the full disease process cannot be observed and recorded. To collect such data repeat observations of cohorts of affected animals in longitudinal studies is required. To achieve this, an experimental study to induce CODD in healthy sheep was devised by replicating on farm factors which drive infection (14); namely, the presence of infected sheep and exposure to damp underfoot conditions (at the feed trough). We successfully induced CODD in 15 of the 30 experimental sheep. Clinically, all these sheep showed the same sequential pattern of lesion development which was from the healthy state via the development of ID lesions and thence via the development of footrot lesions to CODD lesions. The observed ID and footrot lesions were as described by (40) and the CODD lesions as described by (8).
Therefore, for this experimental study at least, the three disease states described (ID, footrot & CODD) are best considered as stages in a consistent spectrum of ovine infectious foot disease. This pattern of disease pathogenesis for CODD is surprising given that since its emergence CODD has been considered a separate disease from footrot due to its distinct appearance, failure to respond to footrot treatments (1) and proposed distinct treponemal aetiology (11,41), although epidemiological evidence has always pointed to strong links between these diseases (14,42). Whilst we are con dent in our observations and consider this a key paradigm in the pathogenesis of infectious foot disease, there is still a possibility that naturally occurring eld cases could deviate from those observed in this experiment. However, as we did not manually abrade or macerate tissue as described in other digital dermatitis infection models (43)(44)(45) it could be considered that the circumstances of disease induction described here are the most relevant to natural disease induction in the eld thus far.

