Healthy ovine feet were collected from sheep at a local abattoir post-slaughter. Feet with good hoof conformation and with apparently healthy interdigital skin were selected. Environmental contaminants were removed from the foot and interdigital space using 70% ethanol and antimicrobial skin cleanser (Hibiscrub®) before removal of the hair using scissors. The entire ovine interdigital skin was removed using a sterile scalpel. Skin explants were immediately immersed in transport media after collection [DMEM-HAM'S F12 1:1 (Sigma Aldrich®), Penicillin+Streptomycin 0.01 μg/ml (Gibco®), Amphotericin B 0.01 μg/ml (Lonza®), Gentamicin 5 μg/ml (Sigma Aldrich®), L-glutamine 0.01 μg/ml (Gibco®)]. Biopsies were collected aseptically from each interdigital skin fragment using an 8 mm punch biopsy (National Veterinary Service, UK). The surface of the biopsy was gently incised 4 times with a scalpel blade to simulate naturally occurring damage to the skin (required for the invasion of D. nodosus).

Biopsies were immersed in wash medium [DMEM-HAM'S F12 1:1, Penicillin+Streptomycin 0.01 μg/ml (Gibco®), Amphotericin B 0.01 μg/ml (Lonza®), Gentamicin 2 μg/mL (Sigma Aldrich®)], pre-heated to 37°C, and incubated at room temperature for 15 min. The media change and incubation steps were repeated 3 times.

To provide a physical support for the biopsies, an agarose pedestal (500 μl of 1.2% w/v agarose (Thermo Scientific®) in DMEM-HAM'S F12 1:1 (Sigma-Aldrich®) covered with 0.5 cm² sterile surgical gauze was assembled in a 12 well-plate (Thermo Scientific®). The skin biopsy was placed on the top of the surgical gauze and agarose pedestal (Figure 1B).

For the initial assessment of model viability, five time points were evaluated for up to 76 h of mock infection: biopsy collection −8 h, mock infection 0, 28, 52, and 76 h post mock infection. Two biopsies were collected at each time point; one placed into RNAlater (Sigma-Aldrich®) for DNA extraction and the other one was fixed in 10% v/v Neutral Buffer Formalin (NBF) for histological investigation. The model viability was analyzed assessing at each time point the tissue integrity and architecture (scoring of 2 H&E stained sections), and cell viability by TUNEL stain (5 non-overlapping images for each, epidermis and dermis).

Four experiments using explants from different sheep collected in different occasions were performed. A microenvironment with restricted oxygen supply was produced by placing a sterile agarose plug (300 μl of 0.8% w/v agarose in DMEM-HAM'S F12 1:1) on the top of the biopsy, which was placed on the top of the agarose pedestal and surgical gauze, and incubated in a humidified tissue incubator at 37°C (Heracell 150i, Thermo Fisher Scientific®) with 5% CO₂ for 4 h. 1 mL of culture media without antibiotics [DMEM high glucose (Gibco®) + DMEM HAM'S F12 1:1 (Sigma Aldrich®) in a proportion 3:1, Amphotericin B 0.25 μg/ml (Lonza®), L-glutamine 0.01 μg/ml (Gibco®), 10% Fetal Bovine Serum (Gibco®)] was placed in each well. The infection of the explants with D. nodosus was performed using an 8 mm punch biopsy to aseptically collect a Fastidious Anaerobic Agar plug confluent with D. nodosus strain MM261 (aprV2, Clade I, which correlates with virulent strains) or strain MM277 (aprB2, Clade II, which correlates with benign strains). The agar plug with D. nodosus was placed on the top of the biopsy and 1 mL of culture media without antibiotics was added in each well. The plate was incubated anaerobically (Oxoid AnaeroGen, Thermo Scientific®) at 37°C for 4 h. Post 4 h, maintenance of the model included culture media containing antibiotics [DMEM High Glucose (Gibco®) + DMEM HAM'S F12 1:1 (Sigma Aldrich®) in a proportion 3:1, Penicillin+Streptomycin 0.01 μg/ml (Gibco®), Gentamicin 5 μg/ml (Sigma Aldrich®), Amphotericin B 0.25 μg/ml (Lonza®), L-glutamine 0.01 μg/ml (Gibco®), 10% v/v Fetal Bovine Serum (Gibco®)] in a humidified tissue incubator with 5% C0₂for 24 h. The media were changed 4 h post infection and then every 12 h keeping the same agar plug with D. nodosus placed on the top of the biopsy. Culture media from all time points were collected and stored at −80°C for further cytokine measurement. After 28 h of infection with D. nodosus, biopsies were placed into RNAlater for further DNA extraction or 10% v/v NBF for histology. For DNA extraction each tissue was cut into approximately 4 × 4 mm pieces and incubated with 180 μl of tissue lysis buffer (ATL) with the addition of 20 μl of proteinase K (20 mg/ml) (Qiagen) at 56°C for 3 h. DNA was isolated using the QIAamp cador kit according to manufacturer's recommendations and eluted in 50 μl AVE buffer (Qiagen). The DNA was used for D. nodosus quantification by qPCR, targeting the 16S rRNA gene (Forward primer: 5′-CGGGGTTATGTAGCTTGCTATG-3′, Reverse primer: 5′-TACGTTGTCCCCCACCATAA-3′, probe: 5′FAM-TGGCGGACGGGTGAGTAATATATAGGAATC-TAMRA-3′) (Frosth et al., 2012). qPCR data were normalized to pg of D. nodosus DNA present in the total DNA extracted from each biopsy. The cut off of 0.1pg was assigned for negative qPCR results. The D. nodosus load present on an 8 mm agar area was estimated using the total DNA extracted from D. nodosus confluent on the 8 mm surface and confirmed by qPCR.

Interdigital skin biopsies were fixed for 48 h at 4°C using an extended tissue processing protocol (60 min in dH₂O, 4 h in 50% ethanol, 4 h in 70% ethanol, 16 h in 90% ethanol, 4 h in 100% ethanol, 4 h in xylene, 2 h of wax immersion at 60°C). Paraffin wax embedded tissues were soaked in 10% (v/v) ammoniated water while 3 and 6 μm thick sections were cut from each block by microtome (Leica RM2255®) and placed on polystyrene microscope slides (Thermo Scientific®). H&E stain (Sigma-Aldrich®) was used to allow visual assessment of the 6 μm tissue sections (see Supplementary Table 1 for H&E protocol). All slides were mounted with Distyrene Plasticizer Xylene (DPX) (Sigma-Aldrich®) and were analyzed by microscopy (Leica DM 2500®). A board certified veterinary pathologist, blinded to slide identity, evaluated 54 H&E-stained biopsy sections to assess tissue integrity and architecture. A qualitative, semi-quantitative scoring system was developed to grade the tissue integrity and architecture according to histopathological features (Table 1 and Supplementary Figure 2).

The modified DeadEnd Colorimetric TUNEL System (Promega®) was used to detect DNA fragmentation. Assays were performed according to the manufacturer's instructions using 6 μm tissue sections. Cells were stripped of proteins and made permeable by incubation with 20 μg/ml of proteinase K solution for 20 min and DAB 20x chromogen solution incubation was performed for 3 min. The sections were counterstained with haematoxylin solution for 5 s, rinsed in tap and deionized water. All slides were mounted with DPX. Negative controls were obtained by omitting the TdT enzyme from the TdT mix in the reaction (according to manufacturer's instructions). Positive controls were generated by treating a skin section with DNAse (Promega®). In order to assess the number of dead and live cells, 5 non-overlapping images from each, the epidermis and the dermis of each biopsy tissue section, were used (200x total magnification). Images were captured and analyzed by microscopy (CTR500 microscope, Leica Microsystems®). All cells from all images (epidermis and dermis) from each sample were counted using Fiji/ImageJ-win64 software (https://fiji.sc/, October 2016) and the percentage of dead and live cells were calculated for all 10 images/sample based on the total number of dead and live cells obtained. Brown stained nuclei were assessed as dead cells, whereas blue stained nuclei were assessed as live cells due to counterstain with haematoxylin.