Changes in Composition of Sheep Foot Microbiome During the Development of CODD
The foot microbiome demonstrated a clear pattern of reduced taxonomic diversity as the disease process progressed through each stage from healthy to ID, to footrot and CODD (Fig. 2), whilst beta diversity analysis showed distinct clustering of taxa associated with each stage of disease (Fig. 3). These ndings are consistent with dominance of a reduced number of bacterial taxa which are distinct for each disease stage. Exploration of the speci c taxa associated with these changes was undertaken using Gneiss analysis. The Gneiss analysis (Fig. 5), heatmap (Fig. 4), and analysis of ASVs associated with different disease states (Fig. 6 and Table 3) clearly shows a major change in the balance of microbiome composition during the disease process with Moraxellaceae, Corynebacteriaceae Psedomonaceae Saccharimonadaceae Acholeplasmataceae, Flavobacteriaceae, Ruminococcaceae Carnobacteriaceae Aerococcaceae, Family XI the predominant families in the healthy feet and a progressive shift to Porphyromonadaceae, Family XI, Bacterioida Peptostreptococcaceae Fusobacteriaceae Lachnospiraceae, Wohlfahrtiimonadaceae, Ruminococcaceae Veillonellaceae, Acidaminococcaceae Bacteria belonging to the taxa Porphyromonadaceae, Family XI, Veillonellaceae and Fusobacteriaceae were most strongly associated with the diseased feet and their relative abundance increased progressively as the diseased worsened from healthy to ID, to footrot to CODD providing further evidence for signi cance in CODD pathogenesis.
Porphyromonadaceae, and Fusobacteriaceae, have all previously been associated with the pathogenesis of ovine footrot (46,47) The family Porphyromonadaceae is Gram negative anaerobic family of bacteria which form part of the microbiota of the human and animal gastrointestinal tract and oral cavity. However, some species are considered to be pathogens capable of causing disease in animals and humans. Bacteria from this family have been previously cultured from ovine footrot lesions (48) associated with footrot in recent ovine foot disease microbiota studies (46,47) where it is hypothesised that they may stimulate a change in composition of the foot microbiome as part of the aetiology of footrot. This family of bacteria have also been found in bovine foot diseases where polymicrobial aetiologies are proposed. Microbiome studies of bovine digital dermatitis (49), and claw horn and interdigital lesions (50) have identi ed associations between the Porphyromonadaceae family and diseased state, whilst in humans, Porphyromonas gingivalis causes severe periodontal disease (51).
The Fusobacteriaceae family are also Gram negative, anaerobic, non-motile and fermentative. And as with the Porphyromonadaceae family, they are found in the oral and gastro-intestinal tract intestinal tracts of animals and humans (52). Members of this family cause a wide variety of diseases. Fusobacterium necrophorum is particularly associated with necrotic infections of the respiratory and gastrointestinal systems in farm animals and has long been associated with ovine footrot in where it is now considered to secondarily invade and worsen the severity of early footrot lesions caused by the primary aetiological agent, Dichelobacter nodosus (53). It is also considered to be the major aetiological agent in bovine footrot/foul in the foot together with Bacteroides melaninogenicus (54).
The other two families signi cantly associated with the diseased state in this study were Family XI, (a family in Clostridiales also known as Clostridium cluster XI) and Veillonellaceae. Veillonellaceae have been identi ed in previous microbiota sheep footrot studies ( (46). The Veilonellaceae are Gram-negative, anaerobic, or microaerophilic cocci and coccobacilli which can act as opportunistic pathogens in animals and humans, and are usually found in polymicrobial diseases (55). Clostridiales are Gram positive spore forming anaerobes found in the soil and the gastrointestinal tract of ruminants and can be found in wounds and abscesses. Members of this family are commonly cause fatal speci c disease syndromes in ruminants, however, a role for them in infectious foot disease as never previously been suggested (56).
It was surprising that although present in diseased feet samples, bacteria from the family Spirochetaceae were not statistically associated with the diseased foot samples (Table 3), (as collected by swabbing in the experimental study) especially as there is strong evidence (11,12,57) for several species of this family, Treponema medium phylogroup, Treponema phagedenis phylogroup and Treponema pedis, to be associated with CODD (11). Indeed, it is not uncommon to nd bacteria from the treponeme genus in footrot lesions (40,47,58). However, the Spirochetaceae family were highly abundant in the CODD biopsy samples collected in the cross-sectional farm study ( Fig. 14 and Table 6), suggesting that that this family of bacteria are more readily found in deeper rather than super cial tissues and that biopsy of lesions may give a more representative picture of the dysbiosis of the microbiome associated with CODD and quite possibly other foot diseases as well. However, this invasive sampling method is only suitable for crosssectional studies in sheep and therefore can provide only limited evidence for causality.
Pathogens previously identi ed by bacterial culture and PCR studies, as important in the pathogenesis of ID, footrot and CODD were present in Diseased foot samples. One ASV assigned to Dichelobacter nodosus was classi ed as an Intermediate ASV and was found mainly in ID samples. ASVs assigned to Fusobacterium necrophorum were classi ed as Diseased ASVs and were found mainly in ID and footrot samples. These ndings are consistent with current understanding of the role of these bacteria in the pathogenesis of footrot whereby initial colonisation by D. nodosus is followed by secondary invasion F. necrophorum by (53). Six ASVs were identi ed as Treponema with all six classi ed as Diseased ASVs which were present in ID and footrot samples. Three of these Treponema sequences were identical to those of strains from the DD associated T. medium phylogroup, T. phagedenis phylogroup and T. pedis ((11, 13). Whilst a further three treponeme associated sequences are potentially also more diverse strains of T. medium based on current treponemal species 97% cut off (59) although caution must be applied when assigning species taxonomic classi cations when using such a short region of the 16S rRNA gene.
It is interesting to note given these previously DD-associated species are present on the surface in footrot and might be considered invasive within biopsies during the CODD stage that we are were observing progressive invasion starting with footrot. Indeed, future studies might bene t from attempting to differentiate whether CODD-associated treponeme presence on skin surface in footrot can predict progression to CODD development subsequently.
Changes in the Foot Microbiome following Treatment of CODD All sheep affected with CODD were treated simultaneously with two doses of 10 mg/kg long acting amoxicillin (Betamox LA 150 mg/ml, Norbrook UK) given intramuscularly 48 hours apart. The treatment regime chosen was based on laboratory (60) and clinical data (42) on the e cacy of amoxicllin against treponeme bacteria in CODD and FR. Two weeks post treatment all treated sheep had recovered from CODD and their lesion stage was recorded as CODD grade 5. The alpha diversity, beta diversity metrics and Gneiss analysis (of composition) (Figs. 7,8,9 and 10) demonstrate clearly that the microbiome of the treated sheep resembles very closely that of the healthy state of the foot at the start of the study, which is remarkable given the profound clinical and microbiological changes the feet had experienced. The principle differences between the healthy feet at the beginning compared to the post treatment feet the end of the study were in the relative abundances of families Ruminococcaceae, Lachnospiraceae, Carnobacteriaceae, Staphylococcaceae, Pseudomonadaceae Moraxellaceae, Saccharinonadaceae, Micrococcaceae, Microbacteriaceae. Although Ruminococcaceae, Lachnospiraceae and Moraxellaceae are found in diseased microbiomes (Table 3) none of them were identi ed as being signi cantly associated with the diseased feet microbiomes. Therefore, they may not necessarily represent persistence of pathogenic organisms in the treated feet.