FISH analysis was performed as previously described (Rasmussen et al., 2012). The 16S rRNA-targeting oligonucleotide probes for D. nodosus (S-S-D.nodosus-443 5′-CATGCACCGTTCTTCACT-3') and eubacteria domain (S-D-eub-338-alexa 5′-GTCATTCCATCGAAACATA-3′) labeled with fluorescein isothiocyanate (FITC) or Cy3 were used on 3 μm tissue sections. Hybridization was carried out at 46°C with hybridization buffer (100 mM Tris, pH 7.2, 0.9 M NaCl, 0.1% sodium dodecyl sulfate) containing 5 ng/ml of each applied oligonucleotide probe (usually double hybridizations with a FITC and CY3 labeled probe) for 16 h in a Sequenza slide rack (Thermo Shandon, Cheshire, United Kingdom). Slides were then washed with prewarmed (46°C) hybridisation buffer for 3 × 3 min followed by wash with prewarmed (46°C) washing buffer (100 mM Tris, pH 7.2, 0.9 M NaCl) for 3 × 3 min. Finally they were rinsed in water, air dried and mounted with Vectashield (Vector Laboratories Inc.®) for epifluorescense microscopy using an Axioimager M1 epifluorescense microscope. Images were obtained using an AxioCAM MRm version 3 FireWiremonochrome camera (Carl Zeiss®).

IL1β and CXCL8 were measured in the tissue culture media collected from −4 to 0 h, 0 to 4 h, 4 to 16 h, and 16 to 28 h post infection, using specific sandwich ELISAs as previously described (Doull et al., 2015). Recombinant ovine IL1β (Kingfisher®) and recombinant ovine CXCL8 (Doull et al., 2015) were used as quantifiable ELISA standards. Cytokine concentrations were calculated using a polynomial quadratic regression in GraphPad Prism version 7b.

One-way analysis of variance followed by Dunn's multiple comparisons test was applied for D. nodosus load as well as for percentage of TUNEL positive cells. Mann Whitney (non-parametric test) was used for IL1β release between infected and mock-infected biopsies across the time course. Mean, media, standard deviation (SD) and all other analysis were performed using GraphPad Prism version 7b. A P ≥ 0.05 was considered significant.

The assembled 3D skin infection model consisted of an agarose pedestal overlaid with surgical gauze to provide support for the biopsy in the well of a 12 well-plate (Figure 1B). The skin biopsy was then placed upon the surgical gauze. The skin explants were initially maintained during 4 h under atmospheric oxygen, but the biopsy surface area was sealed with a sterile agar plug to simulate the natural ovine feet microenvironment with a moist surface and restricted oxygen supply (Figures 1A,B). For the infection experiment, the sterile agar plug was replaced by an agar plug with confluent D. nodosus placed on the top of the biopsy and incubated under anaerobiosis for 4 h, followed by 24 h of aerobic incubation to ensure skin tissue survival (Figure 1A). For the mock-infection viability experiment, the sterile agar plug was replaced by another sterile agar plug and the model was maintained under the same temperature and incubation conditions as described above. Up to 1 ml of medium was added up to the top level of the explant to avoid flooding of the explant or removal of the agar plug.

A 76 h time course mock-infection experiment was performed to identify optimal conditions for oxygen dependent tissue to remain viable, while also allowing transient anaerobic incubation essential for the pathogen (Supplementary Figure 1A). The viability of explants was assessed by tissue integrity and cell death at five time points: immediately after sample collection −8 h, mock-infection 0, 28 h, 52 and 76 h post mock infection. Using the scoring system developed to grade the tissue integrity and architecture according to histopathological features (Table 1 and Supplementary Figure 2), it was visualized that skin explants maintained the normal tissue architecture for up to 28 h of incubation with early signs of tissue degeneration in biopsies incubated for more than 52 h (Supplementary Figure 1B). The early signs of tissue degeneration included basal cell vacuolisation, subepidermal clefting, sloughed cells in glands and follicular epithelial cells (Supplementary Figure 1B). Additionally, cell viability was investigated through the quantification of DNA fragmentation using the TUNEL stain. The percentage of live cells in the epidermis decreased from −8 h [99% (2497/2524)) to 76 h of incubation (45% (296/660)] (Supplementary Figures 1C, 3). A high percentage of live cells was found in the dermis in all time points, with a slight decrease from −8 h [99.3% (2092/2105)] to 76 h [84% (952/1131)] (Supplementary Figure 1C). As expected, biopsies fixed at the abattoir post slaughter had a low level of dead cells. The cell viability data confirmed the tissue integrity data as skin explants presented a high percentage of live cells (TUNEL) and well preserved tissue architecture (H&E tissue integrity scores) in both epidermis and dermis (Supplementary Figure 1C). As the 28 h incubation showed more than 50% of the epidermal and dermal cells were alive and the tissue architecture was viable (score 1.5) (Supplementary Figure 1C), this time point was used as the implemented cut-off for further infection experiments.