Study Limitations
As with any study, the effects of sampling bias and bias induced by differences in laboratory methods must be considered when interpreting results and comparing with other studies.
During the sample collection stage, differences in microbiome data obtained could occur as a result of animal level variation or variation due to the sampling method. Both these sources of variation were explored in the study. We observed no difference in diversity or relative abundance of bacteria in the microbiota of sheep's feet that did and did not go on to develop CODD when sampled on entry to the study (Figs. 11,12 and13). Importantly, as well as eliminating this as a source of bias in the results, this nding also suggests that the composition of the foot microbiota does not appear to be protective or predictive for development of CODD.
We did nd important differences in the microbiota of CODD lesions depending on whether the sampling was done by lesion biopsy or swabbing ( Fig. 14 and Table 6). However, it is also important to note that the populations of sheep used here for these two sample sets were different. Nonetheless the bacterial families present were broadly similar, with 11/15 families shared by both. Of particular note were the presence of families identi ed in this analysis study as signi cantly associated with the disease state; Porphyromonadaceae, Familiy XI, and Fusobacteriaceae (Table 3) which provides further evidence for the role in the pathogenesis of CODD. The nding that the Spirochaetaceae family (which have previously been strongly associated with CODD lesions), were more abundant in the biopsy samples, suggests that these bacteria maybe found only in deeper tissues of the disease foot and their detection is strongly in uenced by sampling method. The biopsy method can only be used for single sampling of animals and not repeat sampling as required in this longitudinal study.
Other sources of bias to consider when comparing studies are the different DNA extractions methods used between studies. The method chosen here was selected to ensure extraction of DNA from treponeme bacteria. However, it is acknowledged that it is different from previous sheep foot microbiome studies. (46,47). The V4 region of the bacterial 16S rRNA gene was chosen as it is a well-established methodology for investigating the bacterial diversity using the widely available MiSeq Illumina platform (26). However, the use of a greater proportion of the 16S rRNA gene would be preferred as it would allow better resolution of taxa to the species level, especially given the near entire 16S rRNA gene is now widely established as the key taxonomic tool for species designation and is now approaching achievability in gene targeted metagenomics studies (61).

Conclusions
This is the rst study to report the phenotypic lesion presentation and microbiological changes in sheep's feet during the development of CODD lesions. The principle ndings of the study were:-1. A novel pathogenesis of CODD is proposed with three distinct phenotypic lesion stages were observed in the development of CODD lesions. The initial lesion was ID which progressed to a clinical footrot lesion which developed into a CODD lesion. Therefore, better control of the initial ID and footrot lesions on farms should be prioritized to enable better CODD control.
2. Distinct microbiota were observed for each lesion stage in terms of microbial diversity, clustering and composition. Porphyromonadaceae Familiy XI, Veillonellaceae and Fusobacteriaceae were signi cantly associated with the diseased feet. Veillonellaceae and Fusobacteriaceae was most associated with earlier stages of CODD lesion development, namely ID and footrot stages.
3. Following treatment, the microbiota of CODD affected feet showed a strong tendency to return to the healthy state. 4. The composition of the microbiota of healthy feet did not in uence the development of CODD.
5. The sampling method affected the composition of the microbiota detected in CODD lesions.
However, Porphyromonadaceae Family XI, and Fusobacteriaceae families were highly abundant in both sample types providing further evidence for the role in CODD lesion development. The Spirochaetaceae family, although present in both, were more abundant in the biopsy samples, suggests that these bacteria maybe found more readily in deeper tissues of the disease foot.
6. It is possible using the method described to induce CODD experimentally in previously healthy sheep.              Box and whisker plots of alpha diversity as measured by observed ASVs for samples taken from healthy (A_Healthy), CODD affected (B_CODD), antibiotic treated (C_Treated) sheep's feet.   Gneiss analysis Log ratio balances whose log ratio balance was signi cantly different between microbiomes of samples taken from healthy (A_Healthy), CODD affected (B_CODD) and antibiotic treated (C_Treated) sheep's feet. A lower log ratio shows a shift in the balance toward denominator ASVs whilst a higher log ratio shows a shift towards numerator ASVs.

Figure 11
Box and whisker plot of alpha diversity as measured by observed ASVs for samples taken from healthy sheep's feet that did (case) or did not (control) develop CODD.

Figure 12
Principle Coordinate Analysis (PCoA) plot showing differences in weighted UniFrac distances of samples taken from healthy sheep's feet that did (case) or did not (control) develop CODD in the study.

Figure 13
Dendogram heatmap showing log abundance of ASV in the foot microbiota of samples taken from healthy sheep's feet that did (case) or did not (control) develop CODD in the study.