The D. nodosus of the 8 mm agar plug was quantified by qPCR to estimate the copy number used to infect each biopsy (1.13 × 10⁴ of aprV2 D. nodosus DNA, strain MM261, and 7.16 × 10⁴ of aprB2 D. nodosus DNA, strain MM277). The presence of D. nodosus DNA was confirmed in all infected biopsies with aprV2 (2.13 × 10⁴ D. nodosus/biopsy, n = 6) and aprB2 D. nodosus (1.11 × 10⁵ D. nodosus/biopsy, n = 6) using qPCR. In contrast, all mock-infected biopsies (n = 8) were negative for D. nodosus (Figure 2A).

Fluorescent in situ hybridization (FISH) was used to localize D. nodosus within the skin explant. D. nodosus bacterial cells were identified in the positive porcine lung tissue control (Figure 2B) and in all biopsies infected with either aprV2 (n = 4) or aprB2 D. nodosus (n = 4; Figures 2D–G). Negative tissue control (n = 1; Figure 2C) and all mock-infected skin control biopsies (n = 5) were negative for D. nodosus (Table 2). D. nodosus was identified on the skin surface (Figures 2D–E) and also invading the superficial layers of the epidermis (Figures 2F,G). D. nodosus cells were visualized throughout these layers suggesting that the bacteria were moving away from the initial skin laceration site. D. nodosus was not identified in the dermis. A general eubacterial domain probe did not detect the presence of other bacteria in any infected or mock-infected explants. This showed that both aprV2 and aprB2 strains of D. nodosus have the ability to migrate into the damaged skin layers during 28 h of incubation.

The tissue integrity and cell viability from infected and mock-infected skin explants were assessed to investigate the histological effects of bacterial infection on the tissue viability and maintenance. D. nodosus infection had little impact on tissue integrity/architecture or on cell viability after 28 h (Figure 3). Tissue viability of all mock-infected control biopsies was maintained over 28 h of incubation (median histopathology score 2.5), with the exception of one biopsy (score 1.5) (Figure 3A; See Table 1 for description of histopathology scores). Similarly, infected biopsies maintained the tissue integrity (median histopathology score 2), with the exception of two biopsies (scores 1 and 1.5) (Figure 3A). All mock-infected controls and infected explants presented more than 50% (69–98.9%) of epidermal and dermal cells alive (TUNEL stain) after 28 h of bacterial exposure, with the exception of one biopsy infected with the aprV2 D. nodosus strain (Figure 3B). There was no statistically significant difference between infected and mock-infected skin explants for the percentage of live cells.

Skin explants infected with aprV2 and aprB2 D. nodosus released IL1β in the culture media, which accumulated between 4–16 h and 16–28 h post infection (Figure 4A). At 4–16 h, the concentration of IL1β in infected biopsies (600 ± 492 IL1β pg/mL, n = 8) was statistically significantly greater than in mock-infected controls (10 ± 20 IL1β pg/mL, n = 4) (P ≤ 0.01). All media samples from mock-infected and infected biopsies collected at −4 h mock infection and 4 h after infection (4 h of accumulation) were negative or showed very low concentrations of IL1β (Figure 4A). Mock-infected controls and infected skin explants released similar concentrations of CXCL8 in the culture media that were accumulated between 0–4 h, 4–16 h and 16–28h post infection (Figure 4B). CXCL8 was not released in the culture media between −4 and 0 h of mock infection time (Figure 4B).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